Bridging the Green Space Divide: A Big Data-Driven Analysis of Park Accessibility Inequities in Chinese Megacities Using Enhanced 3SFCA Modeling

Abstract

This study enhances our understanding of urban park accessibility and social equity through a novel methodological framework in Chengdu, China. By improving the three-step floating catchment area (3SFCA) method with space syntax metrics and multi-modal transportation analysis, we identify spatial disparities in green space access. Our methodology, validated with Baidu heat map data, demonstrates improved accuracy in estimating population demand patterns. Key findings include: (1) The enhanced 3SFCA method outperforms traditional approaches in predicting park accessibility, providing reliable evidence for urban planning; (2) significant accessibility disparities exist across transportation modes, particularly affecting non-motorized transport users; (3) newly developed areas show greater park access inequities than established neighborhoods; (4) important mismatches exist between park accessibility and vulnerable population distributions. This research provides targeted recommendations for reducing spatial inequities and improving green space access for all residents, particularly benefiting children and elderly populations in rapidly urbanizing contexts.

1. Introduction

With the acceleration of global urbanization, the proportion of the urban population continues to rise. According to data from the World Health Organization (2020), currently, half of the world’s population resides in urban areas, and this proportion is projected to exceed 70% by 2050 [1]. As densely populated spaces for human life and economic activity, cities face significant challenges in achieving the Sustainable Development Goals (SDGs) [2]. Against this backdrop, urban parks, as key elements in the development of sustainable urban landscapes, play an indispensable role in enhancing citizens’ quality of life and maintaining the ecological integrity of cities [3].

Urban parks are a core component of urban green infrastructure, serving key functions in ecology, landscape, and recreation [4]. An increasing number of studies confirm that parks provide urban residents with essential opportunities to engage with natural environments. As green spaces for leisure and physical activity, they have a significant impact on public health [5]. Research indicates that regular interaction with parks can play a positive role in promoting physical activity, improving mental and physical health, reducing obesity, and alleviating stress [6]. However, along with the urbanization process, large cities are generally facing multiple challenges, including the reduction or insufficient supply of urban green spaces [7,8], the surge in park demand driven by population growth [9,10], and the uneven distribution of parks [11,12]. These issues highlight the critical importance of park accessibility in evaluating spatial equity within urban areas.

In recent years, Chengdu, a leading “Park City” initiative pilot city in China, has expanded its parks to integrate ecological concepts into urban development, improving livability and quality of life. However, spatial mismatches persist in park planning: dense urban cores lack adequate green space, while large parks exist primarily in sparsely populated suburban areas. Park development often triggers unregulated real estate growth, causing environmental gentrification that compromises equitable park access. Studying park accessibility and balancing supply with demand is therefore essential for public health, livability, and spatial equity.

This study examines Chengdu’s core urban areas through the 15-min city concept, assessing park accessibility and equity. We combine spatial analysis, statistical methods, and GIS technology to analyze urban park distribution, green space accessibility patterns, and resource allocation challenges for vulnerable populations. Our research provides evidence-based recommendations to optimize park planning and enhance spatial equity.

The paper is structured as follows: Section 2 reviews literature on urban park accessibility and the 15-min city concept. Section 3 describes data sources and methodology. Section 4 examines park accessibility across transportation modes, compares multi-modal versus single-modal accessibility landscapes, and investigates resource matching between accessibility and equity in Chengdu’s core areas. We also analyze spatial inequities affecting vulnerable groups (elderly and children) using clustering methods and propose targeted solutions. Section 5 summarizes findings, discusses methodological contributions, and suggests future research directions. Section 6 presents conclusions.

2. Literature Review

2.1. Evolution of the Concept of Accessibility and Its Measurement Methods

Accessibility, as a key indicator for evaluating spatial equity, has undergone extensive evolution and development in both its concept and measurement methods. The foundation of accessibility studies can be traced back to the “population potential” concept introduced by Stewart [13] based on population gravity models. Later, Hansen [14] refined the definition by describing accessibility as the potential for interaction opportunities, making the concept more systematic and operational. Early accessibility research primarily focused on single dimensions, such as distance costs, time costs, or economic costs associated with travel between two locations [15]. As research progressed, accessibility began to be considered within the framework of social equity, and has since been widely applied to explore spatial equity issues in public service facility distribution [16].

In terms of measurement methods, accessibility research has evolved from simple to more complex approaches. Early methods included the population-to-provider ratio [17] and travel impedance methods [18]. Subsequently, gravity-based methods, such as potential models, Huff models, and kernel density approaches [19] have gained widespread application. Since the 2000s, the cumulative opportunity approach, particularly the two-step floating catchment area (2SFCA) method and its improved variants, has been favored by scholars for its flexibility and accuracy [17,20].

However, the traditional 2SFCA method has certain limitations, such as neglecting distance decay effects and using fixed catchment sizes [19,21]. To overcome these shortcomings, scholars have proposed various improved models, such as the enhanced two-step floating catchment area (E2SFCA) method [19], the kernel density two-step floating catchment area (KD2SFCA) method [22], the enhanced variable width floating catchment area (EV2SFCA) method [23], the three-step floating catchment area (3SFCA) method [24], and Gaussian-based models [25]. These improved models address the limitations of traditional methods to varying degrees, providing new tools for more accurate accessibility assessments.

2.2. Research Progress on Accessibility of Urban Parks

Urban park accessibility, a key area within accessibility research, has received growing scholarly attention. Research has primarily focused on multiple transportation mode integration as an emerging trend. Researchers now incorporate various transportation modes into accessibility models for more comprehensive assessments [26,27,28]. These studies show that analyzing multiple transportation modes more accurately reflects accessibility differences among social groups. For example, Mao [29] developed the multi-modal two-step floating catchment area method (2SFCAM), which integrates public transit and car travel into the traditional model. Similarly, Tao [30] studied walking and driving impacts on spatial equity by incorporating these travel modes into a potential model.

The diversification of park attractiveness indicators has also become a key focus in research. Early studies primarily used park size as the sole indicator [31]. However, as research has progressed, scholars have recognized that park accessibility is influenced by additional factors such as park functionality, type, landscape quality, and available facilities [32]. Some researchers have begun incorporating these factors into accessibility models to provide a more comprehensive assessment. For example, Fu [33] integrated park size and function into the 2SFCA model, while Dong [34] considered the number of park facilities as an attractiveness factor.

The selection of time thresholds and distance measures is another important aspect of research. When considering walking as a mode of transportation, time thresholds of less than 20 min are most commonly used [35,36]. Additionally, studies have shown that network distance is more accurate and realistic compared to Euclidean and Manhattan distances [37]. These findings provide crucial references for more precise assessments of park accessibility.

Studying park accessibility from the perspective of social equity has become a significant trend. Increasingly, research is focusing on the disparities in park accessibility among different socioeconomic groups, defining spatial equity as the degree to which facilities or services are evenly distributed across various populations [38,39]. This research approach helps identify inequalities within urban areas and provides a basis for developing more equitable urban planning policies.

2.3. Research Deficiencies and Innovations

Our literature review identifies several limitations in current urban park accessibility research. First, many studies ignore transportation mode impacts, overlooking differences in residents’ travel capabilities. Second, research frequently employs simplistic 2SFCA models that disregard distance decay effects, potentially biasing evaluation results. Third, the use of straight-line rather than network distances reduces assessment accuracy. Fourth, few studies consider park attractiveness factors, limiting accurate appeal assessment. Fifth, minimal research compares multi-modal with single-modal accessibility models, restricting our understanding of how different travel modes affect accessibility. Lastly, traditional quantitative approaches neglect important non-physical factors like safety, culture, and personal preferences, creating discrepancies between evaluation results and reality.

