Estimating Carbon Stocks and Biomass Expansion Factors of Urban Greening Trees Using Terrestrial Laser Scanning

Abstract

Urban forest carbon sequestration represents an important component of the global forest carbon pool; however, accurate measurements are limited by the inability of existing field stand models to match the specificity of urban greening species. Herein, canopy volume, carbon stock, and the biomass expansion factor (BEF) of 30 Koelreuteria paniculate trees were measured based on terrestrial laser scanning (TLS) and compared to the results of existing wood volume and carbon stock model measurements. The findings revealed that (1) TLS point cloud data were highly reproducible and accurate (root mean square error of tree height and diameter at breast height were ±0.35 m and ±0.33 cm, respectively). (2) Owing to human interference and cluttered urban environments, the BEF of urban greening tree species fluctuated irregularly, considerably different from that of natural forest stands. (3) Leaf carbon stocks were influenced by the size of the voxel. (4) Different tree measurement factors maintained variable degrees of influence on BEF (height under branch, volume of thick branch, crown width, and projected areas of tree-crown produced correlation coefficients of −0.64, 0.54, 0.45, and 0.43, respectively). Accordingly, the carbon stock and BEF of urban greening tree species can be accurately calculated via TLS without damage.

1. Introduction

The continuous development of urban forestry over recent decades has gradually become an important component of carbon sink forestry and terrestrial ecosystems, thereby maintaining a significant role in regulating global climate change [1,2,3]. Biomass expansion factors (BEF), which convert the timber volume (or dry weight) to biomass, are used to estimate the tree biomass and account for the carbon budget on a national and regional scale [4]. Notably, the BEF represents an essential parameter for calculating biomass and carbon stock, and its accuracy and applicability are crucial for measuring regional carbon stocks [5,6,7,8]. Traditionally, vegetation carbon stocks in each built-up area are calculated using binary allometric growth equations constructed from field stands [9,10]; however, owing to the frequent human disturbances and complex urban-environmental conditions, urban green tree species growth varies from that in the natural field, thus widening the differences in growth dynamics and carbon sequestration capacity between it and the field stands [11,12,13]. Therefore, the resulting applicability between urban green tree species and the existing BEF and models is relatively low. Accordingly, how to accurately measure urban carbon stocks without affecting their normal growth patterns is a pressing issue [14,15].

Terrestrial laser scanning (TLS) is a modern measurement technique in the field of surveying and mapping and has recently emerged as a valuable automated measurement tool [16,17,18]. Compared with traditional measurement methods, TLS presents spatial information of a measured object in the form of the point cloud, thereby improving measurement efficiency, while increasing the accuracy and intuitiveness of the results [19,20,21]. Furthermore, it constitutes the primary future technology of digital forestry measurements and maintains a broad application prospect in forestry investigation and research [22,23]. Numerous researchers have conducted relevant studies on tree carbon stocks and biomass using 3D laser scanning technology, and some progress has been made [24,25].

Han et al. [26] revealed that TLS technology could achieve non-destructive and accurate measurements and proposed using regression analysis to establish a binary wood volume equation and subsequent table for practical applications of forestry investigation [27,28,29,30]. Feng [31] examined the relationships between tree canopy and biomass using TLS and proposed a CART (classification and regression tree) model through multiple fittings. Although the model was more accurate than previous biomass models, the method did not account for gaps between standing tree canopy branches [32]. Chen [33] estimated the aboveground biomass (AGB) in China using a non-destructive method based on TLS, in addition to quantitatively analyzing the effects of differences in bark roughness and trunk curvature on DBH estimates via TLS. Brolly [34] extracted high-precision standing tree parameters from the constructed 3D digital models of the tree trunk, canopy surface, and ground features, ultimately proposing a method for efficiently extracting basic tree measurement factors through TLS, which maintained comparable levels of accuracy to traditional methods [35]. Tanhuanpää [36] and Rahman [37] initially measured the AGB of standing trees based on TLS, and assessed the validity of a binary allometric growth equation on standing urban trees, supporting a technique for assessing standing tree biomass via TLS; however, the former only measured trunk wood volume through TLS, and was combined with a harvesting method for modeling, while the latter would cause certain inaccuracies in carbon stock and biomass estimates if leaf volumes used did not match their density. Accordingly, the challenge of accurately and non-invasively measuring urban standing wood biomass is ongoing [38].

Koelreuteria paniculate is a common urban greening tree species in the south of the Yangtze River Basin, China, and is also the preferred tree species for ecological greening in industrial cities in this region. The research here employed TLS combined with a local sampling method for conducting non-destructive measurements of K. paniculate AGB, aiming to achieve accurate carbon stock estimates for this common urban greening tree species [39]. Therefore, this study aimed to obtain the corresponding BEF of urban greening tree species through specific calculations and analyses, thereby providing theoretical and technical support for the future development of forestry, while informing carbon neutrality targets.

