Guangdong China fir man-made forest biomass addition model construction method based on unmanned aerial vehicle laser radar

Abstract

The invention provides a Guangdong China fir man-made forest biomass addition model construction method based on an unmanned aerial vehicle laser radar, and belongs to the field of model construction, and the method comprises the steps of data acquisition, tree actual survey data acquisition, single tree branch biomass data summarization, single tree biomass model structure establishment, simultaneous equation model establishment and model evaluation index establishment. According to the individual tree biomass model structure, four basic models (a logistic model, a linear model, an exponential model and a power function model) are compared and analyzed, and the power function model is determined as the optimal basic model. The method comprises the following steps: introducing age groups as dumb variables, classifying fir forests according to five different age groups, namely young forests, middle-aged forests, near-mature forests, mature forests and over-mature forests, and incorporating the classified fir forests into a biomass model;

Description

Technical Field

[0001] The present invention relates to the field of model construction, and in particular to a method for constructing a biomass addition model of a Guangdong fir plantation based on an unmanned aerial vehicle (UAV) laser radar. Background Art

[0002] Forests are the backbone of terrestrial ecosystems and occupy a vital position in Earth's ecosystems, playing a significant regulatory role. As a fundamental attribute of forest ecosystems, the measurement and estimation of forest biomass has become a hot topic in forest research and forestry production applications. Aboveground biomass is an essential component of forest ecosystems. Exploring the model structure of biomass of individual tree branches is crucial in forest science and forestry practice. Accurately quantifying forest aboveground biomass is of vital importance for carbon cycle research, ecosystem service assessment, and sustainable forestry management. However, there is considerable uncertainty in estimating forest carbon sinks, and accurate biomass estimation can reduce this uncertainty. Therefore, accurately estimating forest biomass is particularly important. Methods for measuring forest biomass can be divided into direct and indirect methods. Direct methods are time-consuming, labor-intensive, and highly destructive, while the most commonly used indirect method is model estimation. Biomass modeling is the primary method for estimating forest biomass and is an effective and relatively accurate survey method. With advances in mathematical modeling technology, methods for establishing biomass models for individual tree branches have evolved from simple least squares regression to more complex and accurate models that include measurement error compatibility models, including seemingly unrelated regression models (SUR), linear or nonlinear joint estimation models, dummy variables, and mixed-effect models. When processing data at the individual tree scale, these models can better reflect the biomass distribution of various tree parts and provide scientific support for carbon storage and growth analysis at the individual tree level. When it comes to obtaining individual tree biomass parameters, traditional manual survey techniques are time-consuming, labor-intensive, inefficient, and have poor timeliness, making them difficult to adapt to the requirements of precise monitoring of digital forest resources under the new situation.

[0003] Chinese fir, with its excellent wood quality and high economic value in forest management, has become a key economic tree species in subtropical regions. Furthermore, it plays an indispensable role in forest ecosystems. Its aboveground biomass, a major component of forest carbon storage, is crucial for studying the global carbon cycle and addressing climate change. Chinese fir biomass not only influences a forest's carbon storage capacity but is also closely linked to biodiversity and soil quality. Therefore, studying the adaptability and biomass allocation strategies of Chinese fir across its growth cycle has become a hot topic in global forestry research. Significant differences in the biomass allocation strategies of Chinese fir occur across different growth stages. With advancing growth stages, especially in the young and middle forest stages, Chinese fir growth accelerates significantly, with biomass increases primarily concentrated in the trunk and branches. This period of biomass growth is particularly significant and lays a crucial foundation for subsequent carbon storage. Accurately assessing aboveground biomass at different developmental stages not only facilitates more accurate calculations of forest carbon sequestration capacity but also provides a scientific basis for forest management and ecological restoration. Analyzing the biomass structure of Chinese fir branches at different growth stages provides insights into the importance of aboveground biomass in revealing how plants adapt to environmental changes. Therefore, in-depth research on the biomass distribution and growth patterns of Chinese fir will not only help improve the carbon storage capacity of forests, but also promote the development of sustainable forestry and provide support for achieving both ecological and economic benefits.

