Assessing Urban Park Accessibility via Population Projections: Planning for Green Equity in Shanghai

Abstract

Rapid urbanization and demographic shifts present significant challenges to spatial justice in green space provision. Traditional static assessments have become increasingly inadequate for guiding park planning, which now requires a dynamic, future-oriented analytical approach. To address this gap, this study incorporates population dynamics into urban park planning by developing a dynamic evaluation framework for park accessibility. Building on the Gaussian-based two-step floating catchment area (Ga2SFCA) method, we propose the human-population-projection-Ga2SFCA (HPP-Ga2SFCA) model, which integrates population forecasts to assess park service efficiency under future demographic pressures. Using neighborhood-committee-level census data from 2000 to 2020 and detailed park spatial data, we identified five types of population change and forecast demographic distributions for both short- and long-term scenarios. Our findings indicate population decline in the urban core and outer suburbs, with growth concentrated in the transitional inner-suburban zones. Long-term projections suggest that 66% of communities will experience population growth, whereas short-term forecasts indicate a decline in 52%. Static models overestimate park accessibility by approximately 40%. In contrast, our dynamic model reveals that accessibility is overestimated in 71% and underestimated in 7% of the city, highlighting a potential mismatch between future population demand and current park supply. This study offers a forward-looking planning framework that enhances the responsiveness of park systems to demographic change and supports the development of more equitable, adaptive green space strategies.

1. Introduction

Urban public service facilities refer to infrastructure and services provided or supported by the government for collective public use. Among these, urban parks play a vital role in fostering inclusive and livable cities. Beyond offering environmental benefits, parks are essential for improving residents’ physical and mental health, supporting social cohesion, and promoting sustainable urban development [1,2,3,4]. Recent global health guidance, such as the WHO’s 2023 recommendation of providing access to green spaces within 300 m of all residences, further reinforces the importance of equitable green space provision for public health [5,6]. However, amid rapid urban and rural transformation, increasing demand for public spaces is constrained by limited land resources, which intensifies the challenge of equitable spatial allocation. Consequently, the spatial layout of parks and green spaces has become a critical issue in urban planning research. Evaluating and improving the spatial and social equity of park distribution is therefore essential not only for maximizing green infrastructure efficiency but also for ensuring just and inclusive access to urban public services—a fundamental principle in contemporary urban planning and governance.

Research on urban park and green space equity in Western contexts began relatively early and has since evolved into a mature theoretical framework. This body of research has progressed through three key stages: (1) a focus on quantitative equality before the 1970s [7,8]; (2) a shift to spatial equity between the 1970s and 1990s [9]; and (3) a turn toward social equity and justice since the early 21st century [10,11]. Over time, the emphasis has shifted from land-centered fairness to people-centered equity, prioritizing alignment between the spatial distribution of public green spaces and population distribution to achieve a balanced supply and demand. Recent studies have further broadened this focus, moving beyond spatial parity to embrace more comprehensive understandings of social equity and environmental justice. These perspectives highlight how different social groups experience unequal access to green spaces based on factors such as age, education, income, race, and residency status [10,12,13,14,15].

Accessibility assessment has become a widely adopted approach for evaluating the fairness and effectiveness of urban park distribution [16,17]. Accessibility reflects how easily and equitably residents can reach park facilities, accounting for spatial constraints [18]. The concept generally encompasses three core components: proximity, quantity (or area), and quality [19]. With the advancement of research, the definition of accessibility has evolved to better capture the real-world conditions under which residents access and use urban parks. Enhancing the precision of these assessments remains a key objective in the field. Park accessibility is usually measured by the following four types of methods, with different studies adopting different methods, as follows [20,21,22]: (1) the statistical unit indicator method, which calculates green space indicators (e.g., area or quantity) within administrative or statistical units; (2) the spatial isolation method, including minimum distance calculations based on Euclidean or network distances to assess proximity [15]; (3) the cumulative opportunity method, such as the service area approach, which evaluates the quantity of accessible green space within specified distance or time thresholds [14,19]; and (4) the spatial interaction method, including gravity models, which introduces the law of gravity potential to calculate the total potential exerted by all facilities within a defined area [10].

