Parks and People: Spatial and Social Equity Inquiry in Shanghai, China

Abstract

Urban parks are essential public resources that contribute significantly to residents’ well-being. However, disparities in the spatial distribution and social benefits of urban parks remain a pressing issue. This study focuses on the central urban area of Shanghai, a representative high-density megacity, and its findings hold significant reference value for similar cities, systematically evaluating urban park services from the perspectives of accessibility, spatial equity, and social equity. Leveraging multi-source big data and enhanced analytical methods, this study examines disparities and spatial mismatches in park services. By incorporating dynamic data, such as actual visitor attendance and residents’ travel preferences, and improving analytical models, such as an enhanced Gaussian two-step floating catchment area method and spatial lag regression models, this research significantly improves the accuracy and reliability of its findings. Key findings include (1) significant variations in accessibility exist across different types of parks, with regional and city parks offering better accessibility compared to pocket parks and community parks. (2) Park resources are unevenly distributed, with neighborhoods within the inner ring exhibiting relatively low overall accessibility. (3) A spatial mismatch is observed between park accessibility and housing prices, highlighting equity concerns. The dual spatial-social imbalance phenomenon reveals the prevalent contradiction in rapidly urbanizing areas where public service provision lags behind land development. Based on these results, this study proposes targeted recommendations for optimizing urban park layouts, including increasing the supply of small parks in inner-ring areas, enhancing the multifunctionality of parks, and strengthening policy support for disadvantaged communities. These findings contribute new theoretical insights into urban park equity and fine-grained governance while offering valuable references for urban planning and policymaking.

1. Introduction

Urban parks, as vital public service resources, not only provide urban residents with spaces for daily leisure and social interaction but also serve as a critical lever for achieving social equity, enhancing living environments, and fostering sustainable urban development. Adequate accessibility to urban parks is increasingly recognized as a fundamental criterion for a livable urban environment [1]. However, with the rapid pace of urbanization, the planning and development of parks have exhibited significant imbalances, resulting in spatial and social inequities in park access and usage. Some areas concentrate park resources, while others face insufficient provision; at the same time, there are notable disparities in park accessibility among residents of different socio-economic classes. These issues have become urgent challenges that need to be addressed in urban renewal. Such imbalanced development is closely related to spatial planning policies during China’s urban transformation. Since the reform and opening-up, park accessibility has been a key goal in urban planning [1]. With nationwide land use reforms and the development of a market economy, large cities such as Shanghai and Hangzhou have prioritized park development in their central urban areas. Although this policy has significantly improved green space infrastructure in the central districts, it has also contributed to inequalities in park resources in suburban and other areas to some extent [2,3,4].

Existing research commonly divides the fairness of urban park recreational services into two dimensions: spatial equity and social equity [5,6]. Spatial equity primarily focuses on the balance of park distribution and service coverage across regions [7], for example, in the central urban area of Shanghai, where parks are numerous and highly accessible, while suburban park resources are relatively scarce [1]. Social equity, on the other hand, reflects the disparities in park access among different demographic groups [8,9]. Vulnerable groups, such as low-income residents and the elderly, often face disadvantages in terms of park resource availability [9,10], whereas higher-income groups and residents in high-priced housing areas typically enjoy greater convenience [11]. This dual inequity in both spatial and social dimensions presents new challenges for urban planning.

Research on urban park equity primarily unfolds through three analytical dimensions: (1) Social stratification: Linking housing attributes to socioeconomic status, studies reveal unequal park distribution, with housing prices positively correlating to accessibility. High-income groups often secure better-located parks through housing markets [12,13]. (2) Spatial efficiency: Integrating road density, walkability, and public transport metrics, research identifies uneven park development in high-density areas and regional disparities [14,15]. (3) Supply–demand justice: Matching green space supply (coverage, size) to vulnerable groups’ needs, findings show higher park accessibility in city centers but lower equity in disadvantaged neighborhoods [9,16]. Current studies are transitioning from static indicators to multi-source data integration, reflecting the inevitable trend toward efficient composite utilization and refined development of green ecological spaces in high-density urban areas [9].

In recent years, with the development of multi-source data technologies and spatial analysis methods, researchers have gradually introduced big data and quantitative evaluation techniques to enhance the precision and scientific rigor of studies on urban park equity [17,18,19]. The traditional two-step floating catchment area method (2SFCA) has been widely used to analyze park accessibility, effectively integrating the supply–demand relationship [20]. However, this method does not adequately account for the diminishing returns of travel costs with increasing distance, which limits its effectiveness in equity studies. The enhanced Gaussian two-step floating catchment area method (EG2SFCA), by incorporating a decay function, can more accurately reflect accessibility, although the improved indicator system still leaves room for further optimization in empirical research. Moreover, studies on park equity often focus on a single dimension (such as spatial distribution or demographic attributes), lacking a comprehensive framework for evaluating both spatial and social factors [21,22,23].

Based on the current state of research, this study takes the central urban area of Shanghai as a case study and proposes an integrated urban park equity assessment framework that focuses on spatial layout and social class disparities. The specific objectives of this study are (1) to analyze the spatial distribution characteristics of park accessibility across different types of urban parks; (2) to assess the accessibility differences between various social classes, highlighting issues of social equity; (3) to propose recommendations for optimizing park layouts, providing insights for mitigating spatial and social inequities.

As a representative city of China’s urbanization, Shanghai has a high population density, complex community structure, and a diverse park system, making it an exemplary case for studying urban park equity. Furthermore, as China’s urbanization enters the stage of stock optimization, urban renewal and park development are shifting from quantity expansion to equitable distribution and quality improvement. The findings of this study will provide critical theoretical and practical guidance for achieving equitable distribution of urban park resources, while also offering valuable insights for advancing socially inclusive, ecologically sustainable urban renewal.

2. Study Area and Data

2.1. Study Area

This study selects the central urban area of Shanghai as the empirical research site for urban park equity. According to Article 31 of the Shanghai Urban Master Plan (2017–2035), the urban–rural system is structured as a hierarchical system of central city, new cities, new towns, and rural areas. The central city encompasses the city center, major urban areas, as well as regions such as Gaoqiao and Gaodong towns that are adjacent to the city center. The total area of the central city is approximately 1161 square kilometers, with a planned resident population of about 14 million [24]. The city center is defined as the area within the outer ring road, covering about 664 square kilometers, with a planned resident population of approximately 11 million (Figure 1a) [25]. The research scope of this study is the central urban area of Shanghai, which refers to the region within the city’s outer ring road, covering about 664 km². This area represents the political, economic, and cultural center of Shanghai, and its core includes Huangpu District, Xuhui District, Changning District, Yangpu District, Hongkou District, Putuo District, Jing’an District, and the inner areas of Pudong New District.

