Introduction

Worldwide, urban societies grapple with the complex interplay between environmental health and socio-economic development1,2. Contemporary ecological challenges emerge from systemic factors including institutional arrangements, economic structures, and governance frameworks that have historically prioritized economic growth over environmental stewardship3. This multifaceted challenge is especially pronounced in mega-urban agglomerations, where multiple cities become interconnected through economic, social, transportation, and ecological systems4,5, creating intricate sustainability dynamics.

These interconnected urban systems leverage spatial concentration of economic activity to drive regional competitiveness and development6. However, these benefits often come with significant environmental costs7,8. Many residents in mega-urban agglomerations now face escalating environmental risks, including rising carbon emissions, habitat fragmentation, increased flood vulnerability, and the deterioration of ecosystem services9,10,11.

Effectively integrating environmental sustainability and socio-economic well-being requires robust quantification and assessment methods. On the environmental front, natural environments provide essential ecosystem services that support human health and well-being2,12,13,14, including air purification, water regulation, climate stabilization, and recreational opportunities. Measuring these services through ecosystem service value (ESV) calculations offers a standardized method for evaluating environmental quality, which is crucial for informing effective ecological policies15. On the socio-economic front, income levels serve as a widely recognized indicator of socio-economic well-being16. Higher incomes generally enhance access to housing, healthcare, and education, thereby improving overall quality of life17.

The relationship between ESV and socio-economic well-being varies spatially within mega-urban agglomerations. Different areas possess unique ecological sensitivities, environmental carrying capacities, and development potentials that affect how economic activities impact local ecosystems18,19,20. Recognizing this spatial heterogeneity is essential for strategically positioning different areas within mega-urban regions to optimize both environmental and economic outcomes21,22.

Although it is widely acknowledged that maintaining environmental integrity while advancing socio-economic progress can be challenging in mega-urban agglomerations23,24,25, significant knowledge gaps hinder effective management. Firstly, we lack quantitative frameworks to analyze the spatial trade-offs between ESV and socio-economic well-being, which limits insights for coordinated policy actions and makes it difficult to illustrate the optimal combinations of the two. Secondly, existing trade-off analyses primarily operate at broad international or regional scales26,27, potentially overlooking the intricate geographic variations within localized regions. Thirdly, insufficient attention has been paid to developing spatially differentiated management strategies that account for the unique characteristics of interconnected urban areas. These gaps collectively limit our ability to formulate targeted policies that can simultaneously enhance environmental sustainability and economic development across heterogeneous urban landscapes.

The production possibility frontier (PPF) concept—originally from economics—offers a promising framework for analyzing trade-offs between competing objectives. The PPF represents the maximum attainable combinations of two desirable outcomes given limited resources and technological constraints. Recent studies have begun applying this concept to ecosystem services28,29, but these applications typically treat regions as homogeneous units, disregarding the spatial variations that characterize complex urban systems.

In this study, we advance the application of the PPF framework to urban domains by introducing spatial differentiation and eco-socio-economic efficiency as a quantitative indicator of trade-offs between ESV and socio-economic well-being. Our approach integrates spatial clustering with PPF analysis to consider heterogeneous ecological and economic conditions across different zones within a mega-urban region. The framework was applied to the Guangdong-Hong Kong-Macao Greater Bay Area (GBA), a representative mega-urban agglomeration undergoing rapid urbanization and economic expansion while facing significant ecological pressures.

Our analysis revealed critical spatial mismatches in ecosystem service supply and demand, particularly in densely populated coastal regions where economic activities have significantly altered natural ecosystems. By fitting zone-specific PPF curves, we identified areas with substantial potential for improving both ecological sustainability and economic efficiency. This spatially explicit approach enabled us to develop tailored management strategies for different zones, including targeted payment for ecosystem services schemes and ecological restoration initiatives. The framework demonstrates how localized, data-driven approaches could enhance strategic urban planning and governance by identifying zone-specific opportunities and constraints, ultimately supporting more sustainable and resilient development across heterogeneous urban landscapes.

