Assessing Inequality in Urban Green Spaces with Consideration for Physical Activity Promotion: Utilizing Spatial Analysis Techniques Supported by Multisource Data

Abstract

Urban green spaces (UGSs) play a significant role in promoting public health by facilitating outdoor activities, but issues of spatial and socioeconomic inequality within UGSs have drawn increasing attention. However, current methods for assessing UGS inequality still face challenges such as data acquisition difficulties and low identification accuracy. Taking Harbin as a case study, this research employs various advanced technologies, including Python data scraping, drone imagery collection, and Amap API, to gather a diverse range of data on UGSs, including photos, high-resolution images, and AOI boundaries. Firstly, elements related to physical activity within UGSs are integrated into a supply adjustment index (SAI), based on which UGSs are classified into three categories. Then, a supply–demand improved two-step floating catchment area (SD2SFCA) method is employed to more accurately measure the accessibility of these three types of UGSs. Finally, using multiple linear regression analysis and Mann–Whitney U tests, socioeconomic inequalities in UGS accessibility are explored. The results indicate that (1) significant differentiation exists in the types of UGS services available in various urban areas, with a severe lack of small-scale, low-supply UGSs; (2) accessibility of all types of UGSs is significantly positively associated with housing prices, with higher-priced areas demonstrating notably higher accessibility compared to lower-priced ones; (3) children may be at a disadvantage in accessing UGSs with medium-supply levels. Future planning efforts need to enhance attention to vulnerable groups. This study underscores the importance of considering different types of UGSs in inequality assessments and proposes a method that could serve as a valuable tool for accurately assessing UGS inequality.

1. Introduction

Rapid urbanization in China has brought various challenges to public health, despite improving people’s living standards. For instance, high urbanization rates may lead to a decrease in residents’ physical activity levels, indirectly contributing to the onset of various chronic diseases [1]. Urban green spaces (UGSs) serve as primary spaces for residents’ outdoor physical activities, facilitating such activities through the provision of facilities and natural environments [2]. Therefore, ensuring that all residents have convenient access to UGSs is of paramount importance for comprehensively safeguarding their right to health.

Numerous studies have shown that there is inequality in the distribution of UGSs in urban areas. Research on this inequality primarily focuses on two dimensions: spatial and social. Spatial inequality of UGSs mainly refers to the mismatch between the distribution of UGSs in cities and the population [3,4]. Studies conducted in different countries and cities have revealed diverse spatial inequality patterns and causes [3,5,6,7,8]. Contradictions between construction costs and local finances are one of the main reasons for spatial inequality [9]. On the one hand, due to limited urban land resources for UGSs and profit-driven land allocation policies, UGS construction often lags behind other land uses such as commercial and residential areas [10]. Considering land prices, UGSs are often forced to be constructed in low-cost but sparsely populated suburbs [11], resulting in spatial mismatch with the population. On the other hand, local governments may view UGS construction as an important means to attract investment and achieve economic growth in urban planning, potentially leading to an uneven distribution of UGSs across different urban areas and further exacerbating issues such as gentrification [4,12,13]. From this perspective, how to optimize the spatial layout of UGSs at a low cost and high efficiency is an important challenge in addressing spatial inequality in UGSs.

The unreasonable distribution of UGSs may further result in unequal distribution among different social groups, namely, the socioeconomic inequality of UGSs [3,14]. Numerous studies have shown that there is inequality in access to UGSs among different income groups [12], races [14,15], and occupations [16]. Much of the research has focused on income-based UGS inequality, with lower-income groups often having fewer opportunities to access UGSs [7,12]. Particularly in developing countries like China, UGSs significantly influence housing prices in communities, making it difficult for lower-income groups to afford high-quality living environments and thus excluding them. Additionally, there is growing attention toward UGS inequality among different age groups. Due to physical limitations, children and the elderly are more vulnerable to environmental threats [17]. On one hand, issues like obesity, anxiety, and depression pose potential threats to children’s mental and physical health, which could be mitigated by access to UGSs [18,19]. On the other hand, with urban areas facing an increasingly aging population, the elderly become an important user group of UGSs [20]. Access to UGSs is crucial for reducing the incidence of certain diseases and enhancing the well-being of the elderly [20,21].

