Spatial Green Space Accessibility in Hongkou District of Shanghai Based on Gaussian Two-Step Floating Catchment Area Method

Abstract

Green space in urban areas is one of the most critical infrastructures for the well-being of residents, and its spatial accessibility configuration is a key indicator of urban planning layout and ecological civilization construction. Using the Gaussian two-step floating catchment area (G2SFCA) method, K-means cluster analysis, and Kriging spatial interpolation, this study examines the spatial characteristics of green space accessibility in the Hongkou District, focusing on the relationship between “green space-community” supply. The findings indicate the following: (1) green space accessibility in Hongkou District decreases as the level of green spaces decreases. Higher levels of accessibility are associated with smaller variations in spatial distribution. (2) The green space accessibility in Hongkou District is affected by the surrounding large green space. Compared to other areas, the community green spaces near Lu Xun Park and Peace Park have higher accessibility. (3) The green space accessibility in Hongkou District is impacted by the mismatch between population density and green spaces. An overlay analysis of accessibility and population density reveals that high accessibility areas have average or low-average population density. Based on the results of the green space accessibility analysis, recommendations are proposed to optimize the green space layout in Hongkou District from the perspective of spatial justice. These suggestions are aimed at promoting the equalization of green space services in Hongkou District, improving the overall level of urban green space services, with a view to improving the quality of urban living environment and creating a green and livable urban area.

1. Introduction

Urban green spaces play a crucial role in enhancing the aesthetics of modern cities [1], protecting ecosystems, and providing disaster relief [2]. They serve multiple important functions, including recreation opportunities [3], fostering friendly interactions, and showcasing cultural aspects for residents [4,5]. In 2014, Shanghai introduced the concept of the “15-min community living circle” [6], which emphasizes providing essential services and public spaces within a 15 min walking distance. Shanghai has made significant progress in constructing a multi-level urban green space system that encompasses pocket parks, small and micro public green spaces, and transformative waterfront areas. These endeavors, known as “white and green” livelihood projects, have resulted in a diverse range of functional and culturally significant green spaces with convenient transportation. This has significantly reduced the heat island effect [7], and noise pollution in cities [5], provide shaded shelter for pedestrians [8], improved the living and working environment, satisfying people’s aspirations for a better quality of life, while also strengthening the foundation for sustainable development. In recent years, the demand for open green spaces among urban residents has been continuously increasing, especially for those with convenient, equal, and fair open green spaces [9].

Attractiveness, availability, and accessibility are important evaluation criteria for measuring green space demand [10]. Among them, accessibility is one of the key indicators used in urban planning, landscape architecture, geography, and other fields to assess the rationality, fairness, and effectiveness of public service facility layouts [11], and it is often applied in research related to urban green spaces, elderly facilities, medical facilities, etc. The accessibility of green spaces can not only quantitatively reflect the ease or difficulty of pedestrians from specific locations to activity locations, but also effectively reveal the characteristics of the overall spatial distribution of urban green spaces. In contrast, the attractiveness and availability of green spaces tend to be more qualitatively evaluated. This means that the accessibility of green spaces not only reflects whether a person can freely reach the green space from a given location, but also reflects whether they find the green space attractive and whether the functionality of the green space meets their needs [12]. In general, when assessing the quality of urban green space, green space accessibility, as a widely used indicator, is more reflective of the “people-centered” planning concept than the traditional indicators of green space per capita, green space coverage and green space ratio.