2.2. Study Object

This study is based on data from the Shanghai Greening and City Appearance Bureau, from which 254 parks in the central urban area of Shanghai were selected. After excluding three fee-based parks (Shanghai Zoo, Yuyuan Garden, and Gucun Park), a total of 251 free public parks and 9 elevated parks along the outer ring road were included, resulting in a total of 260 free and publicly accessible urban parks for this study.

According to the park classification outlined in the “Shanghai Urban Master Plan (2017–2035),” national parks and suburban parks within the urban–rural park system, which fall under the “rural” category, were excluded from the analysis. This study focuses on the urban park system within the central urban area, which includes urban parks, district parks, community parks, and pocket parks (equivalent to street green spaces). The “Shanghai Master Plan 2035” specifies the size ranges for each park type: urban parks ≥50 hectares, district parks 4–50 hectares, community parks 0.3–4 hectares, and pocket parks 0.066–1.33 hectares.

Based on these classifications, the 260 parks in the central urban area were categorized by size, resulting in the following park distribution: 11 urban parks, 67 district parks, 114 community parks, and 68 pocket parks. The spatial distribution is shown in Figure 1b, with 92 parks within the inner ring, 91 parks between the inner and middle rings, 68 parks between the middle and outer rings, and 9 parks located along the outer ring road.

2.3. Data Preparation

2.3.1. Data Type and Source

In the equity assessment of parks in the central urban area of Shanghai, three main types of data were utilized (

Table 1. Types, sources, and purposes of data used.

Data Types	Data Subclasses	Data Sources	Research Purposes
Spatial data	Park (Name, Boundary AOI, Area, Category)	Shanghai Bureau of Greening and Green Appearance	Accessibility, equity
	Administrative boundary data (Administrative District, Street, Community)	Gaode map	Accessibility
	Community (name, boundary)	Gaode map	Accessibility
	Path planning data (Walking, driving)	Gaode map API	Accessibility
	100 m × 100 m grid population data	WorldPOP	Accessibility
Park visitor data	LBS visitor data	Position location	Accessibility
	Location positioning	Position location	Accessibility
Residential community resident data	Residential housing price data	Anjuke	Equity
	Population of the community	Population Grid	Accessibility

): The first category is spatial data, which includes park names, boundaries (Area of Interest, AOI), area, and category, as well as Shanghai’s urban road network data, administrative boundary data, including the boundaries of administrative districts and streets, the names and boundaries of residential communities, population data in 100 m*100 m grids, and the shortest route network distance data from residential communities to parks, obtained through path planning via the Amap API. The second category is park visitation data, which includes location-based service (LBS) data of visitor locations within the parks, along with the residential and work locations of these visitors, and demographic profiling labels. The third category is residential data, which includes housing prices and demographic information of the residents in the communities.

2.3.2. Data Cleaning

The first step is to perform grid population adjustment. Using the 100 m × 100 m grid population data for China in 2020 provided by WorldPop, spatial correction of Shanghai’s population data was conducted based on the 7th National Population Census. Population data for each district and residential community were calculated using QGIS3.40, with the total population of the central urban area approximately 10.9 million, which aligns with planning and census data.

The second step is to filter visitor data for those who have visited the parks. Using location-based service (LBS) data from 3 January to 16 January 2022, visitors who stayed in parks for more than 10 min were selected. This resulted in data for 260 parks, with a total of 231,639 visitors and 466,163 visit records.

2.3.3. Data Reliability Verification

Population Data Verification: A paired sample T-test was performed using SPSS 28.0.0.0 to compare the 7th National Population Census data with the adjusted grid population data. The results show that the averages of both datasets are similar, with a T-value of −0.636 and a p-value of 0.267 (>0.05), indicating no significant difference and suggesting that both datasets originate from the same population. Additionally, a correlation test yielded R² = 1 and p < 0.001, demonstrating a significant correlation.

LBS Visitor Data Verification: Spearman correlation analysis in SPSS was used to test the consistency between LBS visitor data for January 2022 and the data from the Greening and City Appearance Bureau. For the 167 shared parks, the correlation coefficient was R² = 0.799 ** and p < 0.001, indicating a significant relationship between the LBS data and the Bureau’s statistics. Thus, the LBS visitor data are deemed reliable.

3. Methods

To better explore the equity of urban park accessibility, an integrated multi-source data framework for urban park equity assessment is proposed (Figure 2). Real multi-source big data are utilized to reflect the usage of urban parks, and the improved Gaussian two-step floating catchment area method (EG2SFCA) is applied to analyze the impact of park supply and demand, as well as distance, on accessibility. Based on the accessibility results, spatial autocorrelation is used to measure spatial data, evaluating the spatial equity of community park accessibility across different regions. Spatial autocorrelation analysis of housing prices and accessibility is also conducted, and a spatial lag regression model is employed for cluster analysis to explore the inequitable relationship between accessibility and socio-economic status. This approach provides a reference for high-precision urban park equity assessment.

3.1. Accessibility Measurement

3.1.1. Calculation of Actual Service Radius Based on Residents’ Real Travel Intentions

In the two-step floating catchment area (2SFCA) method, the selection of a spatial distance threshold d0 for the search area is a key factor influencing the supply–demand calculation. The setting of this threshold d0 can be based on distance or time [11], typically determined by the park service radius [3] or the actual travel intentions of residents [26]. For the accessibility assessment of different types of urban parks in this study, the actual service radius of parks (based on the network distance of residents’ actual travel) was used as the basis for setting d0.

Data on the actual travel distances of park visitors within 14 days in the central urban area of Shanghai were collected (totaling 282,665 records). These data documented the actual network distances between residents’ homes and target parks. Using Python 3.11.0 and the Amap API, the shortest network distance for each “residence-to-park” pair was calculated. The cumulative distribution function (CDF) was then applied to calculate the cumulative probability of residents’ travel distances (as shown in Formula (1)) (Figure 3). The 75% or 80% cumulative probability points were used as reference values for the service radius [26,27], resulting in the following average distance thresholds for the four types of parks: pocket parks (5521 m), community parks (5762 m), district parks (7290 m), and urban parks (9622 m). These thresholds were set as the actual service radius d0 for different types of parks.

$${\mathrm{F}}_{\_}\left(\mathrm{x}\right)=\mathrm{P}(\mathrm{X}\le \mathrm{x})$$

(1)

where x represents the distance variable, and F(x) represents the cumulative probability of the variable X being less than or equal to distance x.

According to the “Shanghai Urban Master Plan (2017–2035)” and the “Technical Guidelines for Pocket Park Construction in Shanghai” (published in 2021), the planned service radii for pocket parks, community parks, district parks, and urban parks are 500 m, 1000 m, 2000 m, and 5000 m, respectively. However, a comparison between the actual and planned service radii reveals that the actual service radii are generally larger than the planned radii, which is consistent with existing research findings [17,26,27]. This study indicates that the coverage of parks in actual use often exceeds the range set by planning, with the service areas being more extensive.

