Remote sensing metering method for carbon reserve variable quantity of highly-closed forest

Abstract

The invention relates to the technical field of graphic attribute analysis, and discloses a high-canopy forest carbon reserve variation remote sensing measurement method, which comprises the following steps of: firstly, carrying out single-tree crown segmentation on a forest, matching trees in different time phases, and then judging the survival state change of the trees in two time phases by utilizing the change of the number of point cloud data points; the method comprises the following steps: classifying trees into normal trees, morning trees, post-morning trees, cut trees and new trees based on survival states of different trees, formulating different carbon reserve variable quantity metering methods one by one, and substituting remote sensing data into a model to calculate the carbon reserve variable quantity of each tree after modeling; the carbon reserve variable quantity of each tree is calculated and then added to obtain accurate carbon reserve variable quantity data, so that the adverse effect of upper canopy shielding on remote sensing is overcome. The ground survey data is only used for modeling, does not need to be obtained from the forest to be measured, is not affected by the terrain of the forest to be measured, and completely gets rid of the ground survey data after modeling is completed.

Description

Technical Field

[0001] The present invention relates to the technical field of graph attribute analysis, and in particular to a remote sensing measurement method for carbon storage variation in high-canopy forests. Background Art

[0002] Since there are a lot of uncommon and complex terms involved, for ease of understanding, here is a glossary: Point cloud: A data set consisting of countless coordinate points, which correspond to points on the surface of the object being measured. Each point usually also contains information such as color. The graphics composed of these points are similar to sand paintings, where dense, sand-like points form an image. Point cloud segmentation: classify the points in the point cloud and determine which points belong to a certain object, for example, which points belong to a certain tree. After point cloud segmentation, the shape of the object that can be recognized by the machine can be obtained (before classification, the outline of these "sand paintings" can be seen by the human eye, but the machine cannot recognize it); Point cloud elevation normalization: The forest surface is uneven, and the collected canopy point cloud data cannot reflect the true height of the tree. Therefore, it is necessary to subtract the elevation of the corresponding ground point from the elevation of the point on the canopy. After point cloud elevation normalization, it is equivalent to planting a forest on a flat ground.

[0003] Canopy density refers to the ratio of the total projected area (crown width) of the tree crowns in the forest under direct sunlight on the ground to the total area of ​​the forest land (stand), which reflects the density of the stand.

[0004] Forest carbon storage: the amount of carbon retained in the forest. If the forest carbon storage increases over a period of time and the increased carbon storage is no longer released, then this process forms a carbon sink.

[0005] Forest carbon storage is an important basis for measuring the ecological value of forests and calculating carbon sinks. It can be calculated by plot survey method, biomass method, or remote sensing estimation method. Among them, remote sensing estimation method is an important and even irreplaceable method for measuring forest carbon storage, because the workload of on-site measurement is huge, and there are even many areas that cannot be measured on-site due to terrain reasons.

[0006] But no matter which method is used, tree height is an indispensable value. Because for a tree, the size of the crown, the tree's diameter at breast height and the tree's height have a great impact on its carbon storage. The tree's diameter at breast height and the size of the crown are usually linearly related. The larger the crown, the larger the tree's diameter at breast height. But the size of the crown and the height of the tree are not one-to-one corresponding. A typical example is Populus tomentosa. In a densely planted environment, Populus tomentosa will have a very small crown and a very high tree height, while in the wilderness, Populus tomentosa will have a very large crown and a much lower tree height.

[0007] In high-density forests, it is difficult to obtain ground points through remote sensing due to the obstruction of the upper forest canopy, and thus it is difficult to obtain tree heights and the growth status of understory vegetation through remote sensing. These missing data can only be obtained through ground surveys, and the advantages of remote sensing estimation methods, such as low workload and no terrain selection, no longer exist. Summary of the invention

[0008] The invention provides a remote sensing measurement method for carbon storage variation in high-canopy forests.

