Applying Stability Theory to Urban Green Space Management: A Case Study in Shanghai, China

Abstract

Landscape structure significantly impacts ecosystem services, yet the stability of ecosystem services in urban green spaces has been insufficiently studied regarding landscape effects. In highly urbanized regions such as Shanghai, it remains unclear which landscape configurations can maintain consistently high and stable regulating services. By calculating the monthly values of regulating services in urban green space sites across five years, we investigated how landscape structure and urbanization influence the temporal stability of regulating service bundles along an urban gradient in Shanghai. Stability was measured as the inverse of the coefficient of the regulating service values, further decomposing it into average regulating service stability and regulating service asynchrony, following the ecological theory. Landscape structure metrics included area, fragmentation, and shape, while urbanization was measured as the proportion of impervious surface surrounding green space sites. The results showed that the stability of regulating service bundles was higher during spring or winter compared to summer and autumn. Overall, we found that fragmentation reduced the stability of regulating service bundles, whereas impervious surfaces had a positive effect, both acting through average regulating service stability. Our study promoted a framework for managing urban green spaces to sustain high and stable ecosystem services, highlighting the importance of preserving contiguous green areas to support sustainable urban planning.

1. Introduction

Urbanization has altered the urban landscape, posing a significant threat to the capacity of ecosystems to provide essential services to humans [1,2]. Landscape structure has significant impacts on ecosystem services, and it is commonly assumed that simplified landscape composition and configuration can lead to reduced service provision [3]. Previous research has largely focused on maximizing ecosystem services or exploring the relationships among multiple ecosystems [4,5], whereas their temporal stability—the consistency with which these services are supplied—has received comparatively less attention. Furthermore, in the context of the pattern–process–service–sustainability paradigm [6], it is crucial to consider not only the high level of ecosystem services but also their temporal stability in relation to landscape structure. Additionally, because multiple ecosystem services (i.e., ecosystem service bundles) often exhibit trade-offs, synchrony, or no clear relationships [7], asynchronous dynamics among different services may buffer fluctuations in individual services and thereby affect the stability of the overall bundle.

Stability is generally defined as the degree to which an ecosystem property fluctuates over time, with smaller variability indicating greater stability, referred to as temporal stability in ecology [8,9]. The overall stability of ecosystem services can be understood as a function of both the temporal dynamics of individual services and the asynchrony among them (Figure 1). ES stability is the average coefficient of variation for a site of all ecosystem services, weighted by their value. ES asynchrony is used to capture the interspecific asynchrony among multiple ecosystem services. Increasing ES asynchrony and the mean ES stability would lead to greater overall ESB stability. ESB stability is equal to the average ES stability under complete synchrony of all ES values, and increases when they become asynchronous. Any factor contributing to ES stability and/or ES asynchrony provides stabilizing effects to ESB dynamics. However, the impact of green space structure on ESB stability remains largely unexplored, which is necessary for balancing ecosystem service maintenance during urban development.

The capacity of urban green spaces to maintain these ecosystem services depends on their landscape structure [10,11]. With a more nuanced understanding of the effects of landscape structure, the landscape can be better planned to sustain the long-term provision of ecosystem services [12,13]. Although maintaining the stability of ES is fundamental for a sustainable landscape [14], limited research has been conducted on the correlation between landscape structure and the stability of ecosystem services, especially in urban green spaces. According to the Shanghai Forestry Bureau, the value of forest ecosystem services—including recreation opportunities, air quality regulation, climate regulation, water flow management, biodiversity protection, nutrient accumulation, soil fertility maintenance, and forest protection—has increased from approximately 1.91 billion USD in 2015 to around 2.20 billion USD in 2019, yet empirical evidence of how landscape structure within green spaces influences the stability of ecosystem services in urban areas remains limited.

Landscape structure (i.e., area and fragmentation) may affect ESB stability through ES stability (Figure S1, adapted from [15]). Multiple ecosystem services with the same trend variation could be provided in large green spaces to achieve mutually beneficial outcomes. For instance, large urban gardens increasing ecological value (i.e., biodiversity conservation) could also generate significant social benefits (i.e., learning and well-being) [16]. Additionally, modifying the composition of urban green space components can easily stimulate the synergy of ecosystem services [17,18]. As such, the landscape structure of urban green spaces may enhance ESB stability via ES stability.