Based on the identified research gaps, this study introduces several innovations. (1) It incorporates a park attractiveness coefficient (based on facility size) as an evaluation metric to more accurately reflect the actual appeal of parks, and integrates topological distance into the park attractiveness competition coefficient, enhancing the accuracy of population estimates at demand points and improving the fit of the 3SFCA model. (2) Various transportation modes (walking, cycling, public transit, and private car) are integrated into the improved 3SFCA model to comprehensively assess park accessibility and equity across different travel modes. (3) The study compares the spatial differences in park accessibility and equity between multi-modal and single-modal approaches, offering a new perspective on how different transportation modes influence these outcomes. (4) Using a four-quadrant clustering method, this study analyzes the relationship between accessibility and the rates of child and elderly populations in different areas, providing optimization suggestions to improve access for vulnerable groups and enhance residents’ travel satisfaction.

This study focuses on the core urban area of Chengdu, China, aiming to enhance the understanding of urban spatial equity by improving the measurement methods of park accessibility. Specifically, the research addresses the following objectives: the landscape characteristics of park accessibility across different scenarios under a multi-modal model; the differences in park accessibility landscapes between multi-modal and single-modal models; whether spatial mismatches exist between parks and population in Chengdu’s core areas; and the policy implications of the findings for Chengdu and other cities. Through these innovative research designs, we aim to provide new insights and methodological references for urban park planning and spatial equity studies.

3. Data and Methods

3.1. Study Area

Chengdu, the capital of Sichuan Province, is renowned as the “Land of Abundance” and serves as a major political, economic, and cultural hub in western China. Located in central Sichuan and the western part of the Sichuan Basin, Chengdu is the central city of the southwestern region and a key component of the Chengdu–Chongqing economic zone. The study designates Chengdu’s core urban districts as the study area (Figure 1), encompassing Jinjiang, Qingyang, Jinniu, Wuhou, Chenghua, Xindu, Pidu, Wenjiang, Shuangliu, Longquanyi, Qingbaijiang, and Xinjin Districts. The study area is delineated by the following boundaries: eastern boundary: the western hillside of Longquan Mountain (incorporating Chenghuan Road and Automobile City Boulevard); southern boundary: the Science City Middle Road–Tianfu International Airport Expressway corridor; western boundary: the Wenpi Avenue–Chengfei Avenue extension zone; northern boundary: the Pihe Ecological Corridor (including Xiangcheng Avenue). The total area encompasses approximately 1057.5 square kilometers with a permanent resident population of 15.4194 million. Chengdu governs 12 districts, five county-level cities, and three counties. According to the seventh national population census, Chengdu has become the fourth mega-city in China with a permanent population exceeding 20 million, having increased by nearly 6 million over the past decade. The core urban area, which is the focus of our study, is defined by Chengdu’s Urban Spatial Master Plan (2021–2035) and includes the districts of Xindu, Wuhou, Wenjiang, Shuangliu, Qingyang, Qingbaijiang, Pidu, Longquanyi, Jinjiang, Jinniu, and Chenghua. These districts consist of a total of 120 subdistricts, covering an area of 124,155.068072 hectares. To account for edge effects and the maximum service capacity of parks, we also include urban parks and green spaces (UPGS) within a 5 km buffer zone around the study area, extending the scope by applying a 5 km buffer to the boundary of the research area.

3.2. Data Collection and Processing

3.2.1. Supply Data Processing

As shown in Figure 2 and

Table 1. Classification of Park Grades and Service Radius.

Category	Park Grade	Target Population	Suitable Area (hm2)	Service Radius (m)	Transportation Mode
Comprehensive park	1	Urban residents	≥38.12	5000	Walking, bicycle, public bus, private car
Specialized park	2		4.82 < 38.12	2500	Walking, bicycle, public bus
Community park	3	Residential area residents	2.85 < 4.82	1000	Walking, bicycle
Miniature garden or pocket park	4	Elderly and children in residential areas	0.5 < 2.8	500	Walking

, the dataset of urban parks and green spaces includes 63 comprehensive parks, 266 specialized parks, 104 community parks, and 409 small green spaces. The dataset encompasses both the scale and distribution of urban parks within the study area. First, we obtained the locations and sizes of urban parks in the study area using the application programming interface (API) provided by Amap (https://lbs.amap.com, accessed on 28 June 2024). We then downloaded basic information on major parks from Chengdu’s Urban Public Data Open Platform (http://data.chengdu.gov.cn/oportal/index, accessed on 28 June 2024), which included park names, classifications, and sizes. By cross-referencing these two datasets, we constructed the final dataset of supply points.

Previous studies [33] show that community parks and small green spaces, which constitute most parks in our study, have shapes that lead residents to use multiple entry/exit points rather than designated entrances. We therefore used geometric centroids of gridded urban park areas as supply points instead of park entrances. Using ArcGIS 10.8.2, we divided park areas of interest into 500 m × 500 m grids, creating 13,304 supply points. This approach better reflects park-to-road proximity, particularly for irregularly shaped parks like linear parks, providing more realistic representation than single center points.

Park service radius varies by type and size. As a “Park City” pilot, Chengdu requires higher evaluation standards and practical recommendations. Based on the “Chengdu Urban Park and Green Space System Plan (2019–2035)” draft and “GBT 51346-2019 Urban Green Space Planning Standards”, we classified parks by calculating average area (dividing total planned park area in Chengdu’s core urban area for 2035 by park quantity). Parks exceeding 38.12 hectares are classified as comprehensive parks (Grade 1) with a 5000 m service radius, accommodating four transportation modes: walking, cycling, public transit, and private cars. Parks between 4.82 and 38.12 hectares are specialized parks (Grade 2) with a 2500 m service radius, supporting three transportation modes: walking, cycling, and public transit. Table 1 provides additional details.

3.2.2. Demand Data Processing

As shown in Figure 2, the population data used in this study represents the residential population at the community level within Chengdu’s core urban area. Due to the difficulty in directly obtaining accurate population figures for each community, we estimated community-level populations by integrating the 7th National Population Census data (released in November 2020 at the district/county level) with community point data from Anjuke’s website (collected in May 2024). The community point data includes information such as name, location, and number of households (filtered to include only valid community points with more than 10 households).

We first used Python 3.11 to scrape community point data from Anjuke’s APP, as recorded in the Amap POI dataset from May 2024. We then performed population correction using ArcGIS 10.8.2. For each community, we calculated a correction coefficient by dividing its household count by the total households in its district/county. This coefficient was multiplied by the district/county’s total population to estimate each community’s population. This method yielded a total population of 12,932,105 across all communities.

3.2.3. Road Network Data Processing

The road network data was downloaded from OpenStreetMap (https://www.openstreetmap.org/, accessed on 28 June 2024) and adjusted using map data provided by “Tianditu” from China’s National Public Geographic Information Service Platform (https://www.tianditu.gov.cn, accessed on 28 June 2024) within ArcGIS 10.8.2. Further calibration was performed by comparing it with the core urban areas outlined in the “Chengdu Urban Spatial Master Plan (2021–2035)” to ensure geographic alignment for the study area. Based on the “Chengdu Urban Transportation 13th Five-Year Plan”, five road types were selected as the foundational road network data: highways and elevated roads, urban expressways, primary roads, secondary roads, and local streets. Figure 1 and Figure 2 show the overview of the study area and the distribution of road networks, urban parks, and population density within Chengdu’s core urban area.

3.2.4. Residents’ Travel Modes and Travel Behavior Management

Travel behavior is a crucial metric in quantifying park accessibility, as it reflects residents’ preferences for modes of transportation and determines the travel cost from their residences to parks. Based on the “Chengdu Green Space System Plan (2013–2020)”, the draft of the “Chengdu Park City Green Space System Plan (2019–2035)”, and the guidelines from the “Shanghai 15-Min Community Life Circle Plan”, as well as considering the optimal time scale for most residents’ travel comfort to parks, this study sets the travel time threshold at 15 min.