Currently, green space accessibility is mainly calculated by quantitative methods, of which GIS is the main calculation platform. GIS-based methods for measuring urban green space accessibility include spatial syntax [13,14], buffer analysis [15,16], network analysis [17], and the two-step floating catchment area (2SFCA) method [18,19]. Compared to the first three methods, which have simpler calculations but higher data precision requirements, the 2SFCA method takes into account supply–demand relationships and introduces the concept of a “spatial threshold”. Continuously optimizing the model establishment and calculation methods enables it to more accurately reflect the supply and demand dynamics of urban green spaces, making it an ideal approach for measuring green space accessibility. For example, Li et al. (2016) conducted a comparative analysis of green space accessibility in Shanghai using the G2SFCA and gridded G2SFCA methods [20]. Ren et al. (2021) proposed an improved two-step floating catchment area analysis model to address imbalances in the supply and demand of parks and green spaces in Huangpu District, Shanghai [21]. Wu et al. calculated urban green space accessibility in Beijing using the Gaussian-based 2SFCA method and assessed the differences in green space accessibility among different income groups [22]. Yang et al. (2021) combined the Two-step Floating Catchment Area Analysis and the Transportation Improvement Quality System (TIQS) to construct a multi-mode two-step floating catchment area model, exploring differences in multi-scale park green space accessibility and equity under three travel modes (walking, public transport, and private car) in Guangzhou [23]. Liu et al. (2021) measured green space accessibility in different racial census tracts in Chicago using the G2SFCA method, revealing spatial differentiation patterns of green space accessibility in areas with different income levels [24]. Therefore, by the Gaussian two-step floating catchment area method quantitatively measuring the ease or difficulty for residents to reach surrounding green spaces considering distance and time costs, it accurately and intuitively reflects the size of urban green space service capacity and the rationality of spatial allocation.

Therefore, based on the above discussion, this study selects Hongkou District in Shanghai as the research object, and calculates and analyzes the accessibility to parks from the perspective of the “15-min community living circle”. The main objectives of the study are as follows: firstly, to quantify the spatial differences in the accessibility of green spaces in Hongkou District, Shanghai, using the Gaussian two-step floating catchment area method. Secondly, to qualitatively investigate the main factors affecting the accessibility of parks. Thirdly, to provide strategies for the management and construction of urban green space systems for Hongkou District, Shanghai, as well as for other urban areas at a similar stage of development. Through these objectives, this study aims to gain an in-depth understanding of green space accessibility in Hongkou District, reveal the influence of different factors on accessibility, and provide a complete theoretical framework/guidance for the implementation of greening strategies in other developing countries, especially in mega-cities such as Shanghai.

2. Data and Methods

2.1. Study Area

Shanghai, as a significant national center city, has been actively pursuing the development of a “park city” for several years. The city has achieved remarkable achievements in this endeavor, with a per capita green space coverage of 8.8 square meters in 2022. This makes Shanghai an exemplary case with valuable lessons and reference significance. The study takes Hongkou District, Shanghai, as the study area and uses streets of the earliest urban centers in Shanghai, located in the northeast of the city, straddling the three inner and outer rings, adjacent to Huangpu District, Zhabei District, Yangpu District, and bordering Baoshan District, and located between 121°27′18″ and 121°30′46″ E and 31°14′38″ and 31°19′50″ N. By 2022, the area of the district will be 23.48 square kilometers, and the population density is 32,261.4 persons per square kilometer, ranking first among all administrative districts in Shanghai. With the completion of Shanghai’s 13th Five-Year Plan, Hongkou District has built 17.27 hectares of new green space of various types, and has made remarkable achievements in urban construction. The study focuses on eight streets in Hongkou District, namely Jiangwan Town, Liangcheng New Village, Quyang Road, Guangzhong Road, Ouyang Road, North Sichuan Road, Jiaxing Road, and the North Bund. This area includes prominent landmarks such as Luxun Park and the North Bund Riverside Green (Figure 1).

2.2. Data Sources and Pre-Processing

2.2.1. Urban Green Space Data

This study selected the green areas in Hongkou District, Shanghai, which are mainly for daily recreation, open to the public, with certain recreational facilities and services, as well as ecological soundness, landscape beautification, science education, and emergency shelter, as the object of accessibility optimization, referred to as “recreational green areas”. Specifically, the study targets 28 urban green areas in the complete urban park system of “District Park-Community Park-Pocket Park” identified in the Three-Year Action Plan for Ecological Protection and Construction of Hongkou District from 2021 to 2023, which is representative of the study. To account for residents’ spatial perception, certain green space patches that are divided by roads and rivers but considered as a cohesive unit were combined during the data post-processing stage [25]. Furthermore, recognizing that residents in the study area may also utilize green spaces located outside of the immediate study area, the data collection scope was expanded to include surrounding green spaces.