Results

Spatial heterogeneities and mismatches in ecosystem service supply and demand

The GBA, located in the Pearl River Delta, includes 62 county-level administrative units. Our analysis revealed pronounced spatial patterns in ecosystem service supply-demand dynamics across the region. Figure 1 illustrates the ecosystem service supply-demand ratio (ESSDR) at the county level for four key ecosystem services (carbon storage and sequestration, habitat quality, urban cooling, and urban flood risk mitigation), while Supplementary Fig. S1 presents the normalized supply and demand for these same services at a higher resolution of 100 m.

Fig. 1: Spatial distribution of ecosystem service supply-demand ratio (ESSDR) across the Guangdong-Hong Kong-Macao Greater Bay Area (GBA).
figure 1

a Carbon storage and sequestration. b Habitat quality. c Urban cooling. d Urban flood risk mitigation. Darker colors indicate higher ESSDR values. Ecosystem service supply and demand intensities were normalized to create dimensionless values for comparability. The calculated ESSDR reflects relative spatial differences rather than absolute supply and demand values.

Full size image

The GBA exhibits significant spatial heterogeneities in ecosystem service distribution, with supply generally concentrated in peripheral areas and demand centered in developed coastal zones. This pattern creates notable spatial mismatches, consistent with previous studies9,10. Ecosystem service supply is highest in peripheral regions and lowest in central areas, while ecosystem service demand is concentrated in the central and southern coastal regions. Consequently, the ESSDR shows marked spatial variation—highest in the peripheral areas and notably low in central and southern coastal regions (Fig. 1).

These spatial patterns reflect regional development disparities. The peripheral, mountainous areas, which remain less developed, maintain abundant natural ecosystems that provide high levels of ecosystem services. In contrast, the economically active central and southern coastal regions, characterized by dense populations, high ecosystem service demand intensity, and extensive built-up areas, exhibit lower ESSDR due to fragmented ecological spaces and limited natural ecosystems.

Zonal clustering based on ecosystem services and socio-economic attributes

To account for spatial heterogeneities in the trade-off relationships between ESV and socio-economic well-being, we classified the county-level administrative units within the GBA into distinct zones using k-means clustering. This classification was based on ESSDRs and socio-economic attributes (Methods). The optimal number of clusters (five) was determined using the elbow and silhouette methods (Supplementary Fig. S2). Each zone exhibits a unique ecological and socio-economic profile, as visualized in Fig. 2, with city-level interpretations provided in Supplementary Table S1.

Fig. 2: Classification of the GBA into five distinct eco-socio-economic zones.
figure 2

a Spatial distribution of zones derived from k-means clustering analysis. b Radar charts showing the characteristic profiles of each zone based on normalized indicators. Zones were determined by analyzing ecosystem service supply-demand relationships and socio-economic factors. Radar axes represent ecosystem service supply (ESS), ecosystem service demand (ESD), supply-demand ratio (ESSDR), average income (Income), and population density (Popu), with all indicators normalized for comparative analysis.

Full size image

The five zones are characterized as follows: (1) Abundantly sufficient zone (ASZ) has the highest ESSDR (0.96), with low population density (70 people/km2) and lowest income levels, indicating minimal environmental pressure. (2) Moderately sufficient zone (MSZ) maintains a balanced ESSDR (0.75), with higher population density and stronger economic activity compared to ASZ. (3) Slightly sufficient zone (SSZ) shows a moderate ESSDR (0.35), typically found in emerging economic areas with medium population density and income levels. (4) High demand zone (HDZ) is densely populated (359 people/km2) and economically vibrant, but has a low ESSDR (0.29) due to high ecosystem service demand exceeding supply. (5) Deficit zone (DZ) faces critical ecosystem service shortages (ESSDR = -0.17) with high population density (305 people/km2) and relatively strong income levels, requiring urgent management interventions.

This zonal classification reflects regional differences in ecosystem service supply-demand dynamics, population distribution, and socio-economic conditions, providing the foundation for our subsequent PPF analysis.

Zonal eco-socio-economic characteristics

Zone-specific PPF curves reveal distinct trade-off relationships between ESV and socio-economic well-being across the five identified zones (Fig. 3). For this analysis, socio-economic well-being was quantified using total labor income, as income levels serve as a widely recognized indicator of access to essential services and quality of life (Methods). Each PPF curve illustrates how increasing socio-economic well-being affects ESV, with varying threshold points and rates of decline that reflect the unique ecological and economic conditions of each zone.

