Quantifying and Optimizing Vegetation Carbon Storage in Building-Attached Green Spaces for Sustainable Urban Development

Abstract

Public building-attached green spaces are increasingly important urban carbon sinks, yet their carbon sequestration potential remains poorly understood and underutilized. This study quantified vegetation carbon storage across three attached green space typologies (green square, roof garden, and sunken courtyard) at a representative public building in Wuhan, China, using field surveys and species-specific allometric equations. Total carbon storage reached 19,873.43 kg C, dominated by the green square (84.98%), followed by a roof garden (12.29%) and sunken courtyard (2.72%). Regression analysis revealed strong correlations between carbon storage and morphological traits, with diameter at breast height (DBH) showing the highest predictive power for trees (r = 0.976 for evergreen, 0.821 for deciduous), while crown diameter (CD) best predicted shrub carbon storage (r = 0.833). Plant configuration optimization strategies were developed through correlation analysis and ecological principles, including replacing low carbon sequestering species with high carbon native species, enhancing vertical stratification, and implementing multi-layered planting. These strategies increased total carbon storage by 131.5% to 45,964.00 kg C, with carbon density rising from 2.00 kg C∙m⁻² to 4.63 kg C∙m⁻². The findings provide a quantitative framework and practical strategies for integrating carbon management into the design of building-attached green spaces, supporting climate-responsive urban planning and advancing sustainable development goals.

1. Introduction

Urban green spaces serve as vital natural carbon sinks, playing a crucial role in mitigating urban carbon emissions that currently account for over 70% of global CO₂ emissions [1]. In China, rapid urbanization—exceeding 60%—generates approximately 3 billion additional tons of carbon emissions per 10% increase in urbanization rate [2], underscoring the urgent need to enhance urban green space carbon storage capacity to achieve carbon neutrality and improve environmental governance. Among various urban green space types, affiliated green spaces attached to developed urban land constitute 30–60% of total urban greenery and function as effective carbon reservoirs [3,4,5]. China’s urban expansion has driven substantial growth in public building infrastructure, with total floor area increasing from 7.8 to 15 billion square meters from 2010 to 2021, representing 6.04% annual growth [6]. Accompanying this construction boom, public building-attached green spaces have expanded proportionally, in compliance with China’s 1994 urban greening policy mandating that 35% of new public construction land must be reserved for green space development [7]. Despite their emergence as key components of urban ecosystems, these public building-attached green spaces remain underutilized, failing to reach their full potential in emissions reduction, microclimate regulation, and ecological service provision. A major limiting factor is the prevalence of inefficient landscape designs dominated by ornamental shrubs with hard paving, which severely constrain carbon sequestration performance [8,9,10]. However, research indicates that optimized vegetation configurations could substantially boost carbon sink potential through strategic plant selection, improved community structure design, and enhanced species complementarity [11]. Implementing such refined planting strategies in public building-attached green spaces would not only enhance carbon storage through biological synergies and more effectively offset emissions, but also elevate overall ecological performance [12]. Therefore, estimating carbon storage and improving carbon sink functionality in these green spaces would provide critical support for urban carbon cycling and neutrality objectives while simultaneously strengthening climate resilience and advancing sustainable urban development.

Attached green spaces function as vital ecological connectors within urban construction lands, complementing independent green spaces to address the spatial limitations of green infrastructure in dense cities [13,14]. These auxiliary vegetated zones provide measurable carbon sequestration, emission reduction benefits, and space-efficient advantages [15,16,17]. Recent studies have primarily investigated two aspects: carbon storage characteristics across types of attached green spaces and efficiency enhancement methodologies through optimized vegetation configuration. Firstly, there are significant differences between the different types of attached green spaces in terms of both carbon storage and carbon density. Chen et al. (2024) found industrial green spaces had higher carbon density than commercial ones, with vegetation structure and plant attributes being key determinants [14]. Tang et al. (2020) established a carbon storage estimation model and found residential attached green spaces had lower carbon density than agricultural or forest counterparts, revealing land use type significantly influences sequestration capacity [18]. Moreover, due to the limitation of land use type, the carbon storage of attached green spaces is generally lower than that of independent green space types. For example, Deng et al. (2023) found that the average carbon density of urban attached green space ecosystems was significantly lower than that of park green space and protective green space [19]. Xu et al. (2025) employed Sentinel-1/2A remote sensing data and RF modeling to estimate that attached green spaces contribute approximately 2.53% to the total carbon storage density when compared with park, protective, and production green spaces [20]. Given the relatively low carbon density observed in urban attached green spaces, optimizing vegetation configuration emerges as a crucial strategy for enhancing their carbon storage potential [21]. For example, Nowak et al. (2002) demonstrated a significant positive correlation between tree cover density and carbon absorption capacity through systematic structure–function analysis [22]. Sun (2019) further revealed significant correlations between green space carbon density and landscape structures such as green space patch coverage and aggregation degree [23]. Ma et al. (2021) developed a carbon sequestration model revealing that variations in plant community structure directly govern carbon sink capacity dynamics [24]. Recent work by Harris et al. (2022) pioneered a three-dimensional green volume index through integrated UAV and terrestrial laser scanning to quantify the relationship between vegetation spatial configuration and carbon sink performance, providing technical support for optimizing green space layout [17]. Separately, Zhang et al. (2024) demonstrated that optimizing plant community structure enhanced carbon sink capacity by quantifying the effect of plant community structure, including density and cover indicators, on carbon storage [16]. Up to now, studies have examined carbon sequestration in attached green spaces along streets, industrial zones, storage areas, residences, institutions, and cultural facilities [14,18,25]. However, research on carbon storage quantification and vegetation optimization in green spaces attached to public buildings still requires further exploration. We aim to provide some insights for the understanding of this field.