In the first step, the Point of Interest (POI) data for green areas in Hongkou District were obtained from the Gaode online map API. This dataset includes information such as latitude and longitude coordinates for the entrances and exits of green park areas, as well as the area of each green space. Next, sub-meter aerial photography image data from the primary monitoring satellite imagery of Hongkou District Traffic were utilized. To ensure calculation accuracy, entrance point coordinates were collected in two scenarios: firstly, for green park areas with fences, the entrance point coordinates were directly collected. Secondly, for green park areas without fences, multiple entrance points were selected, ensuring that the distance between each entrance point did not exceed 50m. After careful comparison, a total of 28 green spaces were ultimately chosen for the study.

2.2.2. Other Data

As the model takes into account the supply–demand relationship, population distribution can greatly influence accessibility outcomes [26]. The number of individuals residing in the vicinity of a park serves as an indicator of potential demand for green spaces. However, acquiring direct population data for each residential community poses challenges. Therefore, this study utilizes the 2020 seventh census data of Hongkou District as a reference for estimating the population of each residential community, using the base street (or township) population number as a basis.

$$P{R}_i=PD\times \frac{S{R}_i}{{{\displaystyle \sum }}_{i=1}^{n}S{R}_i}$$

(1)

In the equation, $P{R}_i$ is the number of people in residential community i; $PD$ is the number of people. $S{R}_i$ represents the number of households in residential community i, and n denotes the number of residential neighborhoods within the street.

The data for accessibility analysis also included the Hongkou district zoning map and transportation network map collected through Python crawling and revised in conjunction with the Shanghai Urban Master Plan (2016–2040) and the Shanghai Comprehensive All spatial data were converted to a unified spatial coordinate system through “Address-GPS Coordinate Converter” to ensure data compatibility.

2.3. Research Methods

2.3.1. Gaussian Two-Step Floating Catchment Area Method

The two-step floating catchment area method addresses the supply of urban green spaces and the demand from residential communities. However, it fails to account for actual resistance and distance decay effects within the search range [24,27]. To address this limitation, Dai (2011) proposed an improved G2SFCA method [28]. This study enhances the commonly used G2SFCA method by incorporating improved data sources for green space and optimizing the origin-destination (OD) costing rules. This approach enables us to measure accessibility across multiple modes of transportation for four levels of green space in Hongkou District.

The calculation process consists of two steps:

In the initial step, we calculate the supply–demand ratio of parkland within residential communities. For each green space j, a spatial distance threshold d₀ is defined to establish a spatial scope. Within this scope, the population of each street $k$ is considered, and a weight is assigned using the Gaussian equation. The weighted populations are then aggregated to determine the total number of potential users for the specific green space j. Next, we divide the size of the green space by the number of potential users to obtain the supply–demand ratio $R_j$ using the following equation:

$$R_j=\frac{S_j}{{{\displaystyle \sum }}_{k\in \left\{d_{kj\le }{d}_0\right\}}G\left(d_{kj}, d_0\right)P_k}$$

(2)

In the equation, $P_k$ represents the population of street $k$ within the spatial scope of green space j when the spatial distance $\left\{d_{kj\le }{d}_0\right\}$; $d_{kj}$falls within the threshold. The spatial distance is measured from the center of street k to the center of green space $j$. The area of green space is denoted as$S_j$, indicating its capacity. The Gaussian equation, $G\left(d_{kj}, d_0\right)$, incorporates the concept of spatial friction and is calculated as follows:

$$G(d_{kj},d_0)=\{\begin{array}{ll}\frac{e^{-(\frac{1}{2})\times {(\frac{d_{kj}}{d_0})}^2}-e^{-(\frac{1}{2})}}{1-e^{-(\frac{1}{2})}},& if d_{kj}\le d_0\\ 0,& if d_{kj}>d_0\end{array}$$