Fig. 3: Trade-off relationships between ecosystem service value (ESV) and socio-economic well-being across different zones.
figure 3

Production possibility frontier (PPF) curves for: (a) Abundantly sufficient zone. b Moderately sufficient zone. c Slightly sufficient zone. d High demand zone. e Deficit zone. Red lines represent fitted PPF curves (sigmoid functions with R² > 0.97), with Pareto-optimal points (red) and non-optimal points (gray) shown for each zone. All zones display characteristic trade-off patterns: initial stability in ESV as socio-economic well-being increases, followed by accelerated ESV decline beyond zone-specific thresholds. Curve equations are shown in each panel.

Full size image

The maximum ESV values for ASZ, MSZ, SSZ, and HDZ are high (0.96, 0.99, 0.97, and 0.94, respectively), indicating substantial ecological capacity. In contrast, DZ shows a significantly lower maximum ESV value (0.41), reflecting its limited ecological potential. In terms of socio-economic well-being, Fig. 3 reveals that as ESV approaches zero, the maximum achievable socio-economic well-being differs across zones. SSZ has the lowest potential, whereas ASZ and DZ can reach moderate levels. MSZ and HDZ demonstrate the highest socio-economic well-being levels in the GBA.

The marginal rate of transformation, represented by the slope of the PPF curves, indicates the rate at which ESV must be sacrificed to increase socio-economic well-being30. The marginal rate of transformation varies significantly along each PPF curve and across different zones. ESV begins to decline noticeably beyond specific socio-economic well-being thresholds: approximately 0.65 for ASZ and MSZ, 0.40 for SSZ, and 0.50 for HDZ and DZ. After these inflection points, ASZ shows the steepest decline in ESV, while DZ experiences the gentlest reduction.

Each zone displays distinctive trade-off characteristics. ASZ possesses abundant ecological resources that support high ESV levels, but socio-economic well-being remains generally low. When socio-economic well-being exceeds 0.65, marginal rate of transformation increases rapidly, leading to a sharp decline in ESV. This pattern reflects ASZ’s mountainous terrain, ecological sensitivity, and limited carrying capacity for intensive economic activities.

MSZ, located in less ecologically sensitive central and southern coastal plains, maintains stable ESV across a broader range of socio-economic well-being. This indicates higher ecological carrying capacity, though ESV eventually declines as socio-economic well-being surpasses this threshold.

SSZ shows an earlier ESV decline (at socio-economic well-being of 0.40), suggesting more immediate trade-offs between ecological and economic objectives in these emerging economic areas.

HDZ and DZ, despite their different ecological baselines, both show ESV decline beginning at moderate socio-economic well-being levels (0.50). Notably, DZ’s generally low ESV across all levels of socio-economic well-being indicates substantial ecological degradation in these highly urbanized areas.

Geographic distribution of efficiency metrics and improvement potential

Pearson correlation analysis (Supplementary Fig. S3) reveals important relationships between our key metrics. ESV shows a moderate negative correlation with eco-socio-economic efficiency and ESV improvement potential. Socio-economic well-being and eco-socio-economic efficiency exhibit a strong positive correlation, while the correlation between socio-economic well-being and ESV improvement potential is not statistically significant.

The spatial distribution of these metrics reveals distinct patterns across the GBA (Fig. 4). ESV is higher in peripheral areas and lower in central and southern coastal regions. Socio-economic well-being shows clear spatial clustering in economically developed central and coastal regions. High eco-socio-economic efficiency is concentrated in the economically advanced regions of southwestern Guangzhou, southeastern Foshan, Dongguan, Shenzhen, and Hong Kong. Among them, the core areas of Guangzhou, Shenzhen, and Hong Kong display significantly higher eco-socio-economic efficiency than other regions. Despite its lower economic development levels, the northwestern part of Zhaoqing demonstrates relatively high eco-socio-economic efficiency, attributed to its high ESV levels. Our analysis identified Zhaoqing, Guangzhou, Foshan, Dongguan, Zhongshan, Zhuhai, and Macao as having the highest ESV improvement potential under current economic conditions.

Fig. 4: Spatial analysis of ecosystem service value and economic metrics.
figure 4

County-level assessment of: (a) Ecosystem service value. b Socio-economic well-being. c Eco-socio-economic efficiency (distance to production possibility frontier). d Ecosystem service value improvement potential. Darker shades indicate higher values for each metric. The maps reveal distinct east-west gradients in efficiency and improvement potential, reflecting regional development patterns and zone-specific ecological-economic relationships.

