Life-Cycle Assessment of Carbon Sink Efficiency in Urban Landscape Spatial Units: Evidence from Luhe Park, Nanjing

Abstract

Urban green spaces are pivotal to enhancing carbon sinks and advancing carbon neutrality goals, yet the structural complexity of green space units often leads to scale mismatches and weak spatial responsiveness in current assessments. This study develops an integrated evaluation framework that combines landscape spatial unit typologies with life-cycle-based carbon flux modeling. We defined 22 landscape spatial unit types based on two-dimensional surface cover and three-dimensional vegetation structure, including waterbodies and vertical greening. A life-cycle carbon model was developed with indicators covering unit carbon sink, unit carbon emission, unit net carbon sink efficiency, and time to carbon balance. Taking Luhe Park in Nanjing as a case study, the carbon sink efficiency indicators were quantified for 108 units over a 50-year cycle. Results indicate that multilayer vegetation structures, high green coverage, and moderate-to-high planting density markedly enhance carbon sink efficiency, whereas extensive built surfaces and high impervious ratios suppress it. K-means clustering classified the spatial units into four types with emphasis on efficiency-driven, structural optimization, functional compatibility, and imbalance compensation, respectively, revealing a clear gradient tied to spatial configuration. To translate diagnosis into design, we report 95% confidence intervals of key structural factors as actionable thresholds. These ranges inform targeted interventions such as maintaining continuity and multilayer structure in high-efficiency areas, modest structural upgrades with native drought-tolerant plants, edge greening with permeable pavements in open spaces, and streamlined vertical systems linked to adjacent high-sink ground units. The framework delivers spatially explicit, life-cycle-aware evidence to support low-carbon planning and design of urban green spaces.

1. Introduction

Against the backdrop of global climate change, urban carbon neutrality strategies are progressively being implemented. Green infrastructure, as a vital component of urban ecosystems, has become a key carbon sink pathway beyond energy transition and industrial transformation [1,2]. In particular, “near-natural” spatial units such as green spaces and waterbodies are increasingly valued for their long-term, stable, and low-cost carbon sequestration potential [3,4].

In recent years, urban carbon sink research has evolved from static assessments to dynamic simulations, and from analyses of individual vegetation types to spatial patterns and configurations, promoting integration across ecology, geography, and urban planning. Various carbon estimation techniques, including biomass-based models, assimilation rate calculations, and micrometeorological observations, have been used, especially in tree-dominated urban landscapes [5,6]. Integrated ecological models such as CASA and InVEST have further enabled large-scale carbon flux simulations [7,8]. Empirical studies reveal the spatiotemporal heterogeneity of carbon sinks in urban green spaces. For instance, Nowak et al. [9] estimated annual carbon storage of urban trees in the United States using biomass and growth models, demonstrating that canopy coverage and age structure strongly influence carbon uptake. Liu & Li [10] applied allometric equations to assess carbon storage across diverse forest types and structures in Shenyang, China, highlighting the sink potential of multilayered vegetation communities with varied age distributions. Xu et al. [11] demonstrated that urban expansion negatively affects vegetative carbon uptake using an enhanced CASA model. Myeong et al. [12] also employed normalized difference vegetation index and regression models to predict urban forest carbon stocks, underscoring the value of remote sensing for large-scale carbon monitoring and management.

To further capture the full carbon budget of urban green spaces, an increasing number of studies have adopted life cycle assessment (LCA) methods to evaluate both carbon emissions and sinks across construction, maintenance, and renewal phases [6]. Strohbach et al. [13] assessed net carbon outcomes in Leipzig under different planting densities and maintenance regimes, revealing that forest-style maintenance outperformed park-style management. Zhang et al. [14] further applied LCA to track the shift in trees from carbon sources to sinks over time, showing that irrigation and pesticide use constitute major carbon emission sources. Perini [15] compared the life-cycle impacts of vertical greening, finding that it markedly reduced global warming potential and energy demand, but increased irrigation water consumption. These findings highlight the importance of assessing carbon performance from a life cycle perspective.

Despite this progress, urban green spaces, as typical components of built environment, are characterized by vertical stratification, blurred spatial boundaries, and complex vegetation structures [16]. These features contribute to spatiotemporal heterogeneity in carbon flux dynamics. Understanding how structural features influence carbon sink efficiency is critical for refined management [17]. To enhance structural interpretability and provide actionable guidance for spatial design, recent studies have incorporated spatial structure and vegetation attributes into carbon performance modeling. Landscape Spatial Units (LSUs) act as a link between spatial structure and carbon processes, converting structural attributes into measurable carbon performance through ecological processes [18]. This structure-process-performance-design framework underpins the carbon sink performance model. Zhao et al. [19] classified urban park habitats in Jinan, China, based on plant species composition and vertical stratification, using the i-Tree model to simulate carbon storage. Fan et al. [20] proposed high-carbon-sequestration planting models in Xi’an, China, highlighting the role of structural factors. Singkran [21] conducted hierarchical sampling and assessment of urban parks in Bangkok, Thailand, proposing that the proportion of green space and tree density jointly determine above-ground carbon sequestration efficiency.

However, current classification systems often focus on individual dimensions and overlook the integration of structural factors with life-cycle carbon flux assessments, limiting the potential for spatial optimization. This study integrates the LSU classification system, which covers both 2D land cover and 3D vertical structure, with a life-cycle carbon flux model, providing a comprehensive framework for evaluating carbon sink efficiency. The system includes 22 LSU types, combining these two dimensions, and incorporates vertical greening and isolated waterbodies as supplementary types. This innovative framework enables a more detailed assessment of carbon sink efficiency and supports modeling analysis. Additionally, an LCA-based life-cycle carbon flux assessment model is developed and combined with LSUs for refined carbon flux evaluation. The model is applied empirically to calculate key performance indicators, identify structural drivers of carbon performance, and establish design reference thresholds, offering guidance for optimizing urban green spaces.

The proposed framework is applied to Luhe Park in Nanjing as an empirical case of urban green space. Specifically, the study aims to: (1) develop and apply a systematic LSU classification for identifying spatial configurations within the park; (2) employ a LCA approach to quantify carbon fluxes and carbon sink performance across different unit types; (3) conduct cluster analysis and mechanism exploration to identify key structural factors; and (4) derive threshold values for factors and propose spatial optimization strategies to enhance the synergy between spatial pattern and ecological efficacy.

2. Materials and Methods

To comprehensively evaluate and optimize the carbon sink potential of urban green spaces, this study establishes an integrated assessment framework composed of three core components: LSU construction, life cycle-based carbon flux accounting, and carbon performance analysis (Figure 1).

2.1. Typological Framework for Landscape Spatial Units

2.1.1. Dimensions and Rationale for Classification

To enable fine-grained evaluation of carbon sink performance in urban green spaces and clarify its responsiveness to spatial configurations and planning design variables, this study establishes an LSU classification framework based on structural recognition. The framework is structured around two principal dimensions including horizontal surface cover composition and vertical vegetation community structure, supplemented by an extension for special spatial types. This classification logic is grounded in a systematic understanding of the multidimensional features of urban green space structure. In the horizontal dimension, different surface elements such as vegetation, impervious surfaces, and waterbodies directly contribute to the dominant carbon sink mechanism [2]. In the vertical dimension, layering complexity governs biomass accumulation and per-area fixation efficiency [20]. Certain urban spaces do not conform to conventional 2D or 3D logic and are therefore classified separately to keep the system complete and adaptable.