Evidence suggests that increased landscape heterogeneity (e.g., the configuration of patches) can intensify the trade-offs between multiple ecosystem services, such as those between carbon sequestration, cooling effect, and water quality regulation [19,20]. Correspondingly, landscape structure may affect ESB stability via ES asynchrony in green spaces (Figure S1). This implies that a decline in certain ecosystem services can be offset by an increase in others, resulting in stable ecosystem services. Furthermore, urbanization could affect the synergies or trade-off relationships between ecosystem services [21], which should be involved in exploring the association between landscape structure and the stability of ecosystem service bundles.

The measurement of ecosystem service value provides an economic perspective on the benefits that ecosystems provide to humans and is a widely used tool in sustainable development [22]. Here, we explore the relationship between landscape structure (indicated by area, shape, and fragmentation) and the stability of ecosystem service value based on monthly data in 12 green space sites along an urban gradient across five years in Shanghai, China. The landscape patterns of the 12 sites included in this study remained unchanged throughout the studied period. We focus on assessing regulating services provided by urban green spaces, which are crucial and widely used in urban planning and green space management research [23,24]. In this context, we focus on how urban green space structure across an urbanization gradient contributes to sustaining stable and high levels of ecosystem services. Specifically, we address two questions: (1) Which aspects of landscape metrics determine the stability of regulating bundles in green spaces: area, fragmentation, or shape? (2) How does landscape structure affect the spatial stability of regulating service bundles, through the regulating service asynchrony or the average regulating service stability?

2. Materials and Methods

2.1. Study Area

Shanghai is located in the Yangtze River Delta, characterized by a flat alluvial plain with an average elevation of about 4 m. The city has a subtropical monsoon climate with hot, humid summers, mild winters, and annual precipitation of around 1499.1 mm in 2024. As a highly urbanized metropolis, rapid land-use change has reduced and fragmented natural habitats. Nevertheless, urban green spaces, such as parks and waterfronts, remain critical for biodiversity conservation and the provision of ecosystem services in this dense urban environment. The study was performed in 12 green spaces sites (Figure 2; Figure S1, Supplementary Materials), i.e., 4 forest park sites, 4 urban park sites, and 4 green belt sites, with each type containing four sites along an urbanization gradient (Figure S2; Table S1, Supplementary Materials), on the mainland of Shanghai, China (114°48′–116°39′ E, 33°42′–34°18′ N). Green spaces in Shanghai have expanded by 32,043.76 hectares between 2014 and 2019, according to the Shanghai Statistical Yearbook of 2015 and 2020.

Green space sites analyzed in this study were categorized based on their functional roles and locations: (1) Forest parks, characterized by a predominant coverage of large trees and the preservation of diverse natural resources, satisfy the recreational demand of citizens while meeting the recreational needs of citizens simultaneously [25]. (2) Urban parks provide recreational opportunities and landscape aesthetics for citizens, featuring diverse and well-manicured vegetation maintained [26]. (3) Green belts, we here define as woodland areas alongside roads that play an important role in mitigating wind, noise, or particulate matter [27].

2.2. Calculating Regulating Services

We here calculated the monthly values of regulating services from March 2014 to February 2019. Regulating service values were calculated based on the unit value-based approach [28]. The value of each regulating service for each land-use (LU) type is calculated as the product of the economic value per unit area of a certain service and its corresponding area (Table S2). We mapped the LU of each green space based on field research combined with Google Earth. Woodland, shrub, lawn, and water were extracted in the following analysis as these LU types are the functional units to provide regulating services (i.e., air quality regulation, climate regulation, waste treatment, and regulation of water flows) in each green space. We calculated air quality regulation, climate regulation, waste treatment, and regulation of water flows, and assessed them as follows:

$${RSV}_{gj}=\sum _1^i{A}_i\times V_{gij}$$

(1)

$$V_{gij}=\sum _1^n{UV}_{in}\times F_{nij}(P_{gj} or R_{gj}){\times A}_i$$

(2)

where RSV_gj is the total value of regulating services in month j in green space site g, A_i is the area of LU type i (i = 4 in this study); V_gij is the unit area value of regulating services of LU type i in month j in site g (USD/ha); UV in is the unit area value of certain kind regulating service n (n = 4 in this research) of LU type i; F_nij is the equivalent dynamic factor for regulating service n in month j. Here, net primary productivity (NPP) and precipitation were used as the dynamic equivalent factors F_nij.