In order to propose reasonable planning strategies, comprehensive assessments of inequality need to be conducted based on multiple spatial data sources using various models and analysis techniques. Recent studies have begun to utilize more advanced technologies to facilitate data acquisition and enhance the accuracy of results. Firstly, Python scripting proves to be a useful tool for efficiently gathering large volumes of data in a short time and has been widely employed in UGS-related research. For example, Liu et al. [22] conducted a cross-cultural comparison of UGS perceived quality using social media data collected through Python. Similarly, Zhang et al. [22] employed Python-based image data collection to assess UGS quality and measure accessibility. Secondly, real-time navigation and route planning have become valuable tools for obtaining accurate travel time data. For instance, Zhang et al. [23] utilized web service APIs from the Amap open platform to collect and analyze travel time data using Python scripts. Chen et al. [7] developed a method for measuring accessibility based on the Amap API, enabling a more precise assessment of UGS accessibility in Shanghai. Furthermore, unmanned aerial vehicle (UAV) observation technology has been employed in studies such as vegetation surveys and green roof observations [24,25], demonstrating significant potential for obtaining high-resolution images to accurately identify UGS structures.

Regarding assessment methods, the spatial inequality of UGSs is typically characterized using accessibility, which measures the opportunity for residents to access public service facilities. The two-step floating catchment area (2SFCA) method [26,27] is a commonly used accessibility measurement model, considering both facility supply and demand size as well as travel distance costs, and has been widely applied in recent research. Scholars have proposed several improved models to enhance the accuracy of results, including measurements of distance attenuation effects [3], improvements to fixed catchment areas [28,29], and enhancements to fixed transportation modes [30,31]. Additionally, other scholars have taken into account the competitive effects among multiple facilities. Prior to the 2SFCA method, they introduced the calculation of selection weights, resulting in the 3SFCA method [32]. Furthermore, Luo [33] considered the service capacity of facilities and distance costs. They introduced the Huff model [34] to further optimize the calculation method of selection weights, resulting in the H2SFCA method.

While considerable progress has been made in studying the inequality of UGSs, there are still some shortcomings in existing research. Firstly, there is limited research focusing on the assessment of accessibility to different types of UGSs. In fact, UGSs, based on differences in internal structure, have varying construction costs and may provide differentiated service functions. A precise assessment of accessibility to different types of UGSs can help to identify the types of services lacking in different areas, thereby achieving low-cost and high-efficiency layout optimization. Secondly, existing 2SFCA models still have several shortcomings: (1) the measurement of UGS supply scale only uses area as a measure, without considering the internal structural elements of UGSs, which may lead to inaccurate measurements of accessibility; (2) existing 3SFCA and H2SFCA models also overlook the impact of UGS structural elements on the calculation of resident selection weights; (3) the use of distance measurement ignores the travel speed of residents when using different modes of transportation; (4) typically, the centroid of a UGS is used as the supply point, ignoring the distance between the centroid of a larger-scale UGS and its entrance. Finally, there is limited research on the association between accessibility and urban spatial structure factors [8]. More in-depth research is needed to explain the causes of accessibility from multiple perspectives.

To address these shortcomings, this study first obtained basic data using advanced technologies such as Python scripting, UAV imagery acquisition, and the Amap API. Then, the 2SFCA model was improved from both supply and demand aspects, and spatial and socioeconomic inequalities of different types of UGSs were assessed separately (Figure 1). In conclusion, this study aims to provide feasible solutions for mitigating UGS inequality at a low cost and high efficiency through more accurate assessment methods of UGS inequality.

2. Materials and Methods

2.1. Study Area

Harbin, located in northeastern China, is a significant large-scale industrial city in the country. This study selected the main urban area of Harbin as the research area (Figure 2). There are several reasons for this selection. Firstly, compared to surrounding areas, the main urban area has relatively complete urban planning and construction. It concentrates a significant number of UGS resources in the region with a high population density, making it highly meaningful for studying UGS accessibility. Second, existing UGSs in this area exhibit significant differences in scale and internal structure, aligning with the objectives of our study.

2.2. Data Sources and Processing

2.2.1. UGS Boundary and Entrance Data

UGS boundary and entrance data were obtained from the Amap Open Platform (https://lbs.amap.com/, accessed on 2 April 2023). UGSs selected for this study met the following criteria: (1) had a certain area of hard surface available for activities, (2) had any type of vegetation, and (3) were free and open to the public. A total of 133 UGS areas were identified within the study area, including parks, greenways, and some greened plaza areas. We developed a Python program to collect the AOI boundary data of UGSs based on the Amap Open Platform API.

High-resolution satellite images of Harbin city were obtained from LocaSpaceViewer (LSV) (using Google Maps 2022 edition) and used as a base map. The entrances of each UGS were manually extracted on the ArcGIS platform. This was carried out because larger UGSs may have multiple entrances, and the distances between these entrances and the centroid can be considerable. This approach improves the accuracy of travel time calculations. For UGSs without clearly defined entrances (such as squares), the centroid was used as a representative supply point.