• Surface cover composition (Type A)

This dimension follows the dominant element principle, in which each spatial unit is classified according to the surface element (vegetation, hardscape, or water) with the highest surface cover proportion (SCP) [22]. For vegetation-dominated units, we distinguish high cover (SCP ≥ 80%) and medium cover (SCP < 80%). Although literature does not universally fix these thresholds, prior studies show diminishing marginal gains or saturation at higher cover ratios [23], with evidence of slight declines beyond 82% [24] and peak storage around 85%–88% in simulations [25]. We therefore adopt 80% as the threshold value for vegetation-dominated high-cover units based on prior research and modeling convenience.

2.

Vertical vegetation structure (Type B)

Based on the vegetation structure literature [20,26] and field vegetation observations in the study area, five structural types are delineated to reflect varying degrees of stratification: tree-shrub-grass multi-layer structures (B1), tree-shrub structures (B2), tree-grass structures (B3), shrub-grass structures (B4), and single-layer vegetation dominated by ground cover (B5). These typologies reflect differing ecological organizations, which contribute to variations in vegetation productivity and carbon potential across spatial units.

3.

Extended elements (Type C)

Some units exhibit carbon pathways or life-cycle profiles not well represented by Type A and Type B, including isolated waterbodies (M21) and vertical greening (M22). Waterbodies are typically regarded as stable long-term carbon sinks primarily via sediment burial, aquatic uptake, and microbial processes rather than terrestrial photosynthesis [27,28]. Vertical greening encompassing green roofs, green facades, and trellis, is vegetation on building fabric [29]. Although vegetated, these systems are based on building structures where building material production, construction processes, and limited soil depth often constrain net carbon sequestration efficiency. Evidence suggests these non-biotic constraints can influence performance more than in ground-level plantings [30]. Accordingly, treating vertical systems as a distinct LSU enables life-cycle–appropriate accounting and supports targeted optimization strategies.

2.1.2. Construction of Landscape Spatial Unit Typology System

By cross-combining Type A with Type B and incorporating two special types, 22 LSU types are defined to support carbon performance assessment (Figure 2).

• Standard composite types (M1–M20)

The core of the classification system comprises 20 standard spatial unit types (M1–M20), generated through the intersection of type A and type B categories. This integrated typology reflects the combined influence of horizontal surface cover and vertical vegetation structure on carbon sink mechanisms and sink potential. It accommodates a broad spectrum of spatial morphologies and ecological gradients—from continuous green expanses to fragmented patches, and from multi-layered canopies to single-layer herbaceous covers.

2.

Extended types (M21–M22)

Two additional unit types complement the standard framework: isolated waterbodies (M21) and vertical greening systems (M22). These units exhibit unique carbon cycling pathways and distinct life-cycle emission profiles that set them apart from conventional green space configurations. Their separate classification helps reduce model bias and prevents typological ambiguity, thereby improving assessment precision.

The 22-type classification system provides a clear structural logic based on surface cover and vegetation structure, representing various urban green spaces relevant to carbon flux assessment. It integrates with life cycle assessment models and supports targeted optimization of spatial configurations and carbon performance, making it a useful tool for low-carbon landscape planning.

2.2. Life Cycle Carbon Flux Assessment Model

To comprehensively assess the carbon sink performance of different LSUs over the full life cycle of urban green space, this study applies the LCA methodology to develop an integrated sink-emission-performance evaluation model (Figure 3). LCA, standardized by ISO 14046:2016 [31], provides a structured framework for assessing the potential environmental impacts arising from resource consumption and emissions throughout the entire life cycle of a product or system.

This study adopts LSUs as the functional unit and sets a 50-year reference life cycle to balance ecological dynamics with practical LCA considerations. This period corresponds to approximately 60 years of tree age, beyond which common urban trees typically stabilize or die, with carbon sequestration rates leveling off and maintenance intensity decreasing [13]. Moreover, beyond this timeframe, uncertainties related to stand dynamics, management regimes, and infrastructure renewal increase significantly, reducing the reliability of long-term predictions [14]. As such, a 50-year life cycle boundary is defined, following a “cradle-to-grave” approach, which encompasses material production, transportation, construction, operation and maintenance, and renewal and replacement. A comprehensive carbon process inventory is developed, including transportation energy consumption, machinery energy use, and material consumption. Using detailed CAD drawing data, a simulation model is created to estimate the life-cycle carbon fluxes based on the structural characteristics of urban green spaces.

2.2.1. Carbon Sink Estimation Model

The carbon sink model accounts for five key landscape components: trees, shrubs, groundcover, waterbodies (including open water, aquatic plants, and sediment), and soil. A hybrid approach combining process-based simulation and coefficient-based estimation is employed. For trees, annual carbon uptake is dynamically simulated using growth models and biomass equations. For shrubs and lower vegetation layers, sink is calculated using area-based empirical coefficients (

Table 1. Carbon sink estimation models and formulas.