Carbon storage represents the accumulated carbon mass within an ecosystem’s carbon pools over time [26]. Its accurate quantification in green spaces is essential for analyzing carbon sink trends, identifying influencing factors, and developing enhancement strategies. Currently, four primary methods quantify vegetation carbon storage: plot surveys for field data, software simulations for modeling, remote sensing for spatial analysis, and assimilation techniques for flux measurements. The sample plot method estimates carbon storage change per unit area and time by establishing monitoring plots in representative forest areas with good growth conditions. Zhang et al. (2019) developed allometric growth equations using tree diameter at breast height, height, and age to quantify forest carbon sequestration capacity [27]. Singkran (2022) similarly employed biomass approaches to assess carbon sequestration in urban park green spaces [28]. This method remains predominant in forest carbon storage calculation due to its direct, explicit, and technically straightforward nature. However, it faces limitations including destructive sampling, discontinuous observation capabilities, and complex processing procedures. Conversely, software simulation utilizes specialized tools like Citygreen, the National Tree Benefit Calculator (NTBC), Pathfinder, and I-Tree Eco to model carbon sinks in urban green spaces and estimate sequestration potential. For instance, Li et al. (2022) employed the NTBC model to quantify carbon sequestration benefits in residential green spaces in Nanjing [29]. Meanwhile, the I-Tree Eco module has been used for ecological assessments across diverse regions such as the United States [30], the United Kingdom [31], Thailand [32], and Hungary [33]. These systems comprehensively estimate carbon sequestration by integrating vegetation characteristics, meteorological data, and local environmental parameters. A significant limitation arises because their underlying physiological, ecological, and climatic parameters are calibrated for U.S. conditions, potentially introducing substantial errors when applied to regions with divergent ecosystems or climates. Alternatively, the remote sensing method enables rapid, real-time, large-scale acquisition of vegetation parameters through satellite platforms to estimate carbon storage. Tang et al. (2020) developed a model using vegetation coverage data from satellite imagery to quantify aboveground carbon storage in urban green spaces [18]. Zhu et al. (2024) used parametrized forest carbon stock models to achieve high-resolution, large-scale mapping and global dynamic monitoring of forest carbon sinks through remote sensing technology [34]. Remote sensing methods for estimating carbon storage in urban green spaces face challenges due to spatial heterogeneity and temporal dynamics, including difficulties in accurately quantifying plant density and delineating overlapping tree canopies [35], while interspecific ecological variations introduce additional uncertainties that may lead to assessment errors. Finally, the assimilation method quantifies carbon sequestration by directly measuring instantaneous CO₂ flux in plant leaves using specialized instruments, deriving calculations from net photosynthetic rate and leaf area. This method provides accurate, plant-specific measurements, enabling direct comparisons of carbon sequestration capacity and benefits across species, as demonstrated in studies of garden plants in West Bengal [36], Rome [37], and Fuzhou [38]. Yang et al. (2024) [39] further applied this method to evaluate urban green space species’ carbon sink potential, supporting the selection of high sequestration trees in regions. The assimilation method’s results usually exhibit significant sensitivity to temporal and spatial factors due to inherent diurnal and seasonal fluctuations in leaf photosynthetic rates. From the above, these methods for estimating carbon storage in urban green spaces present distinct advantages and limitations. Considering these methodological constraints and our specific research objectives, this study adopts the plot survey method for carbon storage quantification.

This study quantitatively analyzed vegetation carbon storage capacity in attached green spaces of a public building in Wuhan. Comprehensive field surveys were first conducted to collect plant community data, systematically measuring and recording key morphological parameters for all vegetation within the green spaces. Carbon storage quantification was then used to compare sequestration benefits among green square, roof garden, and sunken courtyard areas, with correlation and regression analyses revealing relationships between plant traits and carbon capacity. Finally, by integrating empirical data analysis with ecological principles, this study established a comprehensive morphological parameter system for high carbon sequestration plants and developed practical, feasible optimization strategies. This study establishes a theoretical framework for assessing carbon storage in urban building-attached green spaces and provides practical guidance for designing high carbon sequestration plant communities with optimized structures, aligning with the emerging trend of integrating multiple ecological indicators into climate-adaptive urban green infrastructure planning [40].

2. Materials and Methods

2.1. Project Information and Data Source

The selected representative public building was newly completed in 2024 for mixed use and located in the Wuchang District of Wuhan, as illustrated in Figure 1. The complex integrates multiple functions, including government services, commercial offices, business facilities, community health services, and an information center. The 24,017 m² development features a 71.55 m tall structure with 16 aboveground floors and 4 underground floors. The site incorporates three meticulously designed green spaces: a 5707 m² ground-level public square fostering community interaction, a 3727 m² roof garden serving restorative space for office workers, and a 937 m² sunken courtyard providing intimate recreation. These attached green spaces collectively demonstrate innovative urban design by addressing diverse user needs while optimizing the ecological performance of a limited site area through vertical stratification of functional greenery.

This study employed a comprehensive field investigation approach combined with university–industry collaboration to collect original data on vegetation characteristics and carbon storage. Using the plot survey method, we systematically recorded all trees, shrubs, and herbs within the study area, measuring key growth parameters including species identification, DBH, tree height (H), and CD. These field measurements were complemented by official project documentation provided by the construction department, including detailed landscape plans, quantity surveys, and itemized pricing tables. All collected vegetation data, encompassing species composition, population counts, and morphological dimensions, were compiled into a comprehensive plant attribute database to support biomass estimation and carbon storage calculations. This methodological approach ensured reliable data acquisition, forming a solid foundation for subsequent biomass modeling and carbon quantification.

2.2. Quantifying Carbon Storage of Vegetation

The study quantified plant biomass and carbon storage across vegetation strata encompassing trees, shrubs, and herbaceous layers in green spaces attached to the typical public building. Total vegetation carbon storage was estimated with standardized allometric equations in conjunction with biometric data collected from field surveys. The carbon storage capacity of vegetation components within the attached green space ecosystem was systematically assessed through the following methodology.

2.2.1. Carbon Storage of Trees

The carbon storage of trees was determined using the methodology described in reference [41], specifically through the application of Equation (1). Individual tree biomass was first estimated using species-specific allometric equations based on field-measured dendrometric parameters (e.g., DBH, H). The biomass values were then converted to carbon storage by applying a standard carbon conversion factor, which accounts for the proportion of elemental carbon in plant tissues. This two-step calculation offers a quantification of tree carbon storage.

$$C_{pi}=W\times CF$$

(1)

where C_pi is the carbon storage of the i-th single tree and shrub, kg C; CF is the carbon conversion factor, which is uniformly calculated as 0.47 in this paper [42], C; W is the biomass of the i-th single tree and shrub, kg. For a single tree, its biomass can be estimated through allometric equations, and the corresponding biomass allometric models for distinct tree types are presented in

Table 1. Allometric equations for biomass calculation of different tree species.

