Multi-objective optimization of urban blue-green space configuration using NSGA-II for enhanced ecosystem services Authors: Yangyang Yuan, Mingzhu Yang, Siqi Tang, Shangcen Luo, Jingwen Mao, Sidan Yao, Qianyu Hong Published: 2026 ABSTRACT The exacerbation of global climate change has led to the continuous deterioration of the urban ecological environment. Scientifically planning blue-green spaces can mitigate ecological crises and promote sustainable urban development. There exists a close relationship between the ecological benefits of blue-green spaces and their spatial characteristics, and synergistically enhancing the ecological service efficiency of blue-green spaces through these two aspects is an urgent research topic. This study integrates four ecological benefit objectives—carbon sink, cooling, stormwater resilience, and biodiversity—and applies a multi-objective optimization framework based on grid-based spatial decision units to optimize blue-green space configurations. By establishing a multi-ecological benefit evaluation framework, this study combines ecological benefit assessment models with the NSGA-II multi-objective optimization algorithm and the entropy weight method for weight allocation. Ultimately, an optimal scenario and four preference-oriented scenarios for blue-green space layouts are derived. A case study in Nanjing, China, demonstrates the applicability of the integrated optimization framework in generating urban blue-green spatial layouts that balance multiple ecological benefit objectives. The results indicate that carbon sink and cooling show a positive synergy, while biodiversity and stormwater resilience exhibit trade-off tendencies, and the comprehensive scenario achieves a relatively balanced performance across objectives. The spatial layout characteristics of blue-green spaces vary distinctively across different ecological benefit objectives, with the multi-objective comprehensive optimal under the given proxies, constraints and algorithm settings demonstrating relative balance. The proposed multi-objective optimization methodology for blue-green spatial configuration provides scientifically quantifiable decision-making support for urban blue-green space planning. Introduction In the context of global warming, cities face increasingly severe ecological challenges. Between 2003 and 2020, extreme heat exposure in northern Chinese urban agglomerations surged by 67%, with urban cores contributing over 50% of the pollution load1. Simultaneously, the risk of extreme climate events, such as floods, has risen. Between 2000 and 2018, the global population affected by floods increased by 20–24%, and climate change will likely exacerbate this trend2. Urbanization-driven land cover changes have reduced vegetation, sharply decreased biodiversity, and degraded ecosystem services3. In this context, blue-green spaces, as vital ecological infrastructure, are increasingly recognized for their multiple benefits. Scientific planning and dynamic management of these spaces is a key strategy to address urban ecological challenges and support sustainable, high-quality living environments. Urban blue-green spaces comprise all green and blue areas retained or created during urban development, including natural, semi-natural, and artificial lands and waters, and are essential components of urban ecosystems4. Previous studies have mainly focused on green or blue spaces separately, often emphasizing green areas. However, blue-green spaces form a composite ecosystem with strong interconnections and complementarities in ecological services and spatial distribution5. Research indicates that their synergy can amplify ecological benefits6. Urbanization has disrupted natural ecological processes, with blue-green spaces often encroached upon, causing “insufficient space,” “separation,” and “imbalance,” which limit ecological benefits. The spatial configuration of blue-green spaces is closely related to their ecological functions, and different spatial patterns can influence the realization of ecological benefits. A well-designed spatial configuration can, to a certain extent, reflect the outcomes of ecological processes and play an important role in ecosystem service provision. The intrinsic ecological mechanisms of blue-green spaces shape their spatial distribution patterns, while spatial configuration can partially reflect the underlying ecological processes. These two aspects are closely linked rather than independent. Existing studies have mainly examined blue-green spaces from conceptual perspectives or single dimensions7. However, systematic approaches for achieving the coordinated optimization of multiple ecological objectives and spatial configuration within a unified framework remain limited. The integration and coordination of spatial configuration and ecological benefits are essential for enhancing the ecosystem service performance of blue-green spaces and remain an important research priority. Current research on urban blue-green spaces’ ecological benefits mainly addresses four aspects: carbon sink, cooling, stormwater resilience, and biodiversity conservation. Urban green spaces are primary carbon sinks; larger patch areas, higher vegetation coverage, and greater three-dimensional green volume enhance carbon sequestration8,9. Blue spaces also act as carbon reservoirs, with pathways including soil carbon burial, aquatic biota activity, and carbon dissolution in water. Studies demonstrate synergistic cooling effects of blue-green spaces10; coverage of 30% significantly reduces the urban heat island effect, while over 50% markedly enhances cooling11. Blue-green spaces inherently absorb and mitigate stormwater, and higher-density areas correspond to lower flood risk12. They also provide habitats, supporting biodiversity13, with environmental quality of large natural spaces critically influencing species richness. Existing studies have largely focused on single ecological objectives, with limited attention paid to the interactions among multiple ecological objectives and their spatial configuration mechanisms. Research on integrated ecological enhancement mainly focuses on collaborative planning frameworks or multi-benefit evaluation14, while studies on coordinating multiple ecological objectives remain limited. In landscape architecture, algorithm-assisted urban planning and spatial optimization have become a prominent research trend. Yoon et al. applied the Non-dominated Sorting Genetic Algorithm (NSGA-II) to determine urban green space location and type15. Nguyen et al. used artificial neural networks to develop a landscape management model, achieving over 80% optimization efficiency16. Feng et al. employed cellular automata and particle swarm optimization to simulate urban growth and land-use change in Shanghai17. These studies show that optimization algorithms can account for multiple uncontrollable factors, balance diverse objectives, and reduce solution bias. However, existing studies mainly focus on improving single ecological objectives or general land-use optimization, with relatively limited attention given to the coordinated optimization of multiple ecological objectives from a holistic blue-green space perspective18. Over recent decades, heuristic algorithms have been widely applied to multi-objective optimization across disciplines. Optimizing urban blue-green space configuration requires simultaneously enhancing multiple ecological objectives and balancing trade-offs. Multi-Objective Evolutionary Algorithms (MOEAs) have gained attention for effectively addressing complex problems19,20. By simulating natural selection and population evolution, MOEAs manage multiple objectives across generations and identify