Full size image

The region exhibits a development gradient with eco-socio-economic efficiency decreasing from east to west, while ESV improvement potential shows the opposite pattern among the cities surrounding the Pearl River Delta (Fig. 4c, d). Two primary factors contribute to this spatial structure. Firstly, following the Environmental Kuznets Curve theory, attention to ecological sustainability typically increases after economic development reaches a certain threshold and socio-economic output efficiency improves31,32. This phenomenon explains why eco-socio-economic efficiency is higher in the more economically developed eastern cities of the GBA. For instance, Hong Kong and southern Guangzhou, both classified as MSZ types (Fig. 2), have similar ESV levels despite Hong Kong’s higher economic development. This can be attributed to Hong Kong’s highly efficient land use and strict land management system33, which optimize built-up area utilization and mitigate further ESV loss despite intense urbanization. In contrast, southern Guangzhou, though less economically developed, maintains comparable ESV levels due to lower development intensity.

Secondly, trade-off relationships between ESV and socio-economic well-being vary across zones, leading to different optimal combinations of these two factors. As shown in Fig. 2 and Fig. 4a, b, d, Zhuhai (categorized as MSZ) has an average normalized socio-economic well-being of 0.31 and an ESV of 0.23, whereas Shenzhen’s HDZ regions exhibit higher socio-economic well-being (0.49) but lower ESV (0.17). Referring to Fig. 3b, d, at their current socio-economic well-being levels, Zhuhai could theoretically achieve an ESV close to 0.99, while Shenzhen’s HDZ regions would only reach 0.84 under optimal conditions.

Discussion

This study systematically explores the intricate trade-offs between ESV and socio-economic well-being within the GBA. By integrating spatial clustering with PPF analysis, we revealed significant spatial heterogeneity in eco-socio-economic dynamics across the region. Our approach moves beyond traditional trade-off analyses by accounting for local variations in ecological sensitivity and economic development, enabling more targeted and effective management strategies.

The spatial patterns we identified highlight a critical mismatch between ecosystem service supply and demand, particularly in economically vibrant coastal zones. This mismatch reflects the classic sustainability challenge of mega-urban regions: areas with the highest economic output often experience the greatest ecological deficits, creating an uneven distribution of environmental benefits and costs across the landscape.

The distinct PPF curves for each zone type (Fig. 3) reveal important differences in how ecological and socio-economic priorities interact across the region. In ecologically sensitive mountainous areas (ASZ), even modest economic development can trigger rapid ecosystem degradation once beyond specific thresholds (socio-economic well-being > 0.65). In contrast, less ecologically sensitive central plains (MSZ) can sustain higher levels of economic development with more gradual ecosystem impacts. These zone-specific trade-off relationships provide crucial insights for regional planning that conventional, spatially uniform approaches would miss.

Beyond describing current conditions, our framework offers pathways for improvement through two key metrics: eco-socio-economic efficiency and ESV improvement potential (Fig. 4). The distribution of non-Pareto-optimal points across zones reveals differentiated opportunities for enhancement. ASZ and MSZ show the greatest need for improved efficiency, particularly in areas within Jiangmen, Zhaoqing, and Huizhou that exhibit low socio-economic well-being and eco-socio-economic efficiency—a finding consistent with previous studies identifying these three cities as having low development efficiency within the GBA34.

For optimal regional management, our results suggest distinct development pathways tailored to each zone’s unique characteristics. In mountainous, ecologically rich areas (ASZ), development should prioritize economic activities with minimal environmental impact, such as eco-tourism and sustainable agriculture. The high marginal rate of transformation in these zones indicates that conventional development would come at substantial ecological cost, necessitating strict environmental impact assessments and ecological compensation mechanisms.

Areas with balanced ecosystem service supply and demand (MSZ) offer significant potential for simultaneous improvements in both ESV and socio-economic well-being. These zones should focus on environmentally compatible development that preserves existing ecosystem functions while addressing resource efficiency. In contrast, highly developed areas with ecosystem deficits (HDZ and DZ) should optimize land use patterns and economic structures, supporting the observation that larger cities with higher population and industry concentration tend to achieve higher land use efficiency35. For these zones, strategic industrial policies36 that promote high-value, knowledge-intensive sectors while reducing resource-intensive manufacturing could enhance economic output with lower environmental impacts.