2.2.2. Web Photos and High-Resolution Aerial Imagery of UGSs

Web photos were sourced from Baidu Images and Dianping.com, accessed on 22 June 2023. A Python program was used to batch collect UGS photos, totaling 5425 images. For street corner parks, waterfront green spaces, and other areas without image sources, images were obtained from Baidu Street View.

Imagery data were obtained from high-resolution images captured by unmanned aerial vehicles (UAVs) in June 2023. After the UAVs autonomously flew and captured images according to programmed routes, the images were manually stitched together. The resulting image data will be used to assess the types and quantities of plants and water features within each UGS. In compliance with China’s Interim Measures for the Administration of Unmanned Aerial Vehicles, flight permits were not obtained for certain residential, commercial, and industrial areas within the main urban area. For UGSs in these areas, manual identification and assessment were conducted based on Google satellite imagery and street view maps.

2.2.3. Population Data

Block-level analysis units were used in this study instead of administrative districts, facilitating the exploration of micro-level distribution patterns of UGS accessibility and more accurately identifying underserved areas. Basic population data were derived from the 7th National Population Census data (street-level) from 2020. Housing estate data for Harbin City (as of April 2023) were obtained from the Anjuke website (https://www.anjuke.com/, accessed on 6 April 2023) using Python. Utilizing housing estate data as auxiliary information, the population data at the subdistrict level were spatialized to the block units based on the proportion of total households within blocks to the total households within subdistricts.

2.2.4. Road Network and Travel Time Data

Road network data were sourced from the OpenStreetMap (OSM) platform. Walking was selected as the mode of travel in this study, as it is the most common mode of transportation for residents. Walking time was used instead of walking distance to more accurately reflect the real impedance between supply points and demand points. The Amap Web Service provides a routing API that offers walking, public transportation, driving query, and distance calculation interfaces in HTTP format. This method provides optional routes and travel times based on real-time traffic conditions, road networks, and modes of transportation [7]. Leveraging this functionality, walking time data from the starting point (block centroid) to the destination (UGS entrance) were obtained.

2.3. A UGS Supply Adjustment Index (SAI) Considering Physical Activity Promotion Function

2.3.1. Evaluation Framework of SAI

This study primarily considered the physical activity promotion function of UGSs and utilized a supply adjustment index (SAI) to adjust the calculation process of supply size. The evaluation index was derived from summarizing various urban green space quality evaluation tools [22,35,36,37,38] and previous literature on UGS impact on physical activity (

Table 1. Evaluation indicator system of SAI based on multiple data sources.

Dimension	Element	Description	Scoring Method
Facility Conditions	Walkways	Pathways for slow walking in the park	Existence score
	Sports facilities	Fitness areas, dance squares, basketball courts, etc.	Category count
	Leisure facilities	Cafes, pergolas, umbrella seats, etc.	Category count
Natural Conditions	Vegetation	Lawns, dense forests, sparse forests, tree arrays, etc.	Category count
	Water features	Fountains, streams, artificial lakes, etc.	Category count
Safety Conditions	Barrier-free facilities	Ramps, tactile paving, etc.	Existence score
	Lighting facilities	Ground lights, street lights, etc.	Existence score
	Traffic	Separation of pedestrians and vehicles	Existence score
	Security measures	Surveillance cameras, security personnel, etc.	Existence score
Total Score	\	\	Sum of all scores

). It mainly includes three indicator dimensions: facilities [2,39,40], natural environment [41,42], and safety [43,44], as well as nine assessment indicators such as walking trails, physical activity facilities, and recreational facilities. Among these, footpaths and sports facilities are essential prerequisites for promoting physical activity [2,39], while an adequate provision of leisure facilities, vegetation, and water features can indirectly attract residents to engage in physical activity within UGSs [2]. Additionally, sufficient safety conditions are an effective guarantee for physical activity [2,45].

For each indicator, two scoring methods were adopted: existence score (1 point if the content exists, 0 points otherwise) and category count (1–3 points if 1–3 types exist, 4 points if 4 or more types exist, and 0 points otherwise). Three staff members simultaneously observed UGS photos and evaluated and scored them based on the evaluation indicators. For each indicator of each UGS, the average score of all staff members’ scores was taken as the final score for that indicator. After scoring, the scores of indicators in each dimension were summed to obtain the total quality score for each park. The ratio of the total quality score of each UGS to the total score of all indicators was used as the final SAI.