3.1.2. Enhanced Gaussian-Based Two-Step Floating Catchment Area (EG2SFCA)

This study aims to adopt the Enhanced Gaussian-based Two-Step Floating Catchment Area (EG2SFCA) method to calculate supply–demand accessibility, as it comprehensively considers the impact of park supply scale, population demand scale, and distance on accessibility. The traditional Two-Step Floating Catchment Area (G2SFCA) method only calculates the supply–demand ratio for the target area. In contrast, the Gaussian-based Two-Step Floating Catchment Area method further weights the influence of neighboring areas using a Gaussian function [28,29], where the impact of neighboring areas gradually decreases with distance. The Gaussian function is a bell-shaped curve that establishes a mathematical relationship between distance and the degree of influence [29]. Its rate of decay follows a pattern of slow decrease, rapid drop, and then gradual flattening—experiencing a slow drop, followed by a sharp decline, and finally leveling off. Without the Gaussian function, distance has no impact on the supply–demand ratio, meaning the willingness to use urban parks is not influenced by distance. However, after applying Gaussian decay, the closer individuals are to the park, the higher the number of visitors, resulting in a smaller supply–demand ratio (${\mathrm{R}}_{\mathrm{j}}$), indicating a greater supply shortage and poorer accessibility [20]. This method more accurately reflects the supply–demand relationship and accessibility of urban parks [16], offering more reliable data support for assessing park supply–demand conditions.

However, the Gaussian-based Two-Step Floating Catchment Area method has certain limitations, such as not accounting for park type and attractiveness, and the fixed search radius is ${\mathrm{d}}_0$. Therefore, this study summarizes potential improvements to the G2SFCA method in terms of supply indicators, demand indicators, distance indicators, and search radius, to enhance the precision of accessibility calculations (

Table 2. Ways to enhance Gaussian-based two-step floating catchment area.

Indicators	Traditional Method	Refined Method	Proposed in This Study
Supply Indicators	Population distribution grid	Population at residential community level	●
	Population at district level
	Population at street level
Demand Indicators	Park area	Classified urban parks	●
Distance Indicators	Euclidean distance	Actual road network distance	●
		Travel time
		Path planning API	●
		Introduction of distance decay function	●
Search Radius	Fixed radius	Hierarchical/classified thresholds	●

).

This study utilizes the enhanced Gaussian-based Two-Step Floating Catchment Area (EG2SFCA) method to calculate supply–demand accessibility. Compared to traditional methods, EG2SFCA comprehensively considers the park’s supply scale, the population’s demand scale, and the influence of distance on accessibility, thus providing a more realistic reflection of the supply–demand relationship in urban parks. The traditional G2SFCA method primarily assesses accessibility by calculating the supply–demand ratio of the target area, without sufficiently accounting for the impact of distance. The EG2SFCA method, however, introduces a Gaussian function to weight the influence of neighboring areas, reflecting the attenuation of proximity effects with increasing distance. The Gaussian function is a bell-shaped curve, and its decay rate follows a pattern of “slow decrease—rapid drop—gradual leveling,” meaning that proximity to the park contributes more to accessibility. Without the Gaussian function, the distance between individuals and the park is ignored, implying that people’s willingness to use the park is unaffected by distance. When Gaussian decay is applied, the number of residents closer to the park increases, reducing the supply–demand ratio (${\mathrm{R}}_{\mathrm{j}}$) and amplifying the supply–demand imbalance, resulting in poorer accessibility. EG2SFCA provides a more scientifically accurate evaluation of the supply–demand relationship and accessibility, offering more reliable data support for optimizing park layouts. However, this method also has certain limitations, such as not accounting for the attractiveness of different types of parks and the fixed search radius [30,31]. Based on these issues, this study outlines potential improvements in supply indicators, demand indicators, distance indicators, and search radii (see Table 2) to enhance the precision of accessibility calculations.

The calculation of the enhanced Gaussian-based Two-Step Floating Catchment Area (EG2SFCA) method consists of two steps:

Step 1: Calculation of the park supply–demand ratio (${\mathrm{R}}_{\mathrm{j}}$). For each park (j), a spatial distance threshold ${\mathrm{d}}_0$ is set, forming the park’s catchment area, and the supply–demand ratio is calculated. The process is as follows: For each residential community (k) within the catchment area (${\mathrm{d}}_{\mathrm{k}\mathrm{j}}$ ≤ ${\mathrm{d}}_0$), the population is weighted using the Gaussian function to estimate the park’s potential users. Then, the park’s area (${\mathrm{S}}_{\mathrm{j}}$) is divided by the number of potential users to obtain the supply–demand ratio (${\mathrm{R}}_{\mathrm{j}}$) [32]. The formula is as follows:

$${\mathrm{R}}_{\mathrm{j}}={\displaystyle \frac{{\mathrm{S}}_{\mathrm{j}}}{{\sum }_{\mathrm{k}\in \left\{{\mathrm{d}}_{\mathrm{k}\mathrm{j}}\le {\mathrm{d}}_0\right\}}\mathrm{G}\left({\mathrm{d}}_{\mathrm{k}\mathrm{j}},{\mathrm{d}}_0\right){\mathrm{P}}_{\mathrm{k}}}}$$

(2)

where k refers to the residential community within the catchment area, j refers to the park, ${\mathrm{P}}_{\mathrm{k}}$ refers to the population of residential community k, ${\mathrm{d}}_{\mathrm{k}\mathrm{j}}$ refers to the network distance from the centroid of community k to the centroid of park j, ${\mathrm{S}}_{\mathrm{j}}$ refers to the service capacity of park j (measured by area in this study), G(${\mathrm{d}}_{\mathrm{k}\mathrm{j}}$, ${\mathrm{d}}_0$) refers to the Gaussian function considering spatial decay, and ${\mathrm{R}}_{\mathrm{j}}$ refers to the supply–demand ratio, i.e., the ratio of service capacity of supply point j to the population within the search radius ${\mathrm{d}}_0$ [32]. In the calculation, to avoid errors caused by multiple entrances, the centroids of both the residential area and the park are used for distance measurement. Additionally, the ArcGIS Network Analyst Tool is used to replace Euclidean distance with actual road network distance, enhancing the accuracy of the calculation.