[0009] The technical problem to be solved is: in a forest with high canopy density, it is difficult to obtain ground points through remote sensing due to the obstruction of the upper forest canopy, and thus it is difficult to obtain the height of trees and the growth status of understory vegetation through remote sensing, which brings difficulties to the remote sensing method of measuring carbon storage.

[0010] In order to solve the above technical problems, the present invention adopts the following technical solutions: a remote sensing measurement method for carbon storage changes in high-canopy forests, which is used to measure the carbon storage changes in forest land where it is difficult to collect ground point cloud data by remote sensing means due to canopy closure. During the measurement process, the carbon storage changes of understory shrubs and herbaceous plants are recorded as 0 and are not included in the measurement; The measuring method comprises the following steps: Step 1: Collect remote sensing data of the forest land before and after the change. The time when the data was collected before the change is recorded as the pre-phase, and the time when the data was collected before the change is recorded as the post-phase. The pre-phase and the post-phase are the same time on different dates, and the weather when the data was collected is sunny during the vigorous growth period of the forest; Step 2: Use the data collected in step 1 to establish non-elevation normalized point clouds of the front phase and the back phase respectively, and smooth and denoise the point clouds; Step 3: Segment the crowns of individual trees based on the non-elevation normalized point cloud, and then match the trees in the previous phase with those in the next phase, and match the same tree in the previous phase with those in the next phase; Step 4: Determine the tree species and survival status of the segmented trees, and classify the segmented trees into the following categories: Normal tree: The same tree exists in both the previous and the later phases and survives in both phases. Early-death tree: The same tree exists in both the previous and later phases and dies in both phases; Post-mortem tree: The same tree exists in both the previous phase and the later phase, surviving in the previous phase and dying in the later phase; Felled tree: A tree that exists only in the previous phase; Newborn tree: A tree that exists only in the later phase; Step 5: Establish a mathematical model with tree species, crown characteristic values, and tree height as independent variables and carbon storage as the dependent variable, and record it as a new model; A mathematical model was established with tree species, crown characteristic difference, and tree height difference as independent variables and carbon storage change as dependent variable, and recorded as an incremental model, where the crown characteristic difference and tree height difference were the crown characteristic value and tree height value after the change minus the crown characteristic value and tree height value before the change; Step 6: Calculate the carbon storage changes for different types of trees and then sum them up, where: Change in carbon storage of normal trees: Substitute the tree species, crown characteristic differences, and tree height differences into the incremental model to calculate the change in carbon storage; Change in carbon storage of trees that died early: The change in carbon storage was recorded as 0; Changes in carbon storage of late-living trees: Substitute tree species, crown characteristic values, and tree height values ​​into the new model, calculate carbon storage, and then take the inverse; Change in carbon storage of felled trees: Substitute tree species, crown characteristic values, and tree height values ​​into the new model, calculate carbon storage, and then take the inverse; Change in carbon storage of new trees: Substitute tree species, crown characteristic values, and tree height values ​​into the new model to calculate carbon storage.

[0011] Furthermore, in step one, the image data is collected using the UAV oblique photogrammetry method.

[0012] Furthermore, in step 3, the following method is used to determine whether the tree in the previous phase and the tree in the next phase are the same tree: After excluding the false tree vertices, if the tree vertex of tree B in the later phase falls within the tree crown range of tree A in the previous phase, and there is only one tree B, then tree A and tree B are considered to be the same tree; After excluding false tree vertices, if the tree vertex of tree B in the later phase falls within the crown range of tree A in the previous phase, and there is more than one B tree, the B tree with the closest horizontal distance to the tree vertex of tree A is regarded as the same tree as tree A, and the remaining B trees are regarded as new trees.

[0013] Furthermore, in step 4, the following method is used to determine the survival status of the tree: Count the number of points in the point cloud of the same tree in the previous phase and the next phase; If the number becomes larger, the tree survives in both the previous and the later phases; If the number remains unchanged, the tree dies in both the previous and the next phases; If the number decreases, the tree survives in the previous phase and dies in the later phase.