sets of optimal solutions. Among them, NSGA-II is notable for efficient sorting and handling conflicts between objectives21,22. Through iterative comparison and selection, it progressively approximates the Pareto optimal front, yielding balanced solutions. Applying NSGA-II to blue-green space configuration effectively reconciles ecological objectives while improving target benefits, providing scientific and practical support for urban planning. This study builds upon an existing multi-objective optimization framework by integrating multiple ecological objectives, including carbon sequestration, cooling effects, stormwater regulation, and biodiversity, to achieve coordinated optimization of blue-green space configurations, and applies the approach to a case study in the central urban area of Nanjing, China. The goal is to provide a scientific basis for planning, management, and decision-making. The objectives are: (1) clarify the relationship between configuration and four ecological objectives, establishing a multi-objective evaluation framework; (2) define decision variables, objective functions, and constraints, and construct a genetic algorithm–based optimization model; (3) apply the entropy weight method for weight analysis and simulate blue-green space configurations under different objective preferences and a comprehensive optimal scenario. Study area and data Study area Nanjing, located in southwestern Jiangsu Province within the Yangtze River Delta, has a humid subtropical climate with annual precipitation of 1090.4 mm. Its natural landscape features low mountains, gentle hills, and water bodies covering over 11% of the area. The study area encompasses Nanjing’s central districts—Xuanwu, Gulou, Qinhuai, Jianye, and Yuhuatai—along with parts of Qixia, Jiangning, Pukou, and Liuhe (Fig. 1). The study area covers approximately 810.27 km2. As a representative high-density city in China, Nanjing has experienced rapid urbanization over the past two decades, causing fluctuations in blue-green space—initially increasing, then declining, and rising again23. Large natural patches have been fragmented, undermining continuity and integrity. Many were converted for industrial or residential use, leading to compression and degradation of blue-green spaces and adversely affecting the ecological environment. In this context, blue-green spaces in Nanjing face increasing demands for optimization across multiple ecological benefits, including carbon sequestration, cooling effects, stormwater regulation, and biodiversity. Trade-offs and coordination among these ecological objectives have become increasingly prominent. Therefore, selecting Nanjing as the study area facilitates the exploration of multi-objective ecological optimization in a representative context characterized by relatively abundant blue-green space resources and high urban development intensity. The findings are expected to provide useful insights for similar high-density cities. Fig. 1Overview map of study area. (Created by ArcMap, version 10.8, 10.8, https://www.esri.com). Overview map of study area. (Created by ArcMap, version 10.8, 10.8, https://www.esri.com). Data sources and processing This study used remote sensing imagery, land-use classification, blue-green space distribution, and ecological parameter data. Landsat-8 OLI imagery of Nanjing (2020, 30 m resolution) was obtained from the Geospatial Data Cloud Platform (http://www.gscloud.cn/) and processed in ENVI 5.3 for radiometric calibration, atmospheric correction. The imagery was resampled to 15 m to facilitate spatial analysis; however, it should be noted that resampling only changes pixel size and does not increase the actual spatial resolution. Therefore, the analysis remains constrained by the original 30 m resolution of the data. Land-use data followed the Chinese LUCC classification system, dividing the area into cropland, forest, grassland, wetlands, water, built-up land, and unused land. Ecological parameters (e.g., carbon sink and cooling coefficients) were derived from literature to quantify multi-objective functions. Study area boundaries were sourced from OpenStreetMap and clipped using GIS tools to define Nanjing’s central urban district. All spatial data were converted into JSON format for input into the NSGA-II optimization model. Methodology To enhance ecological benefits in high-density urban areas, this study constructs a multi-objective blue-green space configuration optimization model targeting carbon sink (CS), cooling (CI), stormwater resilience (SW), and biodiversity (DIV) (Fig. 2). The first step is blue-green space extraction: land use is classified from high-resolution remote sensing imagery using machine learning and reclassified to extract blue-green spaces. The second part is model construction: an evaluation system for CS, CI, SW, and DIV is developed, indicators quantified, and coupled with an NSGA-II–based multi-objective optimization algorithm for iterative refinement. The third step provides decision support: preference analysis accommodates different ecological needs, entropy-weighting evaluates optimized schemes, and the comprehensive criteria identify the optimal under the given proxies, constraints and algorithm settings for decision-making. Fig. 2Multi-ecological objective optimization process for urban blue-green space configuration. (Created by ArcMap, version 10.8, https://www.esri.com). Multi-ecological objective optimization process for urban blue-green space configuration. (Created by ArcMap, version 10.8, https://www.esri.com). Extraction of blue-green spaces The Chinese Land Use/Cover Remote Sensing Monitoring Classification System (LUCC) was adopted as the classification standard to interpret remote sensing imagery of Nanjing. On ENVI 5.3, training samples for six land-use types were manually selected, and a Support Vector Machine (SVM) performed the classification. To achieve precise regulation of blue-green spaces, the blue space was subdivided into two categories: water and wetlands. The classification results were validated using randomly selected samples in combination with high-resolution imagery. A confusion matrix was constructed to assess classification accuracy. The results show that the overall accuracy reached approximately 85%, with a Kappa coefficient exceeding 0.81, indicating a high level of reliability and meeting the requirements of this study. Based on the classification results, ArcMap 10.8 software was used to reclassify forest and grassland as green spaces, and water and wetlands as blue spaces, generating a blue-green space distribution map of the central urban area of Nanjing (Fig. 1). The area and spatial coordinates of each category were calculated using the raster-to-polygon tool, and an encoding mapping table (Table 1) was constructed, providing structured data support for the subsequent multi-objective optimization model. Establishment and quantification of the multi-objective evaluation system Quantification of the CS index Natural ecological carriers in urban blue-green spaces—water, forests, soil, and wetlands—are functionally and spatially interrelated and complementary24. They synergistically improve plant growth, microclimate, and thermal conditions, enhancing carbon sink benefits25. Forests26, grasslands27, and wetlands28 are the main carbon sink contributors, with coefficients listed in Table 1. City-wide carbon sink benefits are calculated as the product of land-use areas and their respective coefficients29. The objective function for maximizing carbon sinks is the cumulative sum across all land-use types: assuming CK as the carbon sink coefficient, total carbon sink equals the sum of contributions from all land-use types, expressed