(3)

In the subsequent step, we calculate the accessibility of residential communities. For each street $i$, a spatial distance threshold $d_0$ is established to create another spatial domain. Within this domain, the supply ratio $\left(R_i\right)$ of each green space l falling within the area is assigned a weight using the Gaussian equation. These weighted supply ratios $\left(R_i\right)$ are then aggregated to determine the green space accessibility of each street $A_{i }$. The value of $A_{i }$ represents the per capita occupancy of urban green space in square meters per person within the specific domain. The formula for calculating $A_i$ is as follows:

$$A_i={{\displaystyle \sum }}_{l\in \left\{d_{il}\le d_0\right\}}G\left(d_{il}, d_0\right)R_l$$

(4)

In the equation, $R_l$ denotes the ratio of green space provision within the spatial scope of the street $i \left\{d_{il}\le d_0\right\}$; $d_0$ is the walking time between residential point $i$ and the park green space $l$;$A_i$ denotes the park green space accessibility index of settlement$i$. The larger $A_i$ is calculated, the better accessibility of $i$ is indicated [29].

In the equation, $R_l$ represents the ratio of green space provision within the spatial scope of street $i \left\{d_{il}\le d_0\right\}$ when the walking time between residential point $i$ and park green space l falls within the threshold. $A_i$ denotes the park green space accessibility index of settlement $i$. A higher value of $A_i$ indicates better accessibility for settlement $i$ [19,29].

2.3.2. K-Means Clustering Analysis

Clustering analysis can organize data with similar characteristics in a categorical manner and help explore the differences in the inherent structure between data. K-means clustering analysis is an iterative solution clustering analysis algorithm with the advantages of a simple algorithm and efficient handling of large data sets, and the type of differences in accessibility under different combinations of supply and demand patterns can be obtained using K-means clustering analysis [26]. The clustering indexes include indicators that characterize the accessibility of the evaluation unit, such as the comprehensive accessibility of parkland obtained by aggregating the accessibility of different types of parks; the average value of OD time calculated with the center of the evaluation unit as the starting point and each type of park as the destination, which characterizes the traffic conditions of the evaluation unit; the total area of each evaluation unit calculated by distance decay of each type of park within the service threshold, which characterizes the supply of green space of each type of park. K-means clustering analysis is used to obtain the accessibility of parkland in the study area and its combination of supply and demand, analyze the causes of accessibility differences in terms of supply, demand, and traffic, identify the main limiting factors affecting accessibility in different spatial units, and then propose targeted policy recommendations for different zones [30].

2.3.3. Hierarchical Analysis Method

Natural Breaks (Jenks), a natural split method provided by ArcGIS 10.7, was used to classify the green space accessibility rows of each street, (s) [27]. Natural Breaks is a grouping method that leverages the inherent relationships within data to create distinct groups. Its objective is to maximize the differences between groups while optimizing the similarity within each group. This approach ensures the most scientifically and objectively determined categorization. Using this method, we categorized the accessibility of park green spaces in each street in Hongkou District into five classes: low, lower, average, higher, and high.

