Multi-tree forest aboveground biomass remote sensing estimation method and system and storage medium

Abstract

The invention provides a multi-tree forest aboveground biomass remote sensing estimation method and system and a storage medium, and the method comprises the steps: obtaining multi-source remote sensing data based on a target region, including airborne LiDAR point cloud, ground survey data and DEM data; constructing a terrain heterogeneity index DGTHI in combination with DEM data, and quantifying terrain complexity through weight fusion of standardized terrain factors; extracting forest characteristic parameters and calculating the overground biomass of the single tree; performing classification modeling on the sampling region based on DGTHI, respectively establishing multiple linear regression models for different tree species, and optimizing feature parameter selection; and evaluating the precision of the model through cross validation, and popularizing the model to a county scale to perform overground biomass space inversion and mapping. According to the method, terrain heterogeneity classification and tree species specificity modeling are fused, so that the estimation precision of the forest biomass in the complex terrain region is remarkably improved, and a technical support is provided for regional carbon sink monitoring and sustainable forestry management.

Description

Technical Field

[0001] The present invention belongs to the technical field of remote sensing monitoring of forest resources, and in particular relates to a remote sensing estimation method, system and storage medium for multi-species forest aboveground biomass. Background Art

[0002] As a vital component of terrestrial ecosystems, forests possess a significant carbon sequestration capacity and are a key link in the global carbon cycle. my country is rich in forest resources and boasts a high forest coverage rate. Forest carbon sequestration plays an irreplaceable and important role in achieving the "dual carbon" goals. Above-ground biomass (AGB) is a key indicator for assessing forest carbon storage capacity, and its accurate estimation is of great scientific significance for carbon sink accounting, ecological restoration, and forestry management. However, traditional AGB estimation methods primarily rely on quadrat-scale studies, obtaining parameters such as diameter at breast height and tree height through ground surveys and calculating biomass using allometric growth equations. While highly accurate, such methods are labor-intensive and difficult to generalize to large areas with complex terrain. Furthermore, existing remote sensing studies often focus on single tree species or homogeneous terrain, ignoring the differential effects of terrain heterogeneity on tree species growth and canopy structure, resulting in insufficient model universality. For example, in steep slopes or valley areas, topography-driven microclimate (such as light and water distribution) significantly affects the physiological adaptability of tree species (such as the deep root system of Robinia pseudoacacia enhancing drought resistance and the canopy structure of pine trees optimizing light energy utilization), but existing models mostly use global parameters and cannot capture such spatial differentiation patterns.

[0003] In recent years, remote sensing technology has provided new avenues for AGB estimation. Optical remote sensing infers biomass through vegetation indices (such as NDVI and EVI), but is susceptible to terrain shadowing and spectral saturation. Synthetic aperture radar (SAR) relies on backscatter intensity but lacks sensitivity to vertical canopy structure. In contrast, light detection and ranging (LiDAR) technology directly captures structural parameters such as tree height and canopy density through three-dimensional point clouds, providing a high-precision data foundation for AGB estimation. However, existing LiDAR studies have mostly focused on flat areas or single tree species, lacking systematic modeling of multi-species mixed forests in complex terrain. Studies have shown that topographic heterogeneity (such as elevation fluctuations and slope variations) not only affects LiDAR point cloud feature extraction (e.g., canopy shading effects) but also leads to biased selection of model feature parameters, thereby reducing estimation accuracy. For example, on steep slopes, measurement errors in horizontal structural parameters (such as leaf area index) increase, while the explanatory power of vertical structural parameters (such as the vertical distribution of canopy density) significantly increases. However, existing methods do not dynamically optimize for these differences.

[0004] Although previous studies have attempted to incorporate topographic factors (such as slope and aspect) as auxiliary variables, they often use single indicators or simple linear combinations, failing to fully quantify the coupled effects of topographic spatial heterogeneity on the AGB of multiple tree species. For example, digital elevation models (DEMs) can characterize topographic undulations, but the integration and application of comprehensive indicators such as the standard deviation of elevation within grid cells and surface roughness remains unresolved. Furthermore, parameter selection in multi-species models lacks hierarchy, failing to distinguish between topographically driven basic structural parameters (such as tree height and density) and canopy vertical response parameters (such as leaf area density and canopy heterogeneity), limiting the model's applicability in complex terrain areas.