2. Materials and Methods

2.1. Study Area and Instrumentation

The study area was located at Zhejiang A&F University, which maintains the highest level of tree cover among Chinese university campuses. Specifically, the experimental area is dominated by greenery species, such as K. paniculate and Ginkgo biloba linn, with vegetation coverage ≤ 42%, across a wide variety of species. The tree individuals in the study area are artificially cared for and maintained; thus, external factors strongly influence their growth dynamics, making them representative of typical urban green space (Figure 1).

The TLS data were measured and pre-processed by Leica ScanStation C5 terrestrial and its accompanying processing software, Cyclone REGISTER360 (Leica Geosystems AG, Heerbrugg, Switzerland). The trunk sample was collected by Swedish hand-held corer CO250 (HAGLOFS, Dalarna, Sweden). Branch and leaf samples were collected by high branch shear. DBH was collected by tape.

2.2. Research Technique

In this study, 30 K. paniculate individuals were scanned by the terrestrial laser scanner to obtain standing wood point cloud data. Simultaneously, DBH and tree height of these 30 trees were manually measured, and the obtained data were used for comparison, it was found that the data obtained by TLS has high accuracy (see Section 3.1). Therefore, the volume and carbon storage of urban greening tree species calculated based on TLS was used as standard values to compare and analyze the corresponding results calculated by the existing models. The specific steps are as follows: at first, the carbon storage of various organs was calculated by TLS (Section 2.5.1); next, the existing model was used to calculate the trunk volume and carbon storage of various organs (Section 2.5.2). Then, the biomass expansion factor was calculated by using the overall biomass and volume of standing trees (Section 2.5.3), and the correlation analysis was carried out with other tree measurement factors (Section 3.6.3); compared and analyzed the above TLS-based calculation results with the corresponding existing model calculation results, and drew a conclusion. The conceptual diagram of the research is shown in Figure 2.

2.3. Data Acquisition

Using 2 cm diameter steps, starting with 5 cm. Thirty K. paniculate individuals were selected for subsequent study via the diameter step equivalence method according to the “Log Volume Table (GB 4814-84)” (Figure 3).

2.3.1. Single-Stand Point Cloud Data Acquisition

The 30 sample trees selected following pre-experimentation were scanned by the Leica ScanStation C5.

The specific collection methods are as follows: in order to obtain complete single tree point cloud data, each tree was scanned by 3~4 stations, and the layout of the corresponding stations was approximately in an equilateral triangle or rectangular distribution. According to the terrain and intervisibility of the survey stations, three public reference targets were placed around the target tree to ensure that it could be completely scanned at different survey stations. After the deployment of the measuring station and the target site was completed, the instrument was erected and leveled, and the scanning parameters were set as follows: the scanning range adopted single wooden window scanning, that is, the user-defined scanning range and the low resolution was adopted, and the scanning started after setting the corresponding saving folder. The average scanning time of each station in the scanning process was about 10 min. Besides, the TLS accuracy can reach 4.5 mm~7 mm in the range of 50 m.

2.3.2. Sample Collection

(1)

Trunk sample collection

Trunks were sampled by diameter order according to Figure 3. Three standard trees were selected in each of the three diameter steps from 12 to 16 cm (9 in total), and all individuals in the 10 and 18 cm steps were taken (5 in total), as the sample was ≤3 trees.

Drilling and coring of the selected standard trees occurred at the base of the trunk, 1 m and 1.3 m height, using the Swedish hand-held corer CO250 (HAGLOFS, Dalarna, Sweden). The borehole diameters (the units of the borehole diameters are centimeters) were measured, recorded using Vernier calipers, marked, and returned to the laboratory.

(2)

Crown sample collection

The crown sample was taken from the same trees as the trunk samples. To ensure data integrity, the crown of K. paniculate was divided into upper and lower layers for sampling according to crown length (the units of the crown length are meters). Subsequently, the completed samples were divided into branches and leaves to facilitate subsequent processing.

2.4. Data Pre-Processing

2.4.1. Point Cloud Data

Branch and leaf data were processed in Cyclone. The point cloud reflection intensity histogram was obtained, and the corresponding threshold value was set (Figure 4a), while duplicates and points with distances < threshold (default, 1 mm) were discarded, thereby reducing data redundancy and realizing the primary classification of the point cloud data [40]. The results of the primary classification were then manually differentiated and identified to achieve branch and leaf separation (Figure 4b).