UV is the product of the standard equivalence factor (D) and the equivalent coefficient obtained from [28]. We calculated D as follows:

$$D=\sum {S_m\times F}_m$$

(3)

where D is the economic value of the standard equivalent factor; S_m is the percentage of the area of grain m in total crop area (%); F_m is the net profit per hectare of grain m (USD/ha). We extracted all the data based on the China Rural Statistical Yearbook from 2015 to 2019. Finally, we used the mean economic value across five years as the standard equivalent factor. The final economic value of the standard equivalent factor is 256.31 USD/ha.

We used the improved CASA model [29] to calculate the monthly ${NPP}_j$ in site g, which integrated the NDVI, precipitation, temperature, and solar data into this model. We extracted monthly NDVI, precipitation, temperature, and solar data from the Moderate Resolution Imaging Spectroradiometer (MODIS) NDVI product (https://ladsweb.modaps.eosdis.nasa.gov/) and National Earth System Science Data Center in China (http://www.geodata.cn/), respectively. The solar observed site data from China Meteorological Data Service Centre (http://data.cma.cn) was also obtained and interpolated to build a 1 km resolution raster by the IDW spatial interpolation method. Then, we incorporated the above data into the CASA model to assess the monthly NPP value. We extracted the monthly NPP value of each study site for further analysis by using ArcGIS 10.7. We calculated P_gj as follows:

$$P_{gj}={\displaystyle \frac{{NPP}_{gj}}{\overline{NPP}}}$$

(4)

where NPP_gj refers to the NPP (t/ha) of green space site g in month j; and $\overline{NPP}$ refers to the regional monthly average NPP (t/ha).

We use precipitation as a spatial and temporal regulation factor for the regulation of water flow services, and we calculated it as follows:

$$R_{gj}={\displaystyle \frac{W_{gj}}{\overline{W}}}$$

(5)

where W_gj refers to the average precipitation per unit area (mm/ha) at month j in site g; and $\overline{W}$ refers to the monthly average precipitation per unit area (mm/ha). We extracted monthly precipitation from National Earth System Science Data Center in China (http://www.geodata.cn/) with a 1 km resolution.

2.3. Stability of Regulating Services

The stability of ecosystem service bundles (ESB stability) is thus characterized by the interannual variability of ecosystem service values during the study period (e.g., the variation in regulating values in November from 2014 to 2018), following previous studies [30,31,32]. The stability of individual ecosystem services (ES stability) is measured as the invariability of their interannual values, quantified as the inverse of the coefficient of variation in ecosystem service values within a site across years. The asynchrony among multiple ecosystem services (ES asynchrony) is calculated as the inverse of the square root of ecosystem service synchrony (φ) [32,33]. We calculated the monthly stability of regulating service bundles (ST_RSB) as the inverse of the coefficient of variation in the regulating service value across five years. To better understand how landscape structure impacts the ST_RSB in one site, we partitioned the ST_RSB into the average regulating service stability (ST_RS) and the regulating service asynchrony (1/φ) following the theoretical framework in Thibaut & Connolly (2013) [33], as follows:

$${ST}_{RSB}={\left({\displaystyle \frac{{\sigma }_G}{{\mu }_G}}\right)}^{-1}= {ST}_{RS}/\phi$$

(6)

$$1/\phi ={\left({\displaystyle \frac{{\sigma }_G}{{\sum _i\sigma }_i}}\right)}^{-1}$$

(7)

$${ST}_{RS}={\left( \sum _i{\displaystyle \frac{{\mu }_i}{{\mu }_G}}\times {\displaystyle \frac{{\sigma }_i}{{\mu }_i}} \right)}^{-1}$$

(8)

where the ST_RSB is the stability of the regulating service bundles, ST_RS is the average stability of regulating services; ${\mu }_G$ are the mean value of all kinds of regulating services within site G across years and ${\sigma }_G$ is its standard deviation; ${\mu }_i$ is the mean value of regulating service i; ${\sigma }_i$ is the standard deviation of the specific regulating service i. The temporal stability of regulating service bundles within a site increases when the coefficient of variation in regulating service values decreases.