One prominent method of park accessibility is the two-step floating catchment area (2SFCA) method [23,24], which incorporates both supply and demand dimensions in accessibility analysis. Subsequent studies have introduced several refinements to the 2SFCA model [25,26], including distance decay functions, dynamic or multi-scale service radii [27], consideration of diverse transportation modes [28,29,30,31], and indices reflecting park attractiveness [32]. Furthermore, an increasing number of studies have acknowledged variability in park accessibility needs across age groups, noting that preferences and mobility levels differ significantly by demographic characteristics [33,34]. Additionally, recent studies have begun to refine the traditional accessibility method with the actual mobility information extracted from mobile phone big data or online maps [32,35,36,37]. In particular, the most popular refined method is the Gaussian-based two-step floating catchment area (Ga2SFCA) model, which improves accuracy by incorporating a Gaussian distance decay function that aligns with actual travel behavior. This approach offers more realistic estimates of how well park supply matches population demand and is widely recognized as a reliable and robust evaluation framework [38,39].

However, despite these methodological advancements, most accessibility models continue to rely on static population data, limiting their capacity to reflect dynamic demographic changes. In rapidly evolving urban environments, such static assumptions may lead to ineffective planning outcomes, including unintended consequences such as green gentrification [40] and environmental gentrification [41,42,43]. Studies have highlighted that such patterns of redevelopment frequently occur at the expense of social equity, attracting higher-income populations while displacing vulnerable groups [44,45,46,47].

This underscores the pressing need for planning approaches that consider spatial and temporal dimensions in population distribution since both are inherently linked in shaping urban equity outcomes [48,49]. Indeed, Hunter, Cleland, Cleary, Droomers, Wheeler, Sinnett, Nieuwenhuijsen, and Braubach [50] emphasized that the dynamic understanding of park accessibility and the spatial equity effects of existing park interventions is insufficient. Existing static accessibility models often ignore temporal changes in population distribution, which may exacerbate spatial inequities in the near future, thus creating a persistent lack of access to adequate green spaces for certain communities over time. By failing to anticipate these demographic shifts, such models risk institutionalizing spatial injustice through outdated planning decisions. This limitation significantly weakens the models’ utility for long-term, equity-oriented urban planning—particularly in rapidly urbanizing regions.

In recent years, cities across the world have undergone profound social and demographic transformations. According to United Nations Department of Economic and Social Affairs Department [51], the global population is projected to increase to around 8.5 billion in 2030, 9.7 billion in 2050, and 10.4 billion in 2100. Chinese cities have particularly experienced rapid urbanization over the past two decades. As urban population demands continue to rise, achieving a balance between green space provision and population needs has become critical for effective urban spatial planning and for enhancing residents’ quality of life.

In response, China’s national planning framework—the 14th Five-Year Plan for Public Services—has prioritized the development of a more complete and equitable public service system. In 2024, the government further advocated for the creation of “a public service supply mechanism that is responsive to population changes,” aiming to enhance social equity and promote high-quality population development through optimized resource allocation. As China transitions from an era of rapid expansion to one of quality-focused urbanization, initiatives such as the “300 m to greenery, 500 m to a park” policy have significantly expanded park coverage. However, current park planning practices in China still largely rely on population density indicators to determine appropriate park supply. This approach, while useful, falls short of addressing the challenges of equitable and needs-based resource allocation. In this context, aligning park services with dynamic population changes has emerged as a critical concern for urban governance.

To address the current gap in planning tools, this study takes Shanghai, China, as its focal case. As one of the largest and most rapidly developing cities in China, Shanghai illustrates this challenge. Since 2000, the city has experienced dramatic population growth and spatial expansion, resulting in increasing pressure on its urban green space. Studies have identified a mismatch between the spatial distribution of green spaces and population density [10,31,37,52]. As traditional planning methods often overlook evolving population dynamics, park development fails to keep pace with actual demand. This disconnect highlights a structural gap between supply and demand. Moreover, static evaluation methods lack the flexibility to anticipate future demographic shifts, leading to short-sighted and reactive planning decisions.

To address these limitations, this study proposes a dynamic park accessibility evaluation model—the human-population-projection-Ga2SFCA (HPP-Ga2SFCA)—which enhances the conventional Ga2SFCA framework by integrating population projections. Using census data from 2000, 2010, and 2020 at Shanghai’s neighborhood committee level, along with spatial park data, this study performed time-series clustering to identify population change patterns and forecasted population distribution for the year 2030. By comparing park accessibility across different evaluation scenarios, the model revealed new insights into the spatial–temporal mismatches between service provision and population demand. The HPP-Ga2SFCA model links demographic forecasting with accessibility metrics, addressing the temporal limitations of traditional equity assessments and enabling proactive, equity-driven urban planning. The findings support Shanghai’s ambition to become a “City of a Thousand Parks” not only in terms of coverage but also through precision, responsiveness, and social justice in public service provision.