[0004] Forest biomass can be divided into aboveground and belowground biomass. The aboveground biomass is further divided into four subcomponents: trunk, bark, branches, and leaves. Previous studies have tended to model these subcomponents independently to achieve the required accuracy. However, in forest biomass estimation, data are often affected by multiple errors, resulting in the sum of the biomass of the individual aboveground subcomponents not equaling the total aboveground biomass. Traditional models that fail to properly address these errors can lead to biased estimates and reduced statistical power. Summary of the Invention

[0005] The purpose of this patent is to solve the problems mentioned in the above background technology and provide a method for constructing a biomass addition model of Guangdong fir plantations based on drone lidar.

[0006] The technical solution of the present invention provides: a method for constructing a biomass addition model of a Guangdong fir plantation based on an unmanned aerial vehicle laser radar, comprising the following steps:

[0007] S101: Data collection: Data is collected using an airborne lidar. The raw lidar data is first visualized using Corepore 2.0 software. The visualized data is then interpreted using Radar 360 software. The interpreted lidar point cloud is then used to identify radar structural features, excluding trees, using a distance discriminant clustering algorithm.

[0008] S102: Actual tree survey data collection: Conduct field survey data collection on trees. The collected data excludes dead and fallen trees, dead branches, understory trees, fallen trees, missed measurements, incorrect measurements, and duplicate data. The survey information includes tree species, diameter at breast height, tree height, height below branches, crown width, and growth status.

[0009] S103: Summary of branch biomass data of individual trees: The total biomass is obtained by summarizing the biomass of each branch; the biomass of each branch is calculated as follows (1)-(2), and the total biomass data of trunk, bark, branches, leaves and aboveground of Chinese fir trees in the study area are obtained; and the radar parameters processed by lidar are combined with the biomass of the ground survey to model the total biomass;

[0010]

[0011] Where: MA is the estimated value of aboveground biomass; a0, a1, a2 are model parameters; D is the tree diameter at breast height; H is the tree height; M1, M2, M3, M4 are the estimated values ​​of the biomass of the trunk, bark, branches, and leaves; MA is the aboveground biomass; g1, g2, g3 are the ratios of the biomass of the trunk bark, branches, and leaves to the trunk biomass;

[0012] S104: Single Tree Biomass Model Structure: Based on the collected single tree data from the entire growth cycle of a Chinese fir forest, four biologically significant theoretical tree growth models were considered as the basis for fitting the data: the logistic model, linear model, exponential model, and power function model. The data set was divided into a training sample set and a validation sample set in a ratio of 7:3. 70% of the data was used for modeling and 30% for validation. Radar parameter variables were screened through stepwise regression and a VIF collinearity test was performed. Radar characteristic parameter variables with a VIF greater than 5 were eliminated. Radar tree height (LH) and radar crown width (LCD) were selected as model independent variables to fit the biomass components.

[0013] S105: Construction of dummy variable model: Dummy variables convert categorical variables into binary variables so that they can be included in the regression model for analysis. By introducing the above-mentioned optimal basic parameter model into dummy variables, we can better understand the impact of the same developmental stage on Chinese fir biomass, and thus analyze the impact of different age groups on total biomass. When processing data, the categorical variable "age group AG" is converted into a dummy variable. The categorical variable "age group AG" represents the five age groups of young forest, middle-aged forest, near-mature forest, mature forest, and over-mature forest.

[0014] When age group is used as a dummy variable, it is necessary to transform the age group variable into a quantitative variable, which usually takes a value of 0 or 1 in regression analysis. When there are n independent variables with categorical attributes, it is usually necessary to set one category as a reference, so the number of dummy variables is n-1. The formula is:

[0015]

[0016] S106: Simultaneous equation model: Establish a simultaneous equation model to make the models compatible;

[0017]

[0018] Where: AG1-5 are age groups; a, b, c, d, e, f are model parameters; LH is radar tree height; LCD is radar crown width

[0019] S107: Model evaluation index: The fitting results are evaluated using four evaluation indicators: coefficient of determination R2, root mean square error, total relative error, and Akaike information criterion AIC. When the R2 is larger, the model fitting accuracy is higher, the RMSE is smaller, the model prediction result is higher, the TRE is smaller, the prediction performance is better, and the smaller the AIC is, the better the model fitting effect is. The expression is as follows:

[0020]

[0021] AIC=2k-2ln(L)

[0022] The model formula of the logistic model is

[0023]

[0024] The model formula of the linear model is

[0025] BM=aLH+bLCD+c.