Estimation Content	Formula	Parameter Explanation	References
Tree DBH growth prediction	Logistic equation $\mathrm{y} = {\displaystyle \frac{\mathrm{a}}{1 + \mathrm{b} \times \exp \left(-\mathrm{c} \times \mathrm{t}\right)}} + \in$ Gompertz equation $\mathrm{y} = \mathrm{a} \times \exp (-\mathrm{b} \times \exp (-\mathrm{c} \times \mathrm{t}\left)\right) + \in$ Richards equation ${\mathrm{y} = \mathrm{a} \times (1 - \exp (-\mathrm{b} \times \mathrm{t}\left)\right)}^{\mathrm{c} }+ \in$ Mitscherlich equation $\mathrm{y} = \mathrm{a} \times (1 - \mathrm{b} \times \exp (-\mathrm{c} \times \mathrm{t}\left)\right) + \in$ Weibull equation $\mathrm{y} = \mathrm{a} \times (1-\exp \left(-{\left({\displaystyle \frac{\mathrm{t}}{\mathrm{b}}}\right)}^{\mathrm{c}}\right)) + \in$	a: asymptotic maximum DBH; b: initial value parameter; c: growth rate; t: tree age (years); and y: DBH at age t (cm).	[32,33]
Tree biomass calculation	$\mathrm{B}={\mathrm{B}}_{\mathrm{a}}+{\mathrm{B}}_{\mathrm{b}}=\sum \left({\mathrm{B}}_{\mathrm{a}}^{\left(\mathrm{i}\right)}+{\mathrm{B}}_{\mathrm{b}}^{\left(\mathrm{i}\right)}\right)$ ${\mathrm{B}}_{\mathrm{a}}^{\left(\mathrm{i}\right)}={\mathrm{a}}_1^{\left(\mathrm{i}\right)}{\mathrm{y}}^{{\mathrm{b}}_1^{\left(\mathrm{i}\right)}}$ ${\mathrm{B}}_{\mathrm{b}}^{\left(\mathrm{i}\right)}={\mathrm{a}}_2^{\left(\mathrm{i}\right)}{\mathrm{y}}^{{\mathrm{b}}_2^{\left(\mathrm{i}\right)}}$	B: total biomass (kg); Ba(i), Bb(i): the aboveground and belowground biomass of tree i, respectively (kg). a1(i), b1(i) and a2(i), b2(i): the coefficients for estimating aboveground and belowground biomass of species i.	[34]
Tree carbon storage and sink	${\mathrm{CSt}}_{\mathrm{n},\mathrm{pi}}=\left({\mathrm{B}}_{\mathrm{a}}^{\left(\mathrm{i}\right)}\times {\mathrm{r}}_{\mathrm{a}}^{\left(\mathrm{i}\right)}+{\mathrm{B}}_{\mathrm{b}}^{\left(\mathrm{i}\right)}\times {\mathrm{r}}_{\mathrm{b}}^{\left(\mathrm{i}\right)}\right)\times {\mathrm{N}}_{\mathrm{i}}$ ${\mathrm{CS}}_{\mathrm{n},\mathrm{pi}}=({\mathrm{CS}}_{\mathrm{n},\mathrm{pi}}-{\mathrm{CS}}_{\mathrm{n}-1, \mathrm{pi}})\times 44÷12$	CStn,pi, CSn,pi: carbon storage and sink of tree i in year n (kgCO2e); ra(i) and rb(i): the aboveground and belowground carbon content ratios of tree i; Ni: number of individuals of tree i.	[34]
Annual carbon sink of other elements	${\mathrm{CS}}_{\mathrm{m}}={\mathrm{S}}_{\mathrm{m}}\times {\mathsf{\gamma }}_{\mathrm{m}}$	${\mathrm{CS}}_m: \mathrm{annual} \sin \mathrm{k} \mathrm{of} \mathrm{other} \mathrm{elements} \left({\mathrm{kgCO}}_2\mathrm{e}\right); m: \mathrm{the} \mathrm{natural} \mathrm{carbon} \sin \mathrm{k} \mathrm{type}; {\mathsf{\gamma }}_{\mathrm{m}}$: absorption factor for type m.	[5]
Total life cycle carbon sink	$\mathrm{CS}={\displaystyle {\displaystyle \sum }_{\mathrm{i}=1}^{\mathrm{J}}}{\displaystyle {\displaystyle \sum }_{\mathrm{n}=1}^{\mathrm{T}}}{\mathrm{CS}}_{\mathrm{n},\mathrm{pi}}+{\displaystyle {\displaystyle \sum }_{\mathrm{m}=1}^{\mathrm{M}}}(\mathrm{T}\times {\mathrm{CS}}_{\mathrm{m}})$	CS: total carbon sink (kgCO2e); J: the number of tree species; T: the life cycle time (years); M: the number of other sink types.	[5]

). Parameterization details are given in Supplementary Materials Tables S1–S3.

2.2.2. Carbon Emission Estimation Model

Following the LCA principles, this study establishes a comprehensive carbon emission inventory covering five stages: material production, transportation, construction, operation and maintenance, and renewal. Emissions are calculated using formulas based on material consumption, transport distance, construction energy use, and maintenance frequency (

Table 2. Carbon emission estimation models and formulas.

Estimation Content	Formula	Parameter Explanation	Carbon Source Scope	References
Production phase emissions	${\mathrm{CE}}_{\mathrm{SC}}={\displaystyle {\displaystyle \sum }_{\mathrm{i}=1}^{\mathrm{n}}}{\mathrm{M}}_{\mathrm{i}}{\mathrm{F}}_{\mathrm{i}}$	CSSC: carbon emissions in production (kgCO2e); Mi: consumption of material i (kg); Fi: emission factor of material i (kgCO2e/unit).	Granite, permeable concrete, asphalt, and other landscaping materials	[35,36]
Transport & construction emissions	${\mathrm{CE}}_{\mathrm{YS}}={\displaystyle {\displaystyle \sum }_{\mathrm{i}=1}^{\mathrm{n}}}{\mathrm{M}}_{\mathrm{i}}{\mathrm{D}}_{\mathrm{i}}{\mathrm{T}}_{\mathrm{i}}$$=\mathrm{G}\times {\mathrm{CE}}_{\mathrm{SC}}$${\mathrm{CE}}_{\mathrm{JZ}}={\displaystyle {\displaystyle \sum }_{\mathrm{i}=1}^{\mathrm{n}}}{\mathrm{B}}_{\mathrm{i}}{\mathrm{K}}_{\mathrm{i}}{\mathrm{NF}}_{\mathrm{i}}$$=\mathrm{P}\times {\mathrm{CE}}_{\mathrm{SC}}$	CEYS, CEJZ: emissions from transport and construction (kgCO2e); Di: transport distance for material i (km); Ti: transport emission factor [kgCO2e/(t·km)]; Bi: operating hours (machine-shift) of construction; Ki: energy consumed per unit; NFi: carbon emission factor of energy used by equipment i; G, P: proportions of transport and construction emissions relative to CSSC (G: 2%–6%, P: 5%–10%).	Seedling transport, pit excavation, planting, soil transport, and landscaping materials transportation and construction.
Operation & maintenance emissions	${\mathrm{CE}}_{\mathrm{GH}}={\displaystyle {\displaystyle \sum }_{\mathrm{i}=1}^{\mathrm{n}}}{\mathrm{E}}_{\mathrm{i}}{\mathrm{EF}}_{\mathrm{i}}$	CEGH: carbon emissions during operation and maintenance (kgCO2e); Ei: energy consumption of type i; EFi: carbon emission factor of energy i.	Lighting systems, and green space maintenance.
Renewal phase emissions	${\mathrm{CE}}_{\mathrm{GX}}$$={\mathrm{R}}_{\mathrm{i}}\times ({\mathrm{CE}}_{{\mathrm{SC}}_{\mathrm{i}}}+{\mathrm{CE}}_{{\mathrm{YS}}_{\mathrm{i}}}$$+{\mathrm{CE}}_{{\mathrm{JZ}}_{\mathrm{i}}})$	CEGX: carbon emissions during the renewal stage (kgCO2e); Ri: renewal ratio of element i.	Replacement of paving materials and vegetation.
Total life cycle emissions	$\mathrm{CE}$$={\mathrm{CE}}_{\mathrm{SC}}+{\mathrm{CE}}_{\mathrm{YS}}+{\mathrm{CE}}_{\mathrm{JZ}}$$+{\mathrm{CE}}_{\mathrm{GH}}+{\mathrm{CE}}_{\mathrm{GX}}$	CE: total life cycle carbon emissions (kgCO2e).

). Emissions related to labor, interior construction, building operations and demolition, and the construction and demolition of clean energy systems are excluded from the accounting scope of this study. Details are provided in Tables S4–S8 of Supplementary Materials.

2.2.3. Carbon Sink Efficiency Metrics

To systematically evaluate the carbon flux characteristics of urban green space units, four quantitative indicators are developed: unit carbon sink (UCS), unit carbon emission (UCE), unit net carbon sink efficiency (UNCSE), and carbon equilibrium time (CET) (

Table 3. Carbon sink effectiveness metrics.