The remainder of this paper is organized into specific sections. Section 2 introduces the study area, data sources, and research methods. Section 3 presents the empirical results. Section 4 discusses the model’s contributions, planning implications, and limitations, and Section 5 concludes with a summary of the key findings.

2. Materials and Methods

2.1. Study Area

This research selected Shanghai as a representative case study due to the city’s acute land constraints and growing challenges in equitable public service provision amid rapid urbanization. Between 2000 and 2020, Shanghai’s permanent resident population increased from 16.41 million to 24.87 million, representing a 51.6% rise. Spatially, this population growth followed a pattern of differentiation (Figure 1B). While the central districts remained densely populated, the peripheral areas between the Outer Ring Road and the Suburban Ring experienced the most significant growth. By 2020, over 80% of neighborhood committee units reported a population density exceeding 1000 persons per square kilometer. Concurrently, the city’s demographic structure became increasingly diverse, contributing to increased socio-spatial segregation and widespread socio-spatial transformations [53,54,55]. These trends have intensified the need for a more equitable distribution of urban green space.

In response, the Shanghai municipal government launched the “City of a Thousand Parks” initiative, aimed at expanding and optimizing the city’s park system. Between 2000 and 2020, the total park green space increased from 4812 hectares to 21,981 hectares, and the number of parks grew from 122 to 406 (Figure 1A). Correspondingly, per capita park area grew from 2.93 to 8.84 square meters. In terms of spatial distribution, parks built before 2010 were largely concentrated in the city center. In contrast, during 2011–2020, most parks were located in the outer districts, signaling a deliberate strategy of spatial diffusion. This shift not only expanded the overall green space supply but also promoted a more balanced geographical distribution, extending public service benefits to areas beyond the urban core.

2.2. Data Source

Figure 2 illustrates the study area and the spatial distribution of newly constructed parks across two critical phases: 2001–2010 and 2011–2020. A panel dataset covering the period 2000 to 2020 was constructed for this study. Population and demographic data were obtained from China’s Fifth, Sixth, and Seventh National Population Censuses, providing permanent resident figures at the neighborhood committee level. All datasets were spatially standardized and harmonized according to the 2020 census boundaries for comparability. Detailed data on each park’s location and year of establishment were collected from official administrative sources for the years 2000–2020. This comprehensive dataset enabled the analysis of long-term trends in both population dynamics and green space provision at fine spatial resolution.

2.3. Research Methodology

2.3.1. Fuzzy C-Means Clustering (FCM)

To identify temporal evolution patterns in park accessibility, this study employed the Fuzzy C-Means (FCM) clustering algorithm using the Mfuzz package in R (version 4.4.0). Originally developed for analyzing gene expression time series in bioinformatics [56], FCM has since been applied in urban studies to uncover temporal trends, such as urban–rural energy inequality [57]. Compared to traditional clustering methods like K-means, FCM is more suitable for handling time series data. The FCM algorithm minimizes an objective function to determine the cluster centers, formulated as follows:

$$J\left(Y,V\right)=\min {{\displaystyle \sum }}_{c=1}^C{{\displaystyle \sum }}_{n=1}^T{y}_{cn}^m{d}_{cn}^2$$

(1)

Subject to

$$0\le y_{cn}\le 1,{{\displaystyle \sum }}_{c=1}^C{y}_{cn}=1,0\le {{\displaystyle \sum }}_{c=1}^C{y}_{cn}\le T$$

(2)

Here, $Y$ is a matrix of $C\times N$, and $y_{cn}$ indicates the degree to which the neighborhood committee $n$ belongs to the cluster $c$; $V={\left[v_1,v_2,\dots ,v_C\right]}^T\left(1\le c\le C\right)$ represents the matrix of cluster centers; and $m$ is the fuzzy parameter that control the fuzziness of clustering. In our analysis, we used the $mestimate()$ function, which returned the optimal value of $m=3.62$ based on the data. $d_{cn}^2$ represents the Euclidean distance between the neighborhood committee $n$ and the cluster center $c$.

Through iterative testing and adjustments, the optimal number of clusters was determined as $C=5$ to better fit the data. The resulting cluster assignments were mapped in ArcGIS to visualize the spatial distribution of different population evolution types across Shanghai.