by the following formula:1\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\:\begin{array}{c}{f}_{CS}=\sum\limits_{k=1}^{K}{C}_{k} \cdot {A}_{k}\end{array}$$\end{document} where KK is the total number of land use types, Ck is the carbon sink coefficient for the k-th land use type, and Ak is the total area of the k-th land use type. Quantification of the CI index Urban blue-green spaces act as the primary ecological cooling source. Cooling benefits arise from reducing solar radiation, vegetation transpiration or insulation, and increased surface roughness to enhance airflow30,31. These mechanisms interact synergistically within the blue-green ecosystem. In the construction of the objective function for maximizing cooling benefits, the cooling amount of different land use types is used as the measurement indicator. A calculation formula is established by summing the products of the cooling intensity coefficients (Table 1)32 for different land use types and their respective land area, with constraints such as the area of blue-green space and the water surface ratio. To ensure consistency in the optimization direction while preserving the relative cooling capacities among different land-use types, the absolute values of the cooling coefficients were used. Accordingly, cooling intensity values are reported as positive values throughout the manuscript, with larger values indicating stronger cooling effects. The formula is as follows:2\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\:\begin{array}{c}{f}_{CI}=\sum\limits_{k=1}^{K}{|I}_{k}| \cdot {A}_{k}\end{array}$$\end{document} where K is the total number of land use types, Ik is the cooling intensity coefficient for the k-th land use type, and Ak is the total area of the k-th land use type. Quantification of the SW index The total runoff control standards for different land use types are commonly used to assess the stormwater and flood resilience of the built environment. In this study, the annual runoff control rate of urban blue-green spaces is introduced as a proxy indicator of stormwater resilience, used to indirectly represent their potential capacity for runoff mitigation. It should be noted that this proxy indicator provides a simplified representation of stormwater resilience and does not explicitly account for factors such as slope, drainage conditions, and soil properties. Therefore, the results should be interpreted as relative scenario scores for comparing alternative land-use configurations rather than direct measurements of actual stormwater performance, and they have not been validated through field observations. In the construction of the stormwater and flood resilience objective, the annual total runoff control rate is introduced as a key parameter, with corresponding runoff control coefficients (Table 1) set for different land use types33. The total stormwater and flood benefit is calculated using the product of the land use type area and the runoff control coefficient. According to relevant regulations by the Chinese government, the required runoff control rate for urban built-up areas should not be lower than 75%34, which is set as a constraint. The formula is as follows:\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\:{f}_{SW}=\frac{{\sum}_{k=1}^{K}{W}_{k} \cdot {A}_{k}}{{\sum}_{k=1}^{K}{A}_{k}}\:\left(3\right)$$\end{document}\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\:\begin{array}{c}\:{f}_{SW}\in\:\left[0.75,\right.\left.1\right)\end{array}$$\end{document} where K is the total number of land use types, Wk is the stormwater runoff control rate for the k-th land use type, and Ak is the total area of the k-th land use type. Table 1Coefficients of different land use type.Land use typeCarbon sinkCoolingStormwater runoff ControlForest0.870.680.96Grassland0.020.180.91Water0.000.720.00Wetland0.570.701.00 Table 1Coefficients of different land use type.Land use typeCarbon sinkCoolingStormwater runoff ControlForest0.870.680.96Grassland0.020.180.91Water0.000.720.00Wetland0.570.701.00 Coefficients of different land use type. Quantification of the DIV index Since increasing the proportion of blue-green space helps enhance ecological biodiversity levels35, the proportion of blue-green space is used as a proxy indicator for biodiversity in the optimization process. It should be noted that this indicator does not directly measure species richness, habitat quality, or ecological connectivity. Instead, it represents the relative ecological support potential associated with increased blue-green space coverage under different land-use configurations. The formula is as follows:4\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\:\begin{array}{c}{f}_{DIV}=\frac{{A}_{g}+{A}_{w}}{{A}_{sum}}\end{array}$$\end{document} where Ag represents the total area of all green space blocks; Aw represents the total area of all water blocks; and Asum represents the total area of all land use blocks. Multi-objective optimization algorithm: NSGA-II Layout optimization objectives To construct a multi-ecological objective optimization model, it is necessary to balance multiple objectives and use the NSGA-II algorithm for automatic optimization. In each generation, fast non-dominated sorting and reference-point-based sorting identify non-dominated solutions. Iterations continue until termination criteria are met, yielding the optimal Pareto front. Grid cells are used as the basic decision variables, with each cell representing a gene. All grid cells within the study area together constitute a chromosome, which represents a complete spatial configuration of blue-green spaces. Key parameters include population size (60), number of iterations (300), crossover rate (0.8), and mutation rate (0.05). A binary encoding scheme is employed to represent the transformability of grid cells, where “1” denotes that a grid cell is eligible for land-use conversion during the optimization process, while “0” denotes that the grid cell is constrained or remains unchanged in its original land-use type. Specifically, based on land-use types and planning constraints, the study area is divided into fixed regions and adjustable regions. Built-up land and cropland are restricted by planning and policy requirements and remain unchanged during the optimization process, corresponding to a value of “0”. Forest land, grassland, wetland, and part of unused land are considered the main targets for blue-green space optimization and are allowed to be transformed under given constraints, corresponding to a value of “1”. Water bodies, as critical ecological elements, are subject to strict transformation limitations, and their variation is controlled through constraints on the water surface ratio. It should be noted that the binary encoding does not directly represent specific land-use categories. Instead, the binary value indicates whether a grid cell is eligible for land-use conversion during optimization. Cells coded as “1” belong to the adjustable region and may be reassigned among forest land, grassland, wetland, and unused land according to the optimization operators and area constraints, whereas cells coded as “0” remain fixed in their original land-use type throughout the optimization process. Following crossover and mutation, any infeasible solutions violating predefined area or planning constraints are corrected through a constraint-checking procedure to ensure that all generated solutions satisfy the optimization requirements. The initial population is generated based on the existing land-use distribution to ensure a realistic starting point for the optimization process. Crossover and mutation operations are applied to grid cells by randomly exchanging or modifying the transformability states