2.3.4. Kriging Space Interpolation Method

The analysis using the Gaussian two-step floating catchment area method focused on central points like green park areas and community exits. To obtain accessibility results for locations beyond these main areas, the Kriging spatial interpolation method was employed. This method utilizes Arc-GIS 10.7 software to fit a variation function that determines the trend of regionalization variables, enabling estimation through interpolation [6,31]. The key principles are as follows:

Let the observation in a region at the sampling point location S_i be $\hat{\mathrm{z}}\left(M_i\right)$, at the prediction point $M_{0 }$; the estimated value of $\hat{\mathrm{z}}\left(M_0\right)$ can be obtained by a linear combination of the observations at the surrounding n sampling points $\hat{\mathrm{z}}\left(M_0\right)$ to be obtained by a linear combination of the formula.

$$\hat{\mathrm{z}}\left(M_0\right)={\displaystyle \sum }_{i=1}^n{\beta }_i\mathrm{Z}\left(M_i\right)$$

(5)

In the above formula, ${\beta }_i$ is the weight of sampling point $M_{i }$;${\beta }_i$ needs to satisfy the following equation:

$$\beta \left(M_i ,M_0\right),{\displaystyle \sum }_{i=1}^n{\beta }_i={\displaystyle \sum }_{i=1}^n{\beta }_i\gamma \left(M_i, M_j\right)+\mu =1$$

(6)

In the above formula, $\beta \left(M_i, M_j\right)$ is the semi-variance between the observation point$M_0$and$M_i$; $\beta \left(M_i,M_0\right)$ is the semi-variance between the sampling point$M_i$ and the prediction point $M_0$; $\mu$ is the Lagrange multiplier associated with the variance minimization.

Using this method can maximize the utilization of spatial data information from samples by considering spatial autocorrelation characteristics, such as variance function and covariance. It compares the spatial location of points to be estimated with neighboring known points and fits the location data of each neighboring point. This approach has a reasonable basis [32].

3. Results and Analysis

3.1. Accessible Quantity Distribution

To present the accessibility results of Hongkou District, the natural discontinuity method was utilized to calculate the area and proportion of each accessibility level. The findings are summarized in

Table 1. Corresponding area and proportion distribution of accessibility classification.

Accessibility Classification	Area (km2)	Percentage
Low < 1.68	13.92126	59.35%
Lower 1.69–4.48	5.651818	24.09%
Average 4.49–7.84	2.318745	9.88%
Higher 7.85–16.24	1.518239	6.47%
High > 16.25	0.047394	0.20%

, where larger areas correspond to lower accessibility levels. The overall structure exhibits a gradual decrease in accessibility. Specifically, the proportions of areas with low, lower, average, higher, and high accessibility are 59.35%, 24.09%, 9.88%, 6.47%, and 0.20% (Figure 2), respectively. Notably, the combined area of low and lower-grade accessibility accounts for 73.44% of the total, indicating that the majority of green regions in Hongkou District suffer from inadequate accessibility. Only 0.2% of the locations exhibit high-grade accessibility. A comprehensive quantitative analysis reveals significant disparities in green space accessibility across different areas of Hongkou District, demonstrating distinct spatial patterns.

3.2. Spatial Distribution of Accessibility

In this study, we employed the inverse distance weight interpolation method to analyze the green space accessibility and population distribution in Hongkou District. Figure 3 depicts the green space accessibility map overlaid with population point density mapping, where color-stretched gradient values represent the levels of green space accessibility. The figure reveals a spatial pattern characterized by “high in the central part, high in the west, and low in the rest” across Hongkou District. Specifically, areas with high accessibility are primarily concentrated in the Quyang Road Block, North Sichuan Road Block, Ouyang Block Street, and North Bund Block, with Ouyang Road Block exhibiting a smaller scale compared to the larger scales seen in Quyang Road Block, North Sichuan Road Block, and North Bund Block. On the other hand, the low accessibility aggregation area is predominantly located in the northern part of Hongkou District and the region between the central and southern parts (Jiaxing Road Block). The latter exhibits a more pronounced low-value area due to insufficient distribution of green spaces. In the northern region (Jiangwan Town Block, Liangcheng Xincun Block, and Guangzhong Road Block), although the green space scale is relatively small and scattered, the population density is high, averaging around 20,001–20,069 people per square kilometer. This combination contributes to the formation of another noticeable low-value area.