Indicators	Formula	Indicator Description
Unit Carbon Sink	$\mathrm{UCS}={\displaystyle \frac{\mathrm{CS}}{\mathrm{A}}}$	Total carbon sequestered per unit area over the life cycle (kgCO2e/m2).
Unit Carbon Emission	$\mathrm{UCE}={\displaystyle \frac{\mathrm{CE}}{\mathrm{A}}}$	Total carbon emissions per unit area over the life cycle (kgCO2e/m2).
Unit net carbon sink efficiency	$\mathrm{UNCSE}={\displaystyle \frac{\mathrm{CS}-\mathrm{CE}}{\mathrm{A}\times \mathrm{T}}}$	Net annual carbon sink per unit area, representing carbon balance performance (kgCO2e/(m2·year)).
Carbon equilibrium time	$\mathrm{CET}=\min \left\{\mathrm{t}\in \left[0,\mathrm{T}\right]|\mathrm{CS}\left(\mathrm{t}\right)\ge \mathrm{CE}\left(\mathrm{t}\right)\right\}$	Time required for cumulative sink to offset total emissions, indicating carbon neutrality potential (years).

Note. A denotes the area of the landscape spatial unit (m²), and CS(t) and CE(t) denote the cumulative carbon sink and cumulative carbon emission up to year t, respectively.

). UCS and UCE reflect the fundamental carbon flux capacities of each spatial structure. UNCSE as a comprehensive indicator to assess the annual average net carbon sink capacity per unit area, can establish a unified carbon sink efficiency benchmark for different structural units. CET is used to determine whether a spatial unit can achieve carbon neutrality within its defined life span (T). If CET < T, the unit is considered to have carbon balancing potential. This indicator system ensures both horizontal comparability and vertical traceability, supporting performance benchmarking and optimization across diverse spatial configurations.

2.3. Performance Evaluation of Carbon Sink Efficiency

To systematically assess the carbon sink performance of various LSUs and identify their underlying drivers, this study conducts cluster analysis and correlation analysis based on previously established LSU typology and life cycle carbon flux estimates.

First, based on the carbon sink effectiveness index, K-means clustering is performed in IBM SPSS Statistics 27 to classify LSUs with similar performance profiles. Next, to identify the principal factors influencing carbon sink effectiveness, UNCSE is treated as the response variable, while six spatial structural indicators are selected as explanatory variables: Green Coverage Ratio (GCR), Planting Density (PD), Impervious Surface Ratio (ISR), Vertical Structural Complexity (VSC), Water Surface Ratio (WSR), and Building Area (BA). The calculation methods are detailed in Supplementary Materials Table S9. These variables are chosen based on their relevance to spatial configuration and carbon flux dynamics, as supported by previous studies [22,37,38], and their consistency with the dual-dimensional LSU framework developed in this study. They capture key horizontal and vertical structural attributes affecting biomass accumulation, photosynthetic potential, and blue-green interactions. Given that multiple variables deviate from normal distribution and the structure-performance relationships are anticipated to be monotonic rather than strictly linear, Spearman’s rank correlation analysis was employed to examine the associations between structural factors and UNCSE. This nonparametric method demonstrates greater robustness with the present data characteristics and provides conservative estimates of both the strength and direction of relationships. The analysis enables systematic investigation of statistical relationships between these structural factors and carbon performance, thereby facilitating the identification of dominant driving factors.

Finally, by integrating the results of clustering and correlation analyses, the study summarizes the primary structural drivers across LSU types. This reveals how spatial configuration contributes to variation in carbon sink capacity and provides a methodological foundation for spatial optimization and targeted low-carbon design strategies.

2.4. Study Area and Sample Delineation

To evaluate the carbon sink efficacy and applicability of the proposed LSUs in urban green spaces, this study selects Luhe Park in Nanjing, China, as the empirical case area (Figure 4), Located in the northern part of the city, Luhe Park covers 6.6 hectares and serves a mixed-use urban fabric comprising residential, commercial, and educational functions. As a newly developed zone guided by green and low-carbon planning principles, it features diverse spatial configurations and high structural heterogeneity, making it well-suited for testing LSU adaptability and carbon performance differentiation.

This study utilizes detailed CAD-based planning and design documentation as the primary data source. These include information on vegetation composition, paving materials, and infrastructure layout, allowing for precise spatial unit delineation and accurate modeling of carbon flux variables. Such high-resolution input data facilitates full life cycle simulations across structure, process, and performance dimensions. The park also incorporates sustainable infrastructure features such as permeable pavements, rain gardens, and photovoltaic panels, which contribute to stormwater resource utilization and clean energy generation. These facilities are considered in the life cycle carbon accounting to reflect emission reductions from nature-based and renewable systems.

A uniform 25 m × 25 m grid sampling method was employed to divide the study area into standardized units. After excluding cells with less than 50% effective coverage, a total of 108 LSU sample units were retained. These samples provide the basis for unit classification, carbon flux calculation, clustering analysis, and spatial performance optimization.

3. Results

3.1. Spatial Distribution Characteristics of Landscape Spatial Units

A total of 108 sampling units within the study area were identified and classified into 22 types of LSUs (Figure 5).

The classification results demonstrate that composite multi-layer types, such as M1, M3, and M6, predominate within the spatial structure, accounting for approximately 40% of the sampled units. These spatial units are mainly concentrated along the central green corridors and in large green nodes, suggesting that spatial units with high vegetation coverage and complex vertical structure constitute the dominant matrix of the park. In comparison, hardscape-dominated types, such as M12 and M14, are primarily distributed at plaza entrances and functional intersections. Despite their spatial constraints, these areas play a substantial role in enhancing circulation and organizing functional landscape elements. Landscapes dominated by single-layer herbaceous vegetation, including M5, M10, M15, and M20, are frequently observed in lawn activity zones and along road perimeters, often displaying a fragmented and patch-like distribution pattern.

Furthermore, M21 and M22 exhibit distinct spatial distribution characteristics. M21, representing isolated waterbodies, is concentrated in the central waterfront zone, integrating visual aesthetics with ecological functions. M22 corresponds to vertical greening systems and is primarily situated on roof-covered structures such as parking facilities and public service buildings, reflecting the implementation of three-dimensional greening strategies in an urban ecological design framework.

Overall, the spatial configuration of the study area presents a structurally diverse composition. Multilayer green units constitute the dominant spatial matrix, hardscape functional zones are interspersed across the site, and special units such as waterbodies and vertical greening function as ecological modulators and complementary elements.

3.2. Characterization of Carbon Flux Based on LCA

To identify the carbon flux response patterns of LSUs across the full life cycle, this study conducts a systematic assessment from the dual dimensions of carbon sink and carbon emissions. The analysis focuses on understanding the compositional characteristics of carbon fluxes across different LSU structures and their associated variations in carbon sink effectiveness, thereby revealing the process mechanisms by which spatial configurations influence carbon performance.

3.2.1. Temporal Dynamics and Compositional Analysis of Carbon Fluxes

As indicated by the simulation results, the total carbon sink of the park over a 50-year life cycle is 3,923,030.74 kgCO₂e, whereas the total carbon emissions amount to 2,063,841.22 kgCO₂e, resulting in a net carbon sink of 1,859,189.52 kgCO₂e. The results suggest that the current spatial structure may have a positive potential for carbon sink.