2.3.2. Classification of UGSs Based on SAI

Based on the SAI, this study categorizes UGSs into three types to reflect the supply level of elements related to physical activities within UGSs. (1) Low-supply park (LSP; q ≤ 0.3): these UGS provide fewer types of services and are mainly used for residents’ daily activities nearby. They have advantages such as low construction costs, flexible site selection, and convenient use. (2) Medium-supply park (MSP; 0.3 < q ≤ 0.6): these green spaces can offer more diverse functions, such as basketball courts, children’s play facilities, fitness areas, etc., to meet the needs of different age and interest groups. They have higher construction and maintenance costs and can serve a larger range of resident activities. (3) High-supply park (HSP; q > 0.6): these green spaces have the richest functions and facilities. In addition to basic leisure functions, they may also include larger-scale activity spaces, such as large sports fields. They have the highest construction and maintenance costs and may attract a larger range of residents for activities.

It is important to emphasize that the three types of UGSs have different construction costs and focus on functions. However, this does not mean that UGSs providing fewer services are inferior to other UGSs. In fact, it is unrealistic to demand that all UGSs provide rich services in a city with limited resources. The key is to ensure a reasonable spatial distribution of various types of UGSs. At this level, measuring the accessibility of the three types of UGSs separately can help to more accurately identify the types of services lacking in each neighborhood, thereby achieving higher transformation benefits with lower costs as much as possible.

2.4. The Supply–Demand Improved 2SFCA (SD2SFCA) Method Considering Physical Activity Promotion

2.4.1. Improvement of Supply Scale

The traditional 2SFCA model uses area to represent the supply scale, which overlooks the internal components of UGSs and may result in inaccurate measurements of accessibility. In this study, the calculation method of the supply scale was adjusted based on the SAI of the UGS. The formula is as follows:

$$S_j^A{=S}_j\ast q_j$$

where $S_j^A$ is the comprehensive supply scale of UGS $j$, $S_j$ is the total area of $j$, and $q_j$ is the supply adjustment index of $j$.

2.4.2. Improvement of Demand Scale

The traditional 2SFCA model uses population count to represent demand magnitude. Critics argue that this overlooks the competitive interaction between multiple facilities [32], as residents may only choose a facility that they find more satisfactory. When a resident’s demand has been met by a particular facility, it should be subtracted from the overall demand. Accordingly, Wan et al. [32] introduced the calculation of choice weights before the 2SFCA method, resulting in the 3SFCA method. Luo [33] considered the service capacity of facilities and distance costs, introducing the Huff model [34] to further optimize the calculation of choice weights, resulting in the H2SFCA method. This method has been shown to reduce overestimation of accessibility and improve result accuracy [46,47]. However, these improvements still do not consider the influence of internal compositional elements of facility points. Therefore, this study introduces the proposed SAI into the Huff model to measure residents’ choice weights for supply points. Additionally, the model in this study uses travel time based on road networks instead of traditional travel distance, aiding in a more accurate estimation of travel costs. The formula for calculating choice weights is as follows:

$${Prob}_i=\frac{S_j^A{t}_{ij}G\left(t_{ij}{,t}_0\right)}{\sum _{k\in \left\{t_{kj}\le t_0\right\}}{S}_j^A{t}_{ij}G\left(t_{kj}{,t}_0\right)}$$

$$G\left(t_{ij}{,t}_0\right)=\left\{\begin{array}{r}\frac{e^{-\frac{1}{2}\times {\left(\frac{t_{ij}}{t_0}\right)}^2}{-e}^{-\frac{1}{2}}}{1{-e}^{-\frac{1}{2}}},\hspace{1em}{t}_{ij}\le t_0\\ 0,\hspace{1em}{t}_{ij}>t_0\end{array}\right.$$

where ${Prob}_i$ is the probability of resident point $i$ choosing park $j$, $t_{ij}$ is the travel time from $i$ to $j$, $t_0$ is the time threshold for the specified search range, and $G\left(t_{ij}{,t}_0\right)$ is the Gaussian decay function, with other parameters having the same meanings as above.

Next, centered at each supply point, the total population demand within the catchment area is calculated, so the supply–demand ratio of $j$ is

$$R_j=\frac{S_j^A}{{\sum }_{k\in \left\{t_{kj}\le t_0\right\}}{Prob}_{i}G(t_{kj},t_0)D_k}$$

where $R_j$ is the ratio of the facility scale at supply point $j$ to the population served within the search radius ($t_0$) and $D_k$ represents the demand scale at demand point $k$, represented by the total population of $k$, with other parameters having the same meanings as above.

Finally, for each demand point $i$, search all supply points $j$ within the time threshold range $t_0$ of $i$, summing $R_j$ for each demand point, and adjusting the summation process using the selection probability and Gaussian function, resulting in the accessibility $A_i$ at point $i$:

$$A_i=\sum _{j\in \left\{{\mathrm{t}}_{ij}\le {\mathrm{t}}_0\right\}}{Prob}_{i}G\left({\mathrm{t}}_{ij}{,t}_0\right)R_j$$

2.4.3. Improvement on Fixed Travel Time Threshold

Drawing upon previous research, residents may be willing to spend more time traveling to UGSs that offer additional services [28,47]. Conversely, for UGSs with fewer amenities, residents may incur shorter time costs. Therefore, a variable time threshold was used instead of the traditional fixed time threshold. The walking time thresholds for LSPs, MSPs, and HSPs were set to 10 min, 20 min, and 30 min, respectively, to more accurately reflect residents’ travel behavior preferences.