When considering the population point density distribution, it is observed that the high aggregation area comprising North Sichuan Road Street and Ouyang Road Street exhibits an average population density. Conversely, Quyang Road Street and North Bund Street have smaller populations (Figure 4), with overall densities ranging from 3 to 20,000 people per square kilometer. According to Costanza and Fisher’s study [33], we found that the spatial configuration of urban parks green spaces belongs to the typical user movement related (user movement related). From a user needs perspective, this means that areas with lower population densities tend to have a lower demand for green spaces [34]. Consequently, these areas are better equipped to meet the demand for park green space services, thereby improving overall accessibility to green spaces.

Considering the spatial distribution of green spaces (Figure 5), it is observed that Quyang Road Street benefits from large-scale parks like Quyang Park. North Sichuan Road Street and Ouyang Road Street are characterized by prominent green spaces, including Luxun Park, Heping Park, and North Sichuan Road Park. Moreover, North Bund Street features extensive linear green spaces along the Huangpu River. The presence of these large-scale green spaces significantly influences the distribution of green space accessibility in Hongkou District. Generally, areas with larger green spaces in proximity tend to exhibit higher levels of green space accessibility.

4. Discussion

Through the hierarchical and distributional analysis of green space accessibility in Hongkou District, it is evident that there exists significant spatial differentiation and polarization in the accessibility of green spaces. The primary factor contributing to this differentiation is the mismatch between large-scale green spaces, population density, and the distribution of green spaces within the district.

4.1. The Accessibility of Green Space in Hongkou District Shows Prominent Spatial Differentiation Characteristics

The analysis of green space accessibility reveals a clear trend in Hongkou District, where overall accessibility decreases as the grade lowers. Higher grades of green space accessibility correspond to smaller areas in terms of percentage coverage. The calculation results show that over 70% of the study area exhibits below-average levels of green space accessibility. In contrast, only a few sites (0.2%) demonstrate high green space accessibility. An in-depth analysis of the distribution of low-accessibility areas reveals the following two main problems. Firstly, there is little to no distribution of parks within most of these areas, making it difficult for residents living in these areas to get to a park within 15 min. This contributes to the overall lack of park provision within the zones. Secondly, although in a few areas, residents are able to get to a park within the time threshold, the higher density of residents in these areas results in a demand for parks by the regional population that is much higher than the current level of supply, thus failing to meet the demand of residents. These two issues are intertwined and together constrain the level of green space service in low accessibility areas. Without adequate distribution of parks, the residents’ daily needs for leisure and recreation cannot be met. At the same time, even if parks exist in some areas, due to the high population density, the park resources are unable to meet the collective needs of the residents, resulting in a situation of imbalance between the supply and demand of green space. This highlights significant spatial differentiation in the distribution of regions with varying levels of green space accessibility within Hongkou District, as demonstrated by the inability of the number of parks to match the needs of the population and the uneven distribution of parks that currently exist.

4.2. Green Space Accessibility in Hongkou District Is Influenced by Large-Scale Greenspaces in the Surrounding Areas

The spatial distribution of green space accessibility indicates that areas surrounding larger-scale green spaces, such as Quyang Road Block, North Sichuan Road Block, Ouyang Road Block, and the North Bund Block (along the Huangpu River), exhibit higher levels of accessibility. Conversely, the northern part of Hongkou District (Jiangwan Town Block, Liangcheng Xincun Block, and Guangzhong Road Block) and the region between the central and southern parts (Jiaxing Road Block) lack large-scale green spaces, resulting in relatively lower green space accessibility. This emphasizes the significant role played by large-scale green spaces in enhancing overall accessibility within Hongkou District. Achieving a balanced distribution of such large-scale green spaces can promote spatial equity in regional green space accessibility [21]. Therefore, when planning the green space system, careful consideration should be given to the spatial arrangement of large-scale green spaces.