Distinct temporal patterns emerge in the cumulative spatial distribution of carbon emissions and sinks across over the 108 LSUs (Figure 6). Carbon sink increases over time, following a trajectory of overall expansion, localized intensification, and eventual stabilization. High-sink zones, primarily corresponding to multi-layered green units with tree-shrub structures, become prominent after year 30. In contrast, the carbon emission distribution reveals concentrated peaks during the production and construction phases, with high emission values predominantly occurring between years 10 and 20. Over time, emission intensity gradually stabilizes, entering a stage dominated by operation and maintenance emissions. Certain hardened or construction-intensive units consistently exhibit elevated emission levels throughout the life cycle. The temporal divergence between emission and sink patterns is further illustrated by spatial color shifts across LSUs. While high-rigidity, high-density built units exhibit persistent emission pressure and delayed carbon balance, composite green structures demonstrate accelerated carbon accumulation in the mid-to-late stages. These findings highlight the systematic influence of structural configuration on life cycle carbon performance.

Regarding carbon sink composition (

Table 4. Composition of life cycle carbon sink in Luhe Park.

Carbon Sink Types	Cumulative Carbon Sink
	10 Years/(kgCO2e)	20 Years/(kgCO2e)	30 Years/(kgCO2e)	40 Years/(kgCO2e)	50 Years/(kgCO2e)	Average Annual/(kgCO2e/Year)
Trees	289,615.68	766,170.09	1,335,726.09	1,961,493.14	2,633,889.44	52,677.79
Shrubs	62,458.23	124,916.45	187,374.68	249,832.90	312,291.13	6245.82
Groundcover	108,272.86	216,545.73	324,818.59	433,091.46	541,364.32	10,827.29
Waterbodies	71,058.25	142,116.50	213,174.74	284,232.99	355,291.24	7105.82
Soil	16,038.92	32,077.84	48,116.76	64,155.69	80,194.61	1603.89
Total	547,443.94	1,281,826.61	2,109,210.86	2,992,806.18	3,923,030.74	78,460.61

), trees account for 67.1%, serving as the primary contributors to carbon accumulation during the middle and late stages of the life cycle. Groundcover and shrubs contribute 13.8% and 8.0%, respectively, forming the foundation for carbon sink in the early stages. Although waterbodies and soils contribute less overall, their carbon sink remains stable over the long term. These results indicate that the carbon sink contribution is closely linked to the type and quantity of green space elements.

In terms of carbon emissions, the production phase contributes the largest share, followed by the operation and maintenance phase, while the renewal phase accounts for a comparatively minor portion (

Table 5. Composition of life cycle carbon emissions in Luhe Park.

Lifecycle Stage	Carbon Emission Sources		Total Emissions (kgCO2e)	Proportion of Total
Material production	Stage total		1,184,455.94	57.39%
	Material manufacturing		1,187,411.91
	Recycled/renewable materials		−2955.98
Transportation and construction	Stage total		98,148.17	4.76%
	Transportation of hardscape and building materials		23,689.12
	Earthwork transportation		14,432.69
	Hardscape and building construction		59,222.80
	Nursery stock transport and planting		803.52
Operation and maintenance	Stage total		777,080.44	37.65%
	Lighting systems		2,529,166.65
	Plant pruning		24,496.13
	Fertilizer application		19,255.24
	Pesticide application		16,311.12
	Irrigation water use		276,654.81
	Smart irrigation equipment		2459.13
	Rainwater utilization	Lake storage	−42,364.53
		Sunken green spaces	−20,788.01
		Permeable paving infiltration	−75,110.89
	Clean energy use	Solar energy	−1,869,165.09
		Wind energy	−83,834.10
Renewal and Replacement Stage	Stage total		4156.67	0.20%
	Road surface renewal		4132.57
	Vegetation replacement		24.11
Total			2,063,841.22	100%

). The primary sources of operational emissions include lighting systems and plant irrigation. However, the integration of recycled/renewable materials, rainwater utilization, and clean energy use results in an overall emission reduction efficiency of 50.37% across the entire life cycle, highlighting the effectiveness of these strategies in regulating carbon outputs. It should be noted that, within the defined system boundary, the embodied emissions from the manufacturing and construction of clean energy facilities such as photovoltaic panels and small wind turbines were not included. Consequently, the reported emission reductions reflect operational offsets rather than full life-cycle impacts of these systems. Subsequent consideration of the embedded carbon emissions within the energy system may reduce the net carbon benefits of these technologies.

3.2.2. Comparison of Carbon Sink Performance Indicators Across Spatial Units

To further assess the carbon sink performance of different spatial units, this study calculates and compares four key indicators: UCS, UCE, UNCSE, and CET. These indicators are visualized for all 22 LSU types using a column-normalized heatmap (Figure 7) to reveal performance variations.

The results reveal substantial variation across unit types. In terms of UCS, the highest value is observed in M3 (129.01 kgCO₂e/m²), while the lowest occurs in M15 (10.88 kgCO₂e/m²), underscoring the critical role of vegetation configuration in carbon sink accumulation. For UCE, M22 exhibits the highest unit carbon emissions (159.15 kgCO₂e/m²), whereas M21 (6.42 kgCO₂e/m²) and similar low-intervention units perform better due to limited construction and minimal management input. Regarding UNCSE, M3 achieves the best performance (2.19 kgCO₂e/m²·yr), followed by M1, M2, M6, and M8, all of which are characterized by structurally complex, vegetation-rich configurations. Conversely, M22 and M15 exhibited negative UNCSE due to low green coverage or high building density, with M15 unable to achieve carbon neutrality throughout its lifecycle. This further demonstrates the positive role of vertical greening in accelerating carbon balance. Additionally, approximately 64% of LSUs achieve carbon balance within 15 years, with multi-layered green types like M1–M3 demonstrating strong mid-to-long-term sequestration potential. In contrast, hardscape-dominated or high-density built-up units require 30–50 years or more to offset initial carbon emissions. These findings highlight that improving carbon sink capacity alone is insufficient, and controlling emissions is equally vital for achieving carbon neutrality.

3.3. Efficiency Clustering Assessment and Dominant Mechanism Analysis

To uncover typological variations in carbon sink performance and explore their structural determinants, this study applies clustering and correlation analysis methods.

3.3.1. Identification of Carbon Sink Efficiency Clustered Types

Due to the structural heterogeneity of M22 and its consistent emergence as a standalone cluster in preliminary K-means tests (k = 2–4), we treated M22 as an a priori category to avoid forcing heterogeneous processes into a common partition and to improve typological interpretability. K-means was then applied to the remaining ground units (M1–M21). For these units, we standardized UCS, UCE, UNCSE, and CET, and used the elbow method, yielding K = 3 with a silhouette coefficient of 0.6593. This indicates satisfactory clustering quality with reasonable separation. Combining these 3 clusters with the a priori M22 category produced 4 carbon-efficiency types (Figure 8). When M22 was reintroduced into a unified K-means run with k = 4, it again formed a single cluster alongside the Type I–III groupings, confirming that the clustering structure is stable regardless of whether M22 is treated as a prior category. Summary statistics for each cluster are presented in

Table 6. Carbon sink performance metrics for different cluster types.