[0026] Model formula of the exponential model

[0027] BM=ae -bLH-cLCD .

[0028] The model formula of the power function model is

[0029]

[0030] Beneficial effects of the present invention:

[0031] This application describes a method for constructing an additive biomass model for full-lifecycle Chinese fir forests in Guangdong using UAV radar data. This technology not only improves the efficiency and accuracy of forestry resource surveys but also provides reliable basic information for forest carbon storage capacity and biodiversity conservation. This lays a solid foundation for long-term monitoring and sustainable management of forest ecosystems, and can be extended to estimate biomass across entire forest stands.

[0032] 2. In order to ensure that the sum of the predicted biomass of each branch of a single tree is equal to the total aboveground biomass and to explore the growth status of Chinese fir forests at different developmental stages, this application establishes a comprehensive model system including nonlinear models, dummy variable models and simultaneous equations models, aiming to improve the accuracy of forest biomass estimation by accurately measuring the aboveground biomass (including trunk, bark, branches and leaves) of Chinese fir at different growth stages, thereby ensuring the consistency of the biomass of each subcomponent with the total biomass. The main objectives of this application are: 1) to use airborne lidar (UAV) data to accurately estimate the biomass of each branch of a single tree in Guangdong Chinese fir forests. 2) to construct a biomass model of Chinese fir at different growth stages, as well as the biomass distribution characteristics of Chinese fir in different growth cycles. 3) to solve the incompatibility problem of the biomass of each branch of a single tree by constructing a compatibility model, thereby ensuring the compatibility between the biomass of each component. BRIEF DESCRIPTION OF THE DRAWINGS

[0033] To more clearly illustrate the technical solutions of the present disclosure, the following briefly introduces the drawings required for use in some embodiments of the present disclosure. Obviously, the drawings described below are only drawings of some embodiments of the present disclosure, and those skilled in the art can also derive other drawings based on these drawings. Furthermore, the drawings described below are schematic diagrams and do not limit the actual dimensions of the products, actual processes of the devices, actual timing of signals, etc. involved in the embodiments of the present disclosure.

[0034] Figure 1 This is a diagram of the airborne radar system.

[0035] Figure 2 This is a radar single tree segmentation processing diagram.

[0036] Figure 3 Prediction graphs for the four basic models.

[0037] Figure 4 It is the residual graph of SUR model. DETAILED DESCRIPTION

[0038] In order to enable those skilled in the art to better understand the technical solutions in this application, the following will provide a clear and complete description of the technical solutions in the embodiments of this application in conjunction with the drawings in the embodiments of this application. Obviously, the embodiments described are only part of the embodiments of this application, not all of the embodiments. Based on the embodiments in this application, all other embodiments obtained by ordinary technicians in this field without making creative efforts should fall within the scope of protection of this application.

[0039] The following will be combined with the drawings in the embodiments of this application to clearly and completely describe the technical solutions in the embodiments of this application. Obviously, the embodiments described are only part of the embodiments of this application, not all of the embodiments. Based on the embodiments in this application, all other embodiments obtained by those skilled in the art without making creative efforts are within the scope of protection of this application.

[0040] The present invention is described in further detail below by way of examples.

[0041] Example

[0042] Data for this study were collected from 133 plots in Lechang City, Yingde City, Heping County, Lianshan County, Longshan County, Yunan County, and Shixing County, Guangdong Province. The province's terrain is primarily composed of mountains, hills, plains, and water, with the terrain rising in the north and falling in the south. Guangdong Province has a subtropical monsoon climate, characterized by warm and humid springs, hot and rainy summers, mild and dry autumns, and cool and dry winters, with significant regional variations. Precipitation in the region exhibits distinct monsoon characteristics. Summers are influenced by the southeast monsoon, resulting in abundant precipitation, while winters are dry and rainless, controlled by dry, cold northwesterly winds.

[0043] Compared to traditional remote sensing, the UAV radar technology used in this invention can penetrate the forest canopy. By emitting laser pulses and receiving reflected signals, UAV radar technology can capture three-dimensional structural information about the forest, providing complete data from treetops to tree bases, which is crucial for estimating tree volume and biomass. This high-precision remote sensing technology captures the detailed three-dimensional structure of the surface and vegetation, providing detailed information about trees and the ground. This information, including tree height, canopy density, and topography, is crucial for estimating biomass. Therefore, radar technology is used to accurately measure the biomass of individual trees and their branches.