2.3.2. Gaussian-Based Two-Step Floating Catchment Area Method (Ga2SFCA)

The 2SFCA method calculates accessibility in two steps that accounts for both the supply of parks and the demand from surrounding populations. Ga2SFCA is an enhanced spatial accessibility model that applies a distance decay function based on a Gaussian kernel to measure accessibility more realistically. The Gaussian decay function, which simulates the effect of spatial friction by exhibiting a fast initial decay that gradually slows, aligns more realistically with residents’ travel behavior. The algorithms are shown as follows [23,24]:

Step 1: Supply-side search for the supply–demand ratio $R_j$

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

(3)

Here, $D_k$ is the population of demand unit $k$ within the search radius (i.e., $d_{kj}\le d_0$); $d_{kj}$ is the distance between demand unit $k$ and park $j$; $d_0$ is the distance threshold for supply; $S_j$ is the area of park $j$; and $G\left(d_{kj},d_0\right)$ is the Gaussian distance decay function, accounting for spatial friction.

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

(4)

Step 2: Demand-side calculation of the park accessibility score $A_i^D$

$$A_i^D={{\displaystyle \sum }}_{j\in \left\{d_{kj}\le d_0\right\}}G\left(d_{ij}\right)R_j$$

(5)

Here, $R_j$ is the supply–demand ratio of park $j$ within the catchment area of demand unit (i.e., $d_{kj}\le d_0$), and $d_{ij}$ is the distance from demand unit $i$ and park $j$.

The choice of travel threshold is a critical parameter in accessibility analysis. However, there is no consensus in the existing literature on the optimal threshold value. Regarding the choice of search radius, there is currently no consensus in the literature. Commonly used thresholds include 1 km, 1.6 km, 2 km, 2.5 km, and others [10,58,59]. In this study, a 2 km Euclidean buffer radius was adopted, corresponding approximately to a 20–30 min walking distance. We offer a sensitivity test using 1 km and 3 km radii in Appendix A.1 and a network-based analysis in Appendix A.2. The findings confirm that while absolute accessibility values vary with the radius, the spatial pattern of inequality remains stable, supporting the robustness of our model results. Park accessibility was calculated using ArcMap (version 10.8.1). The resulting accessibility scores were classified into the following four levels using the quartile method:

• Low Accessibility: Bottom 25%;
• Medium-Low Accessibility: 25–50%;
• Medium-High Accessibility: 50–75%;
• High Accessibility: Top 25%.

2.3.3. Dynamic Park Accessibility Assessment Model Based on Human Population Projection (HPP-Ga2SFCA)

To extend park accessibility evaluations beyond present-day demand, this study proposes an enhanced park accessibility evaluation model—HPP-Ga2SFCA—that integrates population forecasting into the traditional Ga2SFCA method. By incorporating population data, the model extends accessibility assessments beyond current demand to account for anticipated future pressures. This approach enables the spatial quantification of potential mismatches between park service provision and population growth.

Given the limited availability of fine-scale temporal data in China—where population statistics at the neighborhood committee level are only available from decennial censuses (2000, 2010, 2020)—this study adopts a compound annual growth rate (CAGR)-based approach to forecast population changes over the next decade. Although more advanced forecasting techniques exist, they typically require denser time-series data and a broader set of predictors. Such models are thus less applicable in our context of small-area estimation with only three temporal data points.

We selected the year 2030 as the projection target for two main reasons. First, it aligns with the structure of China’s national census, which provides data every ten years, making 2030 a consistent and data-supported projection point. Second, it corresponds with key policy milestones outlined in Shanghai’s Urban Master Plan (2017–2035), which sets planning objectives and targets for the year 2035. As such, 2030 serves as a meaningful mid-term horizon for assessing the potential alignment of park service provision with projected population needs. While it is technically feasible to extend projections to a longer-term horizon such as 2040, doing so would increase forecast uncertainty due to the compounded effects of assumptions and the lack of additional demographic data. Therefore, this study focuses on 2030 as a realistic and policy-relevant timeframe.

The HPP-Ga2SFCA model proceeds through the following steps:

• A CAGR is constructed using historical population data from 2000, 2010, and 2020 at the neighborhood committee level.
• The projected population for each unit over a 10-year horizon is estimated based on the computed growth rate.
• These forecasted population figures replace the current population values in the Ga2SFCA model, allowing for a dynamic simulation of park accessibility under future population scenarios.

