Space-air-ground integrated carbon sink monitoring model generation method, monitoring method and system

Abstract

The application discloses a space-air-ground integrated carbon sink monitoring model generation method, a monitoring method and a system. The method comprises the following steps: acquiring multi-modal monitoring data in multiple monitoring sub-regions in a set monitoring region; calculating the carbon sink of each monitoring sub-region according to the multi-modal monitoring data in each monitoring sub-region to obtain the carbon sink calculation result of each monitoring sub-region; and selecting multiple sample monitoring regions from each monitoring sub-region according to a set screening condition. The preset neural network model is trained by using the satellite monitoring data and the unmanned aerial vehicle monitoring data corresponding to each sample monitoring region, and a trained carbon sink monitoring model is obtained. The method and system of the application realize accurate monitoring of forest carbon sink based on cooperation of unmanned aerial vehicles, satellites, ground and environmental data, overcome the problems of low data precision, low efficiency and high cost of single remote sensing technology, and realize accurate monitoring of forest carbon sink.

Description

Technical Field

[0001] This invention relates to the field of remote sensing monitoring technology, and in particular to a method for generating an integrated air-space-ground carbon sink monitoring model, a carbon sink monitoring method, and a system. Background Technology

[0002] Forest carbon sinks refer to the absorption of carbon dioxide from the atmosphere by forest plants through photosynthesis, which is then fixed in vegetation or soil, thereby reducing the concentration of this gas in the atmosphere. Forest carbon sink monitoring refers to the process of measuring changes in carbon storage in forest vegetation and soil using scientific methods and technical means to assess the forest's ability to absorb and fix carbon dioxide from the atmosphere and its dynamic changes.

[0003] Current technologies typically rely on manual monitoring, which is inefficient. While satellite remote sensing has been disclosed for monitoring forest carbon sinks, this method, relying solely on satellite monitoring, results in coarse-grained data, making it difficult to distinguish plant organs. Furthermore, the lack of ground-based sampling verification leads to parameter biases, preventing the acquisition of high-precision monitoring results and failing to meet target measurement requirements. Moreover, satellite monitoring requires significant hardware investment and data processing resources, hindering large-scale application.

[0004] Therefore, existing technologies need further improvement. Summary of the Invention

[0005] In view of the shortcomings of the existing technology, the present invention aims to provide a method for generating an integrated air-space-ground carbon sink monitoring model, a monitoring method and system, to overcome the shortcomings of the existing technology, such as low efficiency of manual monitoring and single monitoring method, high hardware cost and low monitoring accuracy when using satellite remote sensing to monitor forest carbon sinks.

[0006] The technical solution of the present invention is as follows: Firstly, this application discloses a method for generating an integrated air-space-ground carbon sink monitoring model, which includes: Acquire multimodal monitoring data from multiple monitoring sub-regions within a defined monitoring area; wherein each monitoring sub-region is obtained by dividing the defined monitoring area according to a preset grid; the multimodal monitoring data includes: satellite monitoring data, UAV monitoring data, and ground status and environmental data; Carbon sink calculations are performed based on multimodal monitoring data from each monitoring sub-region to obtain carbon sink calculation results for each monitoring sub-region. Based on the carbon sink calculation results in each of the monitoring sub-regions, multiple sample monitoring regions are selected from each of the monitoring sub-regions according to the set screening conditions; The pre-set neural network model is trained using the multimodal monitoring data corresponding to the monitoring areas of each sample to obtain the trained carbon sink monitoring model.

[0007] Optionally, acquiring satellite monitoring data and UAV monitoring data for the designated monitoring area includes: The SAR and hyperspectral data of forests and grasslands within the designated monitoring area are acquired using a target satellite located in a fixed orbit. The growth status and distribution of forests and grasslands within the area are monitored based on the SAR data, and the species of forests and grasslands and carbon sinks are distinguished and classified based on the hyperspectral data. Satellite monitoring data including forest and grassland growth status and distribution monitoring data, first forest and grassland species distinction data and carbon sink classification results are obtained. The UAV acquires visible light spectral data, multispectral data, and lidar data of forest and grassland vegetation within a designated monitoring area. Based on the visible light spectral data and multispectral data, it obtains second forest and grassland species differentiation data, forest and grassland organ differentiation data, and biomass proportion data of different forest and grassland organs. Based on the lidar data, it obtains trunk parameter measurement data and tree outline detection data. The second forest and grassland species differentiation data, forest and grassland organ differentiation data, forest and grassland organ biomass proportion data, trunk parameter measurement data, and tree outline detection data are combined to form UAV monitoring data.

[0008] Optionally, the step of calculating carbon sinks based on multimodal monitoring data in each monitoring sub-region to obtain carbon sink calculation results for each monitoring sub-region includes: Using the first forest and grassland species differentiation data in the satellite monitoring data of each monitoring sub-region as the basic data to be corrected, and using the second forest and grassland species differentiation data in the UAV monitoring data as the standard data, the first forest and grassland species differentiation data in the satellite monitoring data is corrected to obtain accurate forest and grassland species differentiation data. The carbon sink calculation results for each monitoring sub-region are calculated based on the area of ​​each monitoring sub-region, the biomass proportion data of different forest and grassland organs in each monitoring sub-region, the preset carbon density coefficient, the forest and grassland organ differentiation data, and the carbon sink classification results.

[0009] Optionally, the step of selecting multiple sample monitoring areas from each of the monitoring sub-regions according to the carbon sink calculation results in each of the monitoring sub-regions and according to the set screening conditions includes the following screening conditions: whether the number of different tree species in the area exceeds a preset value, whether the area contains trees in different growth states, whether the carbon sink calculation results exceed a preset carbon sink value, and whether the area ratio of the sampling area to the experimental area exceeds a preset area ratio value.

[0010] Optionally, the preset neural network model includes: a classification module and a regression module; The steps for training a pre-defined neural network model using multimodal monitoring data corresponding to each of the sample monitoring areas to obtain a trained carbon sink monitoring model include: Satellite monitoring data, UAV monitoring data, and ground condition and environmental data of each of the sample monitoring areas are input into the classification module to obtain the classification results of forest and grassland species and forest and grassland organs output by the classification module; The forest and grassland species, forest and grassland organ classification results, and trunk parameter measurement data of each sample monitoring area are input into the regression module to obtain the carbon sink prediction data output by the regression module. The error between the predicted carbon sink data and the actual carbon sink data of each of the sample monitoring areas is calculated, and the parameters of the preset neural network model are adjusted according to the error. This process is repeated multiple times until the preset training conditions are met, thus obtaining a trained carbon sink monitoring model.