The Gaussian function calculation method is as follows:

$$\mathrm{G}\left({\mathrm{d}}_{\mathrm{k}\mathrm{j}},{\mathrm{d}}_0\right)=\left\{\begin{array}{cc}\hfill {\displaystyle \frac{{\mathrm{e}}^{-\left(\frac{1}{2}\right)\times (\frac{{\mathrm{d}}_{\mathrm{k}\mathrm{j}}}{{\mathrm{d}}_0}{)}^2}-{\mathrm{e}}^{-\left(\frac{1}{2}\right)}}{1-{\mathrm{e}}^{-\left(\frac{1}{2}\right)}}}, & \mathrm{if} {\mathrm{d}}_{\mathrm{k}\mathrm{j}}\le {\mathrm{d}}_0\hfill \\ \hfill 0, & \mathrm{if} {\mathrm{d}}_{\mathrm{k}\mathrm{j}}>{\mathrm{d}}_0\hfill \end{array}\right.$$

(3)

Step 2: Calculation of the park accessibility (${\mathrm{A}}_{\mathrm{i}}$) for residential communities. For each residential community (i), a search area is set, and the weighted supply–demand ratios of all parks within the catchment area are summed to calculate the park accessibility. The process is as follows: For each community (i), a spatial distance threshold ${\mathrm{d}}_0$ is set, forming the catchment area of the community. Then, the supply–demand ratios (${\mathrm{R}}_{\mathrm{j}}$) of the parks within this area are weighted and summed using the Gaussian function to obtain the park accessibility Ai for the community [32]. The unit of Ai is square meters per person, reflecting the average park area available to residents within the community’s catchment. The formula is as follows:

$${\mathrm{A}}_{\mathrm{i}}={\sum }_{\mathrm{j}\in \left\{{\mathrm{d}}_{\mathrm{i}\mathrm{j}}\le {\mathrm{d}}_0\right\}}\mathrm{G}\left({\mathrm{d}}_{\mathrm{i}\mathrm{j}},{\mathrm{d}}_0\right){\mathrm{R}}_{\mathrm{j}}$$

(4)

where ${\mathrm{R}}_{\mathrm{j}}$ refers to the supply–demand ratio of park j within the catchment area, ${\mathrm{d}}_{\mathrm{i}\mathrm{j}}$ refers to the network distance from residential community i to park j. Other variables are as described in Equation (2) [32].

3.2. Urban Park Equity Measurement Based on Accessibility Results

3.2.1. Spatial Equity

In the equity assessment of parks in the central urban area of Shanghai, three main types of data were utilized (Table 1): The first category is spatial data, which includes park names, boundaries (Area of Interest, AOI), area, and category, as well as Shanghai’s urban road network data and administrative boundary data, including the boundaries of administrative districts and streets, the names and boundaries of residential communities, population data in 100 m × 100 m grids, and the shortest route network distance data from residential communities to parks, obtained through path planning via the Amap API. The second category is park visitation data, which includes location-based service (LBS) data of visitor locations within the parks, along with the residential and work locations of these visitors, and demographic profiling labels. The third category is residential data, which includes housing prices and demographic information of the residents in the communities.

Spatial autocorrelation, as a spatial statistical method, is used to measure the degree of aggregation or dispersion of geographic phenomena. It can represent patterns of clustering, dispersion, or randomness [33]. In this study, GeoDa software is used to calculate the spatial correlation of park accessibility in residential communities, aiming to describe spatial equity. Spatial autocorrelation is typically assessed through global and local spatial autocorrelation. This study employs both the Global Univariate Moran’s I and Local Univariate Moran’s I for the analysis.

(1)

Global Univariate Moran’s I

The global Moran’s I index is used to measure the overall level of spatial aggregation of a variable [34]. It can detect whether significant spatial autocorrelation exists in spatial data and reveal the strength of spatial clustering or dispersion [35]. The value of Moran’s I ranges from [−1, 1]: positive values indicate positive correlation (clustering), negative values indicate negative correlation (dispersion), and 0 indicates a random distribution. Higher Moran’s I values indicate stronger aggregation, while lower values indicate greater dispersion. Generally, a high global univariate Moran’s I implies a high degree of aggregation, while a low index suggests a high level of dispersion [36]. The formula for the global univariate Moran’s I is as follows:

$$\mathrm{I}={\displaystyle \frac{\mathrm{N}}{\mathrm{W}}}{\displaystyle \frac{{\sum }_{\mathrm{i}=1}^{\mathrm{N}}{\sum }_{\mathrm{j}=1}^{\mathrm{N}}{\mathrm{w}}_{\mathrm{i}\mathrm{j}}({\mathrm{x}}_{\mathrm{i}}-\overline{\mathrm{x}})({\mathrm{x}}_{\mathrm{j}}-\overline{\mathrm{x}})}{{\sum }_{\mathrm{i}=1}^{\mathrm{N}}{({\mathrm{x}}_{\mathrm{i}}-\overline{\mathrm{x}})}^2}}$$

(5)

where I refers to the global univariate Moran’s I, N refers to the total number of residential communities, x refers to the independent variable, $\overline{\mathrm{x}}$ refers to the mean of x, and ${\mathrm{w}}_{\mathrm{i}\mathrm{j}}$ refers to the degree of spatial proximity between the residential community i and community j. The larger the weight value, the stronger the spatial dependence between the study units. W is the sum of all ${\mathrm{w}}_{\mathrm{i}\mathrm{j}}$ ($\mathrm{W}=\sum _{\mathrm{i}=1}^{\mathrm{N}}\sum _{\mathrm{j}=1}^{\mathrm{N}}{\mathrm{w}}_{\mathrm{i}\mathrm{j}}$). In GeoDa, this study uses distance weights (distance weight) instead of adjacency weights, which is more suitable for demand analysis based on the service radius of urban parks. A permutation test with 999 iterations is used, and the significance level is set at p-value < 0.01 to evaluate the significance of the global Moran’s I.

(2)

Local Univariate Moran’s I

The local Moran’s I index is used to evaluate the correlation of attribute values between a geographical spatial unit (e.g., residential community) and its neighboring spatial units. It can identify local spatial clustering or dispersion phenomena. The formula for the local univariate Moran’s I is as follows:

$${\mathrm{I}}_{\mathrm{i}}={\displaystyle \frac{{\mathrm{x}}_{\mathrm{i}}-\overline{\mathrm{x}}}{{\mathrm{m}}_2}}\sum _{\mathrm{j}=1}^{\mathrm{N}}{\mathrm{w}}_{\mathrm{i}\mathrm{j}}({\mathrm{x}}_{\mathrm{j}}-\overline{\mathrm{x}})$$

(6)

where ${\mathrm{I}}_{\mathrm{i}}$ refers to the local Moran’s I for residential community, $\mathrm{N}$ refers to the total number of residential communities, $\mathrm{x}$ refers to the independent variable, $\overline{\mathrm{x}}$ refers to the mean of x, ${\mathrm{m}}_2$ refers to the variance of x (${\mathrm{m}}_2={\displaystyle \frac{{\sum }_{\mathrm{i}=1}^{\mathrm{N}}{({\mathrm{x}}_{\mathrm{i}}-\overline{\mathrm{x}})}^2}{\mathrm{N}}}$), ${\mathrm{w}}_{\mathrm{i}\mathrm{j}}$ refers to the spatial proximity between communities i and j, with a higher weight indicating stronger spatial dependence between units, and $\mathrm{W}$ is the sum of all ${\mathrm{w}}_{\mathrm{i}\mathrm{j}}$ ($\mathrm{W}=\sum _{\mathrm{i}=1}^{\mathrm{N}}\sum _{\mathrm{j}=1}^{\mathrm{N}}{\mathrm{w}}_{\mathrm{i}\mathrm{j}}$).The local Moran’s I can identify regions that exhibit spatial clustering or dispersion, and it complements the global Moran’s I. The relationship between the global and local Moran’s I is as follows: $\mathrm{I}=\sum _{\mathrm{i}=1}^{\mathrm{N}}{\displaystyle \frac{{\mathrm{I}}_{\mathrm{i}}}{\mathrm{N}}}$. By analyzing both the global and local Moran’s I indices, this study is able to comprehensively assess the spatial equity of park accessibility in residential communities, providing valuable reference for the optimization of park resource allocation.