2.5. Association between Block Characteristics and UGS Accessibility

After analyzing the spatial differences in accessibility, we further aimed to identify the specific associations between different block characteristics and UGS accessibility. Drawing on previous research [6,48,49], this study selects eight variables from three dimensions of block characteristics: population age structure, socioeconomic status, and built environment features (

Table 2. Potential influencing factors of accessibility.

Indicator	Variable	Explanation
Age	Proportion of old people (%)	Proportion of population aged 65 and above within each block
	Proportion of children (%)	Proportion of population aged 16 and below within each block
Socioeconomic status	Housing price	Average housing price within each block
Built environment features	Population density	Total population quantity divided by block area
	Building density	Building footprint area divided by block area
	Greenery rate	Total green space area within block divided by block area
	Block age (year)	Median construction year of residential areas within each block
	Distance to the city center	Euclidean distance from block centroid to regional center

).

The selection of these variables is driven by three reasons. Firstly, we consider the proportion of elderly and adolescent populations in blocks to observe whether they face UGS inequality, as they are often considered high-demand groups for UGSs. Secondly, we examine the association between block socioeconomic status and accessibility using average housing prices. Thirdly, block population density, building density, and green space ratio represent land use characteristics, while the age of construction of blocks primarily reflects their relative age, and the distance to the city center reflects block location conditions. These built environment variables can to some extent reflect the level of urban development and may further influence UGS accessibility.

Initially, a multiple linear regression model was established, with the eight block characteristic factors as independent variables and the accessibility of various types of UGSs as the dependent variable, to examine potential influencing factors on accessibility from a global perspective. Subsequently, a focused analysis was conducted on the association between accessibility and socioeconomic status. Housing prices were divided into four levels (low, medium-low, medium-high, and high) based on quartiles, and the accessibility of UGSs and median values of block characteristic variables for each price level were calculated. The Mann–Whitney U test was then employed to compare the accessibility of UGSs at different price levels pairwise. Data analysis was conducted using SPSS 21.0 and Geoda 1.14.0 software.

3. Results

3.1. Classification Results of UGS Based on SAI

Table 3. Calculation results of SAI and classification results statistics.

	N	Minimum	Maximum	Mean	Sum	Std. Devitation
LSP	32	0.14	0.29	0.25	8.00	0.04
MSP	65	0.33	0.52	0.41	26.67	0.07
HSP	36	0.67	0.90	0.70	25.19	0.09

and Figure 3 present the classification results and spatial distribution of the 133 UGS within the area, respectively. Overall, UGS distribution is dense in the western part of the city, with a variety of types, while UGSs in the south and east are relatively sparse. MSPs are the most numerous, accounting for 48.87% of the total, concentrated in the northeast. LSPs have the fewest numbers, accounting for 24.06%, mostly distributed in the west and central areas. HSPs are concentrated in the west, followed by the southeast.

3.2. Measurement Results of UGS Accessibility

3.2.1. Differences in Accessibility of Three Types of UGSs

Figure 4 illustrates the accessibility of overall UGSs (OP) as well as the accessibility results for three different types of UGSs. For the overall UGSs, high accessibility values are concentrated in the western, riverside, southeastern, and eastern areas within the second ring road of the city. Accessibility is generally lower in the southern and central parts of the city. Although these areas have a small number of UGS resources, their overall area and internal facilities are insufficient to match the higher population demands. Accessibility is poorest in the eastern outskirts, where obtaining any services is nearly impossible.

When considering the accessibility of the three types of UGSs separately, the results show that only about 11.72% of blocks can reach LSPs (

Table 4. Descriptive statistics of UGS accessibility.

	Min	Max	Mean	Median	Standard Devitation	Serviced Blocks	Serviced Area/km2	Serviced Population
LSP	0.00	20.50	0.11	0.00	1.04	127	12.14	390,967
MSP	0.00	98.05	1.60	0.18	4.82	674	95.52	2,553,738
HSP	0.00	193.44	5.25	1.33	13.01	795	110.42	2,965,519
OP	0.00	193.44	6.96	2.37	15.43	940	136.74	3,611,683

), mainly distributed in the western and central parts of the city. The proportion of blocks served by MSPs is 62.18%, with accessibility generally decreasing from the suburbs toward the central urban areas, with high values concentrated in the western, riverside, and northeastern parts of the city. HSPs can serve 73.34% of blocks. Their accessibility distribution is more dispersed, with high values concentrated in the western, riverside areas, eastern, and southeastern parts.