Travel speeds were determined according to relevant policy regulations and prior research [40,41,42,43]. For normal adults in urban environments, walking and cycling speeds typically range from 1 to 1.5 m/s and 2.5 to 4 m/s, respectively. Therefore, this study sets the walking speed at 4 km/h and the cycling speed at 10 km/h. According to the “Road Traffic Safety Law of the People’s Republic of China”, the maximum speed limits for buses and private vehicles are 50 km/h and 60 km/h, respectively. However, given that vehicles in congested urban areas rarely reach these speeds, this study adjusts the bus speed to 20 km/h and private vehicle speed to 40 km/h.

To ascertain residents’ transportation modes and behavioral patterns, we conducted random questionnaire sampling at the ten most frequently searched parks on Chengdu’s map (

Table 2. Demographics of respondents and their visits to green spaces.

Type		Num	Percent (%)	Mean
Gender	Male	273	54.6	/(not applicable)
	Female	227	45.4
Age	0–18	16	3.3	Average age 48.19
	18–30	192	39.7
	30–45	126	26
	45–60	104	21.5
	>60	46	9.5
Educational level	Primary school (or below)	34	7	/
	Middle school	101	20.8
	High school	179	36.9
	Bachelor’s degree	145	29.9
	Master’s degree	19	3.9
	PhD	6	1.2
Occupation	Student	38	7.8	/
	Corporate employee	72	14.8
	Government employee	14	2.8
	Public institution Employee	48	9.9
	Retiree	232	47.9
	Freelancer and others	77	15.9
Income (CNY/month)	<3000	183	37.8	Average income 4124.72
	3000–6000	198	40.9
	6000–10,000	67	13.8
	10,000–20,000	29	5.9
	>20,000	4	0.8
Travel time (min)	<10	179	36.9	Average time 12.95
	15–20	116	23.9
	20–30	96	29.8
	30–45	50	10.3
	>45	43	8.8
Travel distance (m)	<500	165	34.1	Average distance 1328
	500–1000	144	29.7
	1000–2000	78	16.1
	2000–4000	58	11.9
	>4000	39	8

). In May 2024, we distributed 500 questionnaires to residents and visitors at these 10 most frequented parks and primary pedestrian centers, successfully collecting 484 valid responses, yielding a response rate of 96.8%. The respondents encompassed diverse demographic characteristics, including gender, age, educational attainment, occupation, income levels, and transportation preferences. The sample size and demographic distribution were methodically designed to ensure representative coverage of the broader population and travel preferences within Chengdu’s core area. This provided foundational data for subsequent analysis.

The travel behavior data presented in

Table 3. Summary Table of Travel Behavior of Chengdu Residents.

Mode of Transportation	Walking	Cycling (Shared Bikes, Personal Bicycles)	Public Transportation (Bus, Excluding Rail Transit)	Private Car (Private Vehicles, Taxis, Ride-Hailing Services)
Proportion of transportation Mode choice	22.20%	10.40%	29.60%	37.80%
Travel time	15 min	15 min	15 min	15 min
Travel speed	4 km/h	10 km/h	20 km/h	40 km/h
Travel distance	1 km	2.5 km	5 km	10 km

(including travel time and distance) was similarly derived from random surveys conducted at the top ten parks in Chengdu’s core area, with validation through previous research findings [41] and the “2023 Chengdu Green Transportation Operation Characteristics Report” to ensure data accuracy and reliability. The four primary transportation modes in Chengdu’s core urban area—walking, cycling, public transit, and private vehicles—account for 22.2%, 10.4%, 29.6%, and 37.8% of residents’ transportation choices, respectively. The typical travel distances for walking, cycling, public transportation, and private vehicles are 1 km, 2.5 km, 5 km, and 10 km, respectively. These data facilitate a comprehensive understanding of behavioral patterns among different groups utilizing various transportation modes to access green spaces.

3.2.5. Baidu Heat Map Data Processing

The spatial vitality of urban parks refers to their ability to support activities and meet user needs within a given space. It encompasses the attractiveness of parks and their capacity to draw crowds, foster activities, and encourage social interaction [44]. Spatial vitality is a crucial indicator of whether space design and construction meet the spatial usage demands and expectations of residents. To measure the spatial vitality of urban parks, this study utilized Baidu heat maps, which outline the geographic location data of Baidu app users at specific time points. These heat maps display user distribution in an area using different colors [45]. Compared to social media check-in data, Baidu heat maps are preferred as they provide real-time data with minimal bias for population dynamics studies. Scholars have compared Baidu heat map data with Weibo check-in data for describing urban park usage, finding that Baidu data offer superior accuracy [46]. Studies indicate that Baidu heat map data can serve as a reliable indicator of population density and have been widely used to measure urban population mobility and vitality [47]. Baidu heat map data are also frequently used to assess urban park usage due to its higher spatial resolution [46,48].

In this study, Baidu heat map data serve as characteristic spatial data. To ensure research representativeness, six weeks of Baidu heat map data were randomly selected throughout the year, deliberately excluding public holidays. The specific time periods include: 16–22 October 2023; 11–17 December 2023; 19–25 February 2024; 15–21 April 2024; 17–23 June 2024; and 5–11 August 2024. Regarding data processing, we utilized ArcGIS 10.8.2 software to establish buffer zones around park green spaces, with the buffer distance equal to the interval between Baidu heat map data points (170 m), thereby ensuring maximum inclusion of park-related population data. Through spatial join operations, we integrated the Baidu heat map information within park green space boundaries and estimated actual population numbers within parks based on heat map data.

3.3. Research Methods

This study presents a comprehensive research framework, as illustrated in Figure 3, comprising four sequential stages. In the first stage, we collect and preprocess data from diverse sources. We use open network big data, complemented by field survey calibration, to acquire detailed information on urban park green spaces, road networks, and population distribution. We employ advanced spatial modeling techniques, including space syntax and transportation network analysis, to construct an integrated database. The second stage involves comparing two methodologies: the two-step floating catchment area (2SFCA) and three-step floating catchment area (3SFCA). We validate these methods using Baidu heat map data. Our results confirm that the 3SFCA method performs better in terms of accuracy and scale. We then apply this improved method to evaluate the spatial distribution and accessibility of urban park green spaces across multiple transportation modes (walking, cycling, public transit, and driving) within a 15-min travel framework. In the third stage, we analyze social resource inequality in urban green spaces. We employ several metrics: the Lorenz curve, Gini coefficient, and equity index to quantify supply–demand imbalances. We use Moran’s I to characterize spatial inequality patterns, providing a multidimensional perspective on social equity. We also conduct comparative analyses across administrative divisions and transportation modes to reveal disparities in resource allocation. The fourth stage integrates our findings to propose targeted optimization strategies. Using the four-quadrant clustering method, we recommend specific improvements for different regions and age groups. These include developing pocket parks and recreational facilities for the elderly, and creating miniature parks with child-friendly amenities. Our recommendations aim to enhance accessibility and address the needs of vulnerable groups, providing a basis for evidence-based policy interventions and equitable urban green space planning.

3.3.1. Calculation of Angle Step Depth Based on Space Syntax Line Segment Model

Space syntax is a widely applied and effective method in the field of urban spatial analysis, and its validity has been well-established through numerous case studies and empirical research [49,50,51,52,53,54,55,56,57]. In this study, space syntax analysis is employed as a methodological tool for pre-processing topological distances, aiming to explore the intrinsic topological relationships within the urban road network’s spatial structure. Among the core metrics, angular step depth is used to represent spatial topological relationships by measuring the shortest path between spatial units. Therefore, this study utilizes the angular step depth metric from the space syntax segment model to quantify spatial topological relationships.