[0005] In summary, current remote sensing estimation of forest AGB faces two major bottlenecks: (1) The impact of terrain heterogeneity has not been systematically quantified: existing methods have not constructed a comprehensive terrain index, making it difficult to accurately divide heterogeneous areas and optimize feature selection; (2) Multi-species modeling lacks a differentiated strategy: a single model is difficult to adapt to the response mechanism of different tree species to terrain factors, resulting in fluctuations in estimation accuracy. Summary of the Invention

[0006] To address the above problems, the present invention proposes an AGB estimation method that integrates terrain heterogeneity classification with multi-tree species-specific modeling. Regional classification is achieved by constructing a terrain heterogeneity index (DGTHI). The three-dimensional structure parameters of the canopy are extracted using LiDAR point clouds, and a tree species adaptive regression model is established, providing a technological breakthrough for the precise monitoring of forest carbon sinks in complex terrain areas.

[0007] In order to solve the above technical problems, the present invention provides the following technical solutions: A remote sensing estimation method for aboveground biomass of a multi-species forest comprises the following steps: Step 1. Based on the target area, obtain multi-source remote sensing data, including airborne LiDAR point cloud, ground survey data and DEM data; Step 2. Calculate the terrain heterogeneity index based on the DEM data. By integrating the four indicators of elevation standard deviation, terrain relief, surface roughness, and average slope, the target area is divided into three types of heterogeneity: low, medium, and high using the natural breakpoint method. Step 3. Based on the ground survey data obtained in step 1 and the heterogeneity type areas divided in step 3, the aboveground biomass of individual trees of each species and each heterogeneity type area is calculated using the allometric growth equation; Step 4. Preprocess the airborne LiDAR point cloud and extract forest characteristic parameters for each heterogeneous type area; Step 5. Establish a multiple linear regression model between the aboveground biomass of individual trees in each heterogeneity type area and the forest characteristic parameters as a biomass estimation model; Step 6. Screen the optimal feature combination through Pearson correlation analysis and stepwise regression, verify and evaluate the accuracy of the biomass estimation model, and select the optimal estimation model; Step 7. Apply the optimal estimation model to the LiDAR point cloud data of the entire region to complete the spatial inversion and mapping of aboveground biomass of multiple tree species in the entire region.

[0008] Furthermore, in step 1, the ground survey data includes individual tree locations, tree species, diameter at breast height, tree height and coordinate information.

[0009] Furthermore, in step 2, the terrain heterogeneity index is:

[0010] in, Indicates the i Topographic heterogeneity index of each grid cell; Indicates the i The standard deviation of the elevation of each grid cell 、 They represent the minimum and maximum values ​​of the elevation standard deviation in the neighborhood respectively; Indicates the i The topographic relief of each grid cell, 、 Respectively represent the minimum and maximum values ​​of terrain relief in the neighborhood; Indicates the i The surface roughness of each grid cell, 、 represent the minimum and maximum values ​​of surface roughness in the neighborhood respectively; Indicates the i The average slope of the grid cells, 、 represent the minimum and maximum values ​​of the average slope in the neighborhood, respectively; i =1, 2, 3, ..., N Indicates the number of each grid cell; N Indicates the total number of grid cells; 、 、 、 are the contribution weights of elevation standard deviation, terrain relief, surface roughness, and average slope to the terrain heterogeneity index.

[0011] Furthermore, in step 2, the natural breakpoint method is used to divide the target area into three types of heterogeneity: low, medium, and high. The terrain heterogeneity index of the low heterogeneity type area ranges from 0.015 to 0.073; the terrain heterogeneity index of the medium heterogeneity type area ranges from 0.073 to 0.146; and the terrain heterogeneity index of the high heterogeneity type area ranges from 0.146 to 0.290.

[0012] Furthermore, the aboveground biomass of a single tree in step 3 is:

[0013] in, is the aboveground biomass; is the trunk biomass, is the branch biomass, is the leaf biomass.

[0014] Furthermore, in step 4, the forest characteristic parameters include: point cloud point count related variables, height related variables, density related variables and vertical structure variables.

[0015] Furthermore, step 6 includes the following sub-steps: Step 6.1. Screen the characteristic parameters significantly correlated with aboveground biomass using Pearson correlation analysis; Step 6.2. Use backward stepwise regression to eliminate multicollinearity and retain parameters with variance inflation factor (VIF) < 5. Step 6.3. Evaluate the accuracy of the biomass estimation model through five-fold cross-validation and select the coefficient of determination R 2 The model with the highest value and the smallest root mean square error (RMSE) is taken as the optimal estimation model.