Cluster Number	Number of Modes	UCS (kgCO2e/m2)	UCE (kgCO2e/m2)	UNCSE (kgCO2e/m2·Year)	CET (Years)
Type I	3	123.90 ± 6.80	18.70 ± 0.66	2.10 ± 0.13	6–10
Type II	6	58.20 ± 10.07	15.72 ± 2.69	0.85 ± 0.17	11–15
Type III	12	24.46 ± 7.48	13.87 ± 4.64	0.21 ± 0.20	16–20
Type IV	1	59.77 ± 0.00	159.15 ± 0.00	−1.99 ± 0.00	46–50

Note. (1) UCS, UCE, and UNCSE are expressed as mean ± standard deviation (Mean ± SD). (2) Type IV contains only one sample, resulting in a standard deviation of 0.00 for all indicators. (3) CET is presented as a time interval derived from life cycle carbon flux simulation instead of Mean ± SD, due to its discrete and phased characteristics.

, including mean values and standard deviations for the key performance indicators.

• Type I. High-efficiency units: high sink, low emission, rapid balance

This cluster exhibits the highest UNCSE (2.10 kgCO₂e/m²·year) with relatively low emissions, achieving carbon balance within 10 years. It includes LSU types M1–M3, characterized by high vegetation coverage, multilayer vertical structure, and minimal built elements. These units represent optimal configurations for carbon performance.

• Type II. Medium-efficiency units: moderate sink, low emission, mid-term balance

Type II features a cluster-average UNCSE of 0.85 kgCO₂e/m²·year and reaches carbon balance within 11–15 years. It includes LSU types such as M6, M7, and M8, which are characterized by moderately complex vegetation structures and limited hardscape intrusion. These units represent a structurally optimized configuration that balances ecological function and spatial vitality.

• Type III. Low-efficiency units: low sink, moderate emission, long-term type

This cluster contains the largest number of LSU types, including M4, M5, and M9, representing common green space configurations with fragmented or simplified vegetation structures. These units show relatively low carbon sink capacity and require more than 15 years to achieve carbon balance, indicating limited carbon performance under current configurations.

• Type IV. Negative-efficiency units: moderate sink, high emission, delayed balance

Type IV is represented solely by M22, which exhibits moderate sink potential but extremely high emissions, resulting in negative net carbon performance. Due to substantial material input and structural dependency, this type is highly likely to achieve carbon neutrality within its 50-year life cycle and poses challenges for sustainable implementation.

Overall, the clustering results reveal that LSUs with multilayer vegetation and minimal construction disturbance tend to deliver higher carbon efficiency and quicker balance, whereas units with extensive impervious surfaces or intensive construction show lower or even negative net performance. These findings support the use of differentiated planning strategies and typology-specific design interventions for enhancing carbon sink outcomes in urban green spaces.

3.3.2. Analysis of Factors Influencing Carbon Sink Efficiency

To avoid potential interference caused by differing carbon cycle mechanisms, the M21 samples, which represent pure waterbody units, were excluded from the analysis. Spearman correlation tests were then conducted separately for 82 standard ground-level green space samples and 13 M22 vertical greening samples, aiming to identify the key structural factors influencing carbon sink efficiency (Figure 9).

In ground-level green spaces, carbon sink efficiency showed significant correlations with multiple structural variables. Combining the correlation coefficient and p-value results (see Table S10 in Supplementary Materials for details), it can be observed that VSC, GCR and PD exhibit significant correlations. The PD was the most influential positive factor, indicating that dense vegetation arrangements effectively enhance carbon accumulation. Green coverage ratio and vertical structural complexity were also positively correlated, suggesting that continuous vegetation coverage and stratified vertical structure are beneficial for long-term sink potential. In contrast, impervious surface ratio and water surface ratio exhibited negative correlations, implying that excessive non-vegetated or aquatic surfaces could weaken carbon sequestration capacity. Additionally, a notable negative correlation was observed between water area and green coverage, suggesting a spatial trade-off between blue and green infrastructure under land constraints.

For the M22 units, building area had a strongly negative correlation with carbon sink efficiency, making it the dominant limiting factor. Conventional green space metrics such as planting density and green coverage showed minimal influence, highlighting the overwhelming impact of built form in vertical greening systems.

The clustering relationships of spatial structure factors in Figure 9 help clarify the associations between various factors. These structural factors can be categorized into 3 groups: vegetation-related factors, building-related environment factors, and other land cover factors. Within each functional category, the factors exhibit consistent effects on carbon sink efficiency. Based on these relationships, differentiated design strategies can be developed for optimizing carbon sink efficiency.

4. Discussion

4.1. Mechanisms of Pattern Structure Factors on Carbon Sink Efficiency

This study demonstrates that spatial structure systematically influences the carbon sequestration efficacy of urban green spaces. Within ground-level LSUs (M1–M20), planting density plays a key role in governing biomass accumulation per unit area. Moderate increases in density raise above-ground biomass, thus enhancing carbon uptake. However, excessive crowding can lead to competition and mortality among plants, which erodes the net gains in carbon sequestration over time [39,40]. The green coverage ratio affects carbon sink capacity by defining the extent of vegetated surfaces. When coverage is low to moderate, expanding green area enhances storage; once coverage is ample, gains depend more on structural quality than on further expansion. This aligns with the findings of Sun et al., who demonstrated that green areas with coherent distribution tend to have higher carbon sequestration performance [38]. Vertical complexity reflects the volume of photosynthetic tissues and the layering of functional strata. Maintaining these variables, LSU types with tree-shrub-grass structures, such as M1, M6, and M11, demonstrate the highest carbon sequestration capacity, which corroborates Fan’s research [20].

Hard surfaces and waterbodies further impact LSU efficiency. A higher impervious surface ratio reduces available space for vegetation and soil carbon storage, and may alter microclimatic conditions, limiting the overall carbon sink capacity [41]. In contrast, blue-green space interactions can help optimize natural irrigation, regulate the microclimate, and support long-term ecological functions. Research by Probst, Yuan, and others shows that highly aggregated and well-connected blue-green spaces contribute significantly to carbon sequestration [42,43]. This highlights the strategic importance of designing blue-green spaces to enhance carbon sink benefits.

Vertical greening systems (VGS) are subject to various environmental conditions that can either directly or indirectly affect their carbon sequestration potential. Building constraints on substrate depth, load capacity, and water supply often lead to limited rooting zones and water stress, which suppress net primary productivity and reduce biomass accumulation per unit area [44]. Plant choice is further constrained toward drought-tolerant, shallow-rooted, small-crown taxa, reducing long-term carbon sink potential. Nevertheless, vertical greening can influence building microclimate and energy consumption through boundary layer disturbance and surface albedo changes, thereby indirectly affecting carbon budgets. The magnitude of this effect depends on factors such as orientation, building envelope construction, climate zone, and vegetation coverage [45]. Therefore, the carbon sequestration function of vertical greening is limited by system boundary conditions. As Rowe emphasizes, standardizing boundary conditions, data inputs, and incorporating the potential benefits of VGS into LCA are essential for improving the comparability and scientific rigor of VGS research [46].