[0014] Furthermore, in step 4, the crown characteristic values ​​include crown width, crown area, and crown volume; and a sample library is constructed using data of trees measured on site, and a new model and an incremental model are fitted based on the sample library; In the sample library of the new model, tree species, crown characteristics, and tree height values ​​of multiple trees measured on site were used as independent variables, and carbon storage measured by the biomass method was used as the dependent variable; In the sample library of the incremental model, the tree species, crown characteristic differences, and tree height differences of multiple trees measured on site are used as independent variables, and the carbon storage changes measured by the biomass method are used as dependent variables.

[0015] Furthermore, in step four, a new model and an incremental model are fitted using a random forest method, a support vector machine method, or a deep learning method.

[0016] Furthermore, the height of the late-lived trees is calculated using the following method: In the later phase, the ground points of the later tree within the canopy range of the previous phase are collected and compared with the tree top of the later tree in the previous phase to calculate the height of the later tree; The height of felled trees is calculated using the following method: In the later phase, the ground points of the felled trees within the tree crown range of the previous phase are collected and compared with the tree tops of the felled trees in the previous phase to calculate the height of the felled trees; The height of new trees is calculated using the following method: In the former phase, ground points where the newborn trees are located within the crown range of the latter phase are collected and compared with the tree apex of the newborn trees in the latter phase to calculate the height of the newborn trees.

[0017] Compared with the prior art, the remote sensing measurement method for carbon storage changes in high-canopy forests of the present invention has the following beneficial effects: In the present invention, the forest is first segmented into individual tree canopies and trees in different time phases are matched, and then the change in the number of point cloud data points is used to determine the change in the survival status of the trees in the two time phases; Different carbon storage change measurement methods are developed for each tree based on the survival status of each tree. After modeling, remote sensing data is substituted into the model to calculate the carbon storage change of each tree. For normal trees that survived in both phases and whose heights could not be obtained by remote sensing, the difference in crown characteristics and tree height was used to calculate the change in carbon storage. Since the ground elevation of the same tree was the same, the difference between the two could eliminate the influence of the ground elevation, thus obtaining the change in carbon storage without the need for ground points. For understory shrubs and herbaceous plants, as well as premature trees that died in both phases, their carbon storage changes were directly ignored (after growing to a certain extent, understory shrubs and herbaceous plants in high-density forests tend to be stable, fluctuating within a fixed range each year; and the carbon of premature trees can be regarded as part of litter, and the total amount of litter is stable); For the old trees, felled trees, and new trees that do not block the ground in at least one phase, their tree heights are calculated by combining the remote sensing data in two phases to calculate the carbon storage changes; After the calculation of the carbon storage change of each tree is completed, the accurate carbon storage change data can be obtained by adding them together, thus overcoming the adverse effects of the upper canopy on remote sensing. Ground survey data is only used for modeling, and does not need to be obtained from the forest to be measured. It is not affected by the terrain of the forest to be measured (it is entirely possible to find a piece of easily accessible forest land to collect data for modeling), and the modeling is completely free of ground survey data (after the modeling is completed, all subsequent measurement processes can be directly called). BRIEF DESCRIPTION OF THE DRAWINGS

[0018] Figure 1 This is a flow chart of a remote sensing measurement method for carbon storage changes in high-canopy forests according to the present invention; DETAILED DESCRIPTION

[0019] like Figure 1 As shown in the figure, a remote sensing measurement method for carbon storage changes in high-canopy forests is used to measure the carbon storage changes in forest lands where it is difficult to collect ground point cloud data through remote sensing due to canopy closure. During the measurement process, the carbon storage changes of understory shrubs and herbaceous plants are recorded as 0 and are not included in the measurement. After the understory shrubs and herbaceous plants in high-density forests grow to a certain size, they tend to stabilize and fluctuate within a fixed range each year, and their contribution to carbon storage changes can be ignored.