2.4. Landscape Structure

We used landscape metrics to measure landscape structure from size (TA, the total area of green space patches), shape (CIRCLE, the mean of the related circumscribing circle of green space patches [34]), and fragmentation (DIVISION, the probability that two randomly chosen green space patches in the landscape are in different patches of the corresponding patch type [35]) (Table S3, Supplementary Materials). All these variables were calculated in R v4.2.2 using the landscapemetrics package (version 1.5.2). We calculated the metrics of the three aspects for all green space patches: the woodland patches, the shrub patches, the lawn patches, and the water patches.

We calculated the mean impervious area within a 500-m buffer around the focal site from 2014 to 2018 to capture the urbanization around each site using the GAIA database, an annual map of the global artificial impervious area with a high spatial resolution (30 m) from 1985 to 2018 [36]. The analysis was done in ArcGIS 10.8.

2.5. Statistical Analyses

Linear mixed models were used with all combinations of multiple predictors to test whether ST_RSB was related to landscape metrics (i.e., the total area of green space patches [TA], the related circumscribing circle of green space patches [CIRCLE], the division of green space patches [DIVISION], and the impervious surface within 500 m buffer), and the month was treated as a random effect. All landscape metrics were lowly correlated (Spearman |ρ| < 0.4, Figure S3, Supplementary Materials). We used the function model.avg from the MuMIn package in R to select all models and perform model averaging for those with ΔAICc < 2. All response variables were log-transformed to improve the normality, and landscape metrics were standardized before analyses to make the interpreted parameter estimates comparable.

We further disentangled the effects of landscape metrics on ST_RSB and its two components, i.e., the average regulating service stability (ST_RS) and the regulating service asynchrony (1/φ), using the structural equation model (SEM). In the SEM, we set three linear models to calculate the effects of landscape structure on the mean ST_RS (model 1), on the 1/φ (model 2), and the effects of the mean ST_RS and 1/φ on ST_RSB (model 3). Additionally, the month was treated as the random effect in the SEM. We examined the SEM using the ‘piecewiseSEM’ and lme4 packages in R. Non-significant Fisher’s C values (p > 0.05) indicated that the SEM provided a good fit. Variables were log-transformed to achieve normality before analysis.

3. Results

3.1. Temporal Stability of Regulating Service Bundles

The regulating service values exhibited temporal instability from March 2014 to February 2019 across the studied green space sites (Figure S4, Supplementary Materials). The stability of regulating service bundles (ST_RSB) was relatively high in spring or winter, while the lowest ST_RSB usually occurred during summer or fall (Figure S4, Supplementary Materials). The average regulating service stability had the same trends with ST_RSB, while the regulating service asynchrony showed a relatively stable fluctuation (Figure S5, Supplementary Materials).

3.2. Effects of Landscape Metrics on the Stability of Regulating Service Bundles

The monthly stability of regulating services bundles (ST_RSB) was mainly affected by the division index of woodland patches (DIVISION_{woodland patches}) and impervious surfaces within buffer 500 m (Impervious surfaces_{buffer 500 m}). The impact of DIVISION_{woodland patches} was negative (Figure S6, Supplementary Materials), indicating that the green space site with more patches decreased the ST_RSB. The impact of IM_{buffer 500 m} was positive (Figure S6, Supplementary Materials), indicating that ST_RSB in green space sites in the highly developed area was more stable. CIRCLE_{woodland patches} and total green space area showed no significant correlations with the ST_RSB.