The revised accessibility model is expressed as follows:

$$A_i^D={{\displaystyle \sum }}_{j\in \left\{d_{kj}\le d_0\right\}}\left({\displaystyle \frac{S_j}{{{\displaystyle \sum }}_{k\in \left\{d_{kj}\le d_0\right\}}\left(G\left(d_{kj},d_0\right)·P_k^{forecast}\right)}}·G\left(d_{ij}\right)\right)$$

(6)

Here, $r$ is the compound annual growth rate, and $P_k^{forecast}$ represents the projected population in demand unit $k$. All other variables and parameters follow the definition provided in Section 2.3.2.

The projected population $P_k^{forecast}$ and annual growth rate $r$ are calculated as follows:

$$r={\left({\displaystyle \frac{P_{end}}{P_{start}}}\right)}^{{\displaystyle \frac{1}{m}}}-1$$

(7)

$$P_k^{forecast}=P_k^{current}\times {\left(1+r\right)}^n$$

(8)

Here, $m$ is the number of years between the base year and the end year, and $n$ is the number of years into the future for which the population is projected.

3. Results

3.1. Clustering Analysis of Population Change Patterns

Based on the time-series clustering analysis, five distinct patterns of population change were identified in Shanghai (Figure 3). These were categorized as follows:

• Population Growth:

  (1)

  Accelerated Growth (Cluster 2): This cluster exhibits a consistent increase in the growth rate over time.

  (2)

  Slowing Growth (Cluster 3): Initially experiencing rapid growth, this cluster shows a gradual decline in growth rate over time, as indicated by its fitted slope values and temporal trajectories.

• Population Decline:

  (1)

  Accelerated Decline (Cluster 1): This cluster is characterized by an increasing rate of population decline over time.

  (2)

  Slowing Decline (Cluster 5): Although it begins with a rapid decline in population, the rate of decline slows in the second decade, as evidenced by its fitted slope values and temporal trends.

• Population Fluctuation:

  (1)

  Initial Increase Followed by Decline (Cluster 4): This pattern features an initial rise in population, followed by a subsequent decrease, as revealed by its temporal trajectory and slope values.

As shown in

Table 1. Number and area proportions of population change clusters (2000–2020).

Category	Cluster	Cluster Name	Number of Units	Proportion (%)	Total Area (km2)	Area Proportion (%)
Population Growth	Cluster 2	Accelerated Growth Type	1456	23.65%	1301.97	13.87%
	Cluster 3	Slowing Growth Type	1922	31.22%	1687.53	17.98%
	Total		3378	54.86%	2989.5	31.85%
Population Decline	Cluster 1	Accelerated Decline Type	904	14.68%	2248.23	23.96%
	Cluster 5	Slowing Decline Type	971	15.77%	1943.69	20.71%
	Total		1875	30.45%	4191.92	44.67%
Population Fluctuation	Cluster 4	Initial Increase Followed by Decline Type	904	14.68%	2203.64	23.48%

, neighborhoods experiencing population growth accounted for 54.86% of all units but only 31.85% of the total land area. In contrast, declining neighborhoods comprised 30.45% of the units yet occupied 44.67% of the area, suggesting that population loss tends to occur in spatially larger zones.

Among all clusters, the Slowing Growth type (Cluster 3) was the most prevalent, representing (31.22%) of neighborhoods and characterized by a decelerating yet positive growth trend. The least common types—Accelerated Decline (Cluster 1) and Initial Increase Followed by Decline (Cluster 4)—each comprised 14.68% of units. In terms of spatial extent, Accelerated Decline neighborhoods occupied the largest area (23.96%), while Accelerated Growth areas were the smallest (13.87%).

The spatial distribution revealed marked population decline in both the urban core and the outermost suburbs (Figure 4). Within the inner ring, Accelerated Decline and Slowing Decline neighborhoods accounted for 3.04% and 5.36% of the total, respectively. Beyond the outer ring, these figures were 5.23% and 4.24%, respectively. Conversely, the area between the inner and outer rings emerged as a major zone of population concentration. Specifically, Slowing Growth neighborhoods in this belt comprised 12.55% of the total, while both Accelerated Growth and Slowing Growth types between the outer ring and suburban ring each exceeded 11%, identifying these zones as key areas of suburban expansion.

3.2. Spatial Patterns of Projected Population Growth

Using the population projection model, this study estimated the distribution of Shanghai’s population in 2030 under two scenarios: a short-term projection scenario, extrapolating growth rates from 2010 to 2020, and a long-term projection scenario, based on the compound annual growth rate from 2000–2020.