When compared with studies of other parks in different regions [10,13,39], the structure-performance relationships identified in this study show certain commonalities across different climates and urban environments. More continuous green spaces, moderate to high planting densities, and more complex vertical structures generally correspond to higher levels of carbon sequestration. In contrast, hardened or simplified structures show lower UCS, relatively higher UCE, and longer CET. Optimization strategies should be tailored to the structural attributes of landscape spatial units, focusing on planting density, vertical stratification, and spatial configuration.

4.2. Application and Optimization of Landscape Spatial Patterns

Cluster analysis distinguishes 4 LSU types with clear structural and performance contrasts. To convert diagnosis into design guidance, 95% confidence intervals (CIs) were computed under a consistent LCA boundary for six structural indicators—GCR, ISR, VSC, PD, WSR and BA. These CIs (Figure 10) serve as referential thresholds for targeted interventions and spatial optimization in areas with climatic conditions or green space configurations similar to those in the study region.

• Type I. High-efficiency units

These units exhibit the highest UNCSE and shortest CET. CI values include GCR between 87% and 93%, ISR between 7% and 15%, near-zero WSR, PD between 280 and 445 trees ha⁻¹, and VSC from 0.79 to 0.93. Such LSUs are suited for ecological core zones, green corridors, and large-scale parks where continuity, biomass accumulation, and multilayered vegetation structures are prioritized.

• Type II. Medium-efficiency units

Type II units combine sink performance with public use. CI values include GCR between 47% and 58%, PD of 163–230 trees ha⁻¹, ISR between 21% and 31%, VSC from 0.81 to 0.90, and WSR between 12% and 28%. Although partial overlap with Type I occurs, these units represent spatial mosaics that harmonize ecological function and public usability, particularly suited to neighborhood parks, linear greenways, and community greens [47]. Moderate planting density and layered canopy structure enhance light interception and leaf area index, which are key factors in above-ground growth [26]. As a result, Type II units maintain relatively high carbon sequestration benefits over a 50-year lifecycle.

• Type III. Low-efficiency units

These LSUs typically serve as lawns, plazas, or sparsely vegetated fields where visual openness and event accommodation are prioritized. Their CI ranges include GCR between 23% and 44%, PD between 22 and 65 trees ha⁻¹, ISR from 11% to 29%, VSC between 0.40 and 0.63, and WSR from 35% to 63%. Type III units, primarily composed of lawns, scattered trees, and interspersed with plazas, paths, or waterbodies, have limited canopy volume and photosynthetic surface area. Fragmented vegetation patches and simple vertical structures significantly reduce carbon sequestration [48]. Additionally, the large WSR further weakens carbon accumulation. These factors explain its lower UNCSE and longer CET compared to Type II.

• Type IV. Negative-efficiency units

Represented by M22, CI values indicate GCR ranges from 74% to 91%, PD from 4.78 to 11.69 trees ha⁻¹, ISR from 7% to 25%, VSC from 0.55 to 0.85, WSR of 0, and BA from 145–333 m² per cell. The enlarged building area, service loads and planting constraints are the dominant limits on UNCSE. Spatially couple vertical greening with adjacent high-efficiency ground-level units (Types I/II) to form local source-sink complexes and provide perimeter shading that reduces building cooling demand [49], while functioning as ecological retrofit nodes, green building landmarks or supplemental green volume in dense districts.

A comparative analysis of studies in different cities and climate zones further supports the results identified in this research. For example, Ou et al. in Taichung City, Taiwan Province found that parks with over 65% soft landscape and under 15% hard landscape show significant carbon reduction potential [50]. Jiang et al. found that CO₂ concentrations exhibited a marked upward trend once road coverage exceeded 30% in Shanghai urban parks [51], whereas in this study, the majority of ISR sites across the four categories remained below this threshold. Yi et al. discovered that plant communities with PD between 300 and 450 trees ha⁻¹ exhibited the highest carbon sequestration benefits in Tianjin urban parks [52], similar to Type I. Conversely, carbon sequestration benefits were lowest at PD below 150 trees ha⁻¹, a range encompassing both Type III and Type IV PDs. Additionally, Wang et al. proposed specific tree and shrub size thresholds for carbon sequestration in Beijing parks, further emphasizing the impact of plant specifications [40]. These results suggest that structural thresholds should be recalibrated based on local climate, vegetation types, and management policies.

In addition to structure, plant selection, material choice, and maintenance significantly affect carbon sink efficiency. To improve the carbon sink efficiency, priority should be given to selecting locally adapted and stress-tolerant species, as they tend to achieve higher and more stable carbon sequestration while reducing maintenance-related emissions. In Nanjing, suitable tree selections include Japanese zelkova, camphor tree, and Chinese quince, while low-maintenance evergreens like maidenhair and pearl plum are recommended for ground layers [53]. Second, life-cycle emissions can be reduced by using permeable pavements, smart lighting, and sensor-based irrigation. For example, substituting conventional cement concrete with recycled aggregate concrete may cut embodied emissions by 29% [54]. Finally, for Type IV units, using lightweight, high-recycled-content media and components together with passive rainwater harvesting, and electrified equipment helps minimize routine energy use and improve net performance [44].

Taken together, these measures provide a targeted pathway to raise park-wide UNCSE and shorten CET without altering site function or character. It is worth noting that many carbon-efficient spatial configurations also bring multiple co-benefits, including air purification, microclimate regulation, and the improvement of physical and mental health [55,56]. These broader ecological services should be considered alongside carbon performance, as carbon-efficient green spaces often serve as multi-functional infrastructure, enhancing the overall quality of urban life.

4.3. Uncertainty and Sensitivity Analysis

Uncertainty pervades all stages of LCA, including goal and scope definition, life cycle inventory analysis, and life cycle impact assessment [57]. To evaluate the robustness of life cycle carbon flux results and LSU clustering, sensitivity analysis was performed on three key parameter groups in the model:

(1)

Biomass carbon coefficients, including the carbon content ratios of aboveground and belowground biomass in trees (r_a, r_b), as well as non-tree carbon coefficients for shrubs, groundcover, waterbodies, and soil (γ_m);

(2)

Material production and construction emission factors, which refer to the emission factors associated with the production, transportation, and construction of major hardscape and structural materials (F_i, T_i, NF_i);

(3)

Operational and maintenance energy emission factors, which relate to the electricity and fuel consumption emissions associated with lighting, irrigation, maintenance, and clean energy (EF_i).

These three categories represent uncertainty related to carbon sink processes, embedded emissions from materials, and long-term operational emissions, respectively. For each factor, while keeping other parameters constant, a ±20% perturbation level was applied to reflect the reasonable range of uncertainty for biomass growth and emission factor parameters, as reported in the literature [57,58]. The model was then re-run for each scenario to recalculate the UNCSE and CET for both the park scale and the four clustering types (

Table 7. The sensitivity analysis result.