[0020] The measurement method includes the following steps: Step 1: Collect remote sensing data of the forest land before and after the change. The time when the data was collected before the change was recorded as the pre-phase, and the time when the data was collected before the change was recorded as the post-phase. The pre-phase and the post-phase were the same time on different dates, and the weather when the data was collected was sunny during the vigorous growth period of the forest; The forest is in its prime of growth, when leaves are still hanging from branches and not yet withered. This period varies in different climates. In a tropical rainforest climate, forests grow vigorously throughout the year; in a tropical monsoon or tropical savanna climate, the rainy season is the peak season for forest growth; in a subtropical or temperate climate, the summer is the peak season for forest growth.

[0021] The reason for choosing this period to collect data is that the identification of different tree species needs to be done when the trees are full of leaves. At this time, the hyperspectral images of trees are very different due to different tree species.

[0022] Step 2: Use the data collected in step 1 to establish non-elevation normalized point clouds of the front phase and the back phase respectively, and smooth and denoise the point clouds; Non-elevation normalized point cloud means that the ground elevation will not be subtracted from the point cloud data here, because the ground points here are difficult to obtain through remote sensing means.

[0023] Step 3: Segment the crowns of individual trees based on the non-elevation normalized point cloud, and then match the trees in the previous phase with those in the next phase, and match the same tree in the previous phase with those in the next phase; The segmentation of single tree crowns in forests is now relatively mature. Even for high-density forests, there are many single tree crown segmentation methods, such as CN202310480182.6 - A layered segmentation method for single tree crowns in high-density artificial forests. In actual operation, the appropriate segmentation method can be selected according to needs.

[0024] In the present invention, since the crown segmentation of individual trees is performed in two different time phases (without segmentation, the crown size cannot be measured later), and then the sizes of the trees in the two time phases are subtracted to overcome the impact caused by the lack of ground points, only the subtraction of the size of the same tree is meaningful, so matching must be performed to ensure that the tree whose size is subtracted is the same.

[0025] Step 4: Determine the tree species and survival status of the segmented trees, and classify the segmented trees into the following categories: Normal tree: The same tree exists in both the previous and the later phases and survives in both phases. This tree is the main body of the forest. Since the same tree exists in both the previous and the next phases and survives, the ground points under it cannot be collected at all. However, this can be overcome by subtracting the size of the tree in the two phases.

[0026] Early-death tree: The same tree exists in both the previous and later phases and dies in both phases; The carbon contained in this tree can be considered as part of the litter on the ground, and the total amount of litter is stable.

[0027] Post-mortem tree: The same tree exists in both the previous phase and the later phase, surviving in the previous phase and dying in the later phase; Felled tree: A tree that exists only in the previous phase; Newborn tree: A tree that exists only in the later phase; The last three types of trees do not exist or are dead in a certain time phase. At this time, there is no obstruction to the ground. At this time, the ground points can be obtained to calculate the tree height.

[0028] Tree species, that is, species in the sense of tree biology, such as poplar, willow, etc. Different species of trees have different carbon storage at the same size, so they need to be calculated separately. The use of remote sensing methods to judge tree species is relatively mature, and the tree species classification method based on hyperspectral images is usually used, such as Hou Chengchao. Research on tree species classification based on airborne hyperspectral images based on deep learning [D]. Chinese Academy of Surveying and Mapping, 2023. Yu Hang, Tan Bingxiang, Shen Mingtan, et al. Identification of dominant tree species from airborne hyperspectral images based on machine learning algorithm[J]. Remote Sensing of Natural Resources, 2024, 36(1):118-127. The remote sensing data used to build the point cloud in step 1 is also a hyperspectral image, which can be used to identify tree species.

[0029] Step 5: Establish a mathematical model with tree species, crown characteristic values, and tree height as independent variables and carbon storage as the dependent variable, and record it as a new model; The new model calculates the carbon storage of a tree from scratch.

[0030] A mathematical model was established with tree species, crown characteristic difference, and tree height difference as independent variables and carbon storage change as dependent variable, and recorded as an incremental model, where the crown characteristic difference and tree height difference were the crown characteristic value and tree height value after the change minus the crown characteristic value and tree height value before the change; The incremental model calculates the carbon storage of a tree as it grows.