Number	Tree Species	Calculation Formula	References
1	C. camphora	$W=0.937+0.037D^{2}H$	[43]
2	G. biloba	$W=-0.684+0.090D^{2}H$	[43]
3	K. paniculata	$W=0.915+0.100D^{2}H$	[43]
4	L. indica	$W=0.895+0.035D^{2}H$	[43]
5	A. palmatum “Atropurpureum”	$W=0.043{\left(D^{2}H\right)}^{0.994}$	[43]
6	M. grandiflora	$W=0.33079D^{1.90957}$	[44]
7	P. × fraseri	$W=0.805{\left(D^{2}H\right)}^{1.051}$	[45]
8	P. fortuneana	$W=0.6246{\left(D^{2}H\right)}^{0.8138}+0.1884{\left(D^{2}H\right)}^{1.1503}$	[45]
9	Z. serrata	$W=0.293\times 0.00666\times {\left({\displaystyle \frac{D}{2.54}}\right)}^{2.36318}\times \left(3.28H\right)0.5519$	[46]
10	C. sinensis	$W=0.207631{\left(D^{2}H\right)}^{0.61612}+0.58219{\left(D^{2}H\right)}^{0.85219}+0.039445{\left(D^{2}H\right)}^{0.65514}+0.117284{\left(D^{2}H\right)}^{0.58219}$	[47]
11	A. palmatum	$W=0.54D^{2.972}$	[48]
12	C. japonica	$\mathrm{W}=0.078{\left({\mathrm{B}\mathrm{D}}^2\mathrm{H}\right)}^{0.934}$	[48]
13	Evergreen trees	$W=0.2210D^{1.4}+0.1010D^{1.21}+0.0123D^{2.12}$	[49]
14	Deciduous tree	$W=0.2030D^{1.95}+0.0042D^{3.24}+0.00085D^{3.23}+0.00085D^{3.23}$	[49]
15	bamboo	$W=0.409759D^{1.0615}{H}^{0.5427}$	[50]

Notes: Hi and Di are the tree height and diameter at breast height of the i-th single tree, respectively.

.

2.2.2. Carbon Storage of Shrubs

The carbon storage of shrubs is calculated by multiplying their total biomass by the carbon conversion factor, following the methodology outlined in Formula 1. The total shrub biomass is determined by multiplying the established shrub biomass density of 1.976 kg∙m⁻² by the total vegetated area occupied by shrubs [51]. This approach ensures consistent quantification of shrub carbon storage across the study area while accounting for the known carbon fraction present in shrub biomass.

2.2.3. Herbaceous Carbon Storage

The carbon storage of herbaceous vegetation is calculated by multiplying the unit area carbon storage value of 0.709 kg C∙m⁻² by the total herbaceous coverage area [52]. This method is suitable for urban green spaces, where herbaceous plants exhibit relatively consistent characteristics.

2.3. Data Analysis

Using biomass estimation techniques and community-level calculation methods, this study quantified plant biomass and carbon storage across three types of public building-attached green spaces: the green square, sunken courtyard, and roof garden. A comprehensive vegetation database was constructed, supporting comparative analysis of plant community structure in Excel. Subsequent correlation and regression analyses in SPSS (version 27) identified key plant traits that influence carbon sequestration capacity. Statistical modeling yielded correlation rankings and predictive regression curves, which, when combined with ecological principles, urban design constraints, and practical implementation considerations, provide the basis for proposing evidence-based plant configuration optimization strategies to enhance the carbon storage capacity of attached green spaces.

3. Results and Discussion

3.1. Carbon Storage Estimation

The carbon storage capacities of the green square, roof garden, and sunken courtyard that are attached to the public building were quantified using the previously established methods. Results presented in

Table 2. Carbon storage in green square, roof garden, and sunken courtyard.