4.3. Green Space Accessibility in Hongkou District Is Affected by the Mismatch between Population Density and Green Space

The analysis of accessibility and population density reveals that high accessibility areas exhibit average or low population densities (ranging from 3 to 20,000 people per square kilometer). These areas are characterized by a greater presence of large-scale green spaces and green areas along the Huangpu River. In contrast, low accessibility areas display high population densities (ranging from 20,001 to 20,069 people per square kilometer), with only a few small-scale green spaces scattered unevenly. The overall mismatch between population density and green space distribution in Hongkou District contributes to the apparent spatial variation in green space accessibility. This finding aligns with the concept of “mismatch” discussed by [35], which highlights the discrepancy between green space accessibility and population density.

An analysis of accessibility and population density shows that areas of high accessibility have average or low population densities (3~20,000 people per square kilometer), and that these areas are characterized by a greater number of large green spaces and greenspaces along the Huangpu River. This is similar to the Pearl River coastline in Guangzhou’s Yuexiu District [36,37], where [38] found that green space accessibility was higher near the Pearl River in Guangzhou’s Yuexiu District. However, unlike the Huangpu River riparian area, the Pearl River riparian area in Yuexiu District has managed to reduce the time it takes people to reach the parks while expanding the coverage of the parks through the construction of subway transportation and strip parks along the line, despite its higher population density.

In addition, we have found that areas with low accessibility had higher population densities (20,001 to 20,069 people per square kilometer), with only a few small pockets of unevenly distributed green space. This is consistent with the findings of [39] that the low-value areas of green space accessibility in Shenzhen’s Nanshan District are concentrated in areas with high population densities and low green space distribution. This highlights the fact that in urban planning and management, how to promote equity in the level of service of green space has become a common challenge faced by similar high-level cities such as Hongkou District in Shanghai and Nanshan District in Shenzhen.

5. Conclusions

As a critical component of urban public resource allocation [31], green space is an indispensable public resource in urban development [40]. Every citizen should have the right to enjoy this resource equally. Taking Hongkou District of Shanghai as an example, this study adopts a Gaussian two-step moving search and hierarchical classification method to evaluate the accessibility of 28 urban green spaces, including district parks, community parks and pocket parks, and further analyzes the accessibility differences and their causes under the weighted population density factors by K-means clustering and Kriging spatial interpolation.

This study finds that in reality, due to the market mechanism, green space resources are gradually being occupied by high-income groups [41]. In contrast, low-income groups have difficulty thoroughly enjoying green space services because they cannot afford the high housing prices. The spatial differentiation of green space accessibility shows an extreme imbalance [42,43], resulting in green space becoming a scarce resource in some areas.

To address the imbalance in green space accessibility within Hongkou District, it is essential to optimize the regional green space layout strategy from the perspective of spatial justice, aiming to promote equalization of green space services [25]. Based on the results obtained from the analysis of green space accessibility, the following three spatial optimization suggestions are proposed for the specific context of Hongkou District: (1) in areas such as Jiaxing Road Block, Jiangwan Town Block, Liangcheng New Village Block, and Guangzhong Road Block, where overall park green space accessibility is the lowest due to site limitations, the flexible placement of small pocket parks could be considered to maximize green space availability at every opportunity [44]. (2) To achieve equitable green space services, the construction of new parks should be aligned with the surrounding population size, built environment, and development goals [45,46]. (3) Given the current scattered layout of parkland in the study area, efforts should be made to establish connections between water systems, natural landscapes, and green spaces. This includes a balanced distribution of small green spaces and enhanced connectivity between them, creating a cohesive network that combines points, lines, and surfaces within the parkland system [47]. For instance, the integration of the Huangpu River riverfront linear green space has significantly improved the accessibility of green space in the southern part of Hongkou District [48].

Accessibility plays a critical role in evaluating the capacity of green spaces to serve communities. The G2SFCA method offers a valuable approach in estimating accessibility by considering the supply of green spaces (park area), the demand from the population (population size), and the relationship between supply and demand (distance). Achieving high accessibility requires a balance between supply and demand, as well as appropriate connectivity within an area. The Gaussian two-step floating catchment area method effectively captures the spatial variation in green space accessibility in Hongkou District, providing a comprehensive and objective assessment. However, it is important to recognize the limitations of the method.