[0044] S101: Data Collection

[0045] The airborne laser radar data was collected in March 2024 using the AS-1300HL laser radar system carried by the Huace BB4 UAV for airborne laser scanning. Figure 1 The scan was conducted using a RigelVUX-1LR scanner with a wavelength of 1500nm, a laser pulse length of 3.5ns, and a divergence angle of 0.5mrad. The laser radar's pulse repetition frequency was set to 50kHz, with a maximum scanning angle of 30° and a scanning frequency of 49Hz. A tic-tac-toe pattern was used for flight, ensuring a 50% lateral overlap of the point cloud data. The flight altitude was maintained at 200 meters, with an average flight speed of 10 meters per second, ensuring an average point cloud density of 110 points per square meter for the sample site.

[0046] The raw lidar data obtained was first visualized and interpreted using Corepore2.0 software, and then the visual data was interpreted using Radar 360 software. The interpreted radar point cloud used a distance discriminant clustering algorithm to propose radar structural features excluding trees. The spacing threshold refers to a parameter used when automatically detecting single trees to define the minimum acceptable distance between two trees. When the canopies of trees are close or partially overlapped, setting an appropriate spacing threshold can help distinguish adjacent trees and avoid misidentifying them as one tree. The minimum spacing rule is used in single-tree segmentation to deal with areas with high tree clustering density to ensure that each tree can be identified independently. This is usually done by setting a "forbidden zone" around the detection point of each tree, and no new trees are identified in this area (such as Figure 2 This method was used to segment individual trees within the plot and calculate the key parameters for calculating the biomass of each tree, such as tree height, crown width, and diameter at breast height.

[0047] S102: Actual tree survey data collection:

[0048] Field survey data for trees were collected simultaneously with airborne radar data in March 2024. Ground survey data were collected across 133 150m x 150m plots. Data on fallen trees, dead branches, understory, fallen trees, missed or incorrect measurements, and duplicates were excluded. A total of 20,836 Chinese fir trees were collected, including information on species, diameter at breast height (DBH), height, height below branches, crown width, and growth status. All trees in the study area were planted in Chinese fir forests. Data collection and processing strictly adhered to standard field survey procedures, ensuring data accuracy and reliability.

[0049] S103: Summary of branch biomass data for individual trees:

[0050] Total biomass is calculated by summing the biomass of each branch. The biomass of each branch was calculated using ground-based survey data according to the Chinese national standard "Standing Biomass Model and Carbon-Based Parameters for Major Tree Species" (GB / T43648-2024), as shown in the following table. Total trunk, bark, branch, leaf, and aboveground biomass data were obtained for 20,836 Chinese fir trees within the study area. Modeling was then performed using lidar-processed radar parameters and ground-based biomass data.

[0051]

[0052]

[0053] Where: MA is the estimated value of aboveground biomass; a0, a1, a2 are model parameters; D is the tree diameter at breast height; H is the tree height; M1, M2, M3, M4 are the estimated values ​​of the biomass of the trunk, bark, branches, and leaves; MA is the aboveground biomass; g1, g2, g3 are the ratios of the biomass of the trunk bark, branches, and leaves to the trunk biomass;

[0054] Table 1 Overview of modeling sample data

[0055]

[0056] S104: Single tree biomass model structure:

[0057] From the perspective of individual tree biomass model structure, the model structure is the foundation for determining model construction. This study used individual tree data collected throughout the growth cycle of Guangdong fir forests and considered four biologically significant theoretical tree growth models as the basis for fitting the data: the logistic model, linear model, exponential model, and power function model. This study divided the dataset into a training sample set and a validation sample set in a 7:3 ratio. 70% of the data was used for modeling and 30% for validation. Radar parameter variables were screened through stepwise regression and a VIF collinearity test was performed. Radar characteristic parameter variables with a VIF greater than 5 were removed, and radar tree height (LH) and radar crown width (LCD) were selected as model independent variables to fit the biomass components.