3.2.2. Social Equity

This study investigates the spatial correlation between park accessibility in residential communities and housing prices through spatial autocorrelation analysis. The aim is to evaluate whether there is spatial dependence and clustering characteristics between park accessibility and socio-economic conditions. Additionally, spatial cluster analysis of park accessibility and housing prices across different communities is conducted using GeoDa software to reveal spatial mismatches.

(1)

Global Bivariate Moran’s I Analysis

First, this study employs the Global Bivariate Moran’s I index to test the spatial correlation between park accessibility and housing prices (socio-economic levels). The Global Bivariate Moran’s I is a statistical measure of the correlation between two variables in geographic space [37]. Its value ranges from [−1, 1]: a value of 0 indicates no spatial correlation, a positive value represents a positive correlation (clustering characteristics), and a negative value indicates a negative correlation (dispersed characteristics). The closer the absolute value of the index is to 1, the stronger the degree of clustering or dispersion. This analysis aims to explore the spatial relationship between park accessibility and housing prices in residential communities and to uncover any underlying connections. If spatial autocorrelation is present, it suggests that park accessibility and housing prices in adjacent communities are spatially dependent. The formula for the Global Bivariate Moran’s I is as follows:

$${\mathrm{I}}^{\prime }={\displaystyle \frac{\mathrm{N}{\sum }_{\mathrm{i}}^{\mathrm{N}}{\sum }_{\mathrm{j}\ne \mathrm{i}}^{\mathrm{N}}{\mathrm{W}}_{\mathrm{i}\mathrm{j}}{\mathrm{Z}}_{\mathrm{i}}{\mathrm{Z}}_{\mathrm{j}}}{(\mathrm{N}-1){\sum }_{\mathrm{i}}^{\mathrm{N}}{\sum }_{\mathrm{j}\ne \mathrm{i}}^{\mathrm{N}}{\mathrm{W}}_{\mathrm{i}\mathrm{j}}}}$$

(7)

where ${\mathrm{I}}^{\prime }$ refers to the Global Bivariate Moran’s I, ${\mathrm{Z}}_{\mathrm{i}}$ refers to the standardized housing price of community i, ${\mathrm{Z}}_{\mathrm{j}}$ refers to the standardized accessibility of community j, and ${\mathrm{W}}_{\mathrm{i}\mathrm{j}}$ refers to the spatial correlation between communities i and j.

(2)

Local Bivariate Moran’s I Analysis

Next, this study employs the Local Bivariate Moran’s I index for spatial cluster analysis. The local Moran’s I index can identify the correlation between the attribute values of a specific spatial unit (e.g., a residential community) and its neighboring units, further revealing the spatial heterogeneity of local regions [38]. Compared to the global index, the local index provides more detailed spatial relationship information. In this study, the spatial clusters of park accessibility and housing prices are classified into four patterns: high–high (high housing prices–high accessibility), low–low (low housing prices–low accessibility), low–high (low housing prices–high accessibility), and high–low (high housing prices–low accessibility). The formula for the Local Bivariate Moran’s I is as follows:

$${{\mathrm{I}}^{\prime }}_{\mathrm{i}}={\mathrm{Z}}_{\mathrm{i}}{\sum }_{\mathrm{j}=1}^{\mathrm{N}}{\mathrm{W}}_{\mathrm{i}\mathrm{j}}{\mathrm{Z}}_{\mathrm{j}}$$

(8)

where ${{\mathrm{I}}^{\prime }}_{\mathrm{i}}$ refers to the Local Bivariate Moran’s I for community i, ${\mathrm{Z}}_{\mathrm{i}}$ refers to the standardized housing price of community i, ${\mathrm{Z}}_{\mathrm{j}}$ refers to the standardized accessibility of community j, ${\mathrm{W}}_{\mathrm{i}\mathrm{j}}$ refers to the spatial correlation between communities i and j.

(3)

Spatial Lag Regression Model Analysis

Considering the potential spatial mismatch and spatial dependence issues between park accessibility and housing prices, this study adopts a Spatial Lag Regression model to address the spatial autocorrelation of the dependent variable. The model incorporates a lag term (S-lag) to reflect that the dependent variable is influenced not only by its own value but also by the variables in neighboring regions [1,39]. The equation for the Spatial Lag Regression model is as follows:

$$\mathrm{Y}=\mathsf{\alpha }+\mathrm{p}\mathrm{W}\mathrm{Y}+\sum _{\mathrm{k}=1}^{\mathrm{N}}{\mathsf{\beta }}_{\mathrm{k}}{\mathrm{X}}_{\mathrm{i}}+\mathsf{\epsilon }$$

(9)

where $\mathrm{Y}$ refers to the park accessibility in the community (dependent variable), ${\mathrm{X}}_{\mathrm{i}}$ refers to the average housing price of community (i) (independent variable), ${\mathsf{\beta }}_{\mathrm{k}}$ refers to the regression coefficient for the independent variable (k), N refers to the total number of independent variables, $\mathsf{\epsilon }$ refers to the random error, p refers to the coefficient of spatial lag, $\mathsf{\alpha }$ refers to the constant, and $\mathrm{W}$ refers to the N x N spatial weight matrix.

4. Results

4.1. Accessibility Calculation Results

Based on the actual service radii of urban parks, reflecting residents’ real preferences, this study uses the enhanced Gaussian-based two-step floating catchment area (EG2SFCA) method to calculate the accessibility of four types of parks. The results show the accessibility distribution of residential communities under the actual service radii of parks (Figure 4). Accessibility is classified into five levels, with darker colors indicating higher accessibility. The specific results are as follows:

(1)

Pocket Parks: There are 68 pocket parks with an actual service radius of 5521 m, covering 9225 residential communities. As shown in Figure 3a, the accessibility of residential communities within the inner ring is significantly higher than those in the middle and outer rings. This distribution pattern reflects the concentrated layout of parks in the inner ring.