In the segment model of space syntax, angular step depth is a way to measure the depth of nodes within a spatial configuration. It focuses on the number of angular steps (turns or changes in direction) required to travel from one node to another along a path. Specifically, angular step depth takes into account the number of turns needed when moving from an origin node to a destination node along various paths. The calculation of angular step depth is typically expressed using the following formula:

$$ASD(i)=\sum _j \left({\displaystyle \frac{1}{d_{ij}}}\cdot \mathrm{s}\mathrm{t}\mathrm{e}\mathrm{p}(i,j)\right)$$

(1)

The angular step depth ($ABS\left(i\right)$) of node $i$ refers to the angular depth from node $i$. Node $j$ represents other nodes associated with node $i$. The distance between node $i$ and node $j$ is denoted as $d_{ij}$, while $Step\left(i,j\right)$ refers to the number of steps, i.e., turns, required to move from node $i$ to node $j$.

The specific steps are as follows: First, using ArcMap 10.8.2, the road network data downloaded from OSM is vectorized and converted into road centerlines. The road network is then calibrated and corrected using Google satellite imagery to fix any errors. Second, the road segments closest to the supply and demand points (a total of 2087 segments) are marked. Third, the data is imported into Depthmap+ Beta 1.0 to build a segment model and calculate the angular step depth from the nearest supply point to all other segments globally. Finally, network analysis and spatial joining are performed in ArcMap 10.8.2 to link the supply and demand points to the nearest road network segments (a total of 783,352 connections), yielding the topological distances used in this study.

3.3.2. The Improvement of the Basic Gravity Model Is the Two-Step Floating Catchment Area (2SFCA) Method

The basic gravity model was improved into the two-step floating catchment Area (2SFCA) method, first proposed by Radke and Mu [20], modified by Luo and Wang [17], and further refined by Luo and Qi [19]. The basic 2SFCA model consists of two steps. In the first step, for each supply location of a park, all demand locations (such as residential areas) within the catchment threshold distance are identified, and the supply-to-demand ratio is calculated. The supply-to-demand ratio represents the ratio of park area to the total demand population, and its equation is as follows:

$$R_j={\displaystyle \frac{S_j}{{\sum }_{i\in \left\{d_{ij}\le d_0\right\}}^k D_i\times G\left(d_{ij}\right)}}$$

(2)

In this equation, $Di$ represents the population at demand location $i$, whose centroid is within the catchment area of supply location $j \left(dij\le d0 \right).$ $S_j$ is the total park area at supply location j, and $d_{ij}$ denotes the distance between the population at demand location $i$ and the supply location $j$. The distance decay function $G\left(d_{ij}\right)$ is used to account for the friction of distance.

$$G\left(d_{ij}\right)=\left\{\begin{array}{c}{\displaystyle \frac{e^{-\left(1/2\right)\times {\left(d_{ij}/d_0\right)}^2}-e^{-\left(1/2\right)}}{1-e^{-\left(1/2\right)}}},d_{ij}\le d_0\hfill \\ 0,d_{ij}>d_0\hfill \end{array}\right.$$

(3)

For each location within the threshold distance $d0$, the accessibility $Ai$ of the catchment area is calculated as the sum of the supply-to-demand ratio $R_j$ multiplied by the distance decay function, as follows:

$$A_i=\sum _{i\in \left\{d_{ij}\le d_0\right\}}^m R_j\times G(d_{ij})$$

(4)

3.3.3. Enhanced 3SFCA Method for Multimodal UGSA

The improved 3SFCA method integrates multiple modes of transportation (walking, cycling, public transit, and private car) into a Gaussian-based two-step floating catchment area (2SFCA) model to comprehensively assess park accessibility across different travel modes. Secondly, a park attractiveness coefficient is introduced as an evaluation metric to more accurately reflect the actual appeal of parks. Additionally, a competition coefficient is incorporated as an evaluation metric, factoring in topological distance and path decay distance to enhance the precision of distance measurement.

Step 1:

$$G\left(d_{ij}\right)=\left\{\begin{array}{c}{\displaystyle \frac{e^{-\left(1/2\right)\times {\left(d_{ij}/d_0\right)}^2}-e^{-\left(1/2\right)}}{1-e^{-\left(1/2\right)}}},d_{ij}\le d_0\hfill \\ 0,d_{ij}>d_0\hfill \end{array}\right.$$

(5)

$$T_{ij}={\displaystyle \frac{Ln\left(S_{iJ}\right)G_{iJ}}{T{d}_{ij}+1}}$$

(6)

$$C_{ij}={\displaystyle \frac{T_{iJ}}{\sum _{k\in \left\{\mathrm{Dist}\right(i,j)<d_0\}} T_{ik}}}$$

(7)

In the equation, $i$ represents the demand point (community), $j$ represents the supply point (green space), and $k$ refers to any variable supply point within the range of the demand point. $d_{ij}$ is the travel cost (in time or distance) between demand point $i$ and supply point $j$, while $d0$ represents the threshold value. $G\left(d_{ij}\right)$ is the Gaussian distance decay coefficient, $T_{ij}$ is the attraction index between demand point $i$ and supply point $j$, and $S_j$ represents the total park area (in square meters) at supply point $j$. $Td$ denotes the topological distance, and $C$ is the competition coefficient.

Step 2:

$$R_j={\displaystyle \frac{S_j}{\sum _{i\in \left\{\mathrm{Dist}\right(i,j)<d_0\}} C_{ij}{D}_i{G}_{ij}{P}_n}}$$

(8)

In the equation, $D_i$ represents the population at demand point $i$, whose centroid is located within the catchment area of supply point $j$ $\left(d_{ij}\le d0\right)$. $P$ denotes the proportional coefficient representing the share of residents choosing different modes of transportation. $n$ represents the different transportation options (for example, $P1$ represents the proportion of people walking, $P2$ represents the proportion of people cycling, $P3$ represents the proportion of people using public transit, and $P4$ represents the proportion of people driving private cars.

Step 3:

$$A_i=\sum _{i\in \left\{\mathrm{Dist}\right(i,j)<d_0\}} C_{ij}{R}_j{G}_{ij}{P}_n$$

(9)

In the equation, $A_i$ represents the accessibility index.

3.3.4. Construct an Equity Index to Evaluate the Spatial Equity of UGSA (Urban Green Space Accessibility) Supply

To explore whether there is spatial inequity in park provision across different communities, we introduced a standardized absolute spatial equity measurement method [58] to analyze the fairness of park accessibility. Based on the accessibility scores, the equity score can be calculated as follows:

$$E_i={\displaystyle \frac{\max \left(R_j\right)}{\max \left(A_i\right)}}\times \left(A_i\right)$$

(10)

In this equation, $R_j$ represents the supply-to-demand ratio, and max$\left(R_j\right)$ refers to the maximum supply-to-demand ratio between urban parks and population demand. $a_i$ is the accessibility value of each residential unit, and max$\left(A_i\right)$ represents the maximum accessibility value across all residential units. $E_i$ is the standardized equity score for community, reflecting the balance between supply and demand.

According to the classification standards of [58], the adequacy score can be divided into six levels: (a) Level VI ($E_i$ > 1), indicating an oversupply or abundant park provision; (b) Level V (1 ≥ $E_i$ > 0.75), indicating very good park provision; (c) Level IV (0.75 ≥ $E_i$ > 0.50), indicating good park provision; (d) Level III (0.50 ≥ $E_i$ > 0.25), indicating weak park provision; (e) Level II (0.25 ≥ $E_i$ > 0), indicating very weak park provision; (f) Level I ($E_i$ = 0), indicating no supply or no park provision (

Table 4. Accessibility and Equity Classification Standards.

Class	Range of Ei Value	Supply and Demand Status	Spatial Equality
I	$Ei$ = 0	No supply	High inequality
II	0.25 > $Ei$ > 0	Very weak	High inequality
III	0.5 > $Ei$ ≥ 0.25	Weak	Relative inequality
IV	0.75 > $Ei$ ≥ 0.5	Good	Equality
V	1 ≥ $Ei$ ≥ 0.75	Very good	Relative equality
VI	$Ei$ > 1	Oversupply	High inequality

).