[0058] Basic model formula

[0059]

[0060] Table 24 Detailed parameters and fitting accuracy of basic models

[0061]

[0062]

[0063] S105: Construction of dummy variable model:

[0064] Dummy variables are an important tool for processing categorical variables. They convert categorical variables into binary variables, allowing them to be included in regression models for analysis. By incorporating dummy variables into the optimal basic parameter model described above, we can better understand the impact of the same developmental stage on Chinese fir biomass, thereby analyzing the impact of different age groups on total biomass. When processing the data, the categorical variable "Age Group (AG)" was converted into a dummy variable. The categorical variable "Age Group (AG)" represents five age groups: young forest, middle-aged forest, near-mature forest, mature forest, and overmature forest.

[0065] When age group is used as a dummy variable, it is necessary to transform the age group variable into a quantitative variable, which is usually taken as 0 or 1 in regression analysis. When there are n independent variables with categorical attributes, it is usually necessary to set one category as a reference, so the number of dummy variables is n-1. The formula is:

[0066]

[0067] Table 3 Dummy variable parameters and fitting accuracy

[0068]

[0069] Note: A1 is young forest; A2 is middle-aged forest; A3 is near-mature forest; A4 is mature forest; A5 is over-mature forest

[0070] S106: Simultaneous equation model:

[0071] Accurately estimating the biomass of individual tree branches is crucial in forestry research, particularly in assessing biomass allocation and ecological function in forest ecosystems. When errors are present in both the independent and dependent variables when building a biomass model, traditional modeling methods are no longer suitable for model fitting. To ensure that the predicted value of individual tree biomass equals the combined predicted values ​​for each branch, the additivity or compatibility between the biomass models for each branch must be considered. Ensuring that the sum of each component is compatible with the total is crucial for building a biomass model system. Additive models also account for the inherent relationships between each component and the total. Therefore, it is necessary to establish a system of simultaneous equations to ensure compatibility between the models.

[0072] This study used the seemingly uncorrelated regression (SUR) model to independently model each component of Chinese fir aboveground biomass and jointly solve for model parameters. This approach ensured compatibility between individual tree trunks, bark, branches, and leaves in Chinese fir forests, thereby resolving model incompatibility issues. This approach not only ensured compatibility between biomass components but also resulted in more optimized parameter estimates, making the model more stable.

[0073]

[0074] Where: AG1-5 are age groups; a, b, c, d, e, f are model parameters; LH is radar tree height; LCD is radar crown width

[0075] Table 4 SUR model form and fitting accuracy

[0076]

[0077]

[0078] S107: Model evaluation indicators:

[0079] The fitting results were evaluated using four evaluation indicators: coefficient of determination (R2), root mean square error, total relative error, and Akaike Information Criterion (AIC). A larger R2 indicates a higher model fitting accuracy, a smaller RMSE indicates a higher prediction result, a smaller TRE indicates better prediction performance, and a smaller AIC indicates a better model fitting effect. The expression is as follows:

[0080]

[0081] AIC=2k-2ln(L)

[0082] Result Analysis

[0083] Optimal model selection

[0084] Based on the measured and predicted values ​​of four basic models ( Figure 3 ), the power function model was determined to be the optimal model through comprehensive consideration of the Akaike Information Criterion (AIC), coefficient of determination (R2), root mean square error (RMSEA), and total relative error (TRE). Because the logistic model and power function model had similar validation parameters, the power function model with fewer parameters was selected as the optimal model. The fitting results of the aboveground biomass models showed that, with the exception of leaves and branches, the coefficients of determination (R2) for all other components were above 0.7. The trunk model had the best fitting accuracy, with an R2 of 0.7133; the bark biomass model was second, with an R2 of 0.7052; and the leaf biomass model had the worst fitting accuracy, with an R2 of only 0.5234. The fitting accuracy was as follows: trunk > bark > branches > leaves. Detailed parameters for each model are as follows.

[0085]

[0086]

[0087]

[0088]

[0089] Where: Msg is the trunk, Msp is the bark, Msz is the branches, Msy is the leaves, LH is the radar tree height, and LCD is the radar crown width.

[0090] Dummy variable model

[0091] In this study, the qualitative factors of five age groups, namely young forest, middle-aged forest, near-mature forest, mature forest and over-mature forest, were converted into quantitative factors through the dummy variable method, and introduced into the optimal model of the biomass of each branch of Chinese fir trunk, bark, branch and leaf respectively. By analyzing Chinese fir at different developmental stages, we can better understand the influence of the same developmental stage on the biomass of Chinese fir.