It is evident that the number and distribution of LSPs within the region are severely unreasonable, and a significant portion of the high values in overall accessibility largely depend on HSPs. However, the number of HSPs is limited and not sufficient to serve a wider range of blocks. If only the overall accessibility of UGSs is evaluated, it will be difficult to identify the specific types of services that blocks truly lack.

3.2.2. Classification of Areas with Inadequate Services

Based on the results of accessibility assessments, further classification of areas with inadequate services was conducted. Firstly, an overall classification was conducted, as shown in Figure 5a.

Class I: Represents blocks with no access to any services, accounting for approximately 13.28% of the total. They are mainly distributed in the central, eastern, and southern suburban areas.

Class II: Represents blocks with access to only one class of UGS service, with relatively low accessibility, accounting for approximately 19.93% of the total. They are mainly distributed in the southern region. These blocks have relatively limited access to UGS classes, which may create significant service pressure and may not meet residents’ diverse usage preferences. Subclassification of class II results is shown in Figure 5b. Most blocks in the south can only access HSPs with limited opportunities, and a very small number of blocks can only access LSPs. In contrast, a few blocks in the east, central, and north can only access MSPs.

Class III: Represents blocks with access to two or more classes of UGS services, but with relatively low accessibility. This is mainly due to the limited number of UGS classes, and the overall supply scale does not match the higher population demand. This class accounts for approximately 6.83% of the total and is mainly distributed in the southern and northeastern parts of the city.

3.3. Association between UGS Accessibility and Block Characteristics

3.3.1. Potential Influencing Factors of UGS Accessibility

Table 5. Multiple linear regression results.

Independent Variables	VIF	Standardized Coefficient
		OP	LSP	MSP	HSP
Proportion of old people	1.352	0.124 ***	−0.031	0.150 ***	0.094 **
Proportion of children	1.257	−0.057	−0.019	−0.066 *	−0.042
Housing price	1.409	0.378 ***	0.089 *	0.399 ***	0.294 ***
Population density	1.224	−0.113 ***	−0.007	−0.083 **	−0.103 ***
Building density	1.470	0.010	−0.059	−0.015	0.022
Greenery rate	1.223	0.14 ***	−0.016	0.040	0.152 ***
Block age	1.462	−0.066 *	−0.019	−0.022 *	−0.069 *
Distance to the city center	1.378	0.283 ***	0.072 *	0.260 ***	0.234 ***
Adjusted R2	-	0.208	0.017	0.168	0.149
p-value of the model	-	0.000	0.001	0.000	0.000

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

shows that the variance inflation factor (VIF) for each explanatory variable is less than 7.5, indicating the absence of multicollinearity between variables. For overall UGS accessibility, the proportion of elderly population, housing prices, block greenery rate, and distance to the city center are significantly positively associated with accessibility, while block population density and construction year are significantly negatively associated with accessibility.

When considering different quality levels of UGSs, housing prices and distance to the city center have stable positive associations with all quality levels of UGSs. The block greenery rate has a significant positive association with HSP accessibility. Block age and population density have weak negative associations with MSP and HSP accessibility. The proportion of elderly population has a significant positive association with MSP and HSP accessibility, while the proportion of children population has a weak negative association only with MSPs. The block greenery rate has a significant positive association with HSP accessibility.

This indicates that children may be at a disadvantage in accessing adequate MSP services, while the elderly population may have a more matched accessibility to UGSs. Blocks with higher socioeconomic status, farther distance from the city center, newer construction age, and lower density are more likely to access UGSs. Blocks with higher greenery rates are also more likely to access HSPs.

3.3.2. Disparities in UGS Accessibility and Block Characteristics Based on Housing Price Levels

As shown in

Table 6. Block characteristics variables statistics based on housing price levels.

	(a) Low	(b) Low-Middle	(c) Middle-High	(d) High
Proportion of old people (%)	24.55	25.30	25.18	19.72
Proportion of children (%)	9.29	9.11	9.07	11.29
Population density (person/km2)	27,726	42,947	33,893	25,941
Building density	0.49	0.52	0.53	0.38
Greenery rate	0.13	0.10	0.10	0.13
Block age (year)	23	25	24	18
Distance to the city center (m)	6836.77	4716.20	4179.25	5173.10

, compared to low-priced blocks, high-priced blocks have a lower proportion of elderly population and a higher proportion of children population. However, high-priced blocks exhibit a significantly lower population density and building density. The construction age of high-priced blocks is also generally later, and they tend to have higher greenery rates compared to low-priced blocks. Low-priced blocks may generally be situated farther from the city center. Additionally, the median MSP accessibility in low-priced blocks is slightly higher than in middle-low-priced blocks. This may be due to the fact that low-priced blocks are located in areas farther from the city center, where more UGSs are built, leveraging the surrounding natural resources.