[0031] In step 4, the trees were classified. Different types of trees can collect different data, so different models are used to calculate the change in carbon storage, as follows: Step 6: Calculate the carbon storage changes for different types of trees and then sum them up, where: Change in carbon storage of normal trees: Substitute the tree species, crown characteristic differences, and tree height differences into the incremental model to calculate the change in carbon storage; Change in carbon storage of prematurely dead trees: The change in carbon storage is recorded as 0, that is, the carbon in prematurely dead trees is regarded as part of the dead branches and leaves in the forest, and the carbon storage in both phases is regarded as 0, 0-0=0.

[0032] Since premature trees are dead, it is difficult to determine the tree species. However, since there is no need to calculate the change in their carbon storage, the tree species can be ignored.

[0033] Changes in carbon storage of late-living trees: Substitute tree species, crown characteristic values, and tree height values ​​into the new model, calculate carbon storage, and then take the inverse; The late-dying trees are different from the early-dying trees. The carbon they contain at the beginning is not part of the dead branches and leaves in the forest, so it is a positive value. When it changes from a positive value to 0, the change is a negative value.

[0034] Change in carbon storage of felled trees: Substitute tree species, crown characteristic values, and tree height values ​​into the new model, calculate carbon storage, and then take the inverse; Felled trees are similar to young trees. The carbon they contain is positive at the beginning, and when it changes from positive to 0, the change becomes negative.

[0035] Change in carbon storage of new trees: Substitute tree species, crown characteristic values, and tree height values ​​into the new model to calculate carbon storage.

[0036] In step 1, the image data is collected using UAV oblique photogrammetry.

[0037] In this method, since the images are taken from multiple angles, not only can the crown volume be measured more accurately (the measurement of carbon storage in the present invention has higher requirements for the crown volume measurement), but it is also more convenient to distinguish between ground points, points in the upper canopy, and points in the lower canopy, making it easier to segment single tree crowns without performing elevation normalization.

[0038] In step 3, the following method is used to determine whether the tree in the previous phase and the tree in the next phase are the same tree: After excluding the false tree vertices, if the tree vertex of tree B in the later phase falls within the tree crown range of tree A in the previous phase, and there is only one tree B, then tree A and tree B are considered to be the same tree; After excluding false tree vertices, if the tree vertex of tree B in the later phase falls within the crown range of tree A in the previous phase, and there is more than one B tree, the B tree with the closest horizontal distance to the tree vertex of tree A is regarded as the same tree as tree A, and the remaining B trees are regarded as new trees.

[0039] This judgment method is unique to high-canopy forests, because in high-canopy forests, the growth of trees is equivalent to dense planting. The growth of each tree is constrained by the growth of other surrounding trees and is difficult to grow crooked. After growing taller, the tree apex will still be within the original canopy range. Of course, due to the canopy stratification phenomenon in high-canopy forests, sometimes a new tree will squeeze in between the trees, which will interfere with the matching of the trees. However, the squeezed new tree is located in the gap between the trees and is far away from the tree apex of tree A, so it can be excluded based on this.

[0040] In step 4, the following method is used to determine the survival status of the tree: Count the number of points in the point cloud of the same tree in the previous phase and the next phase; If the number becomes larger, the tree survives in both the previous and the later phases; If the number remains unchanged, the tree dies in both the previous and the next phases; If the number decreases, the tree survives in the previous phase and dies in the later phase.

[0041] In the present invention, the remote sensing data is collected when the leaves are full. If the leaves of the tree are gone, there will be much fewer point cloud data points belonging to the crown after the single tree crown is segmented (at the same time, there will be some more ground points, which are excluded when the single tree crown is segmented), so as to judge whether the tree is dead or alive.