In this study, the improved 3SFCA model incorporates multiple factors, including travel modes, mode choice, travel path distance, topological distance, park attractiveness coefficient, and park competition coefficient. The 3SFCA model is used to assess spatial accessibility influenced by various travel modes.

$$E_{i,n}={\displaystyle \frac{\max \left(R_{j,n}\right)}{\max \left(A_{i,n}\right)}}\times \left(A_{i,n}\right)\times P_n$$

(11)

In this context, $E_{i,n}$ represents the comprehensive equity index, with $n$ indicating the different travel modes ($n$ = 1 for walking, $n$ = 2 for cycling, $n$ = 3 for using public transit, and $n$ = 4 for driving a private car). $Pn$ refers to the proportion of residents choosing each mode of travel (for example, $P1$ represents the proportion of walking, $P2$ the proportion of cycling, $P3$ the proportion of using public transit, and $P4$ the proportion of driving).

To identify spatial clustering patterns of urban park accessibility and equity, this study applied spatial autocorrelation analysis for park accessibility and equity. Moran’s I is commonly used to measure the spatial autocorrelation of variables, revealing their spatial clustering patterns. The Global Moran’s I reflects the overall level of spatial autocorrelation, while the Local Moran’s I (Anselin Local Moran’s I) [59] measures the strength of the association between a variable and its surrounding areas. To explore whether there is spatial autocorrelation in the accessibility and equity of green spaces in Chengdu’s core area, this study first used the Global Moran’s I to assess the degree of spatial clustering and then analyzed the spatial clustering characteristics of different areas through local autocorrelation tools. There are four types of local indicators of spatial association (LISA) clusters: high–high clusters, high–low clusters, low–high clusters, and low–low clusters. High–high and low–low clusters represent areas with high and low values in geographic space, respectively, indicating the hot spots and cold spots of certain attributes in the study area. Using GeoDa 1.14.0 software, Local Moran’s I was applied to generate LISA cluster maps. Since accessibility tends to be affected when traveling to parks on foot or using multiple modes of transportation, but is less impacted by the other three modes, the spatial clustering of walking and multi-modal travel accessibility was further analyzed.

The Lorenz curve and Gini coefficient analysis are commonly used as quantitative indicators for evaluating social equity performance. Initially proposed by [60] to explore income distribution equity, the method has since been widely applied in the field of environmental justice due to the similarity between income distribution and public resource allocation.

The Gini coefficient is the ratio of the area between the line of absolute equality and the Lorenz curve to the area between the line of absolute equality and the line of absolute inequality in the Lorenz diagram. To reflect the equity pattern of urban green space distribution in Chengdu’s core area, this study constructs a green space equity model based on the Gini coefficient [61]. It measures the differences in green space accessibility across different districts and visualizes and analyzes the spatial patterns. The calculation formula for the model is as follows:

$$G{E}_u=1-\sum _{k=1}^n \left(P_k-P_{k-1}\right)\left(C_k+C_{k-1}\right)$$

(12)

$$C_k={\displaystyle \frac{{\sum }_{i=1}^k A_i{r}_i}{{\sum }_{i=1}^n A_i{r}_i}}$$

(13)

In this equation, $GEu$ represents the green space equity index for geographic unit $u$ (such as a district); $n$ is the total number of communities within geographic unit $u$; k refers to the $k-th$ community, ranked in ascending order of green space accessibility, where $k$ = 1, 2, …, n; $A_i$ is the green space accessibility value for community $i$; $ri$ represents the population of community $i$; $Ck$ is the cumulative proportion of the product of green space accessibility and population from community 1 to community $k$, where $C0$ = 0 and $Cn$ = 1; and $Pk$ is the cumulative proportion of the population from community 1 to community $k$, where $P0$ = 0 and $Pn$ = 1. According to the mathematical interpretation of the Gini coefficient, when $GE$ ≤ 0.2, it indicates that the spatial distribution of green space resources is highly equitable; when 0.2 < $GE$ ≤ 0.3, it indicates a relatively equitable distribution; when 0.3 < $GE$ ≤ 0.4, the distribution is moderately fair; when 0.4 < $GE$ ≤ 0.5, there is a notable disparity in the spatial distribution of green space resources; and if $GE$ > 0.5, it suggests a significant inequality in the distribution of green space resources, indicating severe inequity.

3.3.5. Construct a Four-Quadrant Model to Evaluate the Resource Matching Between UGSA (Urban Green Space Accessibility) and Equity

The four-quadrant model is a tool used for analyzing spatial data by combining qualitative and quantitative research [62]. In recent years, this model has been widely applied in areas such as water resource value assessment, landscape ecological security, and ecosystem service quality [63,64]. In this study, the four-quadrant model is used to explore the matching relationship between the population of children and the elderly and the accessibility of green spaces. In this model, the population of children and the elderly, along with green space accessibility, are first standardized using z-scores. Then, a two-dimensional coordinate system is constructed, with standardized accessibility on the $x-axis$ and the standardized population of children and the elderly on the $y-axis$. Based on the quadrant division, eight matching patterns can be identified: high accessibility–high child population, low accessibility–low child population, low accessibility–high child population, high accessibility–low child population, high accessibility–high elderly population, low accessibility–low elderly population, low accessibility–high elderly population, and high accessibility–low elderly population. These matching patterns are spatially visualized, helping to identify areas where there is a mismatch between the distribution of green spaces and areas with concentrated populations of children and the elderly. This provides guidance for optimizing urban green space planning. The standardization formula used in the model is as follows:

$$x=\left(x_i-\overline{x}\right)/s$$

(14)

In this case, $x$ represents the standardized variable (supply or demand), $xi$ is the original supply or demand value for area $i$, $\overline{x}$ is the average value for the area, and $s$ is the standard deviation.

To eliminate differences in the dimensions of supply and demand, the original data has been normalized. The normalization formula is as follows:

$$A_{ij}={\displaystyle \frac{\left(\begin{array}{c}{A}_{ij}-min{A}_j\end{array}\right)}{\left(\begin{array}{c}max{A}_j-min{A}_j\end{array}\right)}}$$

(15)

Accessibility and elderly population rate values are calculated by summing the normalized values. The formula is as follows:

$$A_i=\sum _{j=1}^s A_{ij}$$

(16)

In these equations, $A_i$ represents the accessibility and elderly population rate in the study area, while $A_{ij}$ refers to the specific value for region $i$. By comparing the balance and matching of supply and demand, this model illustrates the dynamic relationship between green space accessibility and the distribution of the elderly population. It helps to identify areas in the city where there is a mismatch between the elderly population and green space resources, offering targeted planning recommendations.

4. Results

4.1. Analysis of Spatial Disparities in UGSA (Urban Green Space Accessibility)

4.1.1. Comparative Analysis of Population Demand Points Based on Baidu Heatmap Data Validation

This study evaluated model accuracy and reliability by comparing estimated populations at demand points before and after improvements against the Baidu heat map population data. As illustrated in Figure 4, regression analysis results indicate that the improved model demonstrates significant advantages in explaining actual populations at demand points. The original model (Y1 = 5.28x + 2265.090, R² = 0.223, RMSE = 11,689.21, MAE = 5525.46) exhibited weak correlation, thus inadequately reflecting the spatial distribution of population at demand points. In contrast, the improved model (Y2 = 3.834x − 723.965, R² = 0.404, RMSE = 7100.26, MAE = 3091.26) demonstrated superior fitting with significantly increased regression coefficients and reduced RMSE and MAE values, indicating more accurate capture of population variations at demand points and better alignment with actual park population dynamics within Chengdu’s core area.