of selected cells, thereby adjusting the spatial configuration and generating new candidate solutions. The optimization process terminates when the predefined number of iterations is reached, and the non-dominated solution set is retained as the final Pareto front for subsequent multi-objective trade-off analysis and decision-making. The optimization objectives for the blue-green configuration are as follows: Maximize: CS. Maximize: CI. Maximize: SW. Maximize: DIV. Constraints on decision variables Area Constraints Area Constraints Area Constraints Area constraints include total area, vegetation cover, landscape diversity, and urban water surface ratio. The total area constraint ensures that the sum of all land-use types equals the study area, maintaining the validity of optimization results. The constraint formula is as follows:5\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\sum\limits_{k=1}^{7}{A}_{K}=S$$\end{document} where k refers to the count of land use types, AK denotes the area (in m²) of land use type k, and S represents the total area (in m²) of the study area. The vegetation cover constraint requires that optimized vegetation coverage not fall below pre-optimization levels, maintaining greening, supporting biodiversity, mitigating urban heat, and enhancing ecological resilience. Vegetation coverage is calculated using the “ecological greening equivalent” method36, with coefficients of 1 for forests, 0.49 for grasslands, and 0.46 for croplands. The constraint formula is as follows:6\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\:\frac{{A}_{2}+{0.49A}_{3}}{S}\ge\:\frac{{S}_{2}+{0.49S}_{3}}{S}$$\end{document} where Ak represents the area (in m²) of the k-th land use type in the optimized scenario; Sk denotes the area (in m²) of the k-th land use type in the unoptimized scenario. k = 2 and k = 3 correspond to forestland and grassland, respectively. The landscape diversity constraint ensures that the combined proportion of forestland, grassland, water, and wetlands in the optimized scenario is not lower than that in the pre-optimization scenario, in order to maintain ecosystem stability and diversity. The constraint formula is as follows:7\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\:\frac{{A}_{2}+{A}_{3}+{A}_{4}+{A}_{5}}{S}\ge\:\frac{{S}_{2}+{S}_{3}+{S}_{4}+{S}_{5}}{S}$$\end{document} where k = 2,3,4,5 correspond to the four land use types: forestland, grassland, water, and wetlands, respectively. The urban water surface ratio constraint requires that, during urban planning and land use optimization, the proportion of urban water areas in the total area must not be lower than the pre-optimization level after adjustments. This condition can be expressed mathematically, where the percentage of water area in the optimized scenario must be greater than or equal to the percentage in the pre-optimization scenario. The constraint formula is as follows:8\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\:\frac{{A}_{4}+{A}_{5}}{S}\ge\:\frac{{S}_{4}+{S}_{5}}{S}$$\end{document} where k = 4,5 correspond to the two land use types: water and wetlands, respectively. (2)Spatial Constraints (2)Spatial Constraints Spatial Constraints Spatial constraints include The Ecological Land Red Line constraint and the urban construction land constraint. The blue-green space within the ecological land red line is protected, and only areas outside the red line may be optimized and adjusted. The urban construction land constraint stipulates that residential, commercial, industrial, and public facility lands must remain stable during the optimization process, with no large-scale adjustments allowed. (3)Land Use and Conversion Constraints (3)Land Use and Conversion Constraints Land Use and Conversion Constraints Land use and conversion constraints include the non-negativity constraint and the plot type conversion constraint. In the optimization of urban blue-green spaces, the area of different land use types must be positive. The model specifies that the area of all land use types must be greater than 0. The expression for the non-negativity constraint is as follows:9\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\:{A}_{k}>0$$\end{document} where k refers to the k-th land use type, and Ak represents the area of the k-th land use type. The plot type conversion constraint restricts the adjustment of construction land, cropland, and water. The possibility of change in construction land is low, so no adjustments are made. Cropland is protected by policy and is not subject to adjustments. Water evolves slowly under natural influences, and their conversion to other land use types is strictly limited in order to maintain ecological stability. Multi-objective empowered decision-making In multi-objective optimization problems, the NSGA-II algorithm typically generates a set of Pareto-optimal solutions, representing trade-offs among different ecological objectives. As no single absolute optimal solution exists, further evaluation of the Pareto solution set is required to identify representative spatial configurations. To this end, the entropy weight method is applied to perform a trade-off analysis of the Pareto front solutions generated by NSGA-II. The resulting weights reflect the informational variation of different ecological indicators within the solution set, rather than their actual importance in practical planning contexts. The entropy-weight method is used to conduct trade-off analysis on the Pareto front solution set generated by the NSGA-II algorithm. By calculating the information entropy of the four ecological objectives—carbon sink, cooling, stormwater resilience, and biodiversity—the weights of each objective are objectively determined. The specific implementation involves three key steps: First, the ecological indicators are normalized, and the formula is as follows:10\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\:\begin{array}{c}{x}_{ij}=\frac{{x}_{ij}-min\left({x}_{j}\right)}{max\left({x}_{j}\right)-min\left({x}_{j}\right)}\end{array}$$\end{document} where xij represents the raw value of the i-th solution for the j-th objective function; min(xj) and max(xj) are the minimum and maximum values of the j-th objective function across all solutions, respectively. In this study, x refers to the blue-green space decision variables, i.e., the spatial configuration encoding. Next, the information entropy of each indicator is calculated to assess its variability. The formula is as follows:11\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\:\begin{array}{c}{\mathrm{p}}_{\mathrm{i}\mathrm{j}}=\frac{{\mathrm{x}}_{\mathrm{i}\mathrm{j}}}{{\sum}_{\mathrm{i}=1}^{\mathrm{n}}{\mathrm{x}}_{\mathrm{i}\mathrm{j}}}\end{array}$$\end{document}12\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\:\begin{array}{c}{\mathrm{E}}_{\mathrm{j}}=-k\sum\limits_{\mathrm{i}=1}^{\mathrm{n}}{\mathrm{p}}_{\mathrm{i}\mathrm{j}}\mathrm{log}\left({\mathrm{p}}_{\mathrm{i}\mathrm{j}}\right)\end{array}$$\end{document}13\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\:\begin{array}{c}k=\frac{1}{\mathrm{log}\left(\mathrm{n}\right)}\end{array}$$\end{document} where Ej is the information entropy of the j-th objective function; pij is the proportion of the normalized value of the i-th solution for the j-th objective function; k is a constant, and n is the total number of solutions. Finally, the weights of each indicator are derived from the information entropy. The lower the information entropy, the higher the weight of the indicator. The formula is as follows:14\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\:\begin{array}{c}{\mathrm{w}}_{\mathrm{j}}=\frac{{1-\mathrm{E}}_{\mathrm{j}}}{{\sum}_{\mathrm{j}=1}^{\mathrm{m}}\left(1-{\mathrm{E}}_{\mathrm{j}}\right)}\end{array}$$\end{document} where wj is the weight of the j-th objective, and m is the total number of objective functions. To assess blue-green space ecological benefits, four single-objective and one multi-objective optimal scenarios are set: carbon sink, cooling, stormwater resilience, biodiversity, and comprehensive optimal. In single-objective scenarios, the target objective weight is 100% and others 0%. In the multi-objective scenario, the sum of the four sub-objective weights is 100%, calculated using the entropy-weight method in Matlab 2018a. Results Algorithm optimization results Fig. 3Pairwise relationships among ecological objectives in the Pareto-optimal solutions.