[0011] Optionally, before the step of training a preset neural network model using multimodal monitoring data corresponding to each of the sample monitoring areas to obtain a trained carbon sink monitoring model, the method further includes: The satellite monitoring data is calibrated using UAV monitoring data within each of the sample monitoring areas, and flux tower data within each of the sample monitoring areas is obtained. The weighting coefficients and carbon conversion coefficients of the satellite monitoring data and UAV monitoring data are adjusted based on the flux tower data.

[0012] Optionally, the preset grid is obtained by uniformly dividing the area according to the preset grid area based on the geographic coordinate system; or the preset grid is obtained by dividing the area according to the terrain based on the geographic coordinate system.

[0013] Optionally, the method further includes: acquiring satellite monitoring data and drone monitoring data at preset time intervals, and using the satellite monitoring data and drone monitoring data to fine-tune the parameters of the carbon sink monitoring model.

[0014] Secondly, this application also provides an integrated air-space-ground carbon sink monitoring method, which includes: Acquire multimodal monitoring data of the target monitoring area; wherein, the multimodal monitoring data includes: satellite monitoring data, UAV monitoring data, and ground condition and environmental data; The multimodal monitoring data of the target monitoring area is input into the trained carbon sink monitoring model to obtain the carbon sink monitoring results of the target monitoring area output by the carbon sink monitoring model; wherein, the carbon sink monitoring model is obtained based on the carbon sink monitoring model generation method.

[0015] Thirdly, this application also provides an integrated air-space-ground forest carbon sequestration monitoring system, which includes: a monitoring data acquisition module and a data processing module; The monitoring data acquisition module is used to acquire multimodal monitoring data of the target monitoring area; wherein, the multimodal monitoring data includes: satellite monitoring data, UAV monitoring data, and ground condition and environmental data; The data processing module is used to input the multimodal monitoring data into the trained carbon sink monitoring model to obtain the carbon sink monitoring results of the target monitoring area output by the carbon sink monitoring model; wherein, the carbon sink monitoring model is obtained based on the carbon sink monitoring model generation method.

[0016] Beneficial effects: This invention proposes an integrated air-space-ground carbon sink monitoring model generation method, monitoring method, and system. The method involves acquiring multimodal monitoring data from multiple monitoring sub-regions within a defined monitoring area; calculating carbon sink results for each sub-region based on the multimodal monitoring data; and selecting multiple sample monitoring regions from each sub-region according to set screening criteria based on the carbon sink calculation results. A pre-set neural network model is then trained using the multimodal monitoring data corresponding to each sample monitoring region to obtain a trained carbon sink monitoring model. The model generation method, monitoring method, and system provided by this invention integrate satellite monitoring data, UAV monitoring data, and ground condition and environmental data to achieve precise monitoring of forest carbon sinks through the coordinated use of UAV, satellite, and ground condition and environmental data. This overcomes the problems of low accuracy, low efficiency, and high cost associated with single remote sensing technologies, thus achieving precise monitoring of forest carbon sinks. Attached Figure Description

[0017] Figure 1 This is a flowchart of the steps in the method for generating an integrated air-space-ground carbon sink monitoring model provided by the present invention; Figure 2 This is a schematic diagram of the multi-pod structure of the UAV provided by the present invention; Figure 3 This is a schematic diagram of the mesh division provided by the present invention; Figure 4 A schematic diagram of the iterative principle in a specific application embodiment of the integrated air-space-ground carbon sink monitoring model generation method provided by the present invention; Figure 5 A flowchart illustrating the steps of a specific application embodiment of the integrated air-space-ground carbon sink monitoring model generation method provided by this invention; Figure 6 This is a schematic diagram of the integrated air-space-ground carbon sink monitoring method provided by the present invention; Figure 7 This is a schematic diagram of the principle of the integrated air-space-ground carbon sequestration monitoring system provided by the present invention. Detailed Implementation

[0018] To make the objectives, technical solutions, and advantages of this invention clearer and more explicit, the invention will be further described in detail below with reference to the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are merely illustrative of the invention and are not intended to limit the invention.

[0019] Those skilled in the art will understand that, unless specifically stated otherwise, the singular forms “a,” “an,” “the,” and “the” used herein may also include the plural forms. It should be further understood that the term “comprising” as used in this specification means the presence of the stated features, integers, steps, operations, elements, and / or components, but does not exclude the presence or addition of one or more other features, integers, steps, operations, elements, components, and / or groups thereof. It should be understood that when we say an element is “connected” or “coupled” to another element, it can be directly connected or coupled to the other element, or there may be intermediate elements. Furthermore, “connected” or “coupled” as used herein can include wireless connections or wireless coupling. The term “and / or” as used herein includes all or any units and all combinations of one or more associated listed items.

[0020] It will be understood by those skilled in the art that, unless otherwise defined, all terms used herein (including technical and scientific terms) have the same meaning as commonly understood by one of ordinary skill in the art to which this invention pertains. It should also be understood that terms such as those defined in general dictionaries should be understood to have the same meaning as in the context of the prior art, and should not be interpreted in an idealized or overly formal sense unless specifically defined as herein.

[0021] Carbon sinks are the processes, activities, or mechanisms by which forest plants absorb carbon dioxide from the atmosphere through photosynthesis and fix it in vegetation or soil, thereby reducing the concentration of this gas in the atmosphere. As the most economical negative carbon pathway globally, carbon sink measurement and monitoring results are the core basis for carbon credit trading. With the restart of the carbon market, forestry carbon sink projects have experienced explosive growth, but "unclear measurement and inaccurate monitoring" have become key bottlenecks restricting the assetization of carbon sinks—traditional technologies result in long project development cycles and low success rates, making it difficult to meet the carbon market's demands for data scientific rigor and timeliness.