Meanwhile, emerging economic areas (SSZ) face a critical transition phase, with moderate ESV improvement potential but increasing development pressures. These areas would benefit from preemptive green infrastructure planning and industrial guidance to avoid the environmental degradation pathways experienced by more developed zones, particularly as they border areas with higher ESV that could be affected by development spillovers37.

The spatial gradient of eco-socio-economic efficiency (higher in the east) and ESV improvement potential (higher in the west) across the GBA necessitates coordinated regional governance mechanisms. This pattern aligns with the Environmental Kuznets Curve theory, but policy interventions can help western areas improve their environmental performance without following the full ‘grow first, clean up later’ trajectory of eastern cities. We recommend incorporating ecological indicators such as gross ecosystem product (GEP) alongside traditional economic metrics in government performance evaluations13. Shenzhen’s pioneering adoption of GEP as the first city in China demonstrates the practical feasibility of this approach38. Such integrative frameworks2,39 can help overcome the common challenge where short-term economic priorities overshadow long-term environmental considerations40.

Payment for ecosystem services (PES) schemes41,42 offer a promising mechanism for addressing the spatial disparities revealed in our analysis. As Fig. 5a illustrates, ASZ regions possess high ESV but low socio-economic well-being, creating an opportunity for ecological compensation from regions with opposite characteristics43,44,45. These compensation arrangements would create economic incentives for conservation while reducing regional inequalities—a particularly important outcome given the stark contrasts in development levels across the GBA46,47,48.

Fig. 5: Relationship between ecological and socio-economic metrics across different zones of the GBA.
figure 5

a Distribution of administrative units based on ecosystem service value (ESV) and socio-economic well-being, revealing distinct zone-specific clustering patterns. b Analysis of eco-socio-economic efficiency versus ESV improvement potential, highlighting opportunities for targeted interventions. Each point represents an administrative unit, colored by zone type with city abbreviations as labels (GZ-Guangzhou, SZ-Shenzhen, ZH-Zhuhai, FS-Foshan, HZ-Huizhou, DG-Dongguan, ZS-Zhongshan, JM-Jiangmen, ZQ-Zhaoqing, HK-Hong Kong, MO-Macao). Dashed lines indicate regional average values.

Full size image

Our analysis suggests a two-tiered implementation approach for PES policies. Firstly, cross-regional compensation could flow from central and southern coastal zones with very low ESSDR (Fig. 2) to peripheral areas with high ESV. Second, within-city pilot programs could target areas with high ESV improvement potential, as shown in Fig. 5b. For instance, while Guangzhou’s central districts face ecosystem service deficits, they could invest in ecological restoration projects in neighboring districts with high improvement potential, creating mutual benefits for both areas. This approach acknowledges the complex scale-dependent nature of PES formulation49,50, which requires careful consideration of stakeholders, payment mechanisms, and socio-ecological impacts across various temporal and spatial scales.

Despite the potential of differentiated management strategies and PES policies, proximity-dependent ecosystem services present persistent challenges in urban centers, including thermal inequity, unequal flood risk exposure, and imbalanced green space accessibility51,52,53,54,55. Our analysis reveals significant imbalances in ecosystem service supply and demand for urban cooling and flood risk mitigation in densely developed areas (Fig. 1c, d), indicating areas of environmental concern across the region’s urban core.

Land scarcity represents a particularly critical barrier to addressing these urban ecosystem service imbalances. Our analysis shows that natural spaces like forests and water bodies provide substantially higher ESV than built-up areas56, yet urbanization continually converts these valuable lands57. Shenzhen exemplifies this challenge, where despite intense development pressure, carefully optimized ecological protection covering just 22% of land area significantly enhances overall ESV (Supplementary Fig. S4). This finding underscores the importance of strategic green infrastructure solutions—including bio-retention systems, green roofs, and urban forests53,56,57 —that can provide critical ecosystem services within space-constrained urban environments.