For the significant effects of DIVISION_{woodland patches} and IM_{buffer 500 m}, we further tested their pathways on ST_RSB, whether through 1/φ or ST_RS. The two metrics affected ST_RSB through the mean ST_RS (Figure 3). DIVISION_{woodland patches} and IM_{buffer 500 m} explained 39% of the variation in ST_RS. ST_RS increased with the Impervious surface_{buffer 500 m} (standardized (std.) effect = 0.62, p < 0.05) and decreased with the DIVISION_{woodland patches} (Std. effect = −0.81, p < 0.05).

4. Discussion

4.1. Landscape Structure and Urbanization Effects on the Stability of Regulating Service Bundles

Our results showed that landscape metrics have a significant impact on the monthly stability of regulating service bundles (ST_RSB). Specifically, landscape fragmentation is the primary driver of monthly variability in these green spaces. In our research, we used the division index to describe the fragmentation of each green space site. Additionally, the monthly ST_RSB was also affected by impervious surfaces. These findings indicate that landscape structure should be taken into account in urban planning to enhance the sustainability of ecosystem services.

The effects of fragmentation of woodland patches on the temporal stability of regulating service bundles (ST_RSB) are in line with previous studies that have identified fragmentation as the key driver of ecosystem services and their stability [37,38]. Easily accessible parks are desired in urban areas to promote human health and well-being; however, this may lead to landscape fragmentation, which can have negative effects on the water quality regulation [39], which may be due to the plant species composition affected by disturbance. [40]. For instance, forest parks, despite their large size (Figure S2) and conservation value, fragmentation from roads and recreational infrastructure may intensify edge effects and disrupt connectivity, which could lead to greater fluctuations in NPP and hydrological processes, thereby reducing the stability of regulating services [41,42]. By contrast, urban parks, though providing important recreational and aesthetic benefits, contain diverse but intensively managed vegetation, which may be more vulnerable to fragmentation-related disturbances that alter vegetation composition and reduce regulating services [43]. Meanwhile, green belts are inherently narrow and linear; thus, fragmentation further amplifies edge effects, increasing exposure to pollutants and reducing their efficiency in keeping regulating services [44].

Additionally, most of the research sites were situated in urban areas where human proximity destabilized the NPP [45]. In this research, an increase in the degree of fragmentation within a green space site is accompanied by a corresponding rise in its patch count. As fragmentation is a key driver of ecosystem services, such as water yield and purification [46]; therefore, it can significantly lead to greater instability in the value of regulating services. Furthermore, we found that there was no significant trade-off between each regulating service (Figure S7, Supplementary Materials). Therefore, landscape metrics affect ST_RSB through the average regulating service stability (ST_RS) in our research.

Interestingly, the presence of impervious surfaces representing urbanization was found to have a positive impact on the ST_RSB. This finding is in contrast to the negative impact of urbanization on ecosystem services (e.g., carbon sequestration), which may constrain sustainable urban development [47,48,49]. This may result from well-maintained vegetation in relatively highly developed areas, such as urban parks [50], and may lead to a relatively high and stable Normalized Difference Vegetation Index (NDVI). In addition, the temperature may be more stable in highly developed areas compared to suburban areas [51]. All these variables are the main drivers of NPP, and their stable dynamics may result in stable fluctuations in regulating service bundles.

4.2. Implications for Green Space Design

Our study provides an empirical study for optimizing stable service provision through managing landscape structure. We explored the effect of landscape structure on the temporal stability of ecosystem service bundles in urban areas according to the stability framework. By examining the correlation between landscape structure and ecosystem service stability, we can enhance our comprehension of ecosystem stability in relation to the diverse effects of landscape characteristics [52,53].