Figure 5a,b, respectively, illustrate the spatial distribution of UGSA before and after the improvements. In Figure 5a, the large blue areas in the central urban area indicate low UGSA, suggesting that residents find it difficult to access parks conveniently by walking or other modes of transportation. This may be due to insufficient parks, inadequate transportation infrastructure, or high population density leading to demand exceeding supply in the central area. In contrast, the peripheral areas show higher UGSA than the central area, especially in the red zones in the southwest and southeast. In Figure 5b, the blue low-accessibility areas in the central urban area have decreased, indicating that through transportation optimization or using more advanced methods to estimate population demand, park accessibility in some regions more closely aligns with actual conditions. However, the issue of resource imbalance in the central area has not been fully resolved. The central urban area still requires additional park supply or improved transportation conditions to enhance UGSA.

4.1.2. Analysis of Spatial Distribution Differences in UGSA Based on the 3SFCA Method

This study systematically compared urban park accessibility across four transportation modes—walking, cycling, public transit, and driving—using the improved 3FSCA method. As shown in the analysis in Figure 6, there are significant differences in park accessibility between different transportation modes, and the spatial distribution is uneven.

In the walking mode, areas with low park accessibility in the central urban area are noticeably fewer compared to other transportation modes. This indicates that the supply of green spaces in the central area is insufficient, making it difficult for residents to access parks via non-walking modes. Interestingly, walking provides a travel advantage in areas where green space supply is lacking. In contrast, the peripheral areas have relatively higher park accessibility, likely due to lower population density and a more abundant distribution of green spaces.

In the cycling mode, the low-accessibility areas in the northwest are significantly reduced, possibly due to a shortage of small green spaces but a higher supply of larger community parks or specialized parks. The public transit mode slightly improves park accessibility in the peripheral areas, although its impact on accessibility in the central area remains limited. The private car mode performs best, substantially increasing park accessibility across the entire city, particularly in suburban areas, significantly enhancing spatial equity.

Overall, the improved 3FSCA model, by integrating multiple transportation modes, effectively reveals the varying impacts of different travel modes on park accessibility. It highlights the pronounced spatial disparities between the central and peripheral urban areas.

4.2. Comparative Analysis of Single-Mode UGSA Equity Based on the 3SFCA Method

To explore the differences in the impact of the four travel modes on UGSA equity, a comparative analysis of UGSA equity for each single travel mode was conducted using the 3SFCA method. Figure 7 and Figure 8 show the frequency distribution and spatial distribution of community equity scores for the four travel modes (walking, cycling, public transit, and driving). The equity score range is divided into six levels, distinguished by different colors. In Figure 8, red indicates communities with better park facilities, light green represents areas with fewer park facilities, and blue denotes areas with no park facilities.

Figure 7a illustrates the frequency distribution of equity scores for the walking mode. There are 782 communities with good or above-average equity scores, while 5285 communities have no park access. This distribution shows that approximately 65.4% of communities cannot reach a park within 15 min on foot, and only 9.6% of communities have easy access to parks. Figure 8a reveals a stark contrast in equity scores between the peripheral and central urban areas under the walking mode. In the peripheral areas, high equity is concentrated in communities near parks, while low equity is found in areas with insufficient or no park supply. In contrast, the variation in high equity scores is more pronounced in the central urban area, indicating that residents living near the city center have easier access to parks on foot. Additionally, the study found that some central communities close to parks did not score well in walking mode, possibly due to the higher population density in these areas, which leads to a greater share of limited park resources.

The distribution pattern of park accessibility equity under the cycling mode is similar to that of the walking mode but shows greater inequality. Figure 7b indicates a slight increase in the frequency of good or above equity scores under the cycling mode compared to walking, suggesting that more communities are identified where supply exceeds demand. This trend indicates that cycling helps residents travel farther and access more park benefits, though some communities located far from parks still experience limited access to park services. Figure 8b shows that high equity scores are more evenly distributed in the central urban area, as residents have more opportunities to cycle to parks. However, low equity score communities remain in the city’s outskirts.

Figure 7c,d, along with Figure 8c,d, illustrate the equity of park accessibility for those using public transit and private cars. From Figure 7c,d, it is clear that there are 1915 communities (public transit) and 1891 communities (private car) with good or above-average equity scores. The equity scores for public transit and private car modes are significantly higher than the average scores for walking and cycling. These trends indicate that more communities benefit from better park services if residents choose to drive or use public transportation to access parks. In both modes, communities in the southern parts of the city show higher equity scores than those in the northern parts.

Figure 8e illustrates further analysis based on the 3SFCA method, showing the spatial distribution of comprehensive equity scores. Overall, the western part of the city has higher equity scores, providing better park services, while most of the eastern area cannot reach a park within 15 min. In the central urban area, the equity distribution pattern for park accessibility is closely aligned with the spatial distribution of public transit and private car usage, as shown in Figure 8c,d. This trend may be linked to the fact that 67.4% of residents prefer to use public transit or drive to parks, as shown in Table 3. In peripheral areas, the equity distribution pattern is more similar to the private car usage pattern, as shown in Figure 8d, since the proportion of car usage increases in the outskirts, especially in the Eastern New District and Tianfu New Area.

“No supply” indicates that residents in these communities cannot access parks within a 15-min walking distance, reflecting a significant deficiency in park service provision in these areas. This phenomenon highlights the lack of equity in urban green space distribution. Consequently, to address this spatial imbalance, urban planners should prioritize these “no supply” communities as key intervention areas, prioritizing the planning and construction of additional parks to enhance park service accessibility and equity, thereby meeting residents’ outdoor activity needs and promoting equitable distribution of urban public resources.

In conclusion, integrating multiple transportation modes is a key strategy for improving the spatial equity of park facilities. Walking, cycling, public transit, and driving each have their advantages and limitations. A combined use of these modes can ensure that more communities, particularly suburban and peripheral areas, enjoy green space resources comparable to those in central areas. Future urban planning should focus on optimizing the integration of transportation networks and park layouts to maximize park accessibility, promote spatial equity, and enhance the quality of life for urban residents.

4.3. Spatial Disparities in Park Accessibility Equity and Travel Behavior

4.3.1. Spatial Clustering Pattern of Park Accessibility Equity Based on the Local Moran’s I Index

Using Local Moran’s I index, we classified park accessibility equity patterns across all 8081 community grids in Chengdu’s core area into four types: HH clusters, LL clusters, HL outliers, LH outliers, and non-significant clusters. Figure 9 illustrates the clustering types differentiated by color. HH clusters (in red) represent communities with high accessibility and high equity scores, surrounded by similar high-accessibility, high-equity communities. LL clusters (in blue) represent communities with low accessibility and low equity scores, surrounded by other low-accessibility, low-equity communities. HL and LH outliers (in orange) indicate communities with high accessibility but low equity scores adjacent to communities with low accessibility but high equity scores. Non-significant areas (in gray) represent locations without significant clustering of accessibility or equity scores.

Our cluster analysis reveals substantial spatial disparities in park accessibility equity across different areas. Figure 9e shows that within the inner ring road (including parts of Jinjiang, Qingyang, Jinniu, Wuhou, and Chenghua districts), spatial differences are relatively minor. Most clusters in this inner area are of the HH type, with no significant clustering, indicating that residents of the inner ring road area enjoy equal access to parks, making it the most ideal residential area. Other districts (such as parts of Xindu, Wenjiang, Shuangliu, Qingbaijiang, Pidu, and Longquanyi) also exhibit minimal spatial disparities. Most communities in these areas fall within the HH cluster, with some showing no significant clustering, though a few communities display HL and LH outliers, suggesting that residents in these areas may face challenges traveling long distances to access parks. In future urban planning revisions, attention should be given to communities with lower equity scores in these six central districts, with an emphasis on increasing the number of parks.

The spatial disparities are more pronounced in four peripheral districts (Pidu, Wenjiang, Shuangliu, and Longquanyi). LL clusters outnumber HH clusters, indicating poorer accessibility and equity in these areas. Park development in the urban-rural fringe was not initially a focus of government efforts, resulting in relatively fewer park services in these four suburban districts. Municipal authorities and policymakers should prioritize the construction of new parks in these lower-equity communities to ensure better access to parks for all residents.