[0016] Furthermore, in step 7, the spatial inversion of aboveground biomass of multiple tree species in the entire region includes: Step 7.1. Extract county-scale forest characteristic parameters based on airborne LiDAR point clouds; Step 7.2. Match the optimal model for the corresponding tree species and heterogeneous regions based on the DGTHI classification results; Step 7.3. Generate a spatial distribution map of aboveground biomass of multiple tree species and calculate the total regional carbon storage.

[0017] In another aspect, the present invention provides a remote sensing estimation system for multi-species forest aboveground biomass, comprising: Data acquisition module. It is used to obtain multi-source remote sensing data based on the target area, including airborne LiDAR point cloud, ground survey data and DEM data; DGTHI classification module. It is used to calculate the terrain heterogeneity index based on DEM data. By integrating four indicators, namely elevation standard deviation, terrain relief, surface roughness, and average slope, the target area is divided into three types of heterogeneity, namely low, medium, and high, using the natural breakpoint method. The aboveground biomass calculation module is used to calculate the aboveground biomass of individual trees of each species and each heterogeneous type area using the allometric growth equation based on the acquired ground survey data and the divided heterogeneous type area; Feature parameter extraction module. It is used to pre-process the airborne LiDAR point cloud and extract forest feature parameters of various heterogeneous areas; A biomass estimation model construction module is used to establish a multiple linear regression model between the aboveground biomass of individual trees of various tree species and various heterogeneous types and the forest characteristic parameters as a biomass estimation model; Model optimization module. It is used to screen the optimal feature combination through Pearson correlation analysis and stepwise regression, verify and evaluate the accuracy of the biomass estimation model, and select the optimal estimation model; Inversion output module. It is used to apply the optimal estimation model to the LiDAR point cloud data of the entire region to complete the spatial inversion and mapping of the aboveground biomass of multiple tree species in the entire region; The multi-species forest aboveground biomass remote sensing estimation system is used to execute the steps in the multi-species forest aboveground biomass remote sensing estimation method according to any one of claims 1 to 8.

[0018] On the other hand, the present invention provides a storage medium storing computer-readable instructions. When the computer-readable instructions are executed by one or more processors, the one or more processors execute the steps of the method for remote sensing estimation of aboveground biomass in multi-species forests.

[0019] Compared with the prior art, the present invention has the following beneficial effects: 1. Existing remote sensing studies of forest biomass often ignore the impact of topographic heterogeneity. This paper innovatively proposes a terrain heterogeneity index (DGTHI). By integrating multiple indicators to quantify terrain complexity, this method enables scientific zoning of the study area and addresses the issue of low estimation accuracy in complex terrain areas. 2. In view of the differences in the responses of different tree species to terrain, the present invention establishes a tree species-specific estimation model, and through the dynamic optimization of characteristic parameter selection strategy, the model R 2 Improved by 0.06-0.21 (the maximum improvement was 0.21 for Robinia pseudoacacia), significantly improving the estimation accuracy; 3. Traditional methods generally have relative errors exceeding 20% ​​at the county level. However, this invention, through terrain heterogeneity classification modeling, controls the relative error of estimates for each tree species to within 15% (as low as 4% for pine trees), providing reliable technical support for accurate monitoring of forestry carbon sinks. 4. The technical framework proposed in this invention can be extended to other mountain forest areas, providing a new methodological reference for global forest carbon cycle research and the realization of the "dual carbon" goals. BRIEF DESCRIPTION OF THE DRAWINGS

[0020] In order to more clearly illustrate the technical solutions in the present invention or the prior art, the following briefly introduces the drawings required for use in the embodiments or the description of the prior art. Obviously, the drawings described below are some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying any creative work.

[0021] Figure 1 Schematic diagram of the process of remote sensing estimation of aboveground biomass of multi-species forests based on terrain heterogeneity classification according to an embodiment of the present invention.

[0022] Figure 2 A scatter plot of predicted values ​​and measured values ​​of a remote sensing estimation model for aboveground biomass of a poplar forest based on terrain heterogeneity classification according to an embodiment of the present invention; Figure 3 A scatter plot of predicted values ​​and measured values ​​of a pine forest aboveground biomass remote sensing estimation model based on terrain heterogeneity classification according to an embodiment of the present invention; Figure 4 A scatter plot of predicted values ​​and measured values ​​of a cypress forest aboveground biomass remote sensing estimation model based on terrain heterogeneity classification according to an embodiment of the present invention; Figure 5 This is a scatter plot of the predicted values ​​and measured values ​​of the remote sensing estimation model for aboveground biomass of Robinia pseudoacacia forest based on terrain heterogeneity classification in an embodiment of the present invention. DETAILED DESCRIPTION

[0023] To make the objectives, technical solutions, and advantages of the present invention more clear, the technical solutions of the present invention will be clearly and completely described below in conjunction with the accompanying drawings. Obviously, the embodiments described are only some embodiments of the present invention, not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without making creative efforts shall fall within the scope of protection of the present invention.