Regions	Items	Name	Numbers	Units	Total Biomass /(kg∙m−2)	Total Carbon Storage
						(kg C)	(kg CO2)
Green square	s1	C. camphora A	3	plant	567.95	266.93	978.76
	s2	C. camphora B	76	plant	8441.83	3967.66	14,548.09
	s3	M. grandiflora	6	plant	228.28	107.29	393.41
	s4	O. fragrans A	9	plant	208.12	97.81	358.65
	s5	O. fragrans B	30	plant	372.66	175.15	642.22
	s6	G. biloba	11	plant	4787.79	2250.26	8250.96
	s7	Z. serrata	20	plant	2.22	1.04	3.82
	s8	C. sinensis	10	plant	1194.11	561.23	2057.84
	s9	C. sinensis “Bush”	3	plant	307.74	144.64	530.34
	s10	K. paniculata A	3	plant	787.92	370.32	1357.85
	s11	K. paniculata B	23	plant	3474.64	1633.08	5987.96
	s12	T. sebifera	6	plant	515.51	242.29	888.40
	s13	P. subg. Cerasus sp. A	3	plant	118.96	55.91	205.00
	s14	P. subg. Cerasus sp. B	6	plant	94.14	44.25	162.24
	s15	E. japonica	3	plant	37.27	17.52	64.22
	s16	M. × soulangeana	25	plant	648.51	304.80	1117.59
	s17	A. palmatum “Atropurpureum”	1	plant	11.13	5.23	19.19
	s18	L. indica	7	plant	58.00	27.26	99.95
	s19	P. cerasifera “Atropurpurea”	13	plant	203.98	95.87	351.52
	s20	A. palmatum A	7	plant	86.94	40.86	149.83
	s21	A. palmatum B	19	plant	168.53	79.21	290.43
	s22	A. palmatum “Atropurpureum”	16	plant	58.29	27.40	100.45
	s23	H. syriacus	3	plant	17.47	8.21	30.11
	s24	P. × fraseri A	4	plant	15.35	7.22	26.46
	s25	Photinia × fraseri Dress B	21	plant	58.21	27.36	100.31
	s26	C. japonica	10	plant	23.81	11.19	41.03
	s27	P. serrulata	16	plant	45.10	21.20	77.72
	s28	P. tobira A	15	plant	42.28	19.87	72.86
	s29	P. tobira B	19	plant	39.10	18.38	67.39
	s30	L. sinense	13	plant	16.25	7.64	28.00
	s31	L. chinense	10	plant	9.44	4.44	16.27
	s32	L. indica “Bush”	1	plant	3.83	1.80	6.61
	s33	P. edulis	26	plant	70.06	32.93	120.74
	s34	F. japonica	103	m2	203.53	95.66	350.75
	s35	P. × fraseri	645	m2	1274.52	599.02	2196.42
	s36	A. japonica “ Variegata”	23	m2	45.45	21.36	78.32
	s37	N. domestica	162	m2	320.11	150.45	551.66
	s38	G. jasminoides	154	m2	304.30	143.02	524.42
	s39	L. japonicum	401	m2	792.38	372.42	1365.53
	s40	L. chinense	169	m2	333.94	156.95	575.50
	s41	E. japonicus	180	m2	355.68	167.17	612.96
	s42	H. monogynum	26	m2	51.38	24.15	88.54
	s43	L. sinense	171	m2	337.90	158.81	582.31
	s44	R. simsii	510	m2	1007.76	473.65	1736.71
	s45	I. crenata	37	m2	73.11	34.36	126.00
	s46	G. jasminoides	200	m2	395.20	185.74	681.06
	s47	H. macrophylla	38	m2	75.09	35.29	129.40
	s48	E. purpurea	10	m2	15.09	7.09	26.00
	s49	S. japonica	33	m2	49.78	23.40	85.79
	s50	T. violacea	102	m2	153.87	72.32	265.17
	s51	O. japonicus	17	m2	25.64	12.05	44.19
	s52	C. indica	8	m2	12.07	5.67	20.80
	s53	A. donax	25	m2	37.71	17.73	64.99
	s54	A. calamus	27	m2	40.73	19.14	70.19
	s55	M. sinensis	19	m2	28.66	13.47	49.39
	s56	I. sibirica	63	m2	95.04	44.67	163.78
	s57	M. capillaris	17	m2	25.64	12.05	44.19
	s58	P. alopecuroides	32	m2	48.27	22.69	83.19
	s59	L. salicaria	14	m2	21.12	9.93	36.40
	s60	C. oshimensis	28	m2	42.24	19.85	72.79
	s61	Hosta	12	m2	18.10	8.51	31.20
	s62	lawn	3146	m2	4650.74	2185.85	8014.77
	Subtotal				35,932.78	16,888.40	61,924.16
Roof garden	rg1	O. fragrans B	4	plant	35.47	16.67	61.12
	rg2	M. rubra	4	plant	35.47	16.67	61.12
	rg3	A. palmatum	4	plant	49.68	23.35	85.62
	rg4	P. mume	3	plant	40.26	18.92	69.38
	rg5	A. palmatum “Atropurpureum”	5	plant	15.84	7.44	27.29
	rg6	P. granatum “Nana”	3	plant	17.19	8.08	29.62
	rg7	L. indica “Bush”	2	plant	9.97	4.68	17.17
	rg8	C. japonica	7	plant	21.45	10.08	36.97
	rg9	P.× fraseri (ball-shaped form)	1	plant	2.22	1.05	3.83
	rg10	P. tobira	1	plant	2.52	1.18	4.34
	rg11	P. fortuneana	1	plant	2.22	1.05	3.83
	rg12	L. chinense var. rubrum	1	plant	2.22	1.05	3.83
	rg13	P. × fraseri	443	m2	875.37	411.42	1508.55
	rg14	N. domestica	20	m2	39.52	18.57	68.11
	rg15	L. japonicum “Howardii”	395	m2	780.52	366.84	1345.10
	rg16	L. chinense var. rubrum	110	m2	217.36	102.16	374.58
	rg17	P. tobira	244	m2	482.14	226.61	830.89
	rg18	R. simsii	316	m2	624.42	293.48	1076.08
	rg19	C. sasanqua	86	m2	169.94	79.87	292.86
	rg20	C. hookeriana	20	m2	39.52	18.57	68.11
	rg21	O. japonicus	88	m2	136.49	64.15	235.22
	rg22	S. lineare	250	m2	387.77	182.25	668.25
	rg23	lawn	783	m2	1214.48	570.81	2092.96
	Subtotal				5198.40	2443.25	8958.58
Sunken courtyard	sc1	O. fragrans B	2	plant	17.73	8.33	30.56
	sc2	P. subg. Cerasus sp.	2	plant	35.91	16.88	61.89
	sc3	A. palmatum	2	plant	24.84	11.67	42.81
	sc4	A. palmatum “Atropurpureum”	2	plant	6.33	2.98	10.92
	sc5	P. tobira A	1	plant	2.52	1.18	4.34
	sc6	P. fortuneana	2	plant	4.45	2.09	7.67
	sc7	F. japonica	64	m2	126.46	59.44	217.94
	sc8	A. japonica	34	m2	67.18	31.58	115.78
	sc9	P. × fraseri	42	m2	82.99	39.01	143.02
	sc10	L. japonicum	60	m2	118.56	55.72	204.32
	sc11	P. tobira	16	m2	31.62	14.86	54.48
	sc12	R. simsii	56	m2	110.66	52.01	190.70
	sc13	H. plantaginea	10	m2	15.09	7.09	26.00
	sc14	O. japonicus	241	m2	363.55	170.87	626.52
	sc15	lawn	96	m2	144.82	68.06	249.57
	Subtotal				1152.71	541.78	1986.51

Notes: The last column of data is obtained as the product of the data in the previous column, with 44/12.

indicate substantial variation in carbon storage among the three green space typologies. The total vegetation carbon storage in the typical public building’s attached green spaces amounts to 19,873.43 kg C, with significant differences across spatial types. Specifically, the green square dominated with 16,888.40 kg C, accounting for 84.98% of the total. Dominance is attributed to its large planting area supporting biodiverse vegetation, where mature trees, including G. biloba and K. paniculata, contributed substantially to biomass accumulation. Notable supplementary contributions from shrubs like A. gramineus further enhanced carbon sequestration, while herbaceous plants showed minimal impact. The roof garden stored 2443.25 kg C, representing 12.29% of the total, with space constraints limiting vegetation to shrub species like L. japonicum “Howardii” and P. × fraseri. The sunken courtyard contributed merely 541.00 kg C at 2.72%, dominated by small shrubs and herbaceous plants like L. japonicum and O. japonicus. Notably, species-level carbon storage exhibited marked spatial variation among different attached green space typologies. This spatial variation was especially evident in O. fragrans, where mature individuals located in the green square demonstrated substantially higher carbon accumulation compared to other spaces. These results highlight the influence of microenvironmental conditions on species performance and carbon sequestration potential.