2.4.2. Tree Measurement Factors

In Cyclone, the elevation values of branch starting positions were obtained, while the minimum and maximum values of the point cloud data for each K. paniculate individual in the X, Y, and Z directions were automatically extracted so that the 3D laser scanned tree height h = Z_max − Z_min, crown width H = (Y_max − Y_min + X_max − X_min)/2, and the height under the branch h_c = Z_c − Z_min. The least-square circle algorithm [41] was used to extract the standing tree DBH by fitting the point cloud to a slice with a 2 cm thickness.

2.4.3. Voxel Size

Leaf point clouds use voxels of different sizes to calculate leaf volume and carbon storage. Plot the curve of leaf volume and carbon storage with the ratio of crown width to pixel side length, and analyze the pixel size that is suitable for calculating the carbon storage of precise leaves.

2.5. Measurement Methods

2.5.1. TLS-Derived Carbon Stock Measurements

• Calculation of trunk carbon stock

The trunk drill core samples were dried at 100 °C to a constant weight, and their total dry weights ($G_{gg}$) were weighed for each diameter class [42].

The total volumes ($V_{gg}$) of drill cores were calculated using the mean diameter values derived from the calipers and corers. The biomass density per unit volume of trunk (${\rho }_g$) was calculated using $G_{gg}$ and $V_{gg}$; however, the obtained TLS point cloud data of the trunk were used to obtain the wood volume (V_S) via the sectional measurement method [19,20]. The carbon stock of the trunks ($C_s$) was measured via Equations (1)–(3):

$${\rho }_g=\frac{G_{gg}}{V_{gg}},$$

(1)

$$V_S={\displaystyle \sum }_{i=1}^{n-1}\frac{1}{3}\pi h_i\left(R_i^2+r_{i+1}^2+Rr\right),$$

(2)

$$C_s=V_s\times {\rho }_g\times 0.48$$

(3)

where $h_i$ represents the segment interval, while $R_i$ and r_i are the radii of the lower and upper cross-sections, respectively.

2.

Calculation of branch carbon stock

Branches were dried to a constant weight, and the total dry weights of the branch samples for each diameter class were weighed as $G_{BS}$ [42]. The total branch volume (V_BS) was obtained referring to the drainage method in “The wood density calculation method(GB/T 1933-2009)”, and the biomass density per unit volume of the branches was obtained as ${\rho }_B$. To minimize error, branch volumes were divided into thick and thin branches to be measured separately (Figure 5b,c), where the former was measured via the sectional measurement method using the classified branch point cloud data (Figure 5a), and the latter were divided into 0.5 cm, 1.0 cm, and 1.5 cm levels by sampling (Figure 5d), before the volume each level was calculated and aggregated to obtain the total thin branch volume (V_TB). The carbon stock of tree branches ($C_B$) were then measured via Equations (4) and (5):

$${\rho }_B=G_{BS}/V_{BS},$$

(4)

$$C_B=\left(V_{CB}+V_{TB}\right)\times {\rho }_B\times 0.48$$

(5)

3.

Calculation of leaf carbon stock

The leaves were killed out at 105 °C for 30 min and dried at 85 °C to a constant weight (48 h), and the total leaf sample dry weights were calculated (M_LS) [42]. The voxel algorithm was used to calculate the effective volume (V_L) according to the 3D point cloud data of the leaves [43,44]; thus, the effective volume of the leaf samples was obtained (V_LS). Combined with M_LS, the leaf biomass per unit volume was acquired (${\rho }_L$). Subsequently, the leaf carbon stock ($C_L$) was measured via Equations (6) and (7):

$${\rho }_L=\frac{M_{LS}}{V_{LS}},$$

(6)

$$C_L=V_L\times {\rho }_L\times 0.48.$$

(7)

2.5.2. Volume and Carbon Storage Model Measurements

• Trunk and canopy volume model measurement methods

The reference value of trunk volume adopted a one-dimensional volume model (V_S1) specific to the hilly mountains of northwest Zhejiang Province, as well as a binary volume model ($V_{SV2}$) common for broad-leaved trees throughout the Zhejiang Province [44]. The reference values of canopy volumes $(V_{CV}$) were calculated via the hemispherical approach of the traditional regular geometrical graphical method [45,46,47], according to Equations (8)–(10):

$$V_{S·1}=5.0479\times {10}^{-5}{D}^{1.91}(\frac{-2053.1848}{D+60}+37.295{)}^{0.8144},$$

(8)

$$V_{SV2}=6.7453\times {10}^{-5}{D}^{1.96}{H}^{0.8144},$$

(9)

$$V_{CV}=\frac{\pi x^{2}y}{6}$$

(10)

where D represents DBH, H represents tree height, x represents crown width, and y represents the crown length.