[0092] Comparing the dummy variable model with the optimal base model identified above, calculations revealed an average reduction of 3% in the RMSE. The reduction was most significant for bark, at 4.2%, followed by branches, at 3.1%, and the lowest for trunks and leaves, both at 2.4%. The R² increased by an average of 2.6%. The leaves and branches showed the most significant improvements, increasing accuracy by 3% and 2.8%, respectively, while the remaining sub-items saw an average improvement of 2%. This demonstrates that adding dummy variables to the base model provides a better fit to the data and can be used to accurately estimate biomass of individual trees across different age groups.

[0093] Y=(∑b 0i ×S i )×(LH b )×(LCD c )

[0094] Where: S i is a dummy variable reflecting different age groups (i = 1, 2, ..., 5); b oi are parameters of different forest types; LH is radar tree height, LCD is radar crown width.

[0095] Biomass model fitting

[0096] The compatible aboveground biomass model jointly models the biomass of the trunk, bark, branches, leaves and other sub-items of standing trees and solves the model parameters jointly, thus solving the contradiction between the incompatibility of the sub-items of biomass and the total biomass in the independent models. Figure 4 , the additive biomass model was fitted using the SUR method, and the results showed that the model had a good fitting effect.

[0097] Depend on Figure 4 It can be known that the biomass of each sub-item of trunk, bark, branches, leaves and the total above-ground biomass R 2 The root mean square error (RMSE) for trunk, bark, branch, leaf biomass, and aboveground biomass was 13.6057, 2.2191, 3.4749, 1.7175, and 20.9835, respectively, representing a 2.6% increase compared to the dummy variable model.

[0098] discuss

[0099] In this example, airborne laser ranging (LIDAR) technology was used to collect data from 20,836 Chinese fir trees across 133 plots in Guangdong Province. The biomass of each branch of each individual tree was calculated. Following preliminary optimization of a basic model, age group was introduced as a dummy variable to model each age group of Chinese fir. Finally, a seemingly unrelated regression (SUR) model was used to ensure compatibility between radar tree height (LH), radar crown width (LCD), and individual tree branch biomass, while also ensuring additivity of branch biomass.

[0100] Establishing a universal or regional forest biomass relative growth equation has always been the direction pursued by the forestry and ecological communities. The combination of multiple models and the application of statistical models can better predict forest biomass. However, due to the influence of factors such as habitat, climate and geography, even the same tree species may have significant differences in biomass in different regions. In addition, there are obvious differences in the biomass accumulation and distribution characteristics of Chinese fir at different developmental stages, which means that a simple basic model may not be enough to accurately reflect the dynamic changes in Chinese fir biomass, which may lead to errors and uncertainties in the model equation. In the process of constructing the biomass model of Chinese fir plantations, the dummy variable model containing age group factors showed higher prediction accuracy than the conventional model, which is consistent with the results of the study. Therefore, on the basis of the basic optimal model, age group was introduced as a classification dummy variable, and simulation and modeling were carried out for Chinese fir of different age groups. Studies have shown that after the dummy variable model introduced the age group indicator, its R 2 , TRE and MSEA are all better than the basic model, which shows that the basic model has certain limitations in establishing a large-scale Chinese fir biomass model. By introducing age group as a dummy variable into the model, the differences in biomass at different developmental stages of Chinese fir can be more comprehensively reflected, thereby improving the model fitting effect. Compared with the basic optimal model, the dummy variable model has a better R after introducing different developmental stages. 2 The average increase was 3.2%, indicating that developmental stage significantly influences the accumulation and distribution of Chinese fir biomass. Therefore, adding age group as a dummy variable to the model can improve the prediction accuracy of individual branch biomass, a point that has been consistently reflected in previous studies. Studies have shown that adding age group as a dummy variable to the model effectively improves model fitting accuracy, particularly for trunks, followed by branches and bark, while leaf biomass is least effective. This study's conclusions are consistent with this finding, suggesting that the poor prediction of leaf biomass may be due to the fact that leaves account for the smallest proportion of aboveground biomass.