[0022] Currently, there are two types of carbon sink standards: international and domestic. International standards, based on the IPCC 2006 guidelines and ISO 14064-2, emphasize the measurement of all elements of the carbon sink (aboveground / belowground biomass, deadwood, and soil organic carbon) and systematic monitoring. Domestic standards, such as the existing GB / T 29345-2013 and NY / T1141-2021 standards, specify the measurement requirements for four types of carbon sinks, including afforestation and forest management, but rely on ground surveys as the core method and limit the measurement frequency to once every 5 years. Taking the domestically dominant CCER methodology as an example, its core principle is based on plot sampling surveys. Carbon storage is estimated by manually measuring parameters such as diameter at breast height (DBH) and tree height, combined with a biomass conversion model, using a "3-5 year periodic monitoring + sampling statistics" model. The process steps are as follows: First, fixed plots are established, for example, 1-4 fixed plots per hectare. Second, the DBH of sample trees is manually measured (accuracy ±0.1cm) and tree height (accuracy ±0.5m). Third, the biomass of individual trees is calculated using the allometric growth equation, and the carbon conversion factor (usually taken as 0.5) is combined to obtain the carbon storage. Finally, the total regional carbon sink is extrapolated according to the sampling ratio. The fundamental flaws of this CCER methodology include the following: 1) Lack of timeliness: The 3-5 year monitoring cycle cannot capture dynamic changes in forests (such as sudden changes in carbon storage caused by pests, logging, and extreme weather), and the data lag leads to a carbon credit valuation deviation of more than 30%.

[0023] 2) Significant sampling error: The number of sample plots is limited (only 1-2 sample plots per 100 hectares on average), making it difficult to cover complex terrain (such as mountains and canyons) and uneven-aged forest stands. Insufficient sampling representativeness leads to measurement errors of 15%-25%.

[0024] 3) Cost and efficiency contradiction: Manual measurement requires a lot of manpower (3-5 people / day per square kilometer), and is limited by terrain and cannot enter dense forest areas. When promoted on a large scale, the unit cost is as high as 50-100 yuan / mu.

[0025] 4) Incomplete carbon pool coverage: Dead wood and soil organic carbon were not included in the measurement, which differs from the IPCC standard and leads to an underestimation of carbon sink by 8%-12%.

[0026] In addition, existing technologies also disclose carbon sequestration measurement methods using satellite remote sensing. However, satellite remote sensing measurement methods have the following drawbacks: 1) They are susceptible to interference from cloud and rain weather (annual effective data rate is less than 60%), cannot penetrate the vegetation canopy to obtain vertical structure information, and have a small-scale monitoring accuracy of less than 70%; 2) LiDAR / UAVs: Equipment procurement costs are high (a single airborne LiDAR unit costs more than 2 million yuan), data processing requires professional algorithm support, and the data processing time per square kilometer exceeds 48 hours, making it difficult to scale up applications; 3) IoT monitoring: The deployment cost per point exceeds 50,000 yuan, maintenance costs are high (15%-20% per year), and it can only represent local point data, making it impossible to achieve regional-scale carbon sequestration accounting.

[0027] Therefore, the carbon sink measurement methods disclosed in the prior art have the following drawbacks: 1. Single satellite monitoring: Although it has a wide coverage area, the data granularity is coarse, making it difficult to distinguish plant organs (leaves, stems, fruits), and the lack of ground sampling verification leads to parameter deviations.

[0028] 2. Single UAV monitoring: Although it has high accuracy, it has limited coverage and high cost, and cannot achieve large-scale macro-monitoring.

[0029] 3. Data fragmentation: Satellite and drone data are not integrated into a unified processing mechanism and lack a closed-loop calibration process, resulting in insufficient accuracy in carbon sink measurement.

[0030] 4. High hardware costs: Traditional LiDAR requires stringent accuracy when used alone, resulting in high hardware investment and data processing pressure.

[0031] This invention addresses the technical problems of insufficient accuracy, low efficiency, high cost, and inaccurate differentiation between tree trunks and branches / fruits in existing forest carbon sequestration measurement technologies. This invention provides an integrated air-space-ground carbon sequestration monitoring model generation method, monitoring method, and system. It realizes carbon sequestration monitoring based on the integration of air, space, and ground, where air corresponds to aviation (UAV), sky corresponds to satellite, and ground corresponds to ground (ground sensor). Through the collaboration of the three parties, accurate measurement of forest carbon sequestration is achieved.

[0032] Furthermore, in the method and system of this invention, the UAV adopts a modular design with multiple payloads, achieving functional synergy by carrying a visible light camera, a multispectral / hyperspectral camera, and an airborne lidar pod: the visible light camera is used for accurate identification of tree species, the airborne lidar completes high-precision measurement of tree trunk thickness based on point cloud data, and the multispectral / hyperspectral camera achieves accurate separation of tree trunks from branches, leaves, and fruits through spectral feature differences; historical data from satellite SAR radar uses publicly available data sources, eliminating the need to purchase additional commercial remote sensing data, effectively reducing data acquisition costs without affecting measurement accuracy, and the forest vertical structure information retrieved from it forms a spatiotemporal dimensional complementary correction with the UAV's multi-pod measured data. This invention effectively overcomes the periodic limitations of traditional manual sampling measurements and the functional shortcomings of single remote sensing technology, achieving precise targeted monitoring of the core measurement object of forest carbon sinks (tree trunks), significantly improving measurement accuracy and data timeliness, shortening the monitoring cycle, reducing overall measurement costs, enhancing large-scale application capabilities, and providing highly reliable measurement basis for carbon credit trading.

[0033] The following example provides a more accurate explanation of the integrated air-space-ground carbon sink monitoring model generation method, monitoring method, and system provided by the present invention.

[0034] Firstly, this application discloses a method for generating an integrated air-space-ground carbon sink monitoring model, such as... Figure 1 As shown, it includes: Step S1: Obtain multimodal monitoring data from multiple monitoring sub-regions within a set monitoring area; wherein each monitoring sub-region is obtained by dividing the set monitoring area according to a preset grid; the multimodal monitoring data includes: satellite monitoring data, UAV monitoring data, and ground status and environmental data.

[0035] In this step, the multimodal monitoring data includes UAV monitoring data obtained from UAVs and satellite monitoring data obtained from satellites. This step involves acquiring visible light spectral data, multispectral data, and lidar data collected by UAVs within the designated monitoring area; acquiring SAR data and hyperspectral data collected by satellites within the designated monitoring area; and utilizing ground condition and environmental data released by ground sensors and weather and environmental platforms. The data acquired by UAVs and satellites are then processed to obtain satellite monitoring data, UAV monitoring data, and ground condition and environmental data.