(2)

Community Parks: There are 114 community parks with an actual service radius of 5762 m, covering 9430 residential communities. From Figure 3b, it can be seen that the accessibility in the inner and middle rings is higher than in the outer ring. This is due to the larger service radius of community parks, as well as the higher number of parks in the inner and middle rings, resulting in these areas having multiple parks covering their residential communities and significantly improving accessibility.

(3)

District Parks: There are 67 district parks with an actual service radius of 7290 m, providing full coverage of 9502 residential communities. From Figure 3c, the communities with high accessibility are mainly distributed in areas where district parks are concentrated, with the accessibility pattern showing dispersion across the inner, middle, and outer rings. This suggests that the accessibility distribution of district parks is significantly influenced by their spatial layout.

(4)

Urban Parks: There are 11 urban parks with an actual service radius of 9622 m, covering all 9502 residential communities. From Figure 3d, the accessibility in the “middle ring to outer ring” region is higher than in the inner ring. This result aligns with the accessibility findings based on the planned service radius. High-accessibility communities are concentrated around urban parks, indicating that urban parks have a significant impact on the accessibility of these communities.

4.2. Spatial Equity Calculation Results

The global univariate Moran’s I index is 0.976 (p < 0.01), indicating that the accessibility of residential communities exhibits significant spatial autocorrelation overall, meaning there is a strong correlation between nearby communities. This suggests that the accessibility of communities is constrained to some extent by spatial association or mutual influence. The clustering map of the local bivariate Moran’s I index (Figure 5) further reveals the spatial clustering characteristics of the 9502 residential communities in central Shanghai. The specific analysis is as follows:

(1)

High–High Clustering (H-H Cluster): A total of 3155 communities belong to the high–high clustering type, which are concentrated in the northeastern and southwestern parts of the central urban area. The northeastern region is primarily located in Yangpu District and Pudong New Area, while the southwestern region covers parts of Xuhui District and Minhang District. High–high clustering areas typically have larger urban parks, which provide higher accessibility to nearby communities. Examples include Century Park, Shanghai Gongqing Forest Park, Qiantan Leisure Park, Shanghai Botanical Garden, Minhang Sports Park, and Minhang Cultural Park. The layout of these large parks significantly improves accessibility in these areas, creating a highly accessible spatial clustering pattern.

(2)

Low–Low Clustering (L-L Cluster): There are 4763 low–low type communities, mainly distributed in two areas. ① Inside the inner ring: Communities in Huangpu, Jing’an, and Hongkou districts, where high population density and insufficient park resources lead to a supply–demand imbalance, resulting in low accessibility clustering. ② Northwest between the inner and outer rings: Communities in Putuo, Changning, Jing’an, and Baoshan districts, where the limited number and uneven distribution of parks further exacerbate the phenomenon of low accessibility. The wide distribution and large number of low–low clustering communities reflect an imbalance between the supply and demand for park resources in the central urban area of Shanghai.

Additionally, 1584 residential communities showed no significant spatial autocorrelation, indicating that their accessibility does not exhibit clear spatial clustering patterns. This could be related to their geographic location and the complexity of the surrounding park resource distribution.

4.3. Social Equity Calculation Results

The global bivariate Moran’s I index is 0.524 (p < 0.01), indicating that there is significant spatial autocorrelation between park accessibility and housing prices across residential communities. This means that communities with similar spatial locations show some correlation, and there exists a spatial relationship or mutual influence between park accessibility and housing prices. The clustering map of the local bivariate Moran’s I index (Figure 6) further reveals the spatial clustering characteristics of the 9502 residential communities in central Shanghai. The specific analysis is as follows:

(1)

High–High Clustering (H-H Cluster): A total of 1010 communities belong to the high–high clustering type, primarily distributed between the inner ring and middle ring along both sides of the Huangpu River, covering parts of Yangpu District, Pudong New Area, and Xuhui District. This area has a superior geographic location in the city center, where housing prices are generally high. Additionally, large parks, such as Century Park, Shanghai Gongqing Forest Park, Qiantan Leisure Park, and Shanghai Botanical Garden, are located in proximity. The concentrated distribution of these park resources leads to a spatial feature of high housing prices and high accessibility in these communities.

(2)

Low–Low Clustering (L-L Cluster): There are 2800 communities in the low–low clustering category, mainly distributed on the northwest side between the inner and outer rings, including parts of Baoshan District, Changning District, Putuo District, and Jing’an District. These areas have a limited number of large parks, with only one large city-level park, Daning Park, and a lack of small or medium-sized parks offering broad coverage, which results in lower park accessibility. At the same time, housing prices in these areas are also low, presenting a dual disadvantage of low housing prices and low accessibility.

(3)

Low–High Clustering (L-H Cluster): A total of 2145 communities belong to the low–high clustering type, primarily located in the northeastern and southwestern parts between the inner and middle rings, as well as some areas along the outer ring. These communities have relatively low housing prices, but are surrounded by several large parks, such as Shanghai Gongqing Forest Park, Expo Park, Jinxi Culture Park, and parks near the outer ring, such as Gaodong Park, Jinhai Wetland Park, and Mianqing Park. The park resources in these areas provide good accessibility services to the residents, resulting in a characteristic of low housing prices but high accessibility.

(4)

High–Low Clustering (H-L Cluster): There are 1963 communities in the high–low clustering type, primarily distributed in the central urban area within the inner ring. These areas are home to a large number of older communities with high population density. However, due to limited land resources, there are not enough large parks, and green space provision is relatively scarce, resulting in lower accessibility. At the same time, these areas are economically active and commercially developed, with relatively high housing prices. This mismatch between resources and demand leads to a typical spatial mismatch of high housing prices and low accessibility.

Additionally, apart from the above four types, there are 1581 communities where the spatial correlation between housing prices and accessibility is not significant.

(5)

Spatial Lag Regression Analysis Results

Based on the results of the spatial lag regression model (

Table 3. Spatial lag regression results.

	Within the Inner Ring Road		Inner Ring to Middle Ring Road		Middle Ring to Outer Ring Road
Price Range (Yuan/Square Meter)	Coefficient	p-Value	Coefficient	p-Value	Coefficient	p-Value
0–50,000	0.5202 **	0.0059	0.2214 *	0.0280	−0.3497 **	0.0070
50,000–70,000	−0.2044	0.1517	0.2548	0.8425	−0.2136 *	0.0173
70,000–90,000	−0.1813 *	0.0713	0.7704 *	0.0129	0.2726	0.3901
90,000–11,000	0.2544	0.0416	−0.5452	0.2757	2.2228 *	0.0206
>110,000	1.2375 ***	0.0001	1.1496 **	0.0052	−0.2858	0.7281

Note: * p < 0.05, ** p < 0.01, *** p < 0.001.

), the complex relationship between housing prices and accessibility is further revealed. The key findings are as follows:

①

Positive Correlation Between Housing Prices and Accessibility: In communities with housing prices ranging from CNY 0 to 50,000 and those exceeding CNY 110,000, a positive correlation between housing prices and park accessibility is observed. This indicates that areas with higher housing prices tend to have better park resources.