Overall, the green square serves as the major carbon sequestration zone among the studied green spaces, a role primarily attributed to the contribution of mature trees. Conversely, the roof garden needs high carbon sequestration shrub species, while the sunken courtyard needs multi-layered planting designs to improve its carbon sequestration potential. In addition to species composition and planting area, the spatial structure and microclimatic conditions of these typologies also play a crucial role in determining carbon storage performance. Sunken courtyards offer stable temperatures due to their enclosed form and low wind exposure, but limited sunlight can constrain photosynthesis and growth. Rooftop gardens, by contrast, face greater wind and solar exposure, causing temperature extremes and evapotranspiration stress, yet potentially extending the growing season for sun-loving species. These structural and microclimatic characteristics should be considered alongside species selection when developing optimization strategies for different attached green space types.

3.2. Plant Carbon Storage Analysis

3.2.1. Correlation Analysis of Plant Carbon Storage

Plant carbon sequestration capacity exhibits significant variation attributable to both interspecific differences and intraspecific morphological traits. To assess the influence of morphological characteristics on carbon storage, this study conducted correlation analyses focusing on key biometric variables, including tree DBH, H, shrub CD, and shrub H. These parameters were systematically evaluated against individual plant carbon storage data to determine their predictive significance. As shown in

Table 3. Correlation between plant biometric measurements and carbon storage by plant type.

No.	Plant Types	Independent Variables	Carbon Storage per Plant	Sample Size
1	Evergreen trees	DBH	0.976 **	8
		Height	0.879 **
2	Deciduous trees	DBH	0.821 **	18
		Height	0.676 **
3	Shrubs	Crown diameter	0.833 **	45
		Height	0.762 **

Notes: ** represents a significant correlation.

, significant positive correlations exist between plant morphological traits and carbon storage for both trees and shrubs. For evergreen trees, DBH showed a particularly strong linear correlation (r = 0.976), significantly exceeding that of tree H (r = 0.879), confirming DBH’s superior predictive utility. This finding aligns with tree physiology; DBH growth directly reflects vascular cambium activity and subsequent xylem biomass accumulation, whereas height growth is primarily constrained by apical meristem regulation [53]. For deciduous trees, the correlation coefficient between DBH and carbon storage (0.821) was significantly higher than that between tree height and carbon storage (0.676), further verifying DBH as a robust and universally applicable predictor for estimating carbon storage in forest trees. However, compared to evergreen trees, the coefficient of determination decreased by approximately 13.7%, primarily due to the distinct phenological cycles of deciduous species, which exhibit greater interannual variability in growth. In shrubs, CD correlated more strongly with carbon storage (r = 0.833) than H (r = 0.762), reflecting their lateral growth strategy. Under spatial constraints, shrubs tend to optimize biomass accumulation through canopy expansion, thereby increasing the canopy photosynthetic surface area.

3.2.2. Regression Analysis of Plant Carbon Storage

This study analyzed the relationships between key plant morphological traits (H and DBH) and individual carbon storage using eight regression models: linear, logarithmic, inverse, quadratic, cubic, power, sigmoidal, and growth-related models (including exponential and logistic). The optimal model was selected based on statistical performance metrics, specifically the coefficient of determination (R²) and the root mean square error (RMSE). The best-fitting models are presented in Figure 2, with 95% confidence intervals and prediction intervals, indicating the reliability and predictive accuracy of the regression analysis.

In terms of plant H, nonlinear regression analyses were constructed to establish fitting curves based on measured data, quantifying the relationship between plant H and per-plant carbon storage across different woody life forms. Results revealed that all life forms exhibited significant power function relationships, though with notable interspecific variation in their functional responses. Specifically, tree species displayed a higher exponent than shrubs, indicating a steeper carbon accumulation trajectory with H. Deciduous trees exhibited lower carbon storage values than evergreen trees at lower heights (x < 5 m) but demonstrated a markedly steeper growth trend beyond 15 m. Evergreen trees ($\mathrm{y}=1.81\times {10}^{-6}{x}^{8.94}$, R² = 0.96) showed an exponent significantly greater than 3, indicating superlinear carbon accumulation during vertical growth. This phenomenon likely stems from the concurrent expansion of support structures like stems as the canopy expands, leading to a geometric increase in trunk biomass relative to height. The exponent value of 8.94 in the above equation implies that a 10% increase in height corresponds to about 23.6% growth in carbon storage, underscoring evergreen trees’ dominant role in carbon sequestration. By comparison, deciduous trees ($y=7\times {10}^{-10}{x}^{12.82}$, R² = 0.68) exhibited a higher exponent, reflecting distinct growth strategies. A 10% height increase resulted in a 35.8% rise in carbon storage, explaining their rapid carbon sink enhancement during early secondary succession. Shrubs ($y=0.57x^{1.19}$, R² = 0.94) demonstrated a near-linear relationship, attributed to their multi-stemmed, clumped growth form. Their carbon storage increment depended more on CD than on H, resulting in a moderate growth trajectory. These findings provide quantitative insights into the basis of carbon storage across woody forms, with implications for vegetation management and carbon sink optimization.