[0036] In detail, the steps for obtaining satellite monitoring data and UAV monitoring data for the designated monitoring area include: Step S11: Use a target satellite in a fixed orbit to acquire SAR data and hyperspectral data of forests and grasslands within the designated monitoring area. Based on the SAR data, monitor the growth status and distribution of forests and grasslands within the area. Based on the hyperspectral data, distinguish forest and grassland species and classify carbon sinks to obtain radar monitoring data including monitoring data of forest and grassland growth status and distribution, first forest and grassland species distinction data, and carbon sink classification results.

[0037] This step utilizes satellites to acquire SAR radar data and hyperspectral data. Because SAR can operate in all weather and all-day conditions, it offers high flexibility and is not limited by weather or lighting conditions, enabling the acquisition of high-quality surface images. Therefore, SAR radar data can provide surface information. Hyperspectral data can capture subtle differences in surface objects across continuous spectral bands. In this embodiment, after acquiring the SAR radar data and hyperspectral data collected by the satellite, edge detection can be performed based on these data to extract the contour information of forests and grasslands, thereby obtaining their shapes. Based on these shapes, the growth status and distribution of forests and grasslands can be determined, and the species can be distinguished based on their shape and distribution. For example, by preprocessing and edge detection on SAR radar data or hyperspectral data, the contour information of tree crowns can be extracted, thus obtaining their shapes. After extracting the crown shapes, the canopy method can be used to determine boundary lines. Canopy-based boundary lines are typically based on the morphological, textural, or spectral characteristics of the crowns, using certain algorithms or rules to separate adjacent crowns, thereby determining the boundary line. Information such as the shape of the forest is determined based on the boundary line.

[0038] Specifically, SAR radar data acquisition utilizes fixed-orbit satellites (such as Sentinel-1 and Gaofen-3) to acquire C-band or X-band SAR images. Leveraging their penetrating power and fixed orbits, centimeter-level monitoring of forest and grassland growth and leaf fall distribution is achieved. Centimeter-level growth monitoring involves measuring the radial growth of trees using interferometric SAR technology (accuracy ≤ 0.5 cm). Leaf fall distribution monitoring involves penetrating the leaf litter layer to obtain the actual morphology and distribution of tree trunks.

[0039] For hyperspectral data acquisition, satellites equipped with hyperspectral sensors (such as Hyperion and Gaofen-5) are used to acquire continuous spectral data in the 400-2500 nm band to achieve primary forest and grassland species differentiation and carbon sink classification. Primary forest and grassland species differentiation involves distinguishing trees, shrubs, grasslands, and specific tree species (such as Pinaceae and Betulaceae) through spectral feature matching (such as vegetation indices and absorption peak positions). Carbon sink classification is based on the differences in spectral responses of different forest and grassland species, combined with growth cycle data, to determine the carbon sink coefficient (such as net primary productivity (NPP) and biomass conversion factor) for each type of forest and grassland.

[0040] After acquiring SAR and hyperspectral data, preprocessing is required. The acquired SAR data undergoes denoising, registration, and interferometry. The hyperspectral data undergoes atmospheric and geometric correction to output standardized image data.

[0041] Step S12: Use a drone to acquire visible light spectral data, multispectral data, and lidar data of forest and grassland vegetation within the designated monitoring area. Based on the visible light spectral data and multispectral data, obtain second forest and grassland species differentiation data, forest and grassland organ differentiation data, and biomass proportion data of different forest and grassland organs. Based on the lidar data, obtain trunk parameter measurement data and forest and grassland outline detection data. Combine the second forest and grassland species differentiation data, forest and grassland organ differentiation data, biomass proportion data of different forest and grassland organs, trunk parameter measurement data, and forest and grassland outline detection data into drone monitoring data.

[0042] Furthermore, in this step, drones are used to acquire lidar data of forest and grassland vegetation within the designated monitoring area. Based on the tree trunk point cloud information extracted from the lidar data, tree trunks can be identified through methods such as cluster analysis, and the proportion or area of ​​different parts of the tree trunk in the vertical direction can be calculated. Thus, the coefficients of each parameter in the biomass estimation formula can be adjusted according to the proportion or area of ​​different parts.

[0043] Specifically, the biomass estimation formula is as follows: ; in, The vegetation biomass per unit area. The normalized vegetation index (calculated from satellite hyperspectral data) is used. The average tree height (obtained from ground sensors or satellite radar data) is represented by a, b, c, d, and e, which are calibration coefficients. In one implementation, a = 0.85, b = 1.2, c = 0.32, d = 1.5, and e = 0.05.

[0044] Furthermore, in order to obtain accurate forest and grassland data for the designated monitoring area, the unmanned system in this step collects forest and grassland data by carrying two pods. For example... Figure 2 As shown, the dual-pod configuration includes a first pod 10 and a second pod 20. The first pod 10 is used to house an RGB camera for acquiring visible light spectra and a multispectral camera for acquiring multispectral data. The second pod is a high-precision LiDAR pod used to house LiDAR sensors.

[0045] In practical implementation, the RGB camera has a resolution of ≥12 megapixels. The multispectral camera includes 4-6 bands, including the near-infrared band. High-precision LiDAR pod: uses an industrial-grade LiDAR sensor with a ranging accuracy of ≤10mm (millimeter level).

[0046] The spectral data captured by the RBG and multispectral cameras in this step can be used for secondary forest and grassland species differentiation, forest and grassland organ classification, and biomass proportion data of different forest and grassland organs. Secondary forest and grassland species differentiation is to supplement the fine-grained classification accuracy of satellite data (such as distinguishing different age groups of the same tree species). Forest and grassland organ differentiation uses high-resolution image segmentation techniques (such as Mask R-CNN) to extract the spatial distribution and area proportion of leaves, trunks, and fruits. Biomass proportion data of different forest and grassland organs is based on the type of forest and grassland organ, statistically analyzing the biomass proportion of different organs.