Figure 5 shows that both scenarios anticipate population concentration along the outer ring, whereas the urban core and the outermost suburban zones are expected to have lower densities. Figure 6A,B illustrate spatial patterns of population change between 2020 and 2030. Under the long-term scenario, continuous clusters of growth are projected between the inner and suburban rings, while the urban core is expected to experience population decline. In contrast, the short-term scenario shows a more dispersed pattern of growth and decline, resulting in a more balanced spatial distribution.

Figure 6C compares the proportion of population change across scenarios. Under the short-term projection, 51.94% of neighborhoods are expected to experience population decline, covering 70.83% of the area, exceeding growth areas in both number and extent. By contrast, the long-term scenario predicts population growth in 65.52% of neighborhoods, which covers 48.78% of the area, indicating a higher number but slightly smaller spatial range. These differences underscore how different projection methods produce diverging outcomes, reflecting the complexity of Shanghai’s demographic evolution over the past two decades.

3.3. Comparative Evaluation of Park Accessibility Models

To evaluate the impact of demographic dynamics on park accessibility, three scenarios were developed:

• Scenario 1: Static baseline, using the Ga2SFCA method with 2020 census data.
• Scenario 2: Short-term projection, applying the HPP-Ga2SFCA method using 2030 population projections based on 2010–2020 growth rates.
• Scenario 3: Long-term projection, applying the HPP-Ga2SFCA method using 2030 population projections based on 2000–2020 compound growth rates.

As shown in

Table 2. Comparison of statistical characteristics of park accessibility results between the traditional and improved models.

Scenario	Accessibility Model	Mean	Standard Deviation	Min.	Max.	Median
Scenario 1: Static baseline	Traditional Ga2SFCA	2.13	7.44	0.00	220.94	0.76
Scenario 2: Short-term projection	Improved HPP-Ga2SFCA	1.30	9.09	0.00	318.61	0.48
Scenario 3: Long-term projection	Improved HPP-Ga2SFCA	1.28	7.38	0.00	255.38	0.46

, the average global accessibility score in the static scenario was 2.13—significantly higher than in the short-term (1.30) and long-term (1.28) scenarios. This suggests that traditional static models may overestimate park accessibility by approximately 40%, introducing systematic bias in conventional evaluation methods.

In terms of spatial distribution across accessibility quartiles (Figure 7), low-accessibility zones (bottom 25%) increased from 53.63% under the static model to over 54.7% under both projection scenarios, indicating improved identification of underserved areas when population dynamics were considered. Middle-low accessibility areas (25–50%) expanded from 12.63% to between 15.96% and 16.95%, reflecting increasing service pressure. Meanwhile, high-accessibility zones (top 25%) decreased from 21.55% (static) to approximately 17% (projected), a reduction of around 24%. This suggests that static models significantly overestimate the extent of well-served areas and that these zones may face service dilution under future demographic pressures.

Figure 8A compares park accessibility outcomes from the static and improved projection models across quartiles, revealing substantial shifts. Many high-accessibility zones in the traditional model downgraded to middle-high levels in the improved model. Simultaneously, a significant number of middle-high zones either upgraded to high or declined to middle-low levels, while some middle-low zones improved to middle-high levels. Figure 8B,C presents the spatial patterns of accessibility change under the projection scenarios. Areas where accessibility improved (shown in red) indicate underestimation in the traditional model, while areas with decreased accessibility (shown in blue) represent overestimation.

Notably, overestimated zones—where the static model reports higher accessibility than the improved model—were predominantly located in the transitional belt between the inner and suburban rings, where population growth is expected to be substantial. In contrast, underestimated zones—where the static model underrepresents future service demand—were more fragmented, scattered within the inner ring (e.g., along Suzhou Creek in Huangpu District) and around the suburban ring (e.g., rural communities in Fengxian District). As shown in Figure 8D, overestimated zones accounted for over 70% of all neighborhood units and approximately 45% of the total urban area, far exceeding the 7% of units and 4% of areas associated with underestimated zones.

These results demonstrate that when population projections are incorporated into park accessibility assessment, both over- and underestimation patterns emerge across the city—patterns that remain hidden in static models. The fact that overestimated areas are spatially extensive and closely aligned with regions undergoing demographic expansion indicates they are not isolated anomalies, but rather critical risk zones for future service mismatches. Conversely, underestimated zones may falsely suggest a need for additional parks, leading to unnecessary investments in areas unlikely to face future service shortages. The coexistence of these two types of misestimations underscores the structural limitations of static models, which are ill-equipped to accommodate dynamic urban conditions. In contrast, the improved HPP-Ga2SFCA model provides a more forward-looking and adaptive framework, enabling planners to better anticipate emerging pressures and implement more equitable, demand-responsive interventions.