4.3.2. Comprehensive Analysis of Park Accessibility Equity Based on the Lorenz Curve and Gini Index

According to Figure 10, the Gini coefficient and Lorenz curve demonstrate the disparities in UGSA distribution across different transportation modes. Results indicate that private vehicle transportation exhibits the lowest UGSA Gini coefficient (0.50), suggesting the most equitable green space accessibility for residents using private vehicles. Conversely, the walking mode shows the highest Gini coefficient (0.81), indicating that 80% of UGSA resources are concentrated among 20% of the population, highlighting the greatest accessibility disparity for those relying on walking access to green spaces. We further compared regional differences within Chengdu’s core area. Results reveal that new districts (such as Longquanyi, Pidu, Shuangliu, and Xindu) exhibit higher Gini coefficients for walking and cycling modes compared to old urban districts (such as Qingyang, Chenghua, Wuhou, and Jinjiang), reflecting more pronounced green space accessibility inequalities in new districts for short-distance travel modes.

4.4. Analysis of Vulnerable Population Density and UGSA Resource Matching Based on the Four-Quadrant Model

4.4.1. Resource Matching Analysis of Children and the Elderly in UGSA

This study utilized the four-quadrant clustering method to analyze the matching relationship between accessibility and the rates of child and elderly populations across different regions. The study did not consider clustering patterns for children in the private car driving mode, but acknowledges that family travel decisions may be indirectly influenced by the accessibility of driving a private car.

As shown in Figure 11, under the walking mode, high accessibility-high child population areas (31.2%) mainly concentrate in the central urban area, reflecting the appeal of well-developed infrastructure and educational resources. Conversely, low accessibility-low child population areas (18.7%) concentrate in the city’s outskirts, likely due to weaker infrastructure. Low accessibility–high child population areas (40%) primarily appear on the city’s edge, where poor walking conditions exist, but the lower cost of living attracts larger families. Under the cycling mode, high accessibility–high child population areas increase to 45.0%, mainly in the inner ring, as well as the eastern and southern parts of the city, highlighting the impact of well-developed cycling facilities. In the public transit mode, high accessibility–high child population areas dominate (50.1%), found within the inner ring and new southern areas (such as Shuangliu), emphasizing the significance of public transportation in family housing choices.

As shown in Figure 12, for the elderly population, under the walking mode, high accessibility-high elderly population areas (25.9%) concentrate in high-density, older communities in the city center and surrounding areas, reflecting the importance of walkability and well-developed social services. However, low accessibility–high elderly population areas (36.2%) primarily exist in older residential neighborhoods on the city’s outskirts, where despite poor walking conditions, traditional living patterns have led to higher proportions of elderly residents. Under the cycling mode, high accessibility-high elderly population areas increase to 36.8%, concentrated in the older central districts, indicating the appeal of good cycling facilities and green environments for the elderly. In the public transit mode, high accessibility–high elderly population areas further increase to 41.5%, focused on areas with advanced public transportation in the core urban region, highlighting the importance of accessible public transit for elderly residents. Under the private car mode, high accessibility–high elderly population areas reach 47.3%, concentrated around major transportation hubs and their surroundings, showing the importance of car travel convenience for some elderly residents.

Overall, the elderly population primarily concentrates in areas with high transportation accessibility, especially under walking, cycling, and public transit modes. This indicates that elderly residents tend to prefer living environments with convenient public transportation and well-developed infrastructure. However, in certain low-accessibility areas, despite poor transportation conditions, the aging population remains high, likely due to lower housing costs or traditional community ties. These findings offer important insights into the relationship between residential choices of different age groups and park accessibility, providing a basis for targeted urban planning and transportation policy development.

4.4.2. Optimization Recommendations for Addressing UGSA Equity Matching Issues for Children and the Elderly

To address the issues of urban park green space accessibility and equity in Chengdu, this study conducted field surveys and questionnaire interviews to better understand the travel difficulties faced by high child and elderly population communities in different areas. The research revealed significant differences in how children and the elderly use park green spaces. We propose optimization recommendations for improving access to park green spaces from the perspectives of children and the elderly. The goal is to enhance the efficiency of urban park use, meet the needs of vulnerable groups, and improve overall well-being and quality of life.

From the perspective of children, surveys with parents and children revealed that while transportation accessibility in central districts (Qingyang, Jinniu, Chenghua, and Jinjiang) is relatively good, the number and distribution of park green spaces are uneven, with a particular shortage of playgrounds for children. Parents consistently highlighted safety and convenience as key factors when taking their children to parks. We recommend adding more children’s play facilities in existing parks and enhancing safety measures, such as installing surveillance cameras, barriers, and dedicated pedestrian and cycling paths for children to ensure safe and convenient access. We also recommend expanding the layout of “micro parks” or community green spaces that provide small-scale play areas within walking distance to ease the travel burden on parents and improve the convenience of park use for both children and parents.

Surveys in newly developed areas of peripheral districts (Wuhou, Shuangliu, Wenjiang) showed notable regional disparities in children’s park accessibility in Shuangliu. We suggest increasing public transport coverage, particularly by introducing school buses or child-friendly transit services, allowing parents easier access to parks. Newly developed residential areas should prioritize the construction of child-focused parks that combine nature education with recreational functions, giving children ample space to play and learn in green environments. In areas like Wenjiang, where public transit accessibility is weaker, building small community parks near residential areas would allow parents to take children to parks on foot, reducing dependence on transportation.

Field surveys in suburban areas (Xindu, Pidu, Longquanyi) revealed sparse green spaces and limited play areas for children, with strong feedback from parents. We recommend prioritizing the development of children’s theme parks in these areas and improving the usage of existing public green spaces. Additionally, creating more green spaces near schools would give students more opportunities to interact with nature after school. To improve park accessibility, the public transportation system should be enhanced, particularly by improving the safety of cycling and walking routes to ensure that both children and parents can safely and easily reach parks.

Field surveys and questionnaires showed that elderly residents in central districts (Qingyang, Jinniu, Chenghua, and Jinjiang) have a high demand for walkable access to parks, especially in older residential areas. Seniors prefer walking to parks in these areas. We recommend enhancing accessibility features in parks in these districts, adding fitness paths, comfortable seating, rest areas, and shaded pavilions to meet the daily exercise and leisure needs of the elderly. The survey also highlighted safety concerns, so we suggest introducing smart technologies such as health monitoring equipment and emergency call devices to ensure elderly residents can safely enjoy park activities. For older neighborhoods with high elderly populations, “pocket parks” should be developed to provide short-distance access to green spaces, improving everyday convenience for seniors.

The survey found that while parks and green spaces are widely distributed in peripheral districts (Wuhou, Shuangliu, Wenjiang), accessibility for the elderly in some newly developed areas is limited, particularly in regions with insufficient public transportation. We recommend establishing senior-friendly bus routes or offering free transit cards through policy initiatives to ensure that the elderly can easily reach parks. Parks should also include more health-focused areas for seniors, such as Tai Chi plazas and rehabilitation fitness equipment, to meet their social and health needs. For Wenjiang, improving public transport coverage, especially between areas with high elderly populations and parks, is essential to make parks more accessible.

The survey results indicated that in suburban areas (Xindu, Pidu, Longquanyi), a lack of green space and public transportation makes it difficult for the elderly to access parks. We recommend introducing “demand-responsive transport” services in suburban areas, offering flexible travel options tailored for the elderly. We also encourage developing “health eco-parks”, combining natural landscapes with healthcare facilities, to offer rehabilitation functions, allowing seniors to enjoy green spaces while maintaining their health. Furthermore, more small green spaces and walking paths should be added within communities to facilitate daily activities for the elderly, improving their quality of life.

By providing targeted optimization recommendations from both the perspectives of children and the elderly, Chengdu can achieve a better balance in green space accessibility and equity, ensuring that different groups enjoy the health, recreational, and social benefits of urban green spaces. These suggestions will help improve the overall quality of life for Chengdu residents and support the sustainable development of the “Park City”.