`. Pairwise relationships among ecological objectives in the Pareto-optimal solutions.`. The analysis of the Pareto-optimal solutions reveals the relationships among the four ecological objectives (Fig. 3). A very strong positive relationship was observed between carbon sink capacity and cooling benefits (r = 0.991, R² = 0.983, p < 0.001), indicating substantial synergy between these objectives. Carbon sink capacity was also strongly associated with stormwater resilience (r = 0.951, R² = 0.904, p < 0.001), suggesting that improvements in carbon sequestration can simultaneously enhance stormwater management performance. Similarly, cooling benefits and stormwater resilience exhibited a strong positive correlation (r = 0.921, R² = 0.849, p < 0.001), demonstrating considerable co-benefits among these ecosystem services. Biodiversity exhibited comparatively weaker relationships with the other objectives. Its association with carbon sink capacity was positive but not statistically significant (r = 0.300, R² = 0.090, p = 0.060), indicating limited evidence for a robust synergy. In contrast, biodiversity showed weak but significant positive correlations with cooling benefits (r = 0.373, R² = 0.139, p = 0.018) and stormwater resilience (r = 0.395, R² = 0.156, p = 0.012). Overall, the results indicate predominantly synergistic relationships among the evaluated ecological objectives, while biodiversity exhibits a relatively independent response compared with the other ecosystem services. Single objective optimal scenarios The outcomes of both single-objective and comprehensive multi-objective optimizations are synthesized in Fig. 4. Subplots a–d present the optimal under the given proxies, constraints and algorithm settings for maximizing each individual ecosystem benefit: carbon sink, cooling, stormwater resilience, and biodiversity, respectively. The integrated solution, which balances the trade-offs between all four objectives to achieve an overall optimum, is presented in the final panel (e). The results for each are detailed as follows: Fig. 4Optimization layouts and area comparisons; (a) carbon sink benefits; (b) cooling benefits; (c) stormwater resilience benefits; (d) biodiversity benefits; (e) comprehensive optimal configuration; (Created by ArcMap, version 10.8, https://www.esri.com). Optimization layouts and area comparisons; (a) carbon sink benefits; (b) cooling benefits; (c) stormwater resilience benefits; (d) biodiversity benefits; (e) comprehensive optimal configuration; (Created by ArcMap, version 10.8, https://www.esri.com). Scenario 1: optimal carbon sink benefits The spatial layout of blue-green spaces under the optimal carbon sink scenario is shown in Fig. 4a, where it can be observed that the proportion of green space in the city has increased. To further explore the underlying patterns before and after optimization, the study compared the areas of different land use patches before and after optimization. Figure 4a illustrates the area changes of four land use types in the blue-green space optimization under the optimal carbon sink scenario. Regarding forest land, the total area of forests decreased by 21.58% after optimization, mainly transitioning to grassland and wetland areas. Meanwhile, the unit patch area of forest increased by 51.06% compared to the pre-optimization scenario. The total area and unit patch area of grassland both showed reductions to varying degrees. Specifically, the total area of grassland decreased by 47.09%, and the unit patch area reduced by 26.45%. As for the blue space, wetlands experienced a significant increase in total area under the carbon sink preference scenario. This increase primarily came from the transformation of forest and grassland areas, with the total wetland area being 8.89 times larger after optimization compared to before. However, the unit patch area of wetlands decreased by 24.84%, a reduction of approximately one-quarter. The total area of unused land was drastically reduced by 92.08%. Scenario 2: optimal cooling benefits Figure 4b shows the layout of land use types after optimization in the scenario where cooling benefits are maximized. In this scenario, the total area of forest land decreased by 17.16% compared to before optimization, with some forest land being converted into wetlands and grasslands. After optimization, the unit patch area of forest land significantly increased, reaching 6.96 times that of the pre-optimization area. The changes in grasslands and forests were similar: the total area increased slightly by 9.29%, but the unit patch area expanded 13.89 times compared to the pre-optimization area, and the proportion of patches also saw a slight increase. As in the carbon sink scenario, wetland area grew substantially, with an increase of 6.69 times the pre-optimization area. Furthermore, the unit patch area of wetlands after optimization decreased by 35.89%. The area of unused land continued to decline, with a total reduction of 69% after optimization. It is noteworthy that, despite the decrease in the total area of unused land, its unit patch area increased by 51.46%. Scenario 3: optimal stormwater resilience Figure 4c reflects the blue-green space layout and area changes before and after optimization in the scenario of optimal resilience to stormwater flooding. From Fig. 4c, it can be observed that the total area of wetlands increased significantly, with the optimized total area increasing 5.79 times. However, the unit patch area decreased by 43.14%. The newly added wetland types are not highly clustered, but rather dispersed in scattered patches. The total area of grasslands increased by 32.99%, mainly due to conversions from forest land and unused land. The area of forest land decreased by 19.45%, but both grassland and forest unit patch areas increased, leading to a significant rise in the dominance of patch area within the overall green space. The total area of unused land decreased by 66.55%, primarily being converted to forest, grassland, and wetland. The unit patch area of unused land increased by 63.4%. Scenario 4: optimal biodiversity Figure 4-d reflects the blue-green space layout and area changes before and after optimization in the scenario of optimal biodiversity. The total area of forest land decreased significantly by approximately 39.14%, with a slight reduction of 6.93% in the average patch area. Forest land was primarily converted into wetlands and grasslands. The increase in wetland area was the most pronounced, with a growth of 11.17 times compared to before optimization, though the unit patch area remained nearly the same before and after optimization. The total area and unit patch area of grasslands both increased, with the total area expanding by 41.73% and the unit patch area approximately 2.43 times that of the pre-optimization value. The total area of unused land drastically reduced to around 31,584.08 m2, with a 69% decrease in unit patch area, showing a trend toward fragmentation. Multi-objective integrated optimal