Regarding DBH, the carbon storage of evergreen trees exhibited a quadratic relationship with DBH ($y=0.2x^2-3x+14.7$, R² = 0.971), indicating that carbon accumulation accelerates as DBH increases. This phenomenon can be attributed to the positive feedback of vascular cambium activity [53]. When DBH exceeds 30 cm, the carbon storage per unit increment of DBH increases by 1.8–2.3 times compared to the initial growth stage, highlighting the pivotal role of large-diameter evergreen trees in long-term carbon sequestration. For deciduous trees, carbon storage follows a cubic growth function ($y=0.02x^3-0.31x^2+2.33x-4.73$, R² = 0.84). As DBH increases, the growth trajectory of carbon storage per deciduous tree becomes steeper than that of evergreen trees, suggesting a more rapid carbon accumulation rate in the later stages of DBH growth. Shrub carbon storage shows an exponential correlation with CD ($y=0.18e^{0.01x}$, R² = 0.78), indicating an accelerated increase in carbon storage as CD expands. This aligns with the biological characteristic of exponential growth in branch and leaf mass following crown expansion, wherein increased CD enhances photosynthetic area and overall biomass. Although individual shrubs store relatively low amounts of carbon, their exponential growth potential makes them highly effective for small-scale ecological restoration. In a mixed forest design, combining evergreen trees (medium- to long-term carbon sinks), deciduous trees (short-term carbon sinks), and shrubs (rapid short-term carbon sequestration) can optimize overall carbon accumulation.

4. Carbon Storage Optimization

4.1. Analysis of Existing Plant Configuration

In urban green space systems, the carbon sequestration capacity of individual plants and their communities plays a crucial role in determining overall carbon sequestration performance. To enhance the carbon sequestration performance of green spaces, this study aims to design diverse, multi-layered plant communities comprising trees, shrubs, and grasses to optimize carbon sequestration allocation. Based on survey data from the public building-attached green spaces, plant species application frequency was statistically analyzed. The results are shown in

Table 4. Frequency analysis of plant usage in attached green spaces.

Frequency/%	Trees	Shrubs	Herbs
$m\ge 8$	C. camphora, O. fragrans, A. palmatum, K. paniculata	P. × fraseri Dress, R. simsii, L. Japonicu “Howardii”	S. lineare, O. japonicus
$3\le m<8$	M. × soulangeana, A. palmatum “Atropurpureum”, Z. serrata, P. cerasifera “Atropurpurea”, G. biloba, P. subg. Cerasus sp.	L. sinense “Variegatum”, P. tobira, E. japonic, F. japonic, N. domestica, G. jasminoides	T. violacea, Hosta, S. japonic, M. capillaris
$0\le m<3$	L. indica, M. grandiflora, T. sebifera, M. rubra, E. japonica, C. sinensis “Bush”, P. mume, P. granatum “Nana”	C. japonica, A. japonica “Variegata”, H. monogynum, C. hookerian, I. crenata	A. gramineus, I. sibirica, C. oshimensis “Evergold”, C. indica, P. alopecuroides

. According to Table 4, the tree layer is dominated by evergreen species, particularly C. camphora and O. fragrans, which contribute significantly to existing carbon storage. However, the species composition exhibits a high concentration, indicating a need to improve ecosystem stability. Medium- and low-frequency species, such as G. biloba and M. grandiflora, demonstrate high carbon storage potential and are recommended for inclusion in future optimization efforts. The shrub layer exhibits a relatively balanced species distribution. High-frequency species are predominantly hedge plants, such as P. × fraseri Dress, while medium-frequency species, including N. domestica and G. jasminoides, offer both ornamental value and ecological benefits. These species have the potential to enhance plant stratification and improve carbon sequestration capacity. The herb layer is primarily composed of ground cover plants, which, despite their relatively low unit carbon storage capacity, provide advantages in terms of ground coverage and landscape continuity. Medium- and low-frequency herbs, such as P. alopecuroides and T. violacea, exhibit strong adaptability and substitution potential, making them suitable for configuration in low-maintenance green spaces. In summary, the current vegetation configuration demonstrates significant structural deficiencies, characterized by carbon storage dominated by monospecific stands and suboptimal community architecture. Given the disproportionately higher carbon storage capacity of trees relative to shrubs and herbaceous species, optimization strategies should prioritize enhancing the proportion and spatial distribution of high-biomass tree species while establishing hierarchical canopy structures. Concurrently, strategic selection of rapid-growth, high-coverage understory species is critical for maximizing carbon storage efficiency per unit area. These structural improvements would concurrently optimize multiple ecosystem services—including biodiversity conservation, microclimate regulation, and aesthetic value—through multi-layered vegetation systems that embody nature-based solutions (NBS) and multifunctional design principles essential for maximizing ecological, climatic, and social co-benefits in space-constrained urban environments. The empirical plant frequency data presented herein provide robust quantitative foundations for evidence-based green infrastructure optimization strategies targeting enhanced urban carbon sequestration capacity.

Figure 3 visually illustrates the data characteristics presented in Table 2. Through visual analysis, high carbon storage plant species can be directly and precisely identified. In the green square, the tree species demonstrating superior carbon sequestration benefits include G. biloba, K. paniculata, C. camphora, Z. serrata, and C. sinensis. The dominant shrub species comprise H. syriacus, P. × fraseri, L. indica “Bush”, P. tobira, and F. japonica. In the sunken courtyard, the most efficient carbon sequestering tree species were identified as P. mume, A. palmatum, and O. fragrans, while prominent shrub species include L. indica “Bush”, P. × fraseri, P. tobira, and C. japonica. The roof garden exhibits high carbon sequestration performance among tree species such as P. subg. Cerasus sp. and A. palmatum, supplemented by shrub species P. tobira and P. fortuneana, among others. The results highlight the spatial variation of plant carbon performance across attached green spaces and provide a valuable basis for species-specific carbon optimization strategies.

4.2. Planting Optimization Strategy to Improve the Carbon Storage

Building on the correlation and regression analyses presented in Section 3.2, this study proposes evidence-based optimization strategies for plant configuration aimed at enhancing carbon sequestration capacity. These strategies integrate the quantitative insights from statistical modeling with ecological principles and practical design considerations, employing a data-driven, expert-informed approach to ensure both scientific rigor and practical feasibility. Specifically, three core strategies are outlined: (1) diversifying tree species composition through strategic selection; (2) replacing low-performance species with high carbon sequestration species; and (3) enriching vertical stratification through multi-layered planting. Implementation of these strategies requires tailored pathways for different types of attached green spaces. The following sections present species selection and spatial configuration guidelines for the green square, rooftop garden, and sunken courtyard, addressing their distinct environmental constraints and functional requirements.