5. Discussion

5.1. The Spatial Inequality of UGSA Distribution and Its Influencing Factors

This study employed the Lorenz curve and Gini coefficient to assess the inequality in urban green space accessibility (UGSA) in Chengdu’s core urban area. This method has been widely applied and validated in recent research [65,66]. Our results show that UGSA values for low-speed travel modes are lower and more unevenly distributed compared to high-speed travel modes. This phenomenon likely results from the limited availability and uneven spatial distribution of green spaces in Chengdu’s core area. With a fixed 15-min travel time threshold, high-speed travel modes can cover a larger search area, resulting in higher UGSA values. In contrast, low-speed travel modes (such as walking) cover a smaller area, leading to more unequal accessibility. This inequality is particularly pronounced in smaller catchment areas, reflecting the core area’s limited green space resources.

We also found that higher UGSA values are primarily concentrated in suburban areas outside the city center, especially for motorized travel modes (public transit and private cars). This likely reflects better-developed transportation networks, larger green spaces, and higher-quality parks in the suburbs. In contrast, the central urban area, particularly the older districts, faces challenges in providing sufficient high-quality green spaces due to traffic congestion and spatial constraints. However, it is noteworthy that residents in older districts enjoy relatively more equitable green space accessibility, likely due to the historical accumulation of green spaces over time. Meanwhile, newly developed areas, despite setting higher UGSA targets, struggle to balance the needs of all residents amid rapid urbanization. These spatial disparities not only reflect differences in population distribution but also in the distribution of resident activities, affecting spatial analysis of supply and demand and the prioritization of improvements.

5.2. Innovations and Optimizations in UGSA Evaluation Methods

This study introduces several innovations to the traditional 2SFCA method, significantly enhancing the accuracy and flexibility of UGSA analysis. While previous studies Blumenberg [67] incorporated supply-demand distribution and travel time factors, they did not address the implementation of attraction coefficients and travel probability. In this research, we first introduce a travel probability and attraction coefficient model that more accurately reflects residents’ preferences for green spaces, avoiding the overestimation issues inherent in conventional approaches. For probability calculations, we developed a more precise model based on the Gaussian function, integrated with green space attraction coefficients, to better capture the impact of distance decay on resident-green space interactions. This improvement surpasses the simplistic method of merely multiplying travel time, providing more realistic probability estimates for green space selection.

In calculating the green space-to-population demand ratio, while previous studies Li [68] incorporated age-group based factors, they did not address the implementation of topological distance and multi-weight methodologies. In this research, we employed topological distance and multi-weight methods to enhance the accuracy of potential user group assessments. By introducing an attraction factor based on green space area and topological distance, we addressed the bias that might arise from using a single area variable. Under the adjustment of the Gaussian function, we set the impedance coefficient for walking distances within 1000 m to 1, better reflecting the real-world effect of distance decay on travel behavior.

More importantly, this study integrates multiple modes of transportation—walking, cycling, public transit, and driving—into a composite transportation system based on real traffic networks. This system not only accounts for transit travel time but also includes the time required to reach transit stations, transfer between stations, and walk from the station to the green space entrance, addressing gaps often overlooked in previous studies. Through these methodological innovations, this study offers a more accurate and comprehensive approach to UGSA spatial supply–demand analysis, providing a more representative assessment of green space accessibility based on multi-modal transportation and actual travel probabilities.

5.3. The Applied Value and Future Development Directions of UGSA Research

This study offers important insights for improving the spatial resource equity in Chengdu’s core area. Our findings highlight the significance of ensuring a more balanced distribution of green spaces in newly developed areas and the necessity of improving transportation infrastructure in peripheral districts. These insights are not only relevant for Chengdu but also provide valuable reference points for green space planning in other highly urbanized regions. However, the study has several limitations, primarily concerning the selection of transportation modes, the setting of travel time thresholds, and the consideration of environmental factors.

Regarding transportation mode selection, while we analyzed various modes, we did not explore residents’ probabilities and preferences when choosing transportation modes. We also did not consider modern options such as subways. In reality, residents may opt for different transportation modes depending on time and conditions, influenced by factors such as economic conditions, travel convenience, and weather. Future research should delve deeper into these complex influencing factors to more accurately assess the impact of transportation modes on green space accessibility.

For travel time threshold, we used a fixed 15-min critical travel time based on the “Community Life Circle Planning Technical Guidelines TD/T 1062-2021”. However, different residents may choose longer or shorter travel times depending on the type of green space and mode of transportation. Future research should examine residents’ preferences for travel time to different types of green spaces, enabling a more comprehensive analysis and optimization of green space accessibility.

Furthermore, while we utilized Baidu heat map data to validate the population data at demand points estimated by the enhanced 3SFCA method, these datasets may still exhibit biases, such as underrepresentation of specific demographic groups or low-activity regions. Although the park visitation data originated from field surveys, they may have limitations in comprehensively capturing seasonal variations in park usage patterns and residents’ habitual visitation behaviors. Additionally, future studies could incorporate seasonal and weather factors to analyze their impact on green space usage patterns and accessibility. Combining socioeconomic factors to further explore the relationship between green space accessibility and social equity is another important research direction. Conducting longitudinal studies to analyze the dynamic changes in green space accessibility during urban development would help us better understand the impact of urbanization on green space accessibility.

6. Conclusions

This study conducted an in-depth exploration of urban green space supply and demand issues in Chengdu’s core area, offering new perspectives in the field of environmental justice research. Through innovatively applying methods and comprehensive data analysis, the study reached several key conclusions:

(1) Regarding methodological innovation and application, we used the improved 3SFCA method, integrating four transportation modes—walking, cycling, driving, and public transit—to estimate UGSA values. This multi-modal transportation analysis provided a more complete and accurate assessment of green space accessibility, overcoming the limitations of traditional single-mode transportation analyses.

(2) We uncovered spatial inequalities using a combination of the Lorenz curve, Gini coefficient, and bivariate global and local Moran’s I spatial statistical methods. Our analysis revealed significant disparities in the UGSA supply-demand relationship across different transportation modes in Chengdu’s core area. Specifically, low-speed travel modes like walking had significantly lower and more uneven UGSA values compared to high-speed modes like driving, highlighting the need to prioritize walking accessibility in green space planning.

(3) We identified clear regional differences, showing that areas with higher UGSA values were mainly in suburban regions outside the city center, particularly for motorized travel modes. In contrast, older urban districts, despite having limited green space resources, had a more evenly distributed accessibility. This insight provides critical guidance for regionally differentiated strategies in urban green space planning.

(4) Through clustering analysis of UGSA supply-demand relationships, we successfully identified communities and regions that require priority improvements. This provides urban planners with clear action points for addressing inequities in green space distribution.

(5) Our policy implications emphasize the importance of ensuring a rational distribution of green spaces in newly developed areas and improving transportation infrastructure in peripheral districts. These findings offer scientific support for crafting differentiated green space planning strategies, helping to improve the overall fairness of urban green space provision.

In conclusion, this study provides a novel research framework and empirical evidence for addressing urban green space supply and demand challenges through innovative methods and comprehensive data analysis. While our analysis focuses on China, the methodological framework, particularly the integration of multi-modal transportation systems, probabilistic accessibility modeling, and attraction coefficients, can be applied to cities worldwide with similar urbanization patterns and green space distribution challenges. These insights offer valuable references for policymakers and urban planners globally, particularly in rapidly urbanizing regions where ensuring equitable access to green spaces remains a critical concern. By emphasizing the spatial alignment of green space supply and demand, this study contributes to the broader discourse on environmental justice and sustainable urban development. The findings and policy recommendations presented here can serve as a foundation for enhancing urban livability and promoting inclusive green space planning in diverse geographical and socio-economic contexts.