2.

Carbon stocks of various organs model measurement methods

According to “ Estimation of Forest Vegetation biomass and Carbon storage in China” [48], the carbon storage of different organs is calculated as presented in Equations (11)–(15):

$$W_1=0.0440{\left(D^{2}H\right)}^{0.9169}\times 0.48,$$

(11)

$$W_2=0.0230{\left(D^{2}H\right)}^{0.7115}\times 0.48,$$

(12)

$$W_3=0.0104{\left(D^{2}H\right)}^{0.9994}\times 0.48,$$

(13)

$$W_4=0.0188{\left(D^{2}H\right)}^{0.8024}\times 0.48,$$

(14)

$$W_S=W_1+W_2+W_3+W_4$$

(15)

where $W_1$ is the trunk model carbon stock, $W_2$ is the bark model carbon stock, $W_3$ is the branch model carbon stock, $W_4$ is the leaf model carbon stock, and $W_S$ is the total aboveground carbon stock.

2.5.3. BEF Measurement

According to “Estimating Biomass Carbon of China’s Forests: Supplementary Notes on Report Published in Science” [49], BEF is the ratio of AGB ($B$) to wood volume ($V_{SS}$) of standing trees (Equation (16)):

$$BEF=\frac{B}{V_{SS}}$$

(16)

2.5.4. Analysis of Variance and Accuracy Assessment Methods

One-way analysis of variance (ANOVA) and t-tests were conducted using SPSS 26.0 (IBMCorp., Armonk, NY, USA) with a significance level set to p = 0.05. Root mean square error (RMSE) and mean absolute error (MAE) were also used to detect errors between TLS measurements and the modeled values (Equations (17) and (18)):

$$RMSE=\sqrt{\frac{{{\displaystyle \sum }}_{i=1}^n\left(y_i-{\widehat{y}}_i\right)}{n}},$$

(17)

$$MAE=\frac{{{\displaystyle \sum }}_{i=1}^n\left|\left(y_i-{\widehat{y}}_i\right)\right|}{n}$$

(18)

where y represents the measured TLS value, and $\widehat{y}$ represents the measured value.

3. Results

3.1. Comparative Analysis of Basic Tree Measurement Factors

Figure 6a,b shows the comparison between 30 K. paniculate DBH and tree heights, respectively, based on TLS estimates and ground measurements. The results produced an RMSE of ±0.35 m for the TLS-based height measurements, with an average relative error of 4.56%; the RMSE of DBH was ±0.33 cm, with an average relative error of 1.96%. The differences between TLS-based and measured data were not significant (p > 0.05), as the accuracies of DBH and tree height were high. Therefore, TLS measurements can be compared to model values as true values (Section 3.2, Section 3.3, Section 3.4 and Section 3.5).

3.2. Analysis of Trunk Volume and Carbon Stocks Measured by Different Methods

The trunk carbon stock calculated here using the wood volume model was derived according to the product of wood volume, trunk biomass density, and carbon content [50].

One-way ANOVAs of the trunk volumes for the 30 trees measured using four different methods revealed no significant differences between the existing one-dimensional volume model and binary volume model (p > 0.05;

Table 1. Comparison of volume and trunk carbon stocks measured via the different methods assessed. Each value represents the sample mean ± standard deviation, and different lowercase letters indicate significant differences between timber volumes measured in different segment lengths (p < 0.05).

Methods	Volume·dm−3	Carbon Stock·kg−1
TLS_0.1 m sectional measurement	37.61 ± 11.01 c	12.26 ± 5.71 d
TLS_1.0 m sectional measurement	40.22 ± 11.49 c	13.11 ± 5.64 cd
One- dimensional volume model	69.12 ± 34.67 ab	21.89 ± 10.67 a
Binary volume model	53.12 ± 30.86 b	16.98 ± 10.11 bc
Carbon stock model	NA	17.89 ± 9.62 bc

); however, both the one-dimensional and binary volume models yielded significantly higher volumes than that obtained by the TLS sectional measurement method (p < 0.05). The highest value (69.12 ± 34.67 dm³) was obtained using the one-dimensional volume model and was significantly different from the results of the other methods, thereby showing the largest error in the volume measured via the one-dimensional volume model. In comparison, the differences between volumes measured by the binary volume model (53.12 ± 30.86 dm³) and the sectional measurement method were considerably smaller.