[0047] In practical applications, high-precision LiDAR sensors can be used for trunk parameter measurement and forest / grassland contour detection. Trunk parameter measurement involves acquiring three-dimensional parameters such as trunk height (measurement accuracy ≤5mm), diameter at breast height (DBH) (measurement accuracy ≤3mm), and crown width (measurement accuracy ≤8mm), accurately reflecting the state of trunk biomass accumulation. Precision-optimized forest / grassland contour detection combines texture features and edge detection results from visible light images to perform targeted denoising, sparsification, and key area completion on LiDAR point cloud data—maintaining millimeter-level measurement accuracy without relying on full-density point clouds. The original point cloud density is optimized from 150 points / cm² to 50 points / cm², reducing data storage pressure (67% reduction in data volume) and subsequent processing time (40% improvement in processing efficiency) while ensuring measurement accuracy, balancing high-precision requirements with engineering practicality. Therefore, high-precision LiDAR sensors can be used to acquire trunk parameter measurement data and forest / grassland contour detection data.

[0048] Furthermore, the preset grid is obtained by uniformly dividing the geographic coordinate system according to the preset grid area. To divide the designated monitoring area into multiple monitoring sub-regions according to the preset grid, such as... Figure 3 As shown, in this embodiment, the method further divides the designated monitoring area into multiple uniform grids based on the geographic coordinate system, and each grid is further divided into a monitoring sub-area according to the preset grid area configuration. In one embodiment, the WGS-84 geographic coordinate system is used to divide the designated monitoring area into uniform grids (the grid size is configurable, preferably km×1km or 500m×500m), with each grid serving as an independent monitoring sub-area.

[0049] In addition, after acquiring radar monitoring data, UAV monitoring data, and ground status and environmental data, based on the grid division results of satellite data, low-altitude flights (50-100m altitude) are conducted in the sampling area and key areas to acquire full-coverage imagery and point cloud data. The imagery and point cloud data are then preprocessed, including point cloud denoising, image stitching, and coordinate registration (unifying the coordinate system with satellite data), to ensure that the spatial overlap error between LiDAR point cloud and visible light imagery is ≤10mm.

[0050] Step S2: Calculate the carbon sink based on the multimodal monitoring data in each monitoring sub-region to obtain the carbon sink calculation results in each monitoring sub-region.

[0051] Once satellite and drone monitoring data for each monitoring sub-region are obtained, this step calculates the carbon sink results for each monitoring sub-region based on the satellite and drone monitoring data.

[0052] Specifically, the step of calculating carbon sinks based on satellite monitoring data and UAV monitoring data in each monitoring sub-region to obtain the carbon sink calculation results for each monitoring sub-region includes: Step S21: Using the first forest and grassland species differentiation data in the satellite monitoring data of each monitoring sub-region as the basic data to be corrected, and the second forest and grassland species differentiation data in the UAV monitoring data as the standard data, the forest and grassland species differentiation data in the satellite monitoring data is corrected to obtain accurate forest and grassland species differentiation data.

[0053] Since the accuracy of forest and grassland species differentiation data in satellite monitoring data is low, this step uses UAV monitoring data with higher accuracy to modify the first forest and grassland species differentiation data in the satellite monitoring data, so as to obtain accurate forest and grassland species differentiation data after correction.

[0054] In detail, the correction of satellite monitoring data using UAV monitoring data as the fine data of the sampling grid includes: 1. Confirmation of tree presence: UAV imagery verifies the vegetation areas identified by the satellite, eliminating misclassifications of non-forest areas (such as buildings and water bodies). 2. Precise tree species classification: The fine classification results from the UAV are used to correct the coarse classification results from the satellite (e.g., if the satellite identifies it as "coniferous forest," the UAV corrects it to "Pinus tabuliformis forest"). 3. Organ-level carbon sequestration calculation: Combining the UAV's millimeter-level organ differentiation data (e.g., precise calculation of trunk volume) with the satellite's carbon sequestration classification parameters, the secondary carbon sequestration amount of "tree species - organ" within the grid is calculated (formula: carbon sequestration amount = biomass × carbon conversion coefficient, where biomass is calculated based on the millimeter-level organ size data and corresponding density, with a calculation error ≤2%).

[0055] Step S22: Calculate the carbon sink measurement results for each monitoring sub-region based on the area of ​​each monitoring sub-region, the biomass proportion data of different forest and grassland organs in each monitoring sub-region, the preset carbon density coefficient, the forest and grassland organ differentiation data, and the carbon sink classification results.

[0056] In one implementation, the carbon sequestration measurement results for each monitoring sub-region are calculated based on a preset carbon sequestration calculation formula, which is as follows: ; Where C represents the total forest carbon content of the monitored sub-region, and B represents the biomass per unit area. Carbon density coefficient, for example: coniferous forests broad-leaved forest R represents the proportion of root biomass. For example, R=0.2 means that root carbon storage accounts for 20% of the total carbon storage, and A represents the area of ​​the target region.

[0057] Step S3: Based on the carbon sink calculation results in each monitoring sub-region, select multiple sample monitoring regions from each monitoring sub-region according to the set screening conditions.

[0058] After obtaining the carbon sink test results in each monitoring sub-region in step S2 above, multiple monitoring sub-regions are selected as sample monitoring regions according to the preset screening conditions.

[0059] Specifically, the step of selecting multiple sample monitoring areas from each of the monitoring sub-regions according to the carbon sink calculation results and setting screening conditions includes the following screening conditions: whether the number of different tree species in the area exceeds a preset value, whether the area contains trees in different growth stages, whether the carbon sink calculation result exceeds a preset carbon sink value, and whether the area ratio of the sampling area to the experimental area exceeds a preset area ratio. If the number of different tree species in the area exceeds the preset value, it indicates that the area is a tree species diversity area, meaning that the area covers all major tree species within the monitoring area. Whether the area contains trees in different growth stages indicates whether the area is a growth stage differentiation area; if so, the area contains young forests, middle-aged forests, mature forests, and over-mature forests. Whether the area ratio of the sampling area to the experimental area exceeds the preset area ratio corresponds to determining whether it is a high-carbon sink area. A high-carbon sink area is a high NPP area (i.e., a high vegetation net primary productivity area) identified by satellite data; the area proportion: the total area of ​​the sampling / experimental area is not less than 5%-10% of the monitoring area.

[0060] Step S4: Use the multimodal monitoring data corresponding to each of the sample monitoring areas to train the preset neural network model to obtain the trained carbon sink monitoring model.

[0061] Once the monitoring areas for each sample are selected, the satellite monitoring data, UAV monitoring data, and ground condition and environmental data corresponding to the sample monitoring areas are input into the preset neural network model to train the preset neural network model, thereby obtaining the trained carbon sink monitoring model.