[0024] Example 1 The present invention will be further described below with reference to the accompanying drawings.

[0025] The present invention provides a remote sensing estimation scheme for aboveground biomass of multi-species forests based on terrain heterogeneity classification, comprising the following steps: acquiring multi-source remote sensing data (airborne LiDAR point cloud, ground survey data, and DEM data) based on a target area, and selecting a representative sampling area from the data; calculating the aboveground biomass of individual trees of each tree species using the ground survey data; preprocessing the LiDAR point cloud and extracting forest characteristic parameters; constructing a terrain heterogeneity index (DGTHI) based on the DEM data, and dividing the study area into three heterogeneity types (low, medium, and high) using the natural breakpoint method by integrating four indicators: elevation standard deviation, terrain relief, surface roughness, and average slope; establishing multiple linear regression models for different tree species (poplar, pine, cypress, and locust) and different DGTHI types, and selecting the optimal feature combination through feature importance analysis and stepwise regression; evaluating the model accuracy using five-fold cross-validation; and finally applying the optimal model to the LiDAR point cloud data of the entire region to achieve spatial inversion and mapping of aboveground biomass of multiple tree species. The present invention solves the technical problem of insufficient accuracy in forest biomass estimation in complex terrain areas by innovatively introducing a terrain heterogeneity classification system and combining it with a multi-tree species-specific modeling method.

[0026] See also Figure 1 When it is necessary to estimate the aboveground biomass of a multi-species forest, the multi-species forest aboveground biomass remote sensing estimation method based on terrain heterogeneity classification provided by the embodiment of the present invention can be used to perform the following steps to achieve accurate analysis: Step S01, based on the target area, select several sampling areas from it, and obtain ground survey data of single trees of multiple tree species in the sampling areas; Furthermore, the individual tree ground survey data include location, diameter at breast height, tree height and tree species type.

[0027] This example surveyed individual tree data from 552 10m×10m plots in Mengyin County, Shandong Province, from March to December 2023. A total of 8,804 tree samples were collected, including four dominant tree species: poplar, pine, cypress, and locust. The specific measurement methods are as follows: Spatial coordinate measurement: Individual tree positioning was performed using a Qianxun RTK device (model: DG50), with a horizontal accuracy of ±2.5mm and a vertical accuracy of ±5mm. Three control points were set up in each plot, using static observation mode, with each observation lasting at least 30 minutes. DBH measurement: Measurements were taken at 1.3m from the trunk using a diameter girder (accuracy of 0.1mm), with two measurements per tree taken and the average taken. For samples located on sloping terrain, measurements were maintained at 1.3m upslope. Tree height measurement: Measurements were taken using a laser altimeter (model: TruPulse360), with a ranging accuracy of ±0.01m. Three measurements were taken from different locations for each tree, with the average taken to ensure data accuracy.

[0028] Then, according to the collection time period of the ground survey data, the UAV laser point cloud and DEM data of the sampling area within the collection time period are obtained; In this embodiment of the present invention, the airborne LiDAR data is collected using a CityMapper-2L sensor mounted on a fixed-wing UAV platform, with a flight altitude of 800m, a heading overlap rate of 70%, and a lateral overlap rate of 50%. Point cloud density > 8pts / m 2 The data contains information such as X / Y / Z coordinates, intensity, and echo count. The DEM data is obtained from the 2023 version of the 30m resolution DEM provided by the National Geographic Information Public Service Platform. It uses the WGS84 coordinate system and has an elevation accuracy of ±5m. Data preprocessing includes projection conversion, outlier removal, and gap filling.

[0029] Step S02, preprocessing the laser point cloud data and extracting feature parameters; This embodiment of the present invention uses the professional point cloud processing software LASTOOLS to systematically process the acquired airborne LiDAR point cloud data. Data preprocessing begins with the following key steps: 1) Point cloud denoising, using a statistical outlier removal algorithm with a neighborhood point count of 50 and a standard deviation multiple of 1.5 to effectively remove flyby and noise points; 2) Ground point classification, using a progressive triangulation algorithm with an initial angle of 8°, a maximum angle of 15°, and a maximum distance of 2.0m to accurately separate ground points from non-ground points; and 3) Height normalization, using a digital elevation model (DEM) as a reference to convert all point cloud height values ​​relative to the ground to eliminate the effects of terrain. The preprocessed point cloud data retains complete vegetation structure information while removing terrain and noise interference.