There were also different degrees of variation between trunk carbon stocks of the 30 individuals according to the five different methods. The highest trunk carbon stock (21.89 ± 10.67 kg) was measured by the one-dimensional wood volume model and was significantly different from all other methods (p < 0.05). Alternatively, there were no significant differences between the trunk carbon stock measured via the binary wood volume model (16.98 ± 10.11 kg), the trunk carbon stock model (17.89 ± 9.62 kg), and the TLS_1.0 m sectional measurement method (13.11 ± 5.64 kg; all p > 0.05); although, all three of these methods were significantly higher than the trunk carbon stock measured by the TLS_0.1 m sectional measurement method (12.26 ± 5.71 kg; all p < 0.05; Table 1).

Accordingly, in the present study, the volume and carbon stocks calculated by section lengths of 0.10 m were taken as the standard values.

3.3. Leaf Volume and Carbon Stock Based on TLS Measurements

Figure 7a,b shows that the leaf effective volumes, as measured based on different image sizes, tended to decrease with increasing K values, which represents the ratio of crown size to the pixel side length. When K is larger, the V_L (effective volume) and V_LS (sample effective volume) are smaller. When K = 200, the curve shows an inflection point, whereas K ≥ 200, leads to more stable V_L and V_LS values. The biomass density per unit leaf volume, except for a few samples, increased almost linearly with the increase in K (Figure 7c).

Figure 7d shows that the leaf carbon stock measured by TLS decreased with increasing K, with clear inflection points occurring at K = 100 and 200. Leaf carbon stocks measured by TLS changed less when K ≤ 100 or K ≥ 200. When K ≥ 200, the calculated leaf carbon storage was considerably less than that when K ≤ 100. Notably, this trend is not consistent with theory, where it is stated that the leaf carbon reserves calculated via K fluctuate about a certain value. A comparison of Figure 7a,b,d revealed that the results were all significantly different when K ≥ 200; therefore, the data when K = 100 were selected as the leaf volume and carbon stock measured by the TLS in this study.

3.4. Analysis of Canopy Volume and Carbon Stock Calculated by Different Methods

Table 2. Comparison of volumes and carbon stocks for canopies as calculated by different methods.

Calculation Method	Canopy Volume·m−3	Canopy Carbon Stock·kg−1
Canopy volume model	20.24 ± 11.87 a	3.52 ± 1.81 b
k = 100 voxel	6.63 ± 1.79 b	2.48 ± 0.86 b
Branch and leaf (k = 100 voxel) were calculated separately	6.63 ± 1.79 b	14.26 ± 8.94 a

Note: The same column of different lower-case letters showed significant difference (p < 0.05).

shows that the canopy volume calculated via the traditional regular geometric model algorithm (20.24 ± 11.87 m³) was significantly higher than that based on TLS, using the voxel method with K = 100 (6.63 ± 1.79 m³; p < 0.05). The canopy carbon stocks (including branches and leaves) measured by the two algorithms (canopy volume model, 3.52 ± 1.81 kg; k = 100 voxels, 2.48 ± 0.86 kg) were not significantly different (p > 0.05), although both were significantly lower than that measured by separating branches and leaves (14.26 ± 8.94 kg).

3.5. Estimating Carbon Stocks of Individual Organs and Individual Trees by Different Methods

Table 3. Comparisons of each organ and the whole carbon stock for the sample wood measured by TLS and existing models.

	Measured Value of TLS (kg)	Model Values (kg)	t	p	RMSE (kg)	Bias (kg)
Trunk	12.26 ± 5.71	17.89 ± 9.62	−2.729	0.008	7.08	−5.53
Branch	12.49 ± 8.22	7.01 ± 4.26	−2.310	0.024	6.55	5.11
Leaf	1.82 ± 0.71	2.96 ± 1.27	−4.102	0.001	1.47	−1.16
Single tree	26.79 ± 13.82	27.91 ± 15.16	−0.309	0.756	4.36	−1.58

shows that the trunk (17.89 ± 9.62 kg) and leaf carbon stocks (2.96 ± 1.27 kg) as measured by the existing model were significantly higher than those measured by the TLS-based model (trunk, 12.26 ± 5.71 kg; leaf, 1.82 ± 0.71 kg); however, the branch carbon stock (7.01 ± 4.26 kg) measured by the model was significantly lower than that measured by the TLS-based model (12.49 ± 8.22 kg). The accuracy of the leaf carbon stocks using the existing model was the highest (RMSE = 1.47 kg), whereas that of trunk carbon stock was the lowest among the organs (RMSE = 7.08 kg). Overall, the differences between the carbon stocks of single-standing trees calculated by the existing model and the results measured via TLS were not significant (p > 0.05), indicating that the sum of the carbon stocks of the trunk, branches, and leaves measured using the TLS model represents a feasible method for calculating the overall carbon stocks of single standing trees.