4.2.1. Optimization of Plants in the Green Square

The green square serves as a key public space for relaxation and recreation. Due to its aesthetic and social functions, it has relatively high plant diversity. However, its carbon sequestration capacity remains inadequate. Four main structural deficiencies have been identified: discontinuous vertical stratification, insufficient carbon storage capacity, unbalanced spatial layout, and fragmented ecological connectivity. Vertically, the tree layer is dominated by C. camphor and K. paniculata trees, with a notable absence of mid-story species, resulting in an incomplete stratification. The shrub layer, largely consisting of P. × fraseri and L. japonicum, fails to form a coherent structure with the tree and ground layers, limiting daylight use to below 35%. The ground cover layer largely consists of turfgrass, covering approximately 12.8% of the total area, with high homogeneity and minimal functional diversity. Regarding species composition and its implications for carbon storage potential limitations, the prevalence of low carbon sequestration tree species is evident. For example, P. subg. Cerasus sp. (9) exhibits a truncated flowering period, thereby constraining its carbon sequestration capacity. Similarly, G. biloba trees (11) demonstrate slow growth and a low leaf area index, further reducing their carbon assimilation potential due to physiological limitations in growth rate and leaf surface area, as well as structural constraints that limit light interception efficiency. In the shrub layer, the biomass accumulation rates of P. × fraseri and L. japonicum “Howardii” are markedly lower than those of native shrub species. Conversely, high carbon sink native tree species, such as L. formosana, M. glyptostroboides, and T. sebifera, remain underrepresented in urban greening initiatives. Currently, only six T. sebifera trees are present, a number insufficient to yield significant ecological benefits. Furthermore, the 3083 m² expanse of traditional lawn lacks seasonal variation, while the limited availability of nectar-producing plants, including L. indica (7) and H. syriacus (3), fails to adequately support insect populations, thereby disrupting the carbon cycle. Consequently, this imbalance perpetuates a vicious cycle characterized by elevated maintenance costs and reduced ecological efficiency.

In response to those identified problems, this study proposes the following optimization strategies, as detailed in

Table 5. Optimization details for the green square.

Community Structure		Plant Configuration
Original	Trees, shrubs, and grasses	C. camphora B (15) + P. subg. Cerasus sp. B (9) + G. biloba (5) + P. × fraseri (323 m2) + Lawn (800 m2)
Optimized	Trees, shrubs, and grasses	L. formosana (10) + T. sebifera (5) +M. spectabilis (20) + M. glyptostroboides (10) + P. × fraseri (160 m2) + H. syriacus (100 m2) + N. domestica (60 m2) + P. alopecuroides (560 m2) + V. philippica (240 m2)

Notes: “+” separates different plant types; numbers in parentheses indicate the number of individuals (for trees) or planting area in m² (for shrubs and herbaceous plants).

. First, structural optimization in the tree layer is achieved through the introduction of native high carbon sequestration tree species. Specifically, 15 C. camphor B trees can be removed and replaced with 10 L. formosana trees and 5 T. sebifera trees at a replacement ratio of 2:1. Additionally, 20 M. spectabilis trees can be introduced into the mid-story space to enhance biodiversity. For the ground cover layer, ecological transformation involves reducing 800 m² (25.9% of the total area) of conventional turfgrass and replacing it with a mixed planting of P. alopecuroides and V. philippica at a ratio of 7:3. Systematic replacement of low carbon storage tree species includes the removal of 9 P. subg. Cerasus sp. B trees and 5 G. biloba trees, to be replaced with 10 M. glyptostroboides trees. Furthermore, 50% of P. × fraseri (323 m²) can be replaced with 200 H. syriacus plants (at a density of two plants/m²) and 240 clusters of N. domestica (at a density of four clusters/m²). These adjustments are expected to improve vertical stratification from two to four layers and raise the proportion of large-diameter trees (DBH > 25 cm) to 12%. Consequently, the carbon storage density of the square green space is projected to increase from 1.14 kg C∙m⁻² to 2.72 kg C∙m⁻².

4.2.2. Optimization of Plants in the Roof Garden

The roof garden, serving as a vital element of urban green infrastructure, plays a significant role in increasing the city’s green coverage rate while offering recreational spaces that enhance both the visual appeal and ecological quality of urban environments. However, the roof garden green spaces within the study site currently exhibit several challenges. The vegetation stratification is relatively simplistic, with an overrepresentation of low carbon storage tree species. The lack of mid-story small trees creates a disrupted tree–shrub secondary layer structure, thereby impairing ecological equilibrium. The shrub layer is dominated by P. × fraseri, while the ground cover layer primarily consists of lawn (783 m², accounting for 31.2% of the area), exhibiting considerably lower carbon sequestration capacity when compared to Sedum and other Crassulaceae species. The species composition demonstrates notably low carbon storage performance, and an ecological compatibility conflict exists between M. rubra—a high water-demand tree species—and the shallow soil conditions characteristic of roof environments.

In response to these identified problems, a set of core optimization strategies was proposed, including lawn reduction, vertical stratification enhancement, density control, and the selection of native plant species, as detailed in

Table 6. Optimization details for the roof garden.

Community Structure		Plant Configuration
Original	Trees, shrubs, and grasses	M. rubra (4) + A. palmatum (4) + P. × fraseri(443 m2) + N. domestica (20 m2) + Lawn (783 m2) + S. lineare (250 m2)
Optimized	Trees, shrubs, and grasses	L. indica (4) + H. syriacus (8) + A. palmatum (7) + P. × fraseri (220 m2) +N. domestica (40 m2) + M. spectabilis (15 m2) + Lawn (300 m2) + S. lineare (490 m2) + D. micrantha (160 m2)

Notes: “+” separates different plant types; numbers in parentheses indicate the number of individuals (for trees) or planting area in m² (for shrubs and herbaceous plants).