[0030] During the feature parameter extraction phase, the present invention systematically extracted 40 feature variables from four dimensions: 1) point cloud count-related variables, including variables such as P_total, P_hv, and P_fe, which are related to the number of vegetation points in the point cloud; 2) height-related variables, including height quantiles such as hp_25, hp_50, H_max, H_min, and H_mean, which are related to vegetation height; 3) density-related variables, including density quantiles such as dp_25, dp_50, CC, and BCC, which are related to vegetation density; and 4) vertical structure variables, including variables such as LAD_mean, LAD_std, LAD_cv, VFP_mean, and VFP_std, which describe the vertical heterogeneity of vegetation canopies. Specific feature parameters are shown in Table 1. All feature parameters were statistically calculated using a 10m×10m grid to ensure consistency with the ground survey plot scale. The calculation method is shown in Table 2. The extracted feature data are stored in a structured database, including the sample number, tree species code and the values ​​of each feature parameter, providing complete data support for subsequent modeling.

[0031] Table 2 Characteristic parameters and their meanings

[0032] Table 3 Characteristic parameter calculation method

[0033] Step S03: construct the terrain heterogeneity index (DGTHI) and perform regional classification. This embodiment of the present invention systematically calculates the terrain heterogeneity index based on 30-meter resolution DEM data using the ArcGIS 10.8 software platform. First, a 3×3 grid analysis window is set up, and four core terrain indicators are calculated within each window: 1) elevation standard deviation (σ), which reflects the degree of elevation fluctuation within the cell; 2) terrain relief (RF = Hmax - Hmin), which characterizes local elevation variations; 3) surface roughness (KR), calculated as the ratio of surface area to projected area; and 4) mean slope (Slope_mean), obtained using the eight-neighborhood algorithm. After each indicator is calculated, it is normalized using the range normalization method to eliminate dimensionality effects.

[0034] The DGTHI is constructed using a weighted fusion method. The optimal weights for each indicator in the embodiment of the present invention are determined using a BP neural network model: elevation standard deviation 24.95%, terrain relief 25.19%, surface roughness 24.68%, and average slope 25.18%. The specific calculation formula is:

[0035] in, Indicates the i Topographic heterogeneity index of each grid cell; Indicates the i The standard deviation of the elevation of each grid cell 、 They represent the minimum and maximum values ​​of the elevation standard deviation in the neighborhood respectively; Indicates the i The topographic relief of each grid cell, 、 Respectively represent the minimum and maximum values ​​of terrain relief in the neighborhood; Indicates the i The surface roughness of each grid cell, 、 represent the minimum and maximum values ​​of surface roughness in the neighborhood respectively; Indicates the i The average slope of the grid cells, 、 represent the minimum and maximum values ​​of the average slope in the neighborhood, respectively; i =1, 2, 3, ..., N Indicates the number of each grid cell; N Indicates the total number of grid cells; 、 、 、 are the contribution weights of elevation standard deviation, terrain relief, surface roughness, and average slope to the terrain heterogeneity index.

[0036] Based on the DGTHI values, the present study used the Jenks natural breakpoint method to categorize the study area into three heterogeneous regions: 1) low heterogeneity (0.015-0.073), primarily consisting of gentle valleys and terraces; 2) moderate heterogeneity (0.073-0.146), corresponding to the hilly transition zone; and 3) high heterogeneity (0.146-0.290), encompassing steep mountainous terrain and canyon areas. The classification results were validated using a 1:10,000 topographic map, achieving an overall accuracy of 89%. The resulting topographic heterogeneity classification map for the study area, containing classification information for 552 sample plots, served as the basis for subsequent zoning modeling.

[0037] Step S04, calculating the aboveground biomass of individual trees using the individual tree ground survey data obtained in step S01; The embodiment of the present invention uses the allometric growth equation as a method for calculating the aboveground biomass of trees. The specific allometric growth equation is shown in Table 3.

[0038] Table 3 Allometric growth equation

[0039] D is the tree's diameter at breast height, H is the tree's height, and the total above-ground biomass = trunk biomass + branch biomass + leaf biomass.