For K. paniculate, carbon stock organ accuracies, and the overall sample trees measured at each diameter ordered according to the existing model were roughly as follows: leaf > whole tree > branch ≈ trunk.

Carbon stock measurement accuracy is high when the diameter order is small (Figure 8) and gradually decreases as the diameter order increases for each organ and single stand. Trunk carbon stock accuracy decreased most rapidly with increasing stumpage diameter in the derived model; however, the accuracy of branch carbon stocks slowly decreased with increasing diameter in the 10–14 cm diameter range and rapidly increased above 14 cm. The accuracy of leaf carbon stocks measured via the derived TLS model decreased linearly with increasing stumpage diameter, whereas that of single stumpage carbon stocks measured remained relatively constant from 10–12 cm DBH and gradually decreased linearly from 12–16 cm DBH. Thus, the accuracy of single stumpage carbon stocks measured via the derived model here decreased linearly with increasing stumpage diameter. The accuracy of leaf carbon stocks measured via the derived TLS model decreased exponentially with increasing diameter; however, the carbon stock of single-stand trees measured remained relatively constant between 10 and 12 cm DBH, before decreasing between 12 and 16 cm. Notably, the carbon stock accuracies for trunks, branches, and single-stand trees measured via the derived TLS model sharply decreased between 16 and 18 cm.

3.6. Measurement and Analysis of BEF by Different Methods

3.6.1. Comparison of Different BEF Measurement Methods

Figure 9 shows BEF1 calculated via TLS, BEF2 calculated based on TLS combined with the volume model, and BEF3 calculated based on the single-standing tree model. Evidently, BEF3 was the most stable, followed by BEF2. BEF1 was highly variable, and its mean value was significantly higher than those of BEF2 and BEF3 (p < 0.05).

3.6.2. Accuracy Comparison of Carbon Stocks Measured by Different BEF

Figure 10 shows a large difference between the accuracy of carbon stocks measured using TLS and those of the model, with the results of the former being more accurate.

3.6.3. Correlation Analysis of BEF and Tree Measurement Factors

Each tree measurement factor maintained different degrees of influence on BEF (Figure 11). Among them, the correlation coefficient between the height under branch and BEF was −0.64, showing a very significant negative correlation. The correlation coefficient between thick branch volume and BEF was 0.54, showing a very significant positive correlation. Alternatively, the crown width and projected areas of tree-crown were significantly positively correlated with BEF, with correlation coefficients of 0.45 and 0.43, respectively. While other factors had no significant correlation with BEF, including the most commonly used basic tree measurement factors in forestry, such as DBH and tree height, as well as the effective volume of leaves and thin branch volume rarely involved in conventional forestry investigations, and the volume required for the measurement of BEF was also not significantly related to BEF.

4. Discussion

4.1. TLS-Based Carbon Stock Measurements

The present study was based on TLS and single-standing tree modeling to achieve non-destructive carbon stock and BEF measurements. As urban greenery trees cannot be felled, obtaining accurate carbon stock estimates is largely infeasible; thus, the highly accurate TLS-based estimates obtained from this study lack certainty. To address this issue, the present study selected a suitable method for measuring the carbon stocks of each organ of standing tree individuals, taking into account the point cloud data collected, as well as the growth characteristics, so as to ensure the overall high accuracy of carbon stock estimates.

By comparing the DBH and heights of 30 K. paniculate individuals extracted via TLS with the field measured values, their average relative errors were 1.96% and 4.56%, respectively. Therefore, it was concluded that TLS data can be used to accurately derive these measurements. Additionally, the results also showed that the TLS-based point cloud data of single-stand trees were highly reproducible and objective, thus meeting the need for subsequent accurate measurements of trunk volume and carbon stock [51]. Similar to the present study, Panagiotidis [52] found that TLS data could be used to measure tree height and diameter with high accuracies, performing very well when using the RANSAC (Random sample consensus) method for stem modeling, with low biases (0.02 and 0.01 for deciduous and coniferous species, respectively), and high accuracies (97.73% and 96.14% for deciduous and coniferous species, respectively). Similarly, Luck [53] found that TLS reconstructed trunks were able to capture the spatial distribution patterns and DBH of single trees with high degrees of confidence when compared to manual measurements. Accordingly, the accuracy of TLS-based estimates may exceed traditional measurement techniques; therefore, the evaluation of volume and carbon storage of urban greening tree species based on TLS was deemed reliable here.