scenario The entropy weight method was used to calculate the weights of the four sub-goals in the multi-objective integrated optimization scenario. The total weight is 100%, with specific weights as follows: carbon sink goal 43%, cooling benefit goal 20%, stormwater resilience goal 25%, and biodiversity goal 12%. In the multi-objective integrated optimal scenario (Fig. 4-e), the area of forest land was significantly reduced, but it still occupies a large area and is widely distributed. The total area of grasslands after optimization is 109,662,048.96 m², which is an increase of approximately 1.09 times compared to before optimization. The total area of wetlands increased significantly, with the highest growth among all land use types, reaching about 7.89 times that of the pre-optimization area. The average patch area of wetlands decreased by 24.84%. The total area of unused land decreased by 92%, and the average patch area was reduced by about half. Discussion Comparison of ecological benefit optimization rates of blue-green spaces under different preference scenarios To assess ecological benefit enhancement under different preference scenarios, four objectives were quantitatively analyzed across five scenarios using NSGA-II results (Fig. 5). Comparison shows that single-objective scenarios improve their target goal but limit others: carbon sink priority boosts carbon sequestration but contributes little to cooling; cooling priority enhances cooling but reduces carbon sink; stormwater resilience priority improves water management but requires long-term evaluation; biodiversity priority increases biodiversity but lowers other benefits. The integrated scenario achieves balanced benefits across all objectives, offering a comprehensive and sustainable solution. These findings suggest that single-objective optimization may compromise overall ecological performance, highlighting the need for a comprehensive strategy to maximize collaborative benefits in blue-green space planning (Table 2). Fig. 5Comparison of ecological benefits of blue-green spaces before and after optimization under different preference scenarios. Comparison of ecological benefits of blue-green spaces before and after optimization under different preference scenarios. Table 2Values of ecological benefits before and after optimization under different preference scenarios.Scenario typeValues of the four ecological objectivesCarbon sink(t)CoolingStormwater resilience(%)BiodiversityCarbon sink priority25109.96001554.997774.810.4376Cooling priority23310.95993777.275475.010.4374Stormwater resilience priority22432.01232061.028475.290.4377Biodiversity priority19662.32071961.028474.830.4380Multi-objective Integrated25043.21542014.764274.980.4378 Table 2Values of ecological benefits before and after optimization under different preference scenarios.Scenario typeValues of the four ecological objectivesCarbon sink(t)CoolingStormwater resilience(%)BiodiversityCarbon sink priority25109.96001554.997774.810.4376Cooling priority23310.95993777.275475.010.4374Stormwater resilience priority22432.01232061.028475.290.4377Biodiversity priority19662.32071961.028474.830.4380Multi-objective Integrated25043.21542014.764274.980.4378 Values of ecological benefits before and after optimization under different preference scenarios. Comparison of blue-green space characteristics under single-preference scenarios Using ArcGIS software, an overlay analysis was conducted on the distribution maps of blue-green spaces in Nanjing’s central urban area before and after optimization (Fig. 6). By comparing the changes in the distribution of blue-green spaces before and after optimization, the study aims to further explore the spatial characteristics of blue-green spaces in Nanjing’s central urban area under single-preference scenarios. Fig. 6Comparison of blue-green space characteristics before and after optimization; (a) carbon sink; (b) cooling; (c) stormwater resilience; (d) biodiversity. (Created by ArcMap, version 10.8, https://www.esri.com). Comparison of blue-green space characteristics before and after optimization; (a) carbon sink; (b) cooling; (c) stormwater resilience; (d) biodiversity. (Created by ArcMap, version 10.8, https://www.esri.com). Carbon sink benefit optimal scenario Figure 6a shows blue-green space configuration in Nanjing’s central urban area before and after optimization under the carbon sink scenario. Both green and blue spaces form large contiguous areas with good integrity and ecological connectivity. Comparison with pre-optimization layout shows significant expansion of blue spaces, especially along the Qinhuai River and surrounding areas, creating a more complex water-land interface and enhancing carbon sink potential37. Some localized reductions occur along the Yangtze River due to urban development, but overall blue space expansion dominates. Green space expansion is more limited, with new patches scattered in built-up areas, notably south of Xixia Mountain and near expanded blue spaces. Cooling benefit optimal scenario Figure 6b compares blue-green space configuration before and after optimization under the cooling scenario. Green spaces remain dominant, with area similar to pre-optimization. Larger new blue patches cluster around existing water, expanding water area and water-land interface length, enhancing evaporation and cooling in surrounding areas38. Blue space reductions are mainly along the Yangtze River, with minimal losses elsewhere. Smaller blue patches are scattered across urban areas, optimizing microclimates and mitigating local heat islands39. This dispersed effect cumulatively enhances overall cooling benefits40. Stormwater resilience benefit optimal scenario Figure 6c shows blue-green space configuration under the stormwater resilience optimal scenario. Post-optimization, urban water area has expanded, with new blue spaces scattered across built-up areas. Water expansion can be achieved via artificial wetlands, widened river channels, rain gardens, and other green infrastructure, enhancing stormwater management, hydrological diversity, aquatic ecosystems, and resilience to extreme events41. Green space coverage remains high, preserving vegetation’s role in reducing runoff, promoting soil and water conservation, and enhancing infiltration42. Large contiguous green spaces provide ecological benefits, including pollutant interception, runoff reduction, groundwater recharge, and reduced reliance on gray infrastructure43. Biodiversity benefit optimal scenario Figure 6d shows blue-green space configuration under the biodiversity optimal scenario. Blue-green spaces exhibit a clustered distribution, with large core habitats such as the Purple Mountain–Xuanwu Lake area surrounded by blue spaces. Previous studies have suggested that continuous green spaces and adjacent blue areas may provide habitats and migration corridors, thereby supporting biodiversity44. The optimization results of this study show a tendency toward the spatial aggregation of green and blue spaces. The spatial aggregation of blue-green spaces observed in this study is broadly consistent with these findings. Wetlands and other blue spaces have expanded, increasing water-land interface areas that enhance ecological habitats for aquatic and terrestrial species. Previous studies have suggested that such spatial configurations may function as “stepping stones” within blue-green networks, facilitating species movement and reducing biodiversity fragmentation45. In this study, similar spatial patterns were observed, which may contribute to connectivity within the overall blue-green spatial structure; however, their specific ecological effects require further validation. Discussion of typical areas In