[0042] In step 4, the crown characteristic values ​​include crown width, crown area, and crown volume; and the sample library is constructed with the data of the trees measured on site, and the new model and the incremental model are fitted based on the sample library; In the sample library of the new model, tree species, crown characteristics, and tree height values ​​of multiple trees measured on site were used as independent variables, and carbon storage measured by the biomass method was used as the dependent variable; In the sample library of the incremental model, the tree species, crown characteristic differences, and tree height differences of multiple trees measured on site are used as independent variables, and the carbon storage changes measured by the biomass method are used as dependent variables.

[0043] The trees in the sample library here can be selected from trees in easily accessible areas rather than trees in the forest to be measured.

[0044] In step 4, a new model and an incremental model are fitted using a random forest method, a support vector machine method, or a deep learning method.

[0045] This is essentially a model, a function of multiple qualitative variables and one dependent variable. There are many mature methods for fitting functions. In addition to those listed here, when the type of tree is relatively single, you can even use Matlab to fit multivariate functions.

[0046] The height of the late-stage trees is calculated using the following method: In the later phase, the ground points of the later tree within the canopy range of the previous phase are collected and compared with the tree top of the later tree in the previous phase to calculate the height of the later tree; The height of felled trees is calculated using the following method: In the later phase, the ground points of the felled trees within the tree crown range of the previous phase are collected and compared with the tree tops of the felled trees in the previous phase to calculate the height of the felled trees; The height of new trees is calculated using the following method: In the former phase, ground points where the newborn trees are located within the tree crown range of the latter phase are collected and compared with the tree apex of the newborn trees in the latter phase to calculate the height of the newborn trees.

[0047] The points in the point cloud model include elevation. The tree height can be obtained by subtracting the elevation of the ground point at the corresponding position from the elevation of the tree vertex. The position of the tree vertex should be based on when the tree is alive. After the tree dies, the position of the tree vertex is not only difficult to obtain accurately, but also may fluctuate.

[0048] This example was applied to the larch and pine plots in Wangyedian Forest Farm in Inner Mongolia, and the carbon sink increments from 2021 to 2022 were tested. The accuracy of carbon sink estimation for different tree species reached more than 90%. This work demonstrates the effectiveness of the proposed technique and shows improvements in accuracy and efficiency compared to existing methods.

[0049] The embodiments described above are merely descriptions of preferred implementation modes of the present invention and are not intended to limit the scope of the present invention. Without departing from the design spirit of the present invention, various modifications and improvements made to the technical solutions of the present invention by ordinary technicians in this field should all fall within the protection scope determined by the claims of the present invention.

Claims

1. A remote sensing measurement method for carbon storage changes in high-canopy forests, used to measure carbon storage changes in forest land where it is difficult to collect ground point cloud data by remote sensing means due to canopy closure, characterized in that: During the measurement process, the carbon storage changes of understory shrubs and herbaceous plants were recorded as 0 and not included in the measurement; The measuring method comprises the following steps: Step 1: Collect remote sensing data of the forest land before and after the change. The time when the data was collected before the change is recorded as the pre-phase, and the time when the data was collected before the change is recorded as the post-phase. The pre-phase and the post-phase are the same time on different dates, and the weather when the data was collected is sunny during the vigorous growth period of the forest; Step 2: Use the data collected in step 1 to establish non-elevation normalized point clouds of the front phase and the back phase respectively, and smooth and denoise the point clouds; Step 3: Segment the crowns of individual trees based on the non-elevation normalized point cloud, and then match the trees in the previous phase with those in the next phase, and match the same tree in the previous phase with those in the next phase; Step 4: Determine the tree species and survival status of the segmented trees, and classify the segmented trees into the following categories: Normal tree: The same tree exists in both the previous and the later phases and survives in both phases. Early-death tree: The same tree exists in both the previous and later phases and dies in both phases; Post-mortem tree: The same tree exists in both the previous phase and the later phase, surviving in the previous phase and dying in the later phase; Felled tree: a tree that exists only in the previous phase; Newborn tree: A tree that exists only in the later phase; Step 5: Establish a mathematical model with tree species, crown characteristic values, and tree height as independent variables and carbon storage as the dependent variable, and record it as a new model; A mathematical model was established with tree species, crown characteristic difference, and tree height difference as independent variables and carbon storage change as dependent variable, and recorded as an incremental model, where the crown characteristic difference and tree height difference were the crown characteristic value and tree height value after the change minus the crown characteristic value and tree height value before the change; Step 6: Calculate the carbon storage changes for different types of trees and then sum them up, where: Change in carbon storage of normal trees: Substitute the tree species, crown characteristic differences, and tree height differences into the incremental model to calculate the change in carbon storage; Change in carbon storage of trees that died early: The change in carbon storage was recorded as 0; Changes in carbon storage of late-living trees: Substitute tree species, crown characteristic values, and tree height values ​​into the new model, calculate carbon storage, and then take the inverse; Change in carbon storage of felled trees: Substitute tree species, crown characteristic values, and tree height values ​​into the new model, calculate carbon storage, and then take the inverse; Change in carbon storage of new trees: Substitute tree species, crown characteristic values, and tree height values ​​into the new model to calculate carbon storage.