[0040] Step S05: Establishing a tree species-specific aboveground biomass estimation model. The embodiment of the present invention establishes differentiated multiple linear regression models for the four dominant tree species of poplar, pine, cypress and locust in the implementation area in different terrain heterogeneity areas (low, medium and high). The modeling process is completed using SPSS 26.0 statistical analysis software. First, the 40 characteristic parameters extracted in step S03 are subjected to Pearson correlation analysis (significance level p < 0.05) to screen out characteristic variables that are significantly correlated with the biomass of each tree species in each terrain heterogeneous area. In order to eliminate the influence of multicollinearity, the stepwise regression method is used for variable selection, and the variance inflation factor (VIF) threshold is set to 5 to ensure the independence of the selected variables. The optimal modeling of each tree species heterogeneous area is shown in the following table. Table 4 Optimal modeling results of multiple linear regression models for each tree species in different regions

[0041] The model parameters were optimized using a grid search method, and the optimal parameter combination was determined through five-fold cross-validation. For each validation, 80% of the samples were randomly selected as the training set, and the remaining 20% ​​were used as the validation set. The calculation was repeated 100 times and the average value was obtained. All models were tested by residual analysis to ensure that the assumptions of normality and homogeneity of variance were met. Among the 12 tree species and region models (4 tree species × 3 heterogeneity types) that were finally established, the coefficient of determination R 2 The range was 0.68-0.91, and the model improvement of Robinia pseudoacacia in the high heterogeneity area was the most significant, R 2 The global model's accuracy improved from 0.36 to 0.57. The model results are stored in a geographic database, including metadata such as model coefficients, accuracy indicators, and applicable conditions, to support subsequent regional-scale biomass inversion applications.

[0042] Step S06: Verify and optimize the accuracy of each tree species model obtained in step S05. The present invention uses a systematic validation method to comprehensively evaluate the 12 established tree species and region models. The validation process is based on an independent validation data set that contains a randomly retained 20% sample (a total of 1761 tree samples) to ensure the independence of the validation data and the modeling data. Validation indicators include the coefficient of determination (R 2 ), root mean square error (RMSE), mean absolute error (MAE) and relative error (RE). All indicators are calculated using the Python scikit-learn library. Taking the pine tree model as an example, the validation results in the low heterogeneity area show that R 2 The RMSE was 2.87 t / ha, the MAE was 2.15 t / ha, and the RE was 12.3%. All indicators met the accuracy standards required by forestry survey specifications.

[0043] To improve the adaptability of the model under complex terrain conditions, the present invention implemented three-level optimization measures: 1) Feature engineering optimization, re-screening feature combinations through recursive feature elimination (RFE) to increase the R2 of the Robinia pseudoacacia model by 0.05; 2) Model structure optimization, introducing the Ridge Regression algorithm for high heterogeneity areas, effectively reducing the risk of overfitting, and reducing the RMSE of the poplar model by 8.7%; 3) Residual spatial autocorrelation analysis, using the Moran's I index to detect residual spatial clustering, and adding spatial lag terms to models with significant spatial autocorrelation (p < 0.05) for correction. The optimized model system performed well in county-scale verification, with an overall R2 of 1.5 for each tree species. 2 The improvement ranged from 0.06 to 0.21, with the Robinia pseudoacacia model showing the most significant improvement. The final model validation results were presented in the form of thematic maps, including a scatter plot of predicted values ​​versus measured values ​​( Figure 2-Figure 5 ), residual space distribution map and error statistics table, providing a complete accuracy reference for model application.

[0044] Step S07: using the optimal estimation model determined in step S06, quantitatively estimate the aboveground biomass of trees in the entire estimation area.

[0045] The above process provides a remote sensing estimation method for multi-species forest aboveground biomass based on terrain heterogeneity classification. It obtains multi-source remote sensing data (including airborne LiDAR point cloud, ground survey data, and DEM data) from the target area, and selects representative sampling areas from them. The ground survey data are used to calculate the aboveground biomass of individual trees of each tree species. The LiDAR point cloud is preprocessed and forest characteristic parameters are extracted. A terrain heterogeneity index (DGTHI) is constructed based on DEM data. By integrating four indicators: elevation standard deviation, terrain relief, surface roughness, and average slope, the study area is divided into three heterogeneity types: low, medium, and high, using the natural breakpoint method. Multiple linear regression models are established for different tree species (poplar, pine, cypress, and locust) and different DGTHI types. The optimal feature combination is selected through feature importance analysis and stepwise regression. The model accuracy is evaluated using five-fold cross-validation. Finally, the optimal model is applied to the LiDAR point cloud data of the entire region to achieve spatial inversion and mapping of the aboveground biomass of multiple tree species. The results obtained are used for regional forest carbon storage analysis, carbon sequestration potential assessment and abnormal warning. For example, when it is detected that the biomass of Robinia pseudoacacia in a certain area is significantly lower than that of similar areas, early warning information is sent to the forestry management department through a preset communication link to assist in forest resource protection and ecological restoration decision-making.