Scenario	Park Level		Cluster Level
	UNCSE Change	CET Change	Clustering Types	UNCSE Change	CET Change (Standardized to 5 Years)
Biogenic carbon coefficients	−42.20%~+38.49%	−18.60%~+24.03%	Type I	−23.41%~+21.38%	−50%~0
			Type II	−27.45%~+25.02%	0
			Type III	−45.71%~+46.38%	0~+50%
			Type IV	−11.04%~+11.89%	0~+10%
Material emission factors	−13.68%~+13.68%	−18.60%~+12.40%	Type I	−0.19%~+0.56%	−50%~0
			Type II	−2.23%~+2.11%	−33.33%~0
			Type III	−5.71%~+7.49%	0~+25%
			Type IV	−29.19%~+28.95%	0~+10%
Energy emission factors	−29.28%~+29.28%	−14.73%~+16.28%	Type I	−8.62%~+8.99%	−50%~0
			Type II	−19.22%~+19.1%	−33.33%~0
			Type III	−75.34%~+77.11%	−25%~+25%
			Type IV	−8.98%~+8.73%	0~+10%

Note. (1) Material emission factors refer to the emission factors associated with landscape and structural materials during their production, transportation, and construction phases. Energy emission factors pertain to emissions linked to electricity and fuel usage during the operational maintenance phase. (2) For each scenario, only the corresponding parameter set was perturbed by ±20%, while all other parameters were held constant. Changes in UNCSE and CET are expressed as percentage deviations from the baseline.

).

Overall, the sensitivity analysis indicates that the life cycle carbon accounting framework demonstrates strong robustness to reasonable uncertainties in parameters. At the park scale, biomass carbon coefficients had the greatest impact, followed by the energy-related emissions factors associated with operations, while material-related emissions factors had a relatively minor effect. This reflects that, over a 50-year period, the cumulative energy consumption from lighting, irrigation, and maintenance may exceed the one-time construction emissions, especially in urban parks with high-intensity management. The extent of this influence is also related to the system boundary settings of the study.

At the clustering mode level, the perturbation of uncertain factors mainly affected the magnitude of UNCSE, but did not alter the relative ranking of the performance of each type or the clustering assignments of the LSU samples. This indicates that the corresponding structural threshold patterns also remained unchanged. Despite uncertainties in parameters such as biomass carbon coefficients, construction material, and energy emission factors, Type I and Type II units belong to high-efficiency configurations, while Type III and Type IV units still require substantial optimization or compensatory measures. Notably, the biomass carbon coefficients have a relatively large impact on Type III and some low-efficiency modes, highlighting the higher sensitivity of near-neutral net carbon balance to parameter variations. For these units, optimizing structural attributes is more effective than simply improving parameter estimates. The influence of material and energy emission factors is particularly significant for Type IV and Type III units, which have higher embodied carbon emissions during construction and operational loads. Practically, this emphasizes the dual importance of both improving the local emission factor data quality and selecting low-carbon materials and energy-efficient equipment during design and management.

It should be noted that this analysis only considers parameter uncertainty from the three factor groups and assumes fixed structural configurations and growth trajectories. Other sources of uncertainty—such as future climate change, disturbance events, or management regime changes—were not explicitly incorporated into the model, which could further impact long-term carbon fluxes. Nonetheless, within the ±20% test range in this study, the core conclusions remain robust: (1) Land-use structure patterns have a dominant impact on carbon sink efficiency; (2) Type I and Type II units consistently outperform Type III and Type IV units; (3) Optimizing vegetation structure and operational maintenance is crucial for maximizing the net carbon benefits of urban park design.

4.4. Limitations and Prospects

This study successfully integrates LSU modeling with life cycle carbon flux simulation, bridging carbon accounting with spatial configuration, and introduces LCA methods to capture the dynamic characteristics of urban carbon sinks. The conceptual framework is adaptable to other parks and urban areas, providing new insights into the relationship between spatial structure and carbon sequestration. However, several limitations must be acknowledged.

First, due to climate and geographical constraints, the findings are primarily applicable to urban green space configurations similar to those in Nanjing’s climate. The lack of cross-city and cross-climatic validation limits the generalizability of the conclusions. Future research should expand to different cities, climate zones, and urban environments, establishing localized carbon factor databases to assess the transferability and robustness of the findings.

Second, while the study incorporates literature-derived and empirical estimates for carbon sink and emission parameters, it does not fully capture the dynamic interactions between vegetation growth, soil carbon, and anthropogenic inputs, such as fertilization, irrigation, and green space management. These complex, nonlinear feedback mechanisms are critical for accurately modeling carbon fluxes in urban areas. The current approach mainly focuses on vegetation biomass and its direct contribution to carbon sequestration, alongside construction- and operation-related emissions. Future research should integrate process-based models that explicitly account for soil carbon dynamics and management-induced feedbacks, providing a more comprehensive understanding of urban carbon cycling.

Additionally, the existing evaluation framework focuses mainly on carbon sequestration and does not fully account for other key ecological functions or the influence of user behaviors. To gain a holistic understanding of the multifunctionality of urban parks, future studies should include ecosystem services such as microclimate regulation, biodiversity, and health benefits in a multi-objective evaluation framework. This would extend the assessment beyond carbon performance to encompass other essential ecological services.

Finally, developing a decision-support platform for urban green space planning, design, and management would be a valuable next step. Such a platform could integrate spatial structure analysis, carbon sink performance evaluation, and spatial optimization strategies, promoting the multi-scale application of green infrastructure development and renewal.

In conclusion, future research should address these gaps and explore additional dimensions of urban parks’ ecological functions, enhancing the adaptability of the model and expanding its utility across diverse urban contexts.

5. Conclusions

This study develops a structure-sensitive framework that links a dual-dimensional LSU typology with life-cycle carbon flux modeling to evaluate and improve the carbon performance of urban green spaces. Using Luhe Park in Nanjing as a case, we quantified sequestration, emissions, and net efficiency for 22 unit types over 50 years and translated results into design-oriented thresholds and strategies. The main conclusions are

• Structure matters for long-term carbon performance. Coupling LSU classification with LCA revealed substantial heterogeneity among 22 types. Units combining high green coverage with multilayer vegetation consistently outperformed hardened or construction-intensive types, confirming the need for structure-aware assessment.
• A compact indicator set enables comparable evaluation. The metrics UCS, UCE, UNCSE, and CET provide a coherent basis for benchmarking units. High-efficiency types such as M1–M3 showed higher UNCSE and shorter CET, while vertical greening with large built surfaces (M22) exhibited low or negative net performance within the assessed life cycle.
• Four robust carbon-efficiency typologies were identified. Treating M22 as a prior category and clustering M1–M21 yielded three ground-level types; recombining with M22 produced four stable classes. CI-based ranges for GCR, ISR, VSC, PD, WSR, and BA align with the ordering of UNCSE and CET, supporting the typology’s interpretability and use in design targets.
• Actionable, type-specific optimization is feasible. Design guidance derived from CI ranges indicates that high-efficiency units should maintain continuity and multilayered vegetation structures to sustain stable carbon sinks, while balanced units can enhance spatial integration and adopt low-carbon surface materials to lower life-cycle emissions. Open spaces may incorporate localized edge or node greening with permeable pavements to cut embodied carbon while preserving openness. Vertical greening should minimize structural load and maintenance through lightweight components, drought-tolerant species, and passive rainwater systems, linking facades with nearby high-sink ground units to boost overall performance.
• The framework is scalable to other parks and districts and supports planning decisions by linking structure-process-performance. Future work should expand localized carbon-factor databases, incorporate additional ecosystem services, and extend uncertainty/sensitivity analyses to further strengthen transferability and policy relevance.