(1)

When evaluating the supply of green space in a park, the current methods only consider the size of the park when calculating the supply and demand ratio, ignoring the complex interaction between the park quality and shape index [49]. While the extent of parkland undoubtedly has a significant impact on accessibility [50], a more holistic consideration, including a wide range of park elements such as vegetation cover and plant diversity [51], would allow for a more nuanced and accurate portrayal of the extent to which parks contribute to promoting equal accessibility. An in-depth analysis of the vegetation cover and the different plant types present in the park environment can reveal the multiple ways in which these factors are dynamically intertwined with accessibility, thereby increasing the accuracy of our overall understanding of how the elements of the park itself influence park accessibility.

(2)

When measuring population demand, this study uses the population of each residential area as a proxy to assess population demand, providing insights into the overall need for green space in the study area [52]. However, since the final accessibility calculation is based on discrete points centered on each residential area, it requires interpolation using inverse distance weights to transform it into the accessibility of green space in Hongkou District. This interpolation approach may overlook population change characteristics at smaller scales. Therefore, future research can explore utilizing extensive data methods such as cell phone signaling to more accurately measure regional population demand [53]. Moreover, this study does not account for variations in park demand and spatial distance tolerance among different demographic groups [54].

For the older adult population, it is worth noting that older adults are likely to spend more time in parks because of recreational and social activities, thus demonstrating a higher demand for parks [42]. In addition, due to their relatively weaker mobility, older people may have less scope for walking compared to younger people. For children’s groups, the needs of children, another important demographic group, for parks are also of concern. Children often need open space for play, exercise and socialization [55]. Parks therefore play a crucial role in meeting the needs of children. Low-income groups may be more reliant on free public resources for leisure and recreational activities [56]. By understanding the needs of these groups and their expectations of green space accessibility, we can help develop more inclusive and equitable urban planning. In future studies, it would be beneficial to quantify the differences in demand among various resident groups and adjust the evaluation model accordingly to better align with the actual situation.

(3)

When measuring the distance relationship between supply and demand, this study adopts the spatial distance based on the current road network in Hongkou District as the criterion for evaluating supply and demand. This approach is more realistic than traditional European distance calculations through buffer zone analysis. Nonetheless, while it considers the connectivity of the road network in real life, it still lacks consideration of the impact of road events [57], such as morning and evening traffic congestion during peak hours [58]. In future research, it would be beneficial to incorporate online map services to measure the distance between supply and demand by obtaining real-time travel time information, thus taking into account the dynamic effects of road conditions.

Consideration should also be given to the fact that pedestrians may prefer more interesting routes when choosing paths to the green space [59]. For example, roads filled with history may be more appealing because they can provide a richer cultural experience for pedestrians. This factor may have an impact on pedestrians’ path choice when measuring supply and demand, so potential associations between path interest and supply and demand could also be explored in future research. By taking this element of interest into account, future studies could provide a more comprehensive measure of the relationship between supply and demand and provide more creative guidance for urban planning and roadway design. This will help to better meet the needs of pedestrians and create a more attractive and enjoyable travel experience.

Considering the aforementioned theoretical considerations, there is room for improvement in the G2SFCA method to enhance the accuracy of measuring green space accessibility. However, when translating these theoretical findings into practical planning strategies, it is crucial to conduct context-specific analyses based on local conditions. For instance, conducting questionnaire research on local parks within areas identified as high-value gathering areas can provide valuable insights and promote effective planning and design ideas that can be applied to other parks [60]. In conclusion, studying green space accessibility based on theoretical models must be combined with the actual situation to ensure improved accuracy and applicability of the G2SFCA method. It is through this integration that the method can effectively capture the complexities and nuances of real-world scenarios.