[0062] Specifically, in order to obtain more accurate prediction results, the preset neural network includes a classification module and a regression module. The classification module is used to classify forest and grassland species and organs. The regression module outputs the final carbon sink prediction result based on the classification results of forest and grassland species and organs output by the classification module. The carbon sink prediction result is compared with the actual carbon sink monitoring value in the sample monitoring area to obtain the error. The parameters of the preset neural network model are adjusted based on the error. The steps of adjusting the model parameters according to the error are repeated multiple times until training is completed, and the carbon sink monitoring model is obtained.

[0063] In detail, the steps for training a pre-set neural network model using satellite monitoring data and UAV monitoring data corresponding to each sample monitoring area to obtain a trained carbon sink monitoring model include: Step S41: Input the multimodal monitoring data of each sample monitoring area into the classification module to obtain the classification results of forest and grassland species and forest and grassland organs output by the classification module.

[0064] Step S42: Input the forest and grassland species, forest and grassland organ classification results and trunk parameter measurement data of each of the sample monitoring areas into the regression module to obtain the carbon sink prediction data output by the regression module.

[0065] Step S43: Calculate the error between the predicted carbon sink data and the actual carbon sink data of each sample monitoring area, adjust the parameters of the preset neural network model according to the error, and repeat the iteration multiple times until the preset training conditions are met to obtain the trained carbon sink monitoring model.

[0066] Combination Figure 4 As shown, the training of the preset neural network model is divided into two stages. The first stage involves inputting satellite monitoring data and UAV monitoring data of the sample monitoring area into the classification module to obtain the forest and grassland classification results and forest and grassland organ classification results output by the classification module. Then, the tree species classification results and forest and grassland organ classification results are input into the regression module to obtain the carbon sink prediction data output by the regression module.

[0067] Based on the error between the predicted carbon sink data and the actual carbon sink data, the parameters of the classification module and the regression module are adjusted, or the parameters of the classification module or the regression module are adjusted, to obtain the preset neural network model with updated parameters.

[0068] The satellite monitoring data, UAV monitoring data, and ground condition and environmental data of the sample monitoring area are then input into the classification module of the preset neural network model after parameter adjustment. Based on the updated classification module, the classification results of forest and grassland species and the classification results of forest and grassland organs are predicted. The predicted tree species classification results and forest and grassland organ classification results are then input into the updated regression model to obtain the carbon sink prediction data output by the updated regression model.

[0069] The process involves repeatedly inputting satellite monitoring data, UAV monitoring data, and ground condition and environmental data of the sample monitoring area into the preset neural network model after parameter adjustment, and adjusting the model parameters based on the error between the carbon sink prediction data output by the preset neural network model and the actual carbon sink data, until the error between the carbon sink prediction data output by the preset network model after parameter update and the actual carbon sink data is lower than the preset threshold, or the number of iterations reaches the final number, thus obtaining the trained carbon sink monitoring model.

[0070] Furthermore, in order to obtain a carbon sink monitoring model with more accurate prediction results, before the step of training a preset neural network model using satellite monitoring data and UAV monitoring data corresponding to each of the sample monitoring areas to obtain the trained carbon sink monitoring model, the following steps are also included: The satellite monitoring data is calibrated using UAV monitoring data within each of the sample monitoring areas, and flux tower data within each of the sample monitoring areas is obtained. The weighting coefficients and carbon conversion coefficients of the satellite monitoring data and UAV monitoring data are adjusted based on the flux tower data.

[0071] In detail, the initial calibration using satellite parameters includes: using millimeter-level fine data from UAVs in the sampling area as the "true value," and performing high-precision calibration of carbon sink parameters (such as the spectral-biomass conversion coefficient and growth measurement model) from the satellite data. For example, by using millimeter-level increments in the diameter at breast height (DBH) of a certain tree trunk measured by UAVs (such as a quarterly growth of 3.2 mm), the correspondence between the interferometric phase of the satellite SAR data and the actual growth can be deduced, improving the accuracy of satellite growth measurement from centimeter-level to millimeter-level correlation accuracy.

[0072] The steps for calibration and optimization using flux towers include: selecting flux towers (such as eddy covariance flux towers) already deployed within the designated monitoring area to obtain measured data such as CO2 flux and net ecosystem productivity (NEP); using flux tower data to verify the carbon sink calculation results of each monitoring sub-area in step S2; adjusting key parameters such as the weighting coefficients and carbon conversion coefficients of satellite and UAV data to ensure that the overall carbon sink calculation accuracy is ≤3%.

[0073] Furthermore, in order to obtain more accurate monitoring data, the method also includes: acquiring satellite monitoring data, UAV monitoring data, and ground condition and environmental data at preset time intervals, and fine-tuning the parameters of the carbon sink monitoring model using the satellite monitoring data, UAV monitoring data, and ground condition and environmental data.

[0074] In practice, satellite data collection can be performed every 1-3 months, and UAV data collection can be performed every 15-30 days (increased to 7-15 days in key areas), while meteorological data (temperature and precipitation) are recorded simultaneously to aid growth analysis.

[0075] like Figure 5As shown, the carbon sink monitoring model generation method provided in this embodiment first obtains satellite data, UAV data, and ground condition and environmental data based on satellite and UAV acquisition. These data are then processed to obtain satellite monitoring data, UAV monitoring data, and ground condition and environmental data, respectively. Coordinate registration is performed on the satellite and UAV monitoring data, and the registered UAV and satellite data are then integrated to calculate the carbon sink measurement results for each monitoring sub-region divided by the grid. Based on the carbon sink measurement results in each monitoring sub-region, multiple monitoring sub-regions are selected as sample monitoring sub-regions. The data from each selected sample monitoring sub-region is then used for calibration. Finally, the satellite monitoring data, UAV monitoring data, and ground condition and environmental data from each calibrated sample monitoring sub-region are used to iteratively train a preset neural network, thereby obtaining the trained carbon sink monitoring model.

[0076] In practice, in order to fine-tune the parameters of the trained carbon sink monitoring model by regularly acquiring real-time satellite monitoring data, drone monitoring data, and ground condition and environmental data, the trained carbon sink monitoring model can achieve more accurate carbon sink prediction.