4.2. Differences in the Accuracy of Trunk Carbon Stocks by Model

The study of trunk carbon stocks based on five different methods here showed that the prediction accuracies varied. Among them: there was a substantial overestimate predicted by the one-dimensional volume model, the carbon stock trunk measured by the binary volume model was not significantly different from that of the 1.0 m segment length, and both of these models were significantly higher than that measured by the 0.1 m segment length model. Thus, the accuracy of the measured volume largely determined the accuracy of the subsequently measured trunk carbon stocks when using the existing volume model in combination with drill core samples. The trunk carbon stocks based on the carbon stock model were significantly higher than those based on both TLS models (1.0 m and 0.1 m differentiated section lengths). Therefore, the significant differences between the different carbon stock measurement methods were not consistent with those between wood volume, indicating that trunk carbon is determined by (primarily) wood volume and carbon stock per unit volume [54].

4.3. Leaf Carbon Stocks Influenced by the Voxel Size

By comparing the effective volumes of leaves and leaf samples measured under different pixel sizes, it was found that both of their effective volumes tend toward stability when the ratio of crown width to pixel side length, K, was ≥200 (i.e., when the pixel side length was ≤0.025 m). This is notably inconsistent with the findings of Wei [55], who found that the canopy volume stabilized when the ratio of crown width to pixel size was ≤1/10. The likely reason for this discrepancy is that the canopy of standing trees is not dense and solid, but is rather porous. If most pore sizes are taken as critical values, then the measured canopy volume decreases until reaching stability when image edge length decreases above the critical value; however, if the pixel side length continues to decrease after the canopy volume has stabilized to levels below the critical value, then the measured canopy volume will continue to decrease. Here, the canopy volume will decrease further with the lowering side length of the unit pixel until it stabilizes again where the decrease is due to pores in the canopy. Thus, theoretically, as the unit image decreases, the canopy volume will gradually approach the effective volume of the standing tree canopy [56]; however, the pixel side length should not be too small, as this could lead to the fragmentation of the leaf model generated by the voxel algorithm, and result in significant underestimation of the effective volume and carbon storage of leaves. In the present study, it was found that the leaf carbon stock calculated based on voxel decreased rapidly when the pixel side length was <0.05 m.

Further, when compared to Tanhuanpää [36] who did not include TLS-based leaf biomass measurements, and Rahman [37], who used a convex packet algorithm to measure leaf biomass, in combination with an uncorrelated existing leaf density, the present study is the first to measure leaf biomass density per unit volume by combining a body image element algorithm with actual sampling, thus achieving a non-destructive, TLS-based measurement of leaf carbon stocks. The leaf carbon stock tended to be stable between pixel side lengths ≥1/50, and ≤1/100 (Figure 7d); whereas the variation of the carbon stock with pixel side length was relatively consistent, indicating that the leaf carbon stock measured via TLS was reliable.

4.4. Differences between Urban Greening Tree Species and Wild Stands

By comparing the accuracies of carbon stocks for various organs and entire trees measured using existing models with the results of TLS-based measurements, a large variation was revealed in existing carbon stock models of urban greening tree species. Herold et al. [57] found that TLS-derived AGB estimates match the destructively harvested reference values better than estimates based on species-specific empirical allometric equations. McHale [58] considered that these allometric growth equations were established for forest trees and they may have significant differences in form and dimension compared with urban greening trees. Accordingly, there are differences between urban greening trees and wild stands. Further, the accuracy of carbon stocks measured via the existing models decreased with increasing standing tree DBH order. Thus, the existing models are limited when measuring carbon stocks of various organs or stand-level carbon of urban greening species. It is therefore important to establish new allometric equations specifically for urban greening tree species.

Under the constant influence of cluttered urban environments and human disturbance, the differences between urban greenery species and wild stands gradually increase with growth, likely a result of the following reasons: (1) The urban environment differs greatly from the stand conditions in the wild, where nitrogen deposition, heat island effects, and other impacts can affect the growth of urban greening tree species to varying degrees. When combined with human interference, such differences in growth can be even more complex and variable [59,60], amplifying throughout tree growth. (2) The trunk is the main carbon stock organ of greening tree species; however, the trunk carbon stock models currently used are generally based on the average trunk biomass density of an individual species; thus, the errors incurred will not only affect the accuracy of trunk carbon stock measurements but that of the whole single tree carbon stock measurements as well. (3) Existing carbon stock models are based on studies of forest stands in the field, and the specificity of urban greening species may lead to models that are suboptimal, and with limited applicability. Overall, the capacity of existing carbon stock models to meet the needs of today’s precision forestry remains to be further investigated and verified.