[0101] Although the fitting results of the model should be kept within a reasonable expected range as much as possible to ensure that its validity and reliability are not affected by deviations, in actual operation, the biomass of each branch of a single tree often shows large variations, and this variation may be due to problems in the data fitting process. In this embodiment, by modeling and analyzing the age group levels of 20,836 Chinese fir trees in Guangdong Province, it was found that the size of the aboveground biomass of Chinese fir trees was ranked as follows: trunk>bark>branches>leaves. The bark and leaf biomass of young forests are relatively small, so in practical applications, it is recommended to establish different model types according to age groups to improve the accuracy of biomass estimation results. The dummy variable model can effectively process the specific data and characteristics of different branches, but relying solely on the dummy variable model to process Chinese fir biomass data may still have limitations, because the dummy variable model may not fully consider the inherent connections and interactions between the branches of the aboveground biomass of a single tree when running independently. This neglect may lead to deviations or inconsistencies in the prediction results, especially when the data set is large.

[0102] By integrating these independent models, the seemingly unrelated regression (SUR) model ensures the mathematical relationship and intrinsic correlation between total aboveground biomass and its branches, i.e., the logical relationship of additivity between the components. Many biomass equations reported in the literature previously lacked additivity, instead constructing independent equations for each component. However, when comparing the fitting accuracy of additive and non-additive biomass models, researchers found that the non-additive approach can lead to significant deviations between the sum of the biomass of individual branches and the total biomass of an individual tree, potentially leading to errors if these models are applied in practice. The SUR model results showed that while the fitting accuracy of individual branches decreased by an average of 2.5%, the RMSE increased by an average of 2.6%. This is likely due to the SUR model's involvement in the joint estimation of multiple equations. This can complicate the calculation of the inverse matrix, particularly when the error terms are highly correlated, thus compromising model performance. In contrast, the dummy variable model may not fully exploit the potential correlations between the equations.

[0103] In summary, this example, through detailed modeling and analysis of Chinese fir biomass, reveals the respective advantages and disadvantages of the dummy variable model and the SUR model when dealing with biomass at different developmental stages and branches. While the dummy variable model excels in analyzing differences between developmental stages, the SUR model addresses the compatibility issue between branches of a single tree. Future research should continue to explore how to further improve the applicability and prediction accuracy of Chinese fir biomass models through model optimization or model combination, thereby providing a more reliable scientific basis for forest management and ecological protection.

[0104] in conclusion:

[0105] This study constructed models for each component of aboveground Chinese fir biomass using both a dummy variable model and a seemingly unrelated regression (SUR) model. The results showed that introducing age group into the dummy variable model significantly improved the prediction accuracy of each component of Chinese fir biomass at different developmental stages, with particularly strong results for branch and bark biomass. The SUR model overcomes the biomass incompatibility issues inherent in traditional models by ensuring compatibility between each component and the total biomass. Although its performance was slightly inferior to that of the dummy variable model in some indicators, it was more stable in addressing the interrelationships between components. Furthermore, data based on UAV radar technology and ground-based measurements provided technical support for the precise estimation of Chinese fir biomass in this study, laying the foundation for long-term monitoring and sustainable management of Chinese fir forest ecosystems.

[0106] The model constructed in this study not only improves the accuracy of natural Chinese fir forest biomass estimation but also provides support for the improvement of tree species biomass models in different regions and the scientific management of forest resources. The results suggest that when constructing aboveground biomass models for different regions, it is important to comprehensively consider the SUR model and introduce dummy variables to improve model accuracy and applicability, thereby addressing the issue of model incompatibility between different regions. Future research should continue to optimize the model or combine multiple model methods to improve the applicability and prediction accuracy of Chinese fir biomass estimation, and consider the impact of different ecological environments or topographic conditions on Chinese fir growth to ensure the applicability and accuracy of the model in different regions.

[0107] The above describes the basic principles and main features of the present invention and the advantages of the present invention. It is obvious to those skilled in the art that the present invention is not limited to the details of the above exemplary embodiments, and that the present invention can be implemented in other specific forms without departing from the spirit or essential characteristics of the present invention. Therefore, from all points of view, the embodiments should be regarded as exemplary and non-restrictive. The scope of the present invention is defined by the appended claims rather than the above description, and it is intended that all changes that fall within the meaning and range of equivalents of the claims are included in the present invention. Any reference signs in the claims should not be construed as limiting the claim to which they relate.