[0077] Secondly, this application also discloses an integrated air-space-ground method for monitoring forest carbon sequestration, such as... Figure 6 As shown, it includes: Step H1: Acquire multimodal monitoring data of the target monitoring area; wherein, the multimodal monitoring data includes: satellite monitoring data, UAV monitoring data, and ground condition and environmental data. This step achieves precise monitoring of forest carbon sinks by acquiring monitoring data from the air, space, and ground.

[0078] Step H2: Input the multimodal monitoring data of the target monitoring area into the trained carbon sink monitoring model to obtain the carbon sink monitoring results of the target monitoring area output by the carbon sink monitoring model; wherein, the carbon sink monitoring model is obtained based on the generation method of the carbon sink monitoring model.

[0079] It is conceivable that the target monitoring area can be the monitoring area set in the above embodiments, or it can be other monitoring areas. When the target monitoring area is not a set monitoring area, given the potentially large differences in forest and grassland species within the area, in order to achieve accurate detection of forest carbon sinks in the target monitoring area, before using the carbon sink monitoring model generated in the above embodiments for carbon sink prediction, multiple eligible sub-regions within the target monitoring area can be selected to retrain the carbon sink monitoring model generated in the above embodiments to obtain a carbon sink monitoring model more suitable for the target monitoring area.

[0080] Thirdly, this application also discloses an integrated air-space-ground carbon sequestration monitoring system, such as... Figure 7 As shown, it includes: a monitoring data acquisition module 710 and a data processing module 720.

[0081] The monitoring data acquisition module 710 is used to acquire multimodal monitoring data of the target monitoring area; wherein, the multimodal monitoring data includes: satellite monitoring data, UAV monitoring data, and ground status and environmental data; its function is as described in step H1.

[0082] The data processing module 720 is used to input the multimodal monitoring data into the trained carbon sink monitoring model to obtain the carbon sink monitoring results of the target monitoring area output by the carbon sink monitoring model; wherein, the carbon sink monitoring model is obtained based on the carbon sink monitoring model generation method, and its function is as described in step H2.

[0083] The method and system disclosed in this invention have the following advantages: 1) By combining satellite monitoring data and drone monitoring data, more accurate carbon sink prediction results can be obtained.

[0084] The acquisition of satellite monitoring data utilized a combination of hyperspectral and SAR radar data for carbon sequestration measurement. Because the satellite orbit is fixed, satellite SAR radar data can clearly measure growth details smaller than centimeters, as well as the condition of tree trunks after leaf fall. The satellite's hyperspectral data clearly provides the absorption spectrum of trees, allowing for the differentiation between different trees and grasslands. Combined with the growth characteristics of different forest and grassland species, a carbon sequestration classification is given for each type of forest and grassland, thus obtaining the carbon sequestration parameters for that species.

[0085] The drone utilizes multiple pods, with visible light and multispectral imaging primarily used to differentiate between forest and grassland types, as well as the condition of fruits, leaves, and tree trunks. This differentiated data clearly reveals the carbon sequestration capacity of different components, providing a foundation for more accurate statistical analysis. The drone's onboard LiDAR camera mainly provides measurement data, depicting tree trunk growth. Combining this with visible light images reduces the precision requirements of the LiDAR, thus controlling overall hardware camera costs and the amount of data processed by subsequent software.

[0086] Because satellite monitoring data is large-scale and coarse-grained, it is more suitable for macro and global perspectives, while drone data is small-scale and more suitable for sampling and confirmation. Therefore, based on the geographic coordinate system, the forest is divided into multiple grids. By using the integrated algorithm of drone and satellite data, it is possible to comprehensively distinguish whether trees exist, tree species, and the trunk, leaves and fruits of a single tree species, so as to more accurately measure carbon sequestration.

[0087] 2) The trained carbon sink monitoring model can output more accurate prediction results.

[0088] When selecting the training dataset, this invention chooses key sampling and experimental areas, increasing periodic observation and data collection, especially the collection and analysis of UAV data, to provide a high-quality training dataset for the overall sample. Furthermore, during the construction of the training dataset, satellite data from the sampling and experimental areas is initialized and calibrated to determine the satellite data parameters, forming a closed loop. Accurate carbon sink monitoring equipment such as flux towers are used to adjust the initially calculated carbon sink data, and the model parameters are adjusted based on real carbon sink data, thereby training a carbon sink monitoring model with more accurate output prediction results.

[0089] This invention proposes an integrated air-space-ground carbon sink monitoring model generation method, monitoring method, and system. The method involves acquiring satellite and UAV monitoring data from multiple sub-regions within a defined monitoring area; calculating carbon sink data based on the satellite and UAV data from each sub-region to obtain carbon sink calculation results for each sub-region; selecting multiple sample monitoring areas from each sub-region according to set screening criteria based on the carbon sink calculation results; training a pre-defined neural network model using the satellite and UAV data corresponding to each sample monitoring area to obtain a trained carbon sink monitoring model; and finally, using the trained carbon sink monitoring model to monitor carbon sinks in the target monitoring area, thereby obtaining carbon sink monitoring results within the target monitoring area.

[0090] The integrated air-space-ground forest carbon sink model generation method, monitoring method, and system provided by this invention integrates satellite monitoring data and UAV monitoring data, achieving precise monitoring of forest carbon sinks through the collaborative use of UAV and satellite data. This overcomes the problems of low accuracy, low efficiency, and high cost associated with single remote sensing technologies, thus enabling accurate monitoring of forest carbon sinks. Furthermore, the method and system provided in this application are applicable to different types of forests (natural forests and plantations), balancing macroscopic monitoring with microscopic millimeter-level verification, and can directly serve scenarios with stringent accuracy requirements, such as carbon trading and ecological assessment.

[0091] It should be noted that the above application scenarios are shown only for the purpose of understanding the present invention, and the embodiments of the present invention are not limited in any way. On the contrary, the embodiments of the present invention can be applied to any applicable scenario.

[0092] It is understood that those skilled in the art can make equivalent substitutions or modifications to the technical solution and inventive concept of the present invention, and all such substitutions or modifications should fall within the protection scope of the appended claims.