4. Discussion

4.1. Main Findings

Building upon existing research on green space equity, this study introduces a dynamic perspective on urban park accessibility, addressing the temporal mismatch between static service provision and evolving population demands. To overcome the limitations of traditional static planning, we developed an enhanced dynamic park accessibility model—HPP-Ga2SFCA—that integrates projected population data into the Ga2SFCA method. Using park supply data and multi-period census data at the neighborhood committee level in Shanghai, we analyzed historical population trends, projected future demographic changes, and compared accessibility outcomes from both static and dynamic models. The key findings are summarized below:

(1)

Population dynamics and spatial patterns (2000–2020): Time-series clustering revealed five distinct types of population change trajectories, grouped into three broader categories—(i) Population Growth (Accelerated Growth, Slowing Growth), (ii) Population Decline (Accelerated Decline, Slowing Decline), and (iii) Population Fluctuation (Initial Increase Followed by Decline). Neighborhood committees in the growth category accounted for 54.86% of all units but only 31.85% of the total area, indicating that growth was concentrated in relatively small urban zones. Conversely, population decline occurred in fewer units but spanned larger spatial areas. The most common pattern was Slowing Growth, characterized by rapid early growth followed by moderation. Accelerated Decline had the widest spatial extent. Spatially, population decline was most evident in both the city center and the peripheral zones, while rapid population aggregation primarily occurred in the area between the inner and suburban rings.

(2)

Future population projections and spatial patterns: Under both short-term and long-term projection scenarios, population hotspots in 2030 are expected to be concentrated near the outer ring, while cold spots will largely be located in the inner ring and beyond the suburban ring. The long-term scenario projects that 65.52% of neighborhood committees (covering 48.78% of the area) will experience population growth, especially in the transitional zone between the inner and suburban rings. In contrast, the short-term scenario, 51.94% of units (covering 70.83% of the area) will experience population decline, reflecting a more spatially balanced distribution. These contrasting outcomes highlight the complex and nonlinear trajectory of Shanghai’s demographic evolution over the past two decades, as well as the sensitivity of population projections to differing temporal frameworks.

(3)

Improved park accessibility assessments: Static assessments that do not account for population dynamics were found to overestimate park accessibility by approximately 40%. By integrating population forecasts, the dynamic model allowed for more accurate identification of future service gaps and highlighted emerging stress on areas currently considered well-served. The model revealed a tendency for service pressure to shift toward zones presently classified as having medium or low accessibility, indicating a potential dilution of service levels under future demographic conditions. Furthermore, the study identified spatial mismatches in traditional model outputs. Around 71% of neighborhood committees may have been underestimated in terms of accessibility, while approximately 45% of the city’s area may face overestimation risks. Overestimated zones were typically located in areas expected to undergo significant population growth, particularly between the inner and suburban rings, whereas underestimated zones were scattered throughout the inner ring and on the suburban fringe.

In China, static planning mechanisms shaped by the legacy of the planned economy often fail to adapt to rapidly changing social structures, leading to persistent mismatches between service supply and population demand. By introducing both short-term and long-term forecast scenarios, our analysis aims to reveal and amplify the potential bias between park provision and future demographic shifts. The short-term forecast is largely shaped by immediate policy impacts and short-cycle urban dynamics, while the long-term forecast reflects the compounded effect of structural population dynamics and policy trajectories across two decades. In both scenarios, the coexistence of park-surplus and park-deficient areas under different scenarios highlights the urgency of adopting dynamic, forward-looking planning approaches. These findings suggest that static models may inadequately support effective park planning interventions and highlight the importance of integrating population projections into accessibility evaluations.

4.2. Practical Planning Insights

This study provides valuable insights for planning practice, supporting the development of a forward-looking and equity-oriented green space service system aligned with evolving demographic trends. Urban park planning should shift from static, supply oriented models to dynamic strategies that prioritize population-responsive equity. We recommend integrating population projections into routine park accessibility assessments, alongside the establishment of a dynamic monitoring system to regularly detect service deficits and surpluses. Based on this, spatially differentiated planning interventions should be implemented in areas with accessibility overestimation or underestimation— ensuring park supply decisions reflect actual and future demand. Such interventions must coordinate population management and facility provision, aligning housing development with green space planning to promote spatial equity.