Figure 5 provides essential insights for implementing these recommendations, showing how administrative units cluster according to zone type in the ESV–socio-economic well-being space (Fig. 5a) and in the eco-socio-economic efficiency–ESV improvement potential space (Fig. 5b). The clear separation between zone types in these dimensions confirms that our classification framework captures meaningful differences in eco-socio-economic conditions. These visualizations highlight opportunities for targeted interventions—areas with similar socio-economic levels but varying ESV values suggest potential for cross-regional learning and policy transfer, while the efficiency-potential relationship reveals where specific management approaches might yield the greatest returns. This evidence-based spatial differentiation provides a foundation for tailored strategies that respond to local conditions rather than applying uniform policies across heterogeneous landscapes.

Looking forward, policymakers and planners should integrate these spatially explicit insights into urban planning and governance. We recommend implementing tailored PES schemes and ecological restoration initiatives aligned with the distinct ecological and economic profiles of different zones. Further research should focus on capturing dynamic changes through real-time monitoring, exploring additional ecosystem services beyond those examined here, and developing more detailed sector-specific policy instruments based on our zonal classifications.

By acknowledging spatial heterogeneity in eco-socio-economic systems and tailoring management strategies accordingly6, mega-urban regions can more effectively balance development aspirations with ecological imperatives, moving closer to truly sustainable urban futures.

Methods

Data sources

This study utilized diverse datasets to quantify the spatial relationships between ESV and socio-economic well-being in the GBA. For ESV estimation, land use data (2021) were obtained from Land Cover Explorer (https://livingatlas.arcgis.com/landcoverexplorer/). Road network data (2021) were acquired from OpenStreetMap (https://www.openstreetmap.org/), and line density of different roads was calculated using ArcGIS Pro 2.5.2. As land surface temperature data (2021) were unavailable, daytime and nighttime land surface temperature data (2020) were used, sourced from the Resource and Environmental Science Data Platform (https://www.resdc.cn/DOI/DOI.aspx?DOIID=98). Soil type data were also obtained from the Resource and Environmental Science Data Platform (https://www.resdc.cn/data.aspx?DATAID=145). Precipitation data were retrieved from the National Earth System Science Data Center (https://www.geodata.cn).

To quantify ecosystem service demand and socio-economic well-being at a high resolution, anonymized location-based service data were utilized to identify residential and working populations, along with their socio-economic attributes. The residential population was defined as individuals whose location data predominantly indicated residential areas during weekday nights or weekends, characterized by single or non-public Wi-Fi connections, and who had resided in a city for over three months. The working population included individuals mainly located in office buildings or other work-related sites on weekdays, as detected by public Wi-Fi connections. We considered socio-economic attributes including income and expenditure levels, with all data fully anonymized and spatially aggregated to ensure privacy and data security.

To address the inherent limitations of location-based service data, which primarily represents mobile device users53, we augmented these datasets using population statistics from the 2021 Population Census and the 2018 National Economic Census, following the established methods58,59. This integration process involved performing Inverse Distance Weighting interpolation to fill missing data and associate population profiles with spatial distribution, thereby creating a comprehensive representation of population patterns and socio-economic characteristics across the study area.

Regional context and ecosystem services selection

The GBA, located in the Pearl River Delta region of southern China, comprises 62 county-level administrative units, including Guangzhou, Shenzhen, Zhuhai, Foshan, Huizhou, Dongguan, Zhongshan, Jiangmen, Zhaoqing, and the Hong Kong and Macao Special Administrative Regions (Supplementary Fig. S5). With a population of approximately 87 million as of 2025, the region faces both significant environmental challenges and substantial socio-economic disparities, as reflected in the variation of per capita GDP across its constituent cities (Supplementary Table S2).

We selected four ecosystem services—carbon storage and sequestration, habitat quality, urban cooling, and urban flood risk mitigation—that are emphasized in regional ecological restoration plans and commonly used to assess overall regional ESV9,60,61. These services enhance urban resilience to extreme climate events and natural disasters, sustain ecosystem health, and help meet residents’ demand for high-quality ecological settings. For analytical purposes, we considered ESV as a measure of ecosystem service supply, focusing on the total potential of ecosystem services provided to humans, regardless of actual consumption62.