Fragmentation resulting from setting too many smaller patches would decrease the stability of ecosystem services, yet it enhances their value in large area sites where fragmentation is consistently high. Our findings demonstrate the significance of the SLOSS debate, which is widely emphasized in conservation ecology [54,55]. In this context, it is crucial to consider whether it is more effective to establish a single large patch of green space or multiple smaller ones with the same total area when designing green spaces. However, in reality, preserving relatively large green spaces within urban areas can be challenging. Our findings indicate that planners should avoid incorporating excessive road infrastructure within green spaces, as it may fragment these areas and compromise their capacity to provide stable regulating services. However, smaller patches, such as private gardens, can significantly contribute to the delivery of ecosystem services in cities [56,57]. Therefore, when designing green space sites, it is challenging to determine whether a single large patch or multiple small patches should be utilized. A combination of larger and smaller patches may prove most beneficial as the former can preserve high ecosystem services value while the latter caters to human needs. This is also relevant to comprehending the land-sparing versus land-sharing debate for wildlife-friendly in order to discern when and why fragmentation hinders or promotes ecosystem service provision.

Another crucial aspect to consider in green space design for achieving sustainable ecosystem services is the urban matrix surrounding green spaces. Ecosystem services in Shanghai, such as gas regulation, climate control, water conservation, and waste management, were detrimentally impacted by urbanization [58]. Plants in forest patches within highly urbanized areas, which play a crucial role in providing ecosystem services, are under significant stress due to heat and pollution, potentially impacting the effectiveness of regulating services. Furthermore, large forest patches in peri-urban areas provide more ecosystem services compared to those in highly urbanized city centers [59]. As such, we should reconsider how to control the construction intensity around the green spaces.

4.3. Limitation

The equivalent factor method was used to calculate the monthly regulating services based on the equivalent values in this study. Although we adjusted the equivalent values according to the characteristics of Shanghai, the method does not accurately reflect the actual ecosystem services. Additionally, our sampled green space sites do not encompass all possible types of green space in urban areas, such as gardens, which may not fully capture all the potential and influential landscape metrics. The present study investigated monthly regulating services across 12 green space sites, with four sites in each green space type. Despite conducting 60 repeated regulating service assessments to account for seasonal variations, the limited number of sampling sites for each green space type may have led to an underestimation of the full range.

Our analysis only considered landscape structure and did not incorporate detailed vegetation information, and we did not include explicit ecological connectivity metrics due to their strong collinearity with fragmentation indicators (e.g., Aggregation and Cohesion Indexes). This may have overlooked important ecological processes influencing the dynamics of ecosystem services. Moreover, most of our sampling sites were located within urban parks and forest parks, where vegetation is regularly maintained and factors such as NPP tend to fluctuate less than in unmanaged areas. This spatial bias might partly explain the unexpected positive effect of impervious surfaces. Additionally, the non-significant effects of green space area and shape may reflect their relatively weak influence on ecosystem service stability compared to other structural factors, as well as limited variability among sampled sites and potential scale.

The analysis may not fully capture the trade-off between various ecosystem services if multiple aspects, such as supporting services, provisioning services, and cultural services, are not considered. Therefore, future research should prioritize field surveys to assess the real stability of ecosystem services and involve citizen programs in data collection to acquire more site-based information, and explore potential non-linear relationships between green space structure and ecosystem service stability using large datasets.

5. Conclusions

This study directly examines the impact of landscape structure on the stability of regulating services in urban green spaces, using the stability concept framework widely employed in research on animal and plant communities [60,61]. Using monthly data over five years from 12 green space sites across Shanghai, we assessed the effects of landscape metrics—area, shape, and fragmentation—on the stability of regulating services. Our results confirm that landscape fragmentation can destabilize the stability of regulating services. Our findings indicate that effective management of the landscape structure in green spaces might promote the stability of ecosystem services, particularly by prioritizing measures to prevent fragmentation. However, our study also reveals that certain landscape metrics, such as area, which are typically associated with the enhancement of regulating services, do not necessarily contribute to greater stability. Nevertheless, this study has a small sample size, reliance on statistical indices without independent ecological validation, and a focus on regulating services without considering other ecosystem services. These constraints may partly explain the unexpected non-significant effects of area and shape. Our findings emphasize the importance of preventing fragmentation when planning and managing urban green spaces to enhance ecosystem service stability. Future research should incorporate larger sample sizes and extend analyses across multiple temporal and spatial scales, while also integrating fine-scale vegetation data and validating results with field-based ecological measurements. Such efforts will strengthen the ecological basis for designing sustainable urban landscapes that support both biodiversity and long-term human well-being.