In peri-urban zones projected to experience rapid population growth, planning frameworks must go beyond land availability to account for anticipated population increases. Our model demonstrates that large-scale housing expansion without proportional park allocation leads to future service imbalances. To avoid such mismatches, early-stage spatial planning should incorporate dynamic population forecasts, enabling timely identification of zones at risk of accessibility shortfalls. Housing and park provision strategies should be jointly designed based on service capacity forecasts, including moderating housing growth rates, facilitating adaptive reuse of existing stock, and tailoring park provision to evolving demographic structures (e.g., aging, child populations). Instead of static long-term plans, flexible, shorter planning cycles and post-implementation feedback mechanisms—such as routine community surveys—are essential for ongoing service adjustment based on actual needs.

Conversely, in inner-city districts projected to experience population decline, planning should focus on optimizing and upgrading existing resources. This includes improving the quality, multifunctionality, and inclusivity of small-scale parks, while safeguarding access for aging and low-income residents. Park projects that remain unimplemented should be reassessed in light of the latest demographic data, with adjustments made to construction timing and intensity to avoid redundancy and inefficiency. Additionally, expanding the supply of affordable housing in these areas is suggested as a possible mechanism to encourage the return of disadvantaged populations, thereby broadening and rebalancing park service coverage [60,61,62].

4.3. Advantages and Limitations

This study enhances the capacity to identify future, population-driven demand pressures in park accessibility evaluations, improving the responsiveness of park provision to evolving demographic conditions. In the context of rapid urban population changes, particularly in megacities, the proposed approach offers a new lens through which to improve current park planning practices.

This study introduces several key innovations across three dimensions: theory, methodology, and data. At the theoretical level, it introduces a population-responsive public service supply framework to park planning, highlighting the importance of proactively identifying supply–demand mismatches and enabling more dynamic decision-making. At the methodological level, it incorporates a dynamic population projection mechanism into the traditional accessibility assessment framework, addressing the limitations of static models and promoting methodological innovation. By comparing multiple future scenarios, this study systematically reveals how demographic changes influence accessibility assessments and the future spatial equity of park provision. At the data level, we employ high-resolution, multi-period population census data at the neighborhood committee level significantly enhances the spatial precision and temporal depth of accessibility evaluations.

In conclusion, this study provides both theoretical and practical tools to improve the timeliness and precision of park planning interventions, contributing to a dynamically equitable urban green space system that aligns with projected demographic changes. However, this study may have several limitations. First, park accessibility was calculated using Euclidean distance buffers rather than network-based distances. We also recommend future research to integrate dynamic population change with park quality evaluation—such as user satisfaction, maintenance levels, and facility diversity. Second, our population projection relied on a basic CAGR model due to the limitations of demographic data in China, including the fine spatial scale and the long temporal interval. While this model is suitable for estimating long-term population trend, it may not fully capture complex urban dynamics. Future research could enhance predictive accuracy by incorporating more advanced models—such as grey prediction models [63,64], neural networks [65], or system dynamics models [66]—if finer temporal data become available. Third, behavioral differences in park use across demographic groups can affect supply-demand dynamics. While our current model assumes uniform park usage, we suggest using mobility or survey data to consider differentiated usage behavior for future mobility-sensitive evaluations. The population projection in the proposed HPP-Ga2SFCA model can also be extended by incorporating socio-demographic indicators (e.g., age, income, or ethnicity) to better assess specific vulnerable groups. This enables more nuanced and multidimensional assessments of dynamic social justice. Finally, we acknowledge the potential impact of the Modifiable Areal Unit Problem (MAUP) [67,68]. A robustness check at the sub-district level confirmed consistent patterns (see Appendix A.3).

Despite these limitations, the integration of population forecasting into park accessibility analysis offers meaningful insights for policymakers and urban planners. It supports the pursuit of dynamic spatial equity in green space planning and encourages more responsive, future-oriented interventions.

5. Conclusions

This study proposed a dynamic evaluation framework—HPP-Ga2SFCA—that integrates projected population changes into park accessibility assessments, thereby addressing the limitations of static planning approaches. By analyzing multi-period demographic and park spatial data in Shanghai, we identified significant spatial–temporal mismatches between projected population demand and current green space provision. Our findings demonstrate that traditional static models fail to match the pace of urban transformation, potentially resulting in future inequities. In contrast, the dynamic model offers a more responsive understanding of accessibility by capturing emerging service pressures in rapidly growing areas. This enables planners to engage in timely adjustment of green space provisions in line with evolving population patterns. The framework provides a novel analytical tool and practical insights for building a more adaptive, equitable, and sustainable urban green space system—especially vital for managing growth in fast-changing urban fringe areas.