[0108] In addition, it should be understood that although this specification is described in terms of implementation methods, not every implementation method contains only one independent technical solution. This narrative method of the specification is only for the sake of clarity. Those skilled in the art should regard the specification as a whole. The technical solutions in each embodiment can also be appropriately combined to form other implementation methods that can be understood by those skilled in the art.

Claims

1. A method for constructing a biomass additive model for Guangdong fir plantations based on UAV laser radar, characterized by: The following steps are involved: S101: Data collection: Data is collected using an airborne lidar. The raw lidar data is first visualized using Corepore 2.0 software. The visualized data is then interpreted using Radar 360 software. The interpreted lidar point cloud is then used to identify radar structural features, excluding trees, using a distance discriminant clustering algorithm. S102: Actual tree survey data collection: Conduct field survey data collection on trees. The collected data excludes dead and fallen trees, dead branches, understory trees, fallen trees, missed measurements, incorrect measurements, and duplicate data. The survey information includes tree species, diameter at breast height, tree height, height below branches, crown width, and growth status. S103: Summary of branch biomass data of individual trees: The total biomass is obtained by summarizing the biomass of each branch; the biomass of each branch is calculated as follows (1)-(2), and the total biomass data of trunk, bark, branches, leaves and aboveground of Chinese fir trees in the study area are obtained; and the radar parameters processed by lidar are combined with the biomass of the ground survey to model the total biomass; Where: MA is the estimated value of aboveground biomass; a0, a1, a2 are model parameters; D is the tree diameter at breast height; H is the tree height; M1, M2, M3, M4 are the estimated values ​​of the biomass of the trunk, bark, branches, and leaves; MA is the aboveground biomass; g1, g2, g3 are the ratios of the biomass of the trunk bark, branches, and leaves to the trunk biomass; S104: Single Tree Biomass Model Structure: Based on the collected single tree data from the entire growth cycle of a Chinese fir forest, four biologically significant theoretical tree growth models were considered as the basis for the data fitting: the logistic model, the linear function model, the exponential function model, and the power function model. The data set was divided into a training sample set and a validation sample set in a ratio of 7:

3. 70% of the data was used for modeling and 30% for validation. Radar parameter variables were screened through stepwise regression and a VIF collinearity test was performed. Radar characteristic parameter variables with a VIF greater than 5 were eliminated. Radar tree height (LH) and radar crown width (LCD) were selected as model independent variables to fit the biomass components. S105: Construction of dummy variable model: Dummy variables convert categorical variables into binary variables, allowing them to be included in the regression model for analysis. By introducing the above-mentioned optimal basic parameter model into dummy variables, we can better understand the impact of the same developmental stage on Chinese fir biomass, and thus analyze the impact of different age groups on total biomass. When processing data, the categorical variable "Age Group AG" is converted into a dummy variable; the categorical variable "Age Group AG" represents the five age groups of young forest, middle-aged forest, near-mature forest, mature forest, and over-mature forest. When age group is used as a dummy variable, it is necessary to transform the age group variable into a quantitative variable, which usually takes a value of 0 or 1 in regression analysis. When there are n independent variables with categorical attributes, it is usually necessary to set one category as a reference, so the number of dummy variables is n-1. The formula is: S106: Simultaneous equation model: Establish a simultaneous equation model to make the models compatible; Where: AG1-5 are age groups; a, b, c, d, e, f are model parameters; LH is radar tree height; LCD is radar crown width S107: Model evaluation index: The fitting results are evaluated using four evaluation indicators: coefficient of determination R2, root mean square error, total relative error, and Akaike information criterion AIC. When the R2 is larger, the model fitting accuracy is higher, the RMSE is smaller, the model prediction result is higher, the TRE is smaller, the prediction performance is better, and the smaller the AIC is, the better the model fitting effect is. The expression is as follows:

2. The device according to claim 1, characterized in that: The model formula of the logistic model is 3. The device according to claim 1, characterized in that: The model formula of the linear function model BM=aLH+bLCD+c.

4. The device according to claim 1, characterized in that: The model formula of the exponential function model BM=ae -bLH-cLCD 。 5. The device according to claim 1, characterized in that: The model formula of the power function model is

Images