We calculated ecosystem service supply and demand based on 100 m resolution land use data, aligning with location-based service data. While natural geographical units like watersheds may better reflect ecological characteristics, our study adopted county-level administrative units for two reasons. Firstly, economic regions based on administrative boundaries often face conflicts between ecological, environmental, and economic interests, making them responsible for addressing externalities63. Policies based on administrative regions are also more practical and feasible. Secondly, administrative regions are governed by local authorities, which act as representatives of regional interests. Given China’s current stage of development, government intervention plays a crucial role in implementing environmental protection policies41. Consequently, grid-level ecosystem service supply and demand analysis offers high-resolution insights, which can be scaled up to local administrative units to support effective policy-making and address the complexities of eco-socio-economic systems64,65.

Ecosystem service supply and demand assessment

To enhance the accuracy of the results at the study area’s periphery, ESV was assessed within both the GBA and a 10 km buffer zone, allowing for the consideration of external environmental influences on ecosystem service supply beyond the GBA’s administrative boundary. Land use, road network, and environmental data were incorporated into the InVEST model, with biophysical values adapted from existing studies through look-up tables (Supplementary Note).

We used high-resolution spatio-temporal location-based service data to quantify residents’ demand for ecosystem services (Supplementary Note). For carbon storage and sequestration demand, we employed household carbon emissions as a proxy66, calculated using the logarithm of household carbon emissions weighted by residential population density and average expenditure levels67. Habitat quality demand was represented by population density68,69, reflecting the intensity of human activities that affect habitat integrity. Urban cooling demand was quantified using normalized land surface temperature as a weight for total population density70,71. Finally, urban flood risk mitigation demand was assessed through potential socio-economic damage during flood events, combining precipitation as a weight for flood occurrence probability with population density and average income level to represent potential economic losses72.

Ecosystem service supply-demand ratio calculation

We calculated county-level ESSDRs to identify the spatial characteristics of ecosystem service supply-demand relationship. The ecosystem service supply and demand intensities were normalized to make them dimensionless and comparable. It is important to note that the calculated ESSDRs reflect the relative differences across different regions of the GBA, rather than absolute values of ecosystem service supply and demand relationships. The formula for calculating ESSDRs is as follows:

$${ES}SD{R}_{i}=\frac{{{ES}}_{{si}}-{{ES}}_{{di}}}{{{ES}}_{{si}}+{{ES}}_{{di}}}$$
(1)

Where \({ES}SD{R}_{i}\) represents the ESSDR in county-level administrative unit \(i\); \({{ES}}_{{si}}\) denotes the total supply of ecosystem services in county-level administrative unit \(i\); and \({{ES}}_{{di}}\) refers to the total demand for ecosystem services in county-level administrative unit \(i\).

Spatial clustering for differentiated trade-off analysis

The trade-off relationships between ESV and socio-economic well-being vary across regions, making zonal clustering a crucial step before fitting the PPF curves in mega-urban agglomerations73,74,75. We compiled a comprehensive dataset incorporating the ESSDRs of four key ecosystem services, along with the average ESSDR, mean and standard deviations of ecosystem service supply and demand intensity, average income level, and average population density. The dataset was normalized and subjected to k-means clustering, enabling a more precise fitting of PPF curves for different zones. The elbow method and silhouette coefficient were employed to evaluate the clustering effectiveness and determine the optimal number of clusters.

Production possibility frontier modeling and performance metrics

The PPF framework illustrates the maximum attainable combination of two goods or services that an economy can produce, given its resources and technological capabilities76. Points along the PPF represent efficient production scenarios, while points below it indicate inefficiencies and potential for improvement. The distance from a point to the PPF serves as a key indicator of resource utilization efficiency and production potential77. A detailed conceptual framework illustrating this methodology is provided in Supplementary Fig. S6.

By plotting the PPF curves, we visualized the trade-offs between ESV and socio-economic well-being. Following established methodologies78 and considering both data volume and result accuracy, we fitted the PPF curves at a 1 km resolution. Total labor income, derived from location-based service data, was used as a proxy for socio-economic well-being. Both ESV and socio-economic well-being values were normalized before fitting the PPF curves to ensure comparability.