2. The remote sensing measurement method for carbon storage changes in high-canopy forests according to claim 1 is characterized by: In step 1, the image data is collected using UAV oblique photogrammetry.

3. The remote sensing measurement method for carbon storage changes in high-canopy forests according to claim 1 is characterized by: In step 3, the following method is used to determine whether the tree in the previous phase and the tree in the next phase are the same tree: After excluding the false tree vertices, if the tree vertex of tree B in the later phase falls within the tree crown range of tree A in the previous phase, and there is only one tree B, then tree A and tree B are considered to be the same tree; After excluding false tree vertices, if the tree vertex of tree B in the later phase falls within the crown range of tree A in the previous phase, and there is more than one B tree, the B tree with the closest horizontal distance to the tree vertex of tree A is regarded as the same tree as tree A, and the remaining B trees are regarded as new trees.

4. The remote sensing measurement method for carbon storage changes in high-canopy forests according to claim 1 is characterized by: In step 4, the following method is used to determine the survival status of the tree: Count the number of points in the point cloud of the same tree in the previous phase and the next phase; If the number becomes larger, the tree survives in both the previous and the later phases; If the number remains unchanged, the tree dies in both the previous and the next phases; If the number decreases, the tree survives in the previous phase and dies in the later phase.

5. The remote sensing measurement method for carbon storage changes in high-canopy forests according to claim 1 is characterized by: In step 4, the crown characteristic values ​​include crown width, crown area, and crown volume; and the sample library is constructed with the data of the trees measured on site, and the new model and the incremental model are fitted based on the sample library; In the sample library of the new model, tree species, crown characteristics, and tree height values ​​of multiple trees measured on site were used as independent variables, and carbon storage measured by the biomass method was used as the dependent variable; In the sample library of the incremental model, the tree species, crown characteristic differences, and tree height differences of multiple trees measured on site are used as independent variables, and the carbon storage changes measured by the biomass method are used as dependent variables.

6. The remote sensing measurement method for carbon storage changes in high-canopy forests according to claim 1 is characterized by: In step 4, a new model and an incremental model are fitted using a random forest method, a support vector machine method, or a deep learning method.

7. The remote sensing measurement method for carbon storage changes in high-canopy forests according to claim 1 is characterized by: The height of the late-stage trees is calculated using the following method: In the later phase, the ground points of the later tree within the canopy range of the previous phase are collected and compared with the tree top of the later tree in the previous phase to calculate the height of the later tree; The height of felled trees is calculated using the following method: In the later phase, the ground points of the felled trees within the tree crown range of the previous phase are collected and compared with the tree tops of the felled trees in the previous phase to calculate the height of the felled trees; The height of new trees is calculated using the following method: In the former phase, ground points where the newborn trees are located within the crown range of the latter phase are collected and compared with the tree apex of the newborn trees in the latter phase to calculate the height of the newborn trees.

Images