. First, canopy layer optimization can be implemented by refining the tree layer structure. Specifically, four high-water-demand M. rubra trees can be removed and replaced with four wind-resistant and drought-tolerant L. indica myrtle plants. Concurrently, eight H. syriacus plants were introduced to the middle layer to bridge the vertical stratification between the O. fragrans trees and the shrub layer. Additionally, three dwarf A. palmatum plants were introduced to enhance structural diversity. Second, shrub layer transformation can be implemented by replacing 50% of the P. × fraseri coverage with N. domestica (150 plants at 4 plants/m²) and H. syriacus (ball-shaped form, 30 plants at 2 plants/m²). Third, regarding ground cover, the lawn area can be reduced from 783 m² to 300 m² and replaced with a 6:4 mixed vegetation community comprising S. lineare and D. micrantha. These modifications are designed to significantly reduce irrigation demands while creating an integrated roof green space system that balances ecological functionality with aesthetic value. The adjustments are projected to increase the carbon storage density of the roof garden from the current 0.43 kg C∙m⁻² to 0.87 kg C∙m⁻².

4.2.3. Optimization of Plants in the Sunken Courtyard

The sunken courtyard embodies a distinctive spatial configuration that integrates underground and outdoor environments through vertical connectivity, fostering spatial interaction between upper and lower levels while achieving holistic integration of interior and exterior spaces. This design approach not only enhances the architectural spatial form but also improves ventilation and daylighting conditions. The unique microclimate of the sunken courtyard, characterized by reduced wind exposure, limited direct sunlight, and potential drainage challenges, requires careful consideration in plant selection. Current assessments of the public building’s sunken courtyard green spaces reveal the following deficiencies: a simplified vegetation stratification system, insufficient carbon sequestration capacity due to suboptimal tree species selection, and limited tree DBH. Particularly, the vertical structure comprises merely six existing trees with crown diameters ranging from 2.5 to 3.2 m. The shrub layer is predominantly composed of A. japonica and F. japonica, lacking intermediate-sized trees to complete the middle canopy layer. Regarding species composition, deciduous varieties such as P. subg. Cerasus sp. and A. palmatum “Atropurpureum” represent a significant proportion of the existing vegetation. These species exhibit seasonal fluctuations in carbon storage capacity, with substantially reduced carbon fixation capacity during the winter defoliation period due to ceased photosynthetic activity. The A. japonica-dominated shrub layer appears to demonstrate lower biomass production compared to alternative native species such as H. syriacus. Furthermore, the ground cover layer exhibits overreliance on O. japonicus and turfgrass, which collectively account for approximately 61% of the total ground coverage.

In response to the limitations of the sunken courtyard’s single vertical structure and suboptimal carbon storage performance, the following optimization strategies are formulated, as detailed in

Table 7. Optimization details for the sunken courtyard.

Community Structure		Plant Configuration
Original	Trees, shrubs, and grasses	P. subg. Cerasus sp. (2) + A. palmatum (2) + F. japonica (64 m2) + A. japonica (34 m2) + Lawn (96 m2)
Optimized	Trees, shrubs, and grasses	P. subg. Cerasus sp. (2) + A. palmatum (2) + M. spectabilis (4) + T. sebifera (3) + M. chapensis (3) + F. japonica (64 m2) + A. japonica (34 m2) + Lawn (50 m2) + D. micrantha (28 m2) + S. lineare (18 m2)

Notes: “+” separates different plant types; numbers in parentheses indicate the number of individuals (for trees) or planting area in m² (for shrubs and herbaceous plants).

. First, a multi-tiered vegetation configuration can be implemented, comprising evergreen small arbors, native broad-leaved trees, and fast-growing arbors. Specifically, four M. spectabilis trees are introduced to compensate for the carbon sequestration deficit during the leaf-abscission phase of P. subg. Cerasus sp. Additionally, two L. formosana and three T. sebifera trees are incorporated to establish a seasonally complementary carbon sequestration system. Furthermore, three M. chapensis plants can be planted to enhance mid-canopy carbon interception. The ground cover will be ecologically upgraded by reducing the lawn area from 96 m² to 50 m² and replacing it with a 7:3 mixture of D. micrantha and S. lineare. This modification enhances seasonal landscape diversity and improves the complexity of ecological service functions. The optimized design is projected to increase the carbon storage density of the sunken courtyard from the current 0.58 kg C∙m⁻² to 0.94 kg C∙m⁻².

A comparative analysis of the carbon storage capacity among three types of attached green spaces before and after optimization indicates that the implementation of vegetation structure adjustment measures substantially enhances carbon storage capacity. The pre-optimization carbon storage capacity across the green spaces is 19,873.43 kg C, which increased to 45,964.00 kg C following optimization. Similarly, the pre-optimization carbon storage per unit area is 2.00 kg C∙m⁻², rising to 4.63 kg C∙m⁻² after optimization, with an increase of 131.5%. This study confirms that substituting low carbon sequestration plant species with higher carbon storage alternatives, coupled with the enhancement of vertical stratification in plant communities, effectively improves the carbon storage capacity of green space systems. Regarding community configuration models, multi-layered structures integrating trees, shrubs, and grasses are recommended over single-layered arrangements. The synergistic effects of species replacement and structural optimization collectively improve the carbon storage capacity of urban attached green spaces.

5. Conclusions

Urban green spaces play a critical role in global carbon neutrality efforts, with public building-attached green spaces emerging as key components of urban green infrastructure due to their significant carbon sequestration potential. This study conducted a quantitative assessment of carbon storage and vegetation structure across three types of attached green spaces—namely, the green square, roof garden, and sunken courtyard—within a representative public building. Key structural variables analyzed included diameter class distribution, tree-to-shrub ratios, and planting density. Results revealed significant differences in vegetation carbon storage across both species and plant community levels. The identified correlations between plant morphological traits and carbon sequestration capacity pinpointed growth patterns conducive to higher carbon uptake potential. These findings provide empirical insights for enhancing carbon sequestration in urban public building-attached green spaces through optimized vegetation strategies. This study offers a scientific basis for climate-responsive urban planning via improved planting design and spatial configuration.

Although this study provides an empirical foundation for optimizing vegetation composition and spatial configurations, its findings have limited generalizability due to site-specific geographical, climatic, and anthropogenic conditions. Future research should further explore the following directions: (1) expanding sampling across diverse geographical and climatic contexts to enhance broader applicability, (2) investigating synergistic effects between multiple urban green space characteristics and carbon sequestration processes, and (3) developing integrated evaluation frameworks that balance carbon sequestration potential with ecological, functional, aesthetic, and economic benefits. These advancements would advance comprehensive urban green infrastructure optimization for climate mitigation.