We selected Pareto-optimal points for each zone based on Pareto-efficient criterion, applying an algorithm that identifies points where no other point has higher values for both ESV and socio-economic well-being (Eq. (2))26. For each sample point (\({E}_{i},{S}_{i}\)) in the dataset, a point is considered Pareto-optimal if no other sample point (\({E}_{j},{S}_{j}\)) satisfies the following conditions:

$$\forall i,\nexists {\rm{j}}\,{\mathrm{suc}{\rm{h}}\,{\rm{t}}{\rm{h}}\mathrm{at}}\left\{\begin{array}{c}\forall k\in \left\{{E},{S\;}\right\},{k}_{j}\ge {k}_{i}\\ \exists k\in \left\{{E},{S\; }\right\},{k}_{j} > {k}_{i}\end{array}\right.$$
(2)

where \(E\) denotes ESV, and \(S\) denotes socio-economic well-being. Under these conditions, the sample point (\({E}_{i},{S}_{i}\)) can be considered the Pareto-optimal point for the trade-off between ESV and socio-economic well-being within its respective zone, given the available environmental resources and technological constraints.

To fit the PPF curves, we tested different functional forms, including exponential, quadratic, and logarithmic functions. Based on R-square values (Supplementary Table S3), the sigmoid function, a specialized form of the exponential function, was selected as the best fit. The sigmoid function is usually expressed as:

$$\sigma \left(x\right)=\frac{L}{1+{e}^{-k(x-{x}_{0})}}$$
(3)

where \(L\) is the maximum value of the function, \({x}_{0}\) represents the midpoint, and \(k\) defines the curve’s slope.

In each clustered zone, non-optimal points share similar underlying conditions with optimal points, making the nearest optimal point on the PPF the most achievable improvement state79. The distance to the PPF serves as a measure of efficiency in producing ecosystem services and economic outputs. To compare the zonal differences in eco-socio-economic efficiency across the GBA and assess the internal differences within each zone, we defined eco-socio-economic efficiency using the following calculation formula:

$${D}_{i}=\sqrt{{({x}_{i}-{x}_{j})}^{2}+{({y}_{i}-{y}_{j})}^{2}}$$
(4)
$${E{fficiency}}_{i}=1-\frac{{D}_{i}}{{D}_{\max }}$$
(5)

where (\({x}_{i}\), \({y}_{i}\)) represents a point not on the PPF, and (\({x}_{j}\), \({y}_{j}\)) denotes the nearest point on the PPF to (\({x}_{i}\), \({y}_{i}\)). \({D}_{i}\) is the Euclidean distance between the point (\({x}_{i}\), \({y}_{i}\)) and point (\({x}_{j}\), \({y}_{j}\)). \({D}_{\max }\) represents the maximum value among all distances from point (\({x}_{i}\), \({y}_{i}\)) to the nearest point (\({x}_{j}\), \({y}_{j}\)) on the PPF. \({{Efficiency}}_{i}\) represents the eco-socio-economic efficiency of the point (\({x}_{i}\), \({y}_{i}\)).

Using the fitted PPF curves, we determined the optimal ESV value corresponding to the current socio-economic well-being level on the PPF. The difference between this optimal ESV and the actual ESV represents the potential for ESV improvement. The specific calculation formula is as follows:

$${y}_{i}^{{curve}}=\sigma \left({x}_{i}\right)=\frac{L}{1+{e}^{-k({x}_{i}-{x}_{0})}}$$
(6)
$${d}_{i}=\left|{y}_{i}^{{curve}}-{y}_{i}\right|$$
(7)
$${{Potential}}_{i}=1-\frac{{d}_{i}}{{d}_{\max }}$$
(8)

where \(L\), \({x}_{0}\), and \(k\) are the fitted parameters of the sigmoid function. For a given point (\({x}_{i}\), \({y}_{i}\)), \({x}_{i}\) represents the ESV, \({y}_{i}\) represents the socio-economic well-being, and \({y}_{i}^{{curve}}\) is the corresponding value on the fitted PPF. \({d}_{i}\) is the absolute distance between the actual ESV \({y}_{i}\) and the fitted PPF value \({y}_{i}^{{curve}}\) for the point (\({x}_{i}\), \({y}_{i}\)). \({d}_{\max }\) is the maximum absolute distance among all points. \({{Potential}}_{i}\) represents the ESV improvement potential of the point (\({x}_{i}\), \({y}_{i}\)), which quantifies the extent to which ESV can be enhanced relative to its optimal value on the PPF.