②

Stronger Positive Correlation in High-Price Communities: For example, in communities within the inner ring where housing prices exceed CNY 110,000, the regression coefficient is as high as 1.2375, significantly higher than that of communities with housing prices between CNY 0 and CNY 50,000 (with a regression coefficient of 0.5202). This trend is similarly significant in the areas between the inner and middle rings.

③

Negative Correlation and Spatial Mismatch: In the region between the middle and outer rings, the regression coefficient for the accessibility of communities with housing prices ranging from CNY 0 to CNY 50,000 is −0.3497, indicating that low-priced areas actually enjoy higher park accessibility. This aligns with the spatial mismatch phenomenon observed in the high–low and low–high clusters.

In conclusion, the spatial mismatch between park accessibility and housing prices primarily arises from the insufficient park resources and high population density in the urban core areas, while the peripheral areas feature abundant park resources but relatively low population density. This spatial imbalance has exacerbated social inequity to some extent, and addressing it requires optimizing the distribution of parks and improving the equity of public resources.

5. Discussion

5.1. Comparison of Park Accessibility Categories

This study provides a detailed comparison of accessibility differences for four types of urban parks (district parks, city parks, pocket parks, and community parks) based on their actual service radii. The statistical characteristics of residential community accessibility within the service areas of these parks are revealed through box plots (see

Table 4. Quantile statistics of accessibility box plot.

	Pocket Park_5521m	Community Park_6762m	Regional Park_7299m	Urban Park_9622m
Lower Bound (Q0)	0	0	0.15	0.02
Lower Quartile (Q1)	0.01	0.18	0.50	0.28
Median (Q2)	0.03	0.24	0.59	0.57
Upper Quartile (Q3)	0.04	0.27	0.67	0.87
Upper Bound (Q4)	0.06	0.38	1.60	2.48

and Figure 7), and the following two key findings are summarized:

(1)

District parks and City parks have better accessibility than Pocket parks and Community parks

There is a significant difference in the accessibility of residential communities within the actual service radii of the four park categories. District parks and city parks have notably better accessibility than pocket parks and community parks. The median accessibility values for district parks and city parks are 0.59 and 0.57, respectively, which are significantly higher than the medians of 0.03 and 0.24 for pocket parks and community parks. Additionally, from an overall distribution perspective, the accessibility values of district parks and city parks concentrate in the higher range, far outperforming the distribution of pocket parks and community parks.

This result indicates that large parks (such as district parks and city parks), with their larger service radii and higher service capabilities, can cover a broader range of residential communities and provide a higher level of service.

(2)

Spatial Distribution and Layout of Park Accessibility Are Highly Consistent

The spatial distribution characteristics of accessibility for different park categories closely align with their spatial layouts, as shown in the following details: ① Pocket parks mainly serve communities within the inner ring and middle ring areas. Due to their smaller size, their service radii and coverage are relatively limited. ② Community parks primarily serve the areas west of the Huangpu River, where the density of community parks is high, and the layout is relatively concentrated. ③ District parks show a more dispersed distribution of high-accessibility communities, covering specific areas in the inner, middle, and outer rings. ④ City parks have the widest service range, with communities between the middle and outer rings generally having higher accessibility than those within the inner ring. This is because city parks, with larger areas and stronger service capacities, can serve communities farther from the city center. Despite some city parks being located near the outer ring, their accessibility is enhanced by Shanghai’s well-developed transportation infrastructure, allowing them to serve communities at a greater distance, demonstrating a broad service effect.

Overall, the high-accessibility areas within the service radii of these four park categories exhibit varying distributions and present certain spatial polarization characteristics. Moreover, the accessibility of residential communities is significantly influenced by the distribution of parks, with park size and quantity being positively correlated with accessibility. While some city parks are located farther from the city center, good transportation conditions enhance their service coverage, providing convenient accessibility to a larger number of communities.

5.2. Spatial Inequity in Accessibility Distribution: Renewal of Micro-Spaces

Based on the results of the spatial equity analysis, the residential communities in the central urban areas of Shanghai exhibit significant spatial imbalances in park accessibility. This inequity manifests in the following three key aspects:

(1)

Lower Accessibility in the Inner Ring Area: Residential communities within the Inner Ring area, which is the city’s core, generally exhibit lower levels of accessibility. This region is characterized by high land value and concentrated population density. The intense land development in this area restricts the planning and implementation of new large-scale urban parks, exacerbating the deficiency in park accessibility. Furthermore, the limited number and uneven distribution of existing parks further constrain the service capacity for these communities, resulting in overall low accessibility. From the perspective of urban development trajectory, inner-ring areas were typically prioritized in early-stage development, when the planning philosophy emphasized concentrated commercial and residential functions while neglecting balanced allocation of public spaces. This historical legacy has resulted in persistent challenges for subsequent park development.

(2)

Significant Spatial Autocorrelation of Accessibility: High-accessibility areas are predominantly concentrated in the northeastern and southwestern parts between the Inner and Outer Ring roads, while low-accessibility areas are more prevalent in the northwestern part of the central urban area. This phenomenon indicates that the park accessibility of residential communities demonstrates a spatial clustering effect, following a distinct and predictable pattern. This clustering is influenced not only by the distribution of parks but also by factors such as population density and the spatial distribution of residential communities. This clustering phenomenon reflects the systemic nature of resource allocation in urban planning, where park distribution is inherently shaped by urban functional zoning, transportation networks, and historical development trajectories.

(3)

Marked Spatial Imbalance in Accessibility Distribution: The regional differences in the distribution of parks and population density are the primary contributors to spatial imbalance. In areas with high population density and scarce park resources (such as certain parts of the Inner Ring area), residential communities tend to exhibit low levels of accessibility. In contrast, areas with abundant park resources and relatively balanced population distribution (such as those near the Outer Ring) generally exhibit higher accessibility for residential communities. In urban planning practices, developers and municipal authorities are inclined to construct large-scale residential communities with correspondingly sized parks in the outer-ring periphery, serving as a strategy to attract population decentralization and alleviate pressure on inner-ring areas. In contrast, due to land scarcity in inner-ring zones, park development faces high costs and technical complexities, while resource allocation tends to prioritize maintaining existing economic functional infrastructure, resulting in comparatively limited park resources.

To improve accessibility in the Inner Ring area, several measures can be implemented, including the addition of small or micro parks, optimizing the layout of the transportation network, and enhancing the service quality of existing parks [6]. Pocket parks can be strategically placed in vacant areas within residential communities or at street corners. Despite their limited size, they are close to people’s living circles and can make use of urban idle or leftover land, thus avoiding the occupation of high-quality land. This makes the construction of green spaces more flexible and efficient, providing convenient recreational spaces for nearby residents and, to some extent, alleviating the problem of low accessibility. Additionally, a more balanced distribution of park resources, taking into account population density, should be pursued, with the aim of achieving a more equitable spatial allocation of public park facilities.