order to discuss the distribution characteristics of blue-green spaces under the four single-ecological enhancement preference scenarios, three representative areas were selected for analysis: the Xuanwu Lake-Purple Mountain area, the Qinhuai Xinhe tributary area, and the Yangtze River-Mufu Mountain area (Fig. 7). The blue-green space pattern in the Xuanwu Lake–Purple Mountain area is optimized, with greater aggregation and dominance of large patches. Green space expansion exceeds that of blue space (Fig. 7a). Under the carbon sink scenario, large forest patches remain clustered, while some fragmented green patches convert to blue spaces. In the cooling scenario, expansion occurs around major patches such as Zhongshan Mountain and the Yangtze River, with increased forest and grassland areas and stronger patch aggregation. Blue space also expands, mainly through wetlands adjacent to existing water, though its extent remains smaller than green space. In the stormwater resilience scenario, forest area decreases and transitions into scattered grassland, while wetlands around water expand, forming stormwater buffers. In the biodiversity scenario, blue-green spaces exhibit more diverse spatial configurations, with some forest land converted to wetlands. This change reflects adjustments in the proportion and composition of different types of blue-green spaces, which may contribute to improvements in biodiversity-related indicators to some extent. In the Qinhuai New River tributary area (Fig. 7b), the distribution of blue-green spaces before and after optimization differs across the four scenarios, but there is a general trend of expansion in blue spaces. In the carbon sink optimal scenario, the newly added wetlands are mainly distributed around the Qinhuai New River, forming blue-green spaces with higher area dominance and greater aggregation. The connectivity between blue-green patches is significantly enhanced, thereby promoting the carbon sink benefits. In the cooling optimal scenario, the newly added blue spaces are less prominent than in the carbon sink scenario, with a more scattered distribution pattern. Notably, the surface water areas begin to expand outward, increasing the dominance of blue patches. The length of the water-land interface surrounding the water increases, and the larger, more winding blue-green spaces provide better cooling effects. In the stormwater resilience optimal scenario, the newly added blue-green spaces are scattered across the developed areas, and the surface water transition into green spaces around the edges, forming natural buffer zones that help mitigate stormwater and flooding. In the biodiversity optimal scenario, newly added blue and green patches are concentrated around the existing natural blue-green patches. This contrasts with the large patch aggregation observed in the Xuanwu Lake-Purple Mountain area, presenting a pattern of smaller blue-green patch aggregation. The area near the Yangtze River-Mufu Mountain region contains large-scale ecological protection zones. Beyond these protected areas, the optimization of blue-green spaces shows distinct spatial layout changes, particularly in terms of spatial differences under various ecological objectives (Fig. 7-c). In the carbon sink and cooling optimal scenarios, the significantly increased blue spaces are mainly composed of large wetland patches, with more scattered blue spaces appearing in the built-up areas, similar to the Xuanwu Lake-Purple Mountain and Qinhuai Xinhe tributary regions. In the carbon sink optimal scenario, the proportion of green space is higher than that in the cooling scenario. In the stormwater resilience optimal scenario, the distribution of blue patches is more dispersed to achieve better stormwater runoff control. In the biodiversity optimal scenario, large aggregations of blue-green patch clusters emerge, with increased connectivity between blue-green patches. Fig. 7Comparison of typical area under four single-preference scenarios; (a) Xuanwu Lake-Purple Mountain Area; (b) Qinhuai Xinhe Tributary Area; (c) Yangtze Riverbank-Mufu Mountain area. (Created using ENVI 5.3 and ArcMap, version 10.8; https://www.nv5geospatialsoftware.com, https://www.esri.com). Comparison of typical area under four single-preference scenarios; (a) Xuanwu Lake-Purple Mountain Area; (b) Qinhuai Xinhe Tributary Area; (c) Yangtze Riverbank-Mufu Mountain area. (Created using ENVI 5.3 and ArcMap, version 10.8; https://www.nv5geospatialsoftware.com, https://www.esri.com). In summary, the blue-green space characteristics under the four single-preference optimal scenarios differ significantly. In the carbon sink optimal scenario, the emphasis is on the aggregation of green spaces, with newly added blue-green patches primarily clustering adjacent to large existing natural blue-green spaces. In the cooling optimal scenario, the surface water areas expand outward, with the shoreline becoming more winding and complex, leading to an increase in the length of the coastline. In the stormwater resilience optimal scenario, blue-green spaces are distributed in a “scattered” layout within the urban areas. In contrast, in the biodiversity optimal scenario, a “clustered” arrangement of blue-green spaces is prominent, where blue spaces surround green spaces. The connectivity between blue-green patches significantly increases, and the edge effect becomes more pronounced. In the multi-objective comprehensive optimal scenario, the spatial characteristics of blue-green spaces combine the optimal features of the four single-preference scenarios, achieving an integrated enhancement of ecological benefits. Blue-green space configuration strategies based on optimization results Fig. 8Comparison of the integrated optimal blue-green space configuration and the current configuration in Nanjing: (a) comprehensive optimal configuration; (b) current blue-green space configuration. (Created by ArcMap, version 10.8, https://www.esri.com). Comparison of the integrated optimal blue-green space configuration and the current configuration in Nanjing: (a) comprehensive optimal configuration; (b) current blue-green space configuration. (Created by ArcMap, version 10.8, https://www.esri.com). To explore the spatial implications of the optimization results, the blue-green space configuration under the multi-objective integrated scenario is compared with the current pattern in Nanjing (Fig. 8). Existing blue-green spaces are dominated by large patches with uneven distribution, particularly within built-up areas where small-scale spaces are relatively insufficient. In contrast, the optimized configuration introduces more small-scale green and blue units, resulting in a more balanced spatial pattern. Therefore, it is advisable to increase small green spaces and small-scale water bodies where appropriate to address local deficiencies. The optimized configuration also shows stronger adjacency between green and blue spaces, especially with localized clustering around water bodies. This suggests that planning practices may further enhance blue-green coordination by improving the spatial arrangement of green spaces near water bodies. In terms of composition, the optimized configuration exhibits more diverse combinations of blue-green space types. Notably, wetlands occupy a higher proportion compared to the current pattern, indicating that their expansion in suitable areas may improve overall spatial composition. In addition, the increase in grassland area suggests its adaptability in multi-objective optimization. Previous studies indicate that grassland vegetation can enhance soil infiltration46; therefore, increasing grassland in appropriate areas may support stormwater regulation. Overall, while