Claims

1. A method for generating an integrated space-air-ground carbon sink monitoring model, characterized in that, include: Acquire multimodal monitoring data from multiple monitoring sub-regions within a defined monitoring area; wherein each monitoring sub-region is obtained by dividing the defined monitoring area according to a preset grid; the multimodal monitoring data includes: satellite monitoring data, UAV monitoring data, and ground status and environmental data; Carbon sink calculations are performed based on multimodal monitoring data from each monitoring sub-region to obtain carbon sink calculation results for each monitoring sub-region. Based on the carbon sink calculation results in each of the monitoring sub-regions, multiple sample monitoring regions are selected from each of the monitoring sub-regions according to the set screening conditions; The pre-set neural network model is trained using the multimodal monitoring data corresponding to the monitoring areas of each sample to obtain the trained carbon sink monitoring model.

2. The method for generating an integrated air-space-ground carbon sink monitoring model according to claim 1, characterized in that, The step of acquiring multimodal monitoring data from multiple monitoring sub-regions within a defined monitoring area includes: The SAR and hyperspectral data of forests and grasslands within the designated monitoring area are acquired using a target satellite located in a fixed orbit. The growth status and distribution of forests and grasslands within the area are monitored based on the SAR data, and the species of forests and grasslands and carbon sinks are distinguished and classified based on the hyperspectral data. Satellite monitoring data including forest and grassland growth status and distribution monitoring data, first forest and grassland species distinction data and carbon sink classification results are obtained. The UAV acquires visible light spectral data, multispectral data, and lidar data of forest and grassland vegetation within a designated monitoring area. Based on the visible light spectral data and multispectral data, it obtains second forest and grassland species differentiation data, forest and grassland organ differentiation data, and biomass proportion data of different forest and grassland organs. Based on the lidar data, it obtains trunk parameter measurement data and tree outline detection data. The second forest and grassland species differentiation data, forest and grassland organ differentiation data, forest and grassland organ biomass proportion data, trunk parameter measurement data, and tree outline detection data are combined to form UAV monitoring data.

3. The method for generating an integrated air-space-ground carbon sink monitoring model according to claim 2, characterized in that, The step of calculating carbon sinks based on multimodal monitoring data in each monitoring sub-region to obtain carbon sink calculation results for each monitoring sub-region includes: Using the first forest and grassland species differentiation data in the satellite monitoring data of each monitoring sub-region as the basic data to be corrected, and using the second forest and grassland species differentiation data in the UAV monitoring data as the standard data, the first forest and grassland species differentiation data in the satellite monitoring data is corrected to obtain accurate forest and grassland species differentiation data. The carbon sink calculation results for each monitoring sub-region are calculated based on the area of ​​each monitoring sub-region, the biomass proportion data of different forest and grassland organs in each monitoring sub-region, the preset carbon density coefficient, the forest and grassland organ differentiation data, and the carbon sink classification results.

4. The method for generating an integrated air-space-ground carbon sink monitoring model according to claim 1, characterized in that, The step of selecting multiple sample monitoring areas from each of the monitoring sub-regions according to the carbon sink measurement results in each of the monitoring sub-regions and according to the set screening conditions includes the following set screening conditions: whether the number of different tree species in the area exceeds a preset value, whether the area contains trees in different growth states, whether the carbon sink measurement results exceed a preset carbon sink value, and whether the area ratio of the sampling area to the experimental area exceeds a preset area ratio value.

5. The method for generating an integrated air-space-ground carbon sink monitoring model according to claim 1, characterized in that, The preset neural network model includes: a classification module and a regression module; The steps for training a pre-defined neural network model using multimodal monitoring data corresponding to each of the sample monitoring areas to obtain a trained carbon sink monitoring model include: Satellite monitoring data, UAV monitoring data, and ground condition and environmental data of each of the sample monitoring areas are input into the classification module to obtain the classification results of forest and grassland species and forest and grassland organs output by the classification module; The forest and grassland species, forest and grassland organ classification results, and trunk parameter measurement data of each of the sample monitoring areas are input into the regression module to obtain the carbon sink prediction data output by the regression module. The error between the predicted carbon sink data and the actual carbon sink data of each of the sample monitoring areas is calculated, and the parameters of the preset neural network model are adjusted according to the error. This process is repeated multiple times until the preset training conditions are met, thus obtaining a trained carbon sink monitoring model.

6. The method for generating an integrated air-space-ground carbon sink monitoring model according to claim 1, characterized in that, Before the step of training a preset neural network model using multimodal monitoring data corresponding to each of the sample monitoring areas to obtain a trained carbon sink monitoring model, the method further includes: The satellite monitoring data is calibrated using UAV monitoring data within each of the sample monitoring areas, and flux tower data within each of the sample monitoring areas is obtained. The weighting coefficients and carbon conversion coefficients of the satellite monitoring data and UAV monitoring data are adjusted based on the flux tower data.

7. The method for generating an integrated air-space-ground carbon sink monitoring model according to claim 1, characterized in that, The preset grid is obtained by uniformly dividing the area according to the preset grid area based on the geographic coordinate system; or the preset grid is obtained by dividing the area according to the terrain based on the geographic coordinate system.

8. The method for generating an integrated air-space-ground carbon sink monitoring model according to claim 1, characterized in that, The method further includes: acquiring satellite monitoring data and drone monitoring data at preset time intervals, and using the satellite monitoring data and drone monitoring data to fine-tune the parameters of the carbon sink monitoring model.

9. A space-air-ground integrated carbon sequestration monitoring method, characterized in that, include: Acquire multimodal monitoring data of the target monitoring area; wherein, the multimodal monitoring data includes: satellite monitoring data, UAV monitoring data, and ground condition and environmental data; The multimodal monitoring data of the target monitoring area is input into the trained carbon sink monitoring model to obtain the carbon sink monitoring results of the target monitoring area output by the carbon sink monitoring model; wherein, the carbon sink monitoring model is obtained based on the carbon sink monitoring model generation method as described in any one of claims 1-8.

10. A space-air-ground integrated carbon sequestration monitoring system, characterized in that, include: Monitoring data acquisition module and data processing module; The monitoring data acquisition module is used to acquire multimodal monitoring data of the target monitoring area; wherein, the multimodal monitoring data includes: satellite monitoring data, UAV monitoring data, and ground condition and environmental data; The data processing module is used to input the multimodal monitoring data into the trained carbon sink monitoring model to obtain the carbon sink monitoring results of the target monitoring area output by the carbon sink monitoring model; wherein, the carbon sink monitoring model is obtained based on the carbon sink monitoring model generation method as described in any one of claims 1-8.

Images