4.5. TLS Accuracy in Measuring Wood Volume, Carbon Stocks, and BEF of Urban Greening Trees

Comparing volumes measured via the four different methods, it was found that the overall timber volumes calculated using the existing one-dimensional and binary volume models were significantly higher than the TLS-based sectional measurements. Thus, the pre-existing volume models were not capable of accurately estimating trunk volume for urban greening species. The accuracy of the wood volume measured by the existing models thus decreased with stand growth, highlighting the accumulated influence of the urban environment and human interference on the growth process of urban green species, resulting in gradually increasing trunk differences. Shi [61] and Ji [62] similarly found that complex environmental conditions and frequent anthropogenic interference in urban areas can either promote or inhibit standing tree growth to varying degrees, resulting in high uncertainty surrounding their growth patterns when compared with the natural stands (e.g., the complex and variable location of trunk bifurcation initiation, as seen in the present study).

The BEF of urban greening species measured via different methods here also varied from the results of the inverse equation relationship between BEF and standing wood volume obtained by Jingyun et al. [63]. For example, the TLS-measured BEF levels were significantly higher than those calculated based on the single-standing tree model and TLS combined with the volume model (Figure 9). Simultaneously, the correlation analysis between tree measurement factors revealed that the BEF for urban greening species was influenced by the height under the branch, the growth trend of thick branches, and crown length. The height under branch and BEF showed a very significant negative correlation (−0.64), it was mainly because the height under the branch determines the volume to a certain extent, when the height under the branch was small, the calculated standing volume was small, and the BEF was large when the overall carbon storage remained unchanged. The thick branch volume had a very significant positive correlation with BEF (0.54), which may be because the thick branch volume directly determines the size of branch carbon storage, and thick branches may affect the growth of the standing canopy. The crown width and projected areas of the tree-crown were positively and significantly correlated with thick branch volume (0.77,0.82) and were significantly correlated with BEF (0.45,0.43), which was also good proof of this. Therefore, the larger the thick branch volume, the larger the carbon storage of standing trees, and the larger the BEF.

Here, a comparative analysis of K. paniculate carbon stock accuracies as measured using different methods showed that the BEF calculated by the model was less accurate when used to measure the carbon stocks of urban greening tree species. This highlights the significant differences between urban greening tree species and field stands, and that the existing model is insufficient when assessing such environments. In contrast, the TLS-derived BEF values maintained high levels of accuracy, indicating its capacity to be applied to urban green species, and thus making it possible to quickly and accurately estimate regional-scale standing wood carbon stocks in cities. Although only 30 trees were collected in this paper, the experimental process was scientific and detailed, and the error caused by TLS measurement was controlled as much as possible. On the other hand, the selected models such as volume and carbon storage were also the best match for Koelreuteria paniculate trees. Therefore, the experimental design and data processing of this paper were scientific and reasonable. In the subsequent research, we can increase the sample size and add other urban greening trees for sampling and analysis, which makes this study more scientific and reasonable.

5. Conclusions

The study here assessed the accuracy of existing models calculating carbon stocks in urban greening species in comparison to organ- and stand-level estimates of TLS-based carbon stocks, ultimately revealing that existing models of carbon stocks for urban greening species perform poorly. Firstly, tree measuring factors could be obtained with high-precision via TLS (RMSE of tree height was ±0.35 m, with an average relative error of 4.56%; RMSE of DBH was ±0.33 cm, with an average relative error of 1.96%). Secondly, it was found that pixel size was affected by the point cloud data resolution, and the carbon stocks of branches and leaves could be accurately measured when the pixel side length was 1% of the crown width. Simultaneously, to address the poor match of existing models to urban greening trees, this study successfully calculated the carbon stocks of urban greening tree species through the local sampling of point cloud data of various organs. Lastly, the BEF of urban greening tree species was significantly influenced by the height under branch, volume of thick branches, crown width, and projected areas of tree-crown (with correlation coefficients of −0.64, 0.54, 0.45, and 0.43, respectively). The TLS-based methods, combined with local sampling methods discussed here offer non-destructive and accurate measurements of urban tree carbon stocks, thereby addressing the physiological differences between urban green tree species and field stands owing to environmental and human interference and resolving the limited ability of existing models to correctly reflect true carbon stocks of these urban tree species. Thus, the findings here can help realize long-term dynamic monitoring of urban green tree species, and offer technical support for assessing the ecological service values of urban greening trees [64,65]. However, there are still some limitations to this study. For example, (1) the research area is located on the campus, so frequent passers-by and driving will greatly increase the difficulty of point cloud data acquisition; (2) fewer sample trees may lead to certain deficiencies in verifying the model simulation results. In the future, these limits can be further improved to better estimate the carbon stock and BEF of urban greening trees.