The results of the Mann–Whitney U test in

Table 7. Mann–Whitney U test results for UGS accessibility based on housing price levels.

	LSP	MSP	HSP	OP
(a) Low	0.00 (d **)	0.15 (d ***)	0.55 (b */c **/d ***)	1.65 (c */d ***)
(b) Low-middle	0.00 (d **)	0.10 (d ***)	0.85 (a */d ***)	2.21 (d ***)
(c) Middle-high	0.00	0.17 (d **)	0.86 (a **/d ***)	2.48 (a */d ***)
(d) High	0.00 (a/b **)	0.43 (a ***/b ***/c **)	1.30 (a/b/c ***)	4.47 (a/b/c ***)

Note: * p < 0.05, ** p < 0.01, *** p < 0.001. The letters before the housing price levels denote the group numbers. The groups with significant differences and their significance levels are indicated in parentheses after each median.

indicate significant differences in the accessibility of MSPs and HSPs between blocks with high housing prices and those with other housing price levels. The median accessibility of UGSs in blocks with high housing prices is consistently the highest. Conversely, the accessibility of HSPs in blocks with low housing prices is significantly lower than that in the other three types of blocks. It is noteworthy that, although blocks with both high and low housing prices have relatively high median green coverage rates, the accessibility in blocks with low housing prices is significantly lower than that in blocks with high housing prices. This could be attributed to blocks with low housing prices having many underutilized potential UGSs, such as affiliated green spaces and protective green spaces.

4. Discussion

4.1. Disparities in Types of UGS Services Obtained across Different Regions

The research findings indicate significant differentiation in the types of UGS services available in different urban areas. Firstly, the western part of the city facilitates access to various types of UGSs, which aligns with the emphasis on UGS development in newly developed urban areas. The service availability in the riverside areas is also adequate, consistent with previous studies [7,8,48], suggesting a universally positive role of proximity to water bodies in UGS development.

Secondly, most blocks in the eastern suburbs and southeastern parts can only access HSP services. However, due to the limited number of HSPs, their service coverage is relatively small. This suggests that the overall accessibility advantage in certain urban areas may not stem from a variety of green space resources but rather from the local advantages of a few large-scale UGSs. Additionally, most blocks cannot reach LSPs within the specified time threshold, and the accessibility level in the southern part of the city is generally poor, with some blocks only able to access HSP services with low accessibility opportunities. These phenomena may stem from a planning approach based on average indicators. Due to the larger scale and diverse service types of HSPs, they can quickly increase the overall green coverage and per capita green space area in the region to achieve the government’s expected goals [48]. This creates a “superficial” perception of service adequacy and may lead planners to overlook other types of UGSs.

These findings highlight a contradiction: whether, with the same expected total area of UGSs, it is preferable to incrementally increase numerous small-scale UGSs with lower supply capabilities in a decentralized manner or to concentrate on fewer large-scale UGSs with higher supply capabilities in a centralized manner. Previous studies have addressed this issue [12,50], indicating that, compared to geographically concentrating resources and initiating several rounds of upscale development of large UGSs, focusing on smaller-scale interventions is more advantageous for UGS equity and can help to prevent gentrification. This study, from the perspective of different supply capabilities of UGSs, supports this viewpoint. Firstly, the construction and maintenance costs of HSPs are high, and they require large land areas as a foundation, which is unrealistic for high-density urban areas severely lacking in UGSs. Secondly, a single type of UGS may not fully meet the diverse needs of residents and may impose greater service pressures on HSPs. Studies have shown that residents prefer to engage in daily activities in smaller UGSs [51,52], possibly due to the disadvantages of larger UGSs such as crowding and difficulty in accessing activity areas. In contrast, LSPs and MSPs have several advantages, including greater flexibility, ease of encouraging residents to engage in physical activities [53], and contributing to strengthening community ties [54,55], among others. Therefore, in future planning, emphasis should be placed on incrementally supplementing more LSPs and MSPs in a decentralized manner to serve more blocks.

4.2. Association between Block Characteristics and UGS Accessibility

The research results indicate that house prices have a significant positive relationship with UGS accessibility. Except for low-supply parks (LSPs), which have a relatively limited accessible range, the median accessibility of various types of UGSs in high-priced blocks is the highest and significantly different from blocks of other price levels. This supports previous views [7,12], indicating significant socioeconomic inequality in UGS accessibility. This may be because high-income groups are generally willing to pay high costs for a better living environment [56] and have more opportunities to participate in decision-making processes related to their interests [57]. This may increase the housing prices in surrounding blocks [58], leading to significant segregation in accessing UGS services for low-income groups.