[0046] Example 2 This embodiment provides a multi-species forest aboveground biomass remote sensing estimation system, comprising: Data acquisition module. It is used to obtain multi-source remote sensing data based on the target area, including airborne LiDAR point cloud, ground survey data and DEM data; DGTHI classification module. It is used to calculate the terrain heterogeneity index based on DEM data. By integrating four indicators, namely elevation standard deviation, terrain relief, surface roughness, and average slope, the target area is divided into three types of heterogeneity, namely low, medium, and high, using the natural breakpoint method. The aboveground biomass calculation module is used to calculate the aboveground biomass of individual trees of each species and each heterogeneous type area using the allometric growth equation based on the acquired ground survey data and the divided heterogeneous type area; Feature parameter extraction module. It is used to pre-process the airborne LiDAR point cloud and extract forest feature parameters of various heterogeneous areas; A biomass estimation model construction module is used to establish a multiple linear regression model between the aboveground biomass of individual trees of various tree species and various heterogeneous types and the forest characteristic parameters as a biomass estimation model; Model optimization module. It is used to screen the optimal feature combination through Pearson correlation analysis and stepwise regression, verify and evaluate the accuracy of the biomass estimation model, and select the optimal estimation model; Inversion output module. It is used to apply the optimal estimation model to the LiDAR point cloud data of the entire region to complete the spatial inversion and mapping of the aboveground biomass of multiple tree species in the entire region; The multi-species forest aboveground biomass remote sensing estimation system is used to execute the steps in the multi-species forest aboveground biomass remote sensing estimation method according to any one of claims 1 to 8.

[0047] Example 3 This embodiment provides a storage medium storing computer-readable instructions. When the computer-readable instructions are executed by one or more processors, the one or more processors execute the steps of the method for remote sensing estimation of aboveground biomass in multi-species forests.

[0048] It should be understood that parts not elaborated in detail in this specification belong to the prior art.

[0049] It should be understood that the above description of the preferred embodiments is relatively detailed and cannot be considered as limiting the scope of protection of the present invention. It is not necessary and impossible to list all embodiments here. Under the guidance of the present invention, ordinary technicians in this field can also make substitutions or modifications without departing from the scope of protection of the claims of the present invention, which fall within the scope of protection of the present invention. The scope of protection of the present invention shall be based on the attached claims.

Claims

1. A remote sensing method for estimating aboveground biomass of a multi-species forest, characterized by: The following steps are included: Step 1. Based on the target area, obtain multi-source remote sensing data, including airborne LiDAR point cloud, ground survey data and DEM data; Step 2. Calculate the terrain heterogeneity index based on the DEM data. By integrating the four indicators of elevation standard deviation, terrain relief, surface roughness, and average slope, the target area is divided into three types of heterogeneity: low, medium, and high using the natural breakpoint method. Step 3. Based on the ground survey data obtained in step 1 and the heterogeneity type areas divided in step 3, the aboveground biomass of individual trees of each species and each heterogeneity type area is calculated using the allometric growth equation; Step 4. Preprocess the airborne LiDAR point cloud and extract forest characteristic parameters for each heterogeneous type area; Step 5. Establish a multiple linear regression model between the aboveground biomass of individual trees in each heterogeneity type area and the forest characteristic parameters as a biomass estimation model; Step 6. Screen the optimal feature combination through Pearson correlation analysis and stepwise regression, verify and evaluate the accuracy of the biomass estimation model, and select the optimal estimation model; Step 7. Apply the optimal estimation model to the LiDAR point cloud data of the entire region to complete the spatial inversion and mapping of aboveground biomass of multiple tree species in the entire region.

2. The method for estimating aboveground biomass of a multi-species forest by remote sensing according to claim 1, characterized in that: In step 1, the ground survey data includes the location of individual trees, tree species, diameter at breast height, tree height and coordinate information.