5.3. Spatial Differentiation of Social Equity: Optimization of Resource Allocation

Based on the analysis of the global and local bivariate Moran indices, the spatial distribution of residential community accessibility and housing prices exhibits four clustering patterns: high–high, low–low, low–high, and high–low. This study reveals the presence of both spatial clustering and spatial mismatch phenomena in the central urban areas of Shanghai.

Spatial Clustering Phenomenon: Spatial clustering is evident in the high–high and low–low types, which are predominantly distributed between the Inner and Outer Ring roads. High–high areas generally benefit from favorable geographic locations and abundant park resources, while low–low areas suffer from a lack of park resources and high population density, resulting in both low accessibility and low housing prices. The former, situated in prioritized urban development zones, benefit from early-stage infrastructure planning that attracts affluent residents, driving up both housing prices and park accessibility. In contrast, the latter areas predominantly feature high-density residential land use with limited space for public parks, perpetuating low accessibility and property values. On the other hand, low–high and high–low clusters are mainly concentrated within the Inner Ring. Low–high areas, although characterized by lower housing prices, benefit from a rich supply of surrounding park resources. Conversely, high–low areas exhibit a spatial imbalance where high housing prices are not accompanied by adequate park accessibility.

Spatial Mismatch Phenomenon: Spatial mismatch refers to the misalignment of housing prices and park accessibility in spatial distribution, primarily concentrated in the high–low and low–high regions. This phenomenon reflects spatial imbalance, especially in high–low and low–high communities, where high housing prices do not necessarily correlate with high park accessibility, and low-priced areas may exhibit relatively high accessibility. In the context of the entire central urban area of Shanghai, a notable mismatch exists between high housing prices and low accessibility in the Inner Ring area. This spatial imbalance is primarily driven by the high population density and intensive land use in the city center, which has led to a shortage of park and green space resources [40], compounded by the uneven distribution of the population, which exacerbates spatial misalignment. For example, in the Huangpu District’s old city area, such as the Laoximen neighborhood, issues such as severe population aging and social inequity are evident [14], with significant disparities in green space and park development conditions. In contrast, low-priced, high-accessibility communities are concentrated in the northeastern and southwestern parts between the Inner and Middle Rings, as well as along parts of the Outer Ring. These areas generally enjoy a relatively abundant supply of parkland and green space, and housing prices are lower, leading to the phenomenon of an inconsistency between housing prices and accessibility.

It is possible to consider optimizing urban spatial planning and increasing the supply of park green spaces in areas with high housing prices and low accessibility, such as expanding green spaces through methods like vertical greening. For areas with low housing prices and high accessibility, on the basis of enhancing accessibility, commercial, educational, and other supporting facilities can be improved to enhance the overall quality of life of residents. Meanwhile, the resource allocation strategy should be adjusted regularly, mixed-use development should be promoted, the social equity of urban parks should be enhanced, and the quality of life of residents should be improved.

The “accessibility-spatial equity-social equity” framework proposed in this study demonstrates cross-city applicability in high-density urban contexts, yet its manifestations are significantly shaped by urban structural differences: In monocentric cities, like Beijing and Shanghai, the monocentric expansion model has led to a “high housing prices coupled with low accessibility” mismatch in core areas, necessitating a focus on social compensation mechanisms and the activation of existing spaces (e.g., miniature green space networks) to address delayed public service provision. In polycentric cities, like Shenzhen, ecological network-based planning mitigates spatial polarization, while park-TOD integrated development balances job-housing demand disparities. Although parameter weighting requires adaptation to urban development stages and spatial configurations, the framework’s core logic—establishing accessibility as the foundation, spatial balance as the pathway, and social justice as the ultimate goal—provides a trans-regional methodological framework for green space equity governance in high-density cities.

6. Conclusions

Urban parks, as a vital means to improve residents’ well-being, play an irreplaceable role in promoting social equity and spatial balance. However, the social benefits and spatial distribution of urban parks still face significant imbalances, highlighting the urgent need for scientific and refined evaluation frameworks. This study focuses on the central urban areas of Shanghai and systematically assesses and explores the differences and mismatches in urban park services from the dimensions of accessibility, spatial equity, and social equity. As a typical representative of high-density megacities, the contradiction between supply and demand and the spatial differentiation law it reveals have significant implications for cities of the same type. By integrating dynamic data sources, such as actual visitor data and residents’ real travel intentions, and combining advanced methodologies like the improved Gaussian two-step floating catchment area (EG2SFCA) method, random forest regression model, and spatial lag regression model, this study effectively overcomes the issues of data distortion and methodological limitations prevalent in traditional research, significantly improving the precision and credibility of the analysis. Additionally, by categorizing urban parks into pocket parks, community parks, regional parks, and urban parks for comparative analysis, this study provides scientific evidence for implementing differentiated governance strategies.

The findings of this study are as follows: (1) Significant differences in accessibility across park types: Regional and urban parks have significantly better accessibility compared to pocket parks and community parks, reflecting differences in service range and capacity among park types. (2) Significant spatial distribution imbalance: The overall accessibility of residential communities within the Inner Ring area is relatively low, while high-accessibility areas are concentrated in the northeastern and southwestern parts between the Inner and Outer Ring roads. In contrast, the northwestern areas suffer from low accessibility due to a lack of park resources. (3) Spatial mismatch between park accessibility and housing prices: In residential areas within the inner ring and the inner–middle ring to outer ring zones, the relationship between housing prices and park accessibility is complex. There is both a mismatch of high housing prices and low accessibility, as well as a unique situation where low-priced areas have high accessibility, reflecting spatial imbalances in urban resource distribution. These findings corroborate the systemic imbalance prevalent under monocentric urban expansion models, where land development is prioritized over the provision of public amenities, a challenge pervasive in rapidly urbanizing regions across China and globally.

This study not only deepens theoretical understanding of the spatial equity and social equity of urban parks but also offers practical recommendations for optimizing urban park layouts. On the one hand, to address the shortage of park resources in the inner ring area, it is recommended to increase the number of small or micro parks and enhance park multifunctionality to improve accessibility for nearby communities. On the other hand, attention should be paid to the social equity of resource allocation, with policies favoring the improvement of park services in disadvantaged residential areas and balancing the spatial relationship between park resources and population distribution. In summary, this research not only provides a new perspective for the scientific assessment of urban parks but also offers data support and implementation references for urban planners and policymakers. Moving forward, as refined governance progresses, the planning and optimization of urban parks will become an essential means to enhance residents’ quality of life, achieve social equity, and promote sustainable urban development, especially in the context of urbanization with a dense population and intensified resource constraints, this transformation path will have broader practical value.