Similar to previous studies, this research also highlights differences in accessing UGSs between the elderly and children [16,22,49]. Areas with a higher proportion of children may exhibit lower accessibility to MSPs, while areas with a higher proportion of elderly residents may have higher accessibility to both MSPs and HSPs. This could be attributed to two factors. Firstly, families with children may choose to reside in older urban areas, which offer abundant commercial and educational resources [49], but lower UGS accessibility. Secondly, with the large-scale industrial transformation in China, the middle-aged and young unemployed population in old industrial areas are forced to migrate elsewhere, leading to an increase in aging population. However, measures such as industrial land replacement and renovation of old communities have also increased green spaces. Therefore, the elderly population may inadvertently benefit from access to urban parks [49].

Urban spatial structural factors partially explain the inequality of UGSs. Areas with high housing prices generally have later construction times and characteristics of low density and high green coverage. This partially explains the accessibility differences between new and old urban areas. New areas have better conditions for UGS construction, while old areas did not emphasize the importance of UGSs in the early stages and lacked comprehensive construction regulations, making it difficult to retrospectively increase UGSs in already built-up areas. However, this planning approach overlooks the lower population demand in new areas, leading to an increase in housing prices in these areas. Consequently, a large number of UGSs are overly concentrated in certain areas, exacerbating spatial inequality in UGSs and creating significant disadvantages for low-income groups in accessing UGSs.

4.3. Urban Planning and Policy Implications

Firstly, research on UGS spatial inequality should not only focus on areas lacking services but also consider the differences in UGS service types obtained in different areas. For example, the lower accessibility in the central urban area is due to the severe shortage of UGS supply, requiring prioritized expansion of the UGS total area. In contrast, for some blocks in the southern region, a more reasonable approach than continuing to increase HSPs is to add more LSPs and MSPs near these blocks and establish complete connections with existing HSPs. This can achieve greater benefits at lower costs.

Secondly, in terms of specific renovation strategies, attention should be paid to adopting stock planning schemes in areas where it is difficult to increase UGSs. There are three specific recommendations: (i) emphasize avoiding large-scale concentrated development and focus on small-scale interventions to supplement more LSPs and MSPs to bridge the gap; (ii) enhance facilities and services in existing LSPs to provide a more diverse range of functions to alleviate inequality [35]; (iii) advocate for open neighborhood planning to maximize the public benefits of closed non-park green spaces [59,60].

Finally, planners should prioritize addressing green inequality issues by actively intervening with policies and allocating funds to safeguard the health rights of low-income and vulnerable groups. At the same time, attention should be paid to increasing corresponding UGS service facilities in areas with a higher proportion of children and elderly populations to meet their specific needs.

5. Conclusions and Future Work

This study’s main contributions are as follows: firstly, we utilized advanced data collection techniques, including Python data collection, drone image capture, and image stitching technology. This reduced the difficulty of obtaining data and improved the accuracy of the research results. Secondly, we considered elements related to physical activity in the measurement of UGS accessibility and improved the 2SFCA model from both supply and demand aspects, making the evaluation of accessibility more accurate. Thirdly, we separately evaluated the accessibility of different types of UGSs, which helps to formulate more precise planning strategies to ensure the maximization of benefits with limited construction costs. Fourthly, we evaluated the association between various urban spatial structure variables and accessibility, further explaining the reasons for UGS inequality based on previous research.

The conclusions of this study indicate the following: (1) there is significant differentiation in the types of UGS services available in different areas of the city. Government reliance on a single indicator for planning may lead to a severe lack of small-scale, low-cost UGSs, which hinders the full satisfaction of residents’ diverse needs. (2) The accessibility of UGSs is significantly positively influenced by housing prices, with accessibility in high-priced areas notably higher than in low-priced areas, demonstrating clear socioeconomic inequality. (3) Children may be at a disadvantage in accessing UGSs, highlighting the need for future planning to pay greater attention to vulnerable groups.

Several limitations of this research framework must be acknowledged. Firstly, this study only used objective indicators to evaluate the sports service capacity of UGSs, without considering residents’ subjective perceptions and needs in the park. In reality, residents’ perceived quality of parks may vary [61] and be related to various factors [62]. Secondly, different modes of transportation were not considered for the accessibility evaluation of different types of UGSs. Lastly, due to the limitations of micro-data acquisition, the unequal differences in various vulnerable groups have not been further studied. Future research should be based on more field investigations, adopting a combination of subjective and objective data acquisition methods to further improve the reliability of the results.