3. The method for estimating aboveground biomass of a multi-species forest by remote sensing according to claim 1, characterized in that: In step 2, the terrain heterogeneity index is: in, Indicates the i Topographic heterogeneity index of each grid cell; Indicates the i The standard deviation of the elevation of each grid cell 、 They represent the minimum and maximum values ​​of the elevation standard deviation in the neighborhood respectively; Indicates the i The topographic relief of each grid cell, 、 Respectively represent the minimum and maximum values ​​of terrain relief in the neighborhood; Indicates the i The surface roughness of each grid cell, 、 represent the minimum and maximum values ​​of surface roughness in the neighborhood respectively; Indicates the i The average slope of the grid cells, 、 represent the minimum and maximum values ​​of the average slope in the neighborhood, respectively; i =1, 2, 3, ..., N Indicates the number of each grid cell; N Indicates the total number of grid cells; 、 、 、 are the contribution weights of elevation standard deviation, terrain relief, surface roughness, and average slope to the terrain heterogeneity index.

4. The method for estimating aboveground biomass of a multi-species forest by remote sensing according to claim 3, characterized in that: In step 2, the natural breakpoint method is used to divide the target area into three types of heterogeneity: low, medium, and high. The terrain heterogeneity index of the low heterogeneity type area ranges from 0.015 to 0.073; the terrain heterogeneity index of the medium heterogeneity type area ranges from 0.073 to 0.146; and the terrain heterogeneity index of the high heterogeneity type area ranges from 0.146 to 0.

290.

5. The method for estimating aboveground biomass of a multi-species forest by remote sensing according to claim 1, characterized in that: The aboveground biomass of a single tree in step 3 is: in, is the aboveground biomass; is the trunk biomass, is the branch biomass, is the leaf biomass.

6. The method for estimating aboveground biomass of a multi-species forest by remote sensing according to claim 1, characterized in that: In step 4, the forest characteristic parameters include: point cloud point count related variables, height related variables, density related variables and vertical structure variables.

7. The method for estimating aboveground biomass of a multi-species forest by remote sensing according to claim 1, characterized in that: The step 6 includes the following sub-steps: Step 6.

1. Screen the characteristic parameters significantly correlated with aboveground biomass using Pearson correlation analysis; Step 6.

2. Use backward stepwise regression to eliminate multicollinearity and retain parameters with variance inflation factor (VIF) < 5. Step 6.

3. Evaluate the accuracy of the biomass estimation model through five-fold cross-validation and select the coefficient of determination R 2 The model with the highest value and the smallest root mean square error (RMSE) is taken as the optimal estimation model.

8. The method for estimating aboveground biomass of a multi-species forest by remote sensing according to claim 1, characterized in that: In step 7, the spatial inversion of aboveground biomass of multiple tree species in the entire region includes: Step 7.

1. Extract county-scale forest characteristic parameters based on airborne LiDAR point clouds; Step 7.

2. Match the optimal model for the corresponding tree species and heterogeneous regions based on the DGTHI classification results; Step 7.

3. Generate a spatial distribution map of aboveground biomass of multiple tree species and calculate the total regional carbon storage.

9. A remote sensing estimation system for aboveground biomass of multi-species forests, characterized by: include: Data acquisition module. It is used to obtain multi-source remote sensing data based on the target area, including airborne LiDAR point cloud, ground survey data and DEM data; DGTHI classification module. It is used to calculate the terrain heterogeneity index based on DEM data. By integrating four indicators, namely elevation standard deviation, terrain relief, surface roughness, and average slope, the target area is divided into three types of heterogeneity, namely low, medium, and high, using the natural breakpoint method. The aboveground biomass calculation module is used to calculate the aboveground biomass of individual trees of each species and each heterogeneous type area using the allometric growth equation based on the acquired ground survey data and the divided heterogeneous type area; Feature parameter extraction module. It is used to pre-process the airborne LiDAR point cloud and extract forest feature parameters of various heterogeneous areas; A biomass estimation model construction module is used to establish a multiple linear regression model between the aboveground biomass of individual trees of various tree species and various heterogeneous types and the forest characteristic parameters as a biomass estimation model; Model optimization module. It is used to screen the optimal feature combination through Pearson correlation analysis and stepwise regression, verify and evaluate the accuracy of the biomass estimation model, and select the optimal estimation model; Inversion output module. It is used to apply the optimal estimation model to the LiDAR point cloud data of the entire region to complete the spatial inversion and mapping of the aboveground biomass of multiple tree species in the entire region; The multi-species forest aboveground biomass remote sensing estimation system is used to execute the steps in the multi-species forest aboveground biomass remote sensing estimation method according to any one of claims 1 to 8.

10. A storage medium, characterized in that: The storage medium stores computer-readable instructions, which, when executed by one or more processors, enable the one or more processors to execute the steps of the method for remote sensing estimation of aboveground biomass of a multi-species forest as described in any one of claims 1 to 8.

Images