maintaining the stability of core blue-green spaces, planning practices may introduce more diverse blue-green space types to enrich spatial structure. These findings are based on model simulations and should be interpreted with consideration of actual development conditions. Limitations and prospects This study has some limitations. First, regarding data accuracy, the remote sensing data used in this study have a spatial resolution of 30 m, which limits the precise identification of small green spaces and water bodies in high-density urban environments. Future studies could employ higher-resolution remote sensing data to improve the identification and characterization of blue-green spaces, thereby enhancing the accuracy of the analysis. Second, the study sample size is limited. This research focused on the blue-green spaces in the central urban area of Nanjing. Future work could expand to studies at larger regional scales, such as urban agglomerations, by increasing the sample size to explore the similarities and differences in the multi-objective optimization processes of blue-green spaces. Third, the model involves certain simplifications. While the multi-objective optimization model developed in this study can balance ecological benefits related to carbon sink, cooling, stormwater resilience, and biodiversity, the ecological benefits of urban blue-green spaces may also be influenced by other factors such as soil types and vegetation species. Future research could further refine the model to improve its applicability. Fourth, in terms of spatial characteristics, this study focuses on the optimization of blue-green space spatial configurations and does not explicitly incorporate spatial morphology metrics. Therefore, the results mainly reflect the spatial reallocation of land-use types rather than the direct optimization of spatial morphology. Future studies may incorporate landscape pattern metrics to further improve the model. Fifth, some ecological benefit indicators are represented using simplified coefficients or proxy variables, which may not fully capture the complexity of ecological processes. Conclusions For a long time, optimizing the layout of urban blue-green spaces while meeting multiple ecological benefit objectives has been a challenge in the fields of urban planning and sustainable design. To address this challenge, we established objective functions and optimization algorithms, using the central urban area of Nanjing as a practical case study. By analyzing the changes in the weightings of different ecological benefit goals, we developed optimization schemes for blue-green spaces in the central urban area of Nanjing, focusing on four individual ecological objectives: carbon sink, cooling, stormwater regulation, and biodiversity. Based on this, a multi-objective integrated optimal scenario was also proposed, identifying the optimal balance between the four individual ecological benefit objectives, thus presenting a blue-green space layout plan that maximizes overall ecological benefits. These optimization schemes explain the synergistic and coupled mechanisms between blue-green space configuration and ecological benefits. They not only provide scientific evidence for the spatial layout of blue-green spaces but also offer decision-making support for urban planners and managers to meet multiple ecological needs. The main research conclusions are as follows: The genetic algorithm effectively optimized the carbon sink, cooling, stormwater resilience, and biodiversity objectives for the central urban area of Nanjing, successfully producing the optimal under the given proxies, constraints and algorithm settings for the multi-objective integrated scenario.Pareto frontier analysis revealed the synergies and trade-offs between the four ecological objectives. Notably, carbon sink and cooling benefits showed a significant synergistic relationship, while biodiversity benefits and stormwater resilience benefits displayed a trade-off. The relationships between other objective combinations were weak.In the carbon sink optimal scenario, a core aggregation pattern of large green spaces with adjacent clustering was observed. In the cooling optimal scenario, there was a preference for the expansion of surface water areas and the increase of winding shorelines. In the stormwater resilience optimal scenario, a dispersed layout of blue-green patches was evident. In the biodiversity optimal scenario, a “clustered” and highly connected structure of blue-green spaces emerged.The multi-objective integrated optimal scenario is not a simple overlay of the characteristics from individual objectives, but rather an organic integration of blue-green space elements. It establishes a dynamic balance between aggregation, scale, and connectivity, achieving a collaborative enhancement of ecological benefits. The genetic algorithm effectively optimized the carbon sink, cooling, stormwater resilience, and biodiversity objectives for the central urban area of Nanjing, successfully producing the optimal under the given proxies, constraints and algorithm settings for the multi-objective integrated scenario. The genetic algorithm effectively optimized the carbon sink, cooling, stormwater resilience, and biodiversity objectives for the central urban area of Nanjing, successfully producing the optimal under the given proxies, constraints and algorithm settings for the multi-objective integrated scenario. Pareto frontier analysis revealed the synergies and trade-offs between the four ecological objectives. Notably, carbon sink and cooling benefits showed a significant synergistic relationship, while biodiversity benefits and stormwater resilience benefits displayed a trade-off. The relationships between other objective combinations were weak. Pareto frontier analysis revealed the synergies and trade-offs between the four ecological objectives. Notably, carbon sink and cooling benefits showed a significant synergistic relationship, while biodiversity benefits and stormwater resilience benefits displayed a trade-off. The relationships between other objective combinations were weak. In the carbon sink optimal scenario, a core aggregation pattern of large green spaces with adjacent clustering was observed. In the cooling optimal scenario, there was a preference for the expansion of surface water areas and the increase of winding shorelines. In the stormwater resilience optimal scenario, a dispersed layout of blue-green patches was evident. In the biodiversity optimal scenario, a “clustered” and highly connected structure of blue-green spaces emerged. In the carbon sink optimal scenario, a core aggregation pattern of large green spaces with adjacent clustering was observed. In the cooling optimal scenario, there was a preference for the expansion of surface water areas and the increase of winding shorelines. In the stormwater resilience optimal scenario, a dispersed layout of blue-green patches was evident. In the biodiversity optimal scenario, a “clustered” and highly connected structure of blue-green spaces emerged. The multi-objective integrated optimal scenario is not a simple overlay of the characteristics from individual objectives, but rather an organic integration of blue-green space elements. It establishes a dynamic balance between aggregation, scale, and connectivity, achieving a collaborative enhancement of ecological benefits. The multi-objective integrated optimal scenario is not a simple overlay of the characteristics from individual objectives, but rather an organic integration of blue-green space elements. It establishes a dynamic balance between aggregation, scale, and connectivity, achieving a collaborative enhancement of ecological benefits.