Submerged plant carbon sink measuring and calculating method based on space remote sensing and morphological modeling

Abstract

The invention relates to a submerged plant carbon sink measuring and calculating method based on space remote sensing and morphological modeling. The method comprises the following steps: performing remote sensing image acquisition and preprocessing operation on a target water body area; submerged plant image recognition training is carried out, and a submerged plant area in the image is recognized; performing classification modeling and three-dimensional volume estimation on the submerged plants; calculating the biomass of each type of submerged plants according to the three-dimensional volumes calculated by the different types of submerged plants, and carrying out weighted average summary according to the proportion or absolute value of the distribution area of the biomass to form the total biomass of the region; and calculating the annual net carbon sink capacity of the submerged plant ecosystem based on the regional total biomass in combination with the plant carbon content coefficient and the litter carbon release parameter. According to the method, a calculation process with parameter adaptability and feedback optimization capability is constructed through remote sensing information identification, typical plant morphology classification, structural parameter estimation and carbon flux deduction, and quantitative estimation and dynamic evaluation of the carbon sequestration capability of the submerged plants are realized.

Description

Technical Field

[0001] The present invention relates to the technical field of carbon sink assessment of aquatic ecosystems, and in particular to a method for calculating carbon sinks of submerged plants based on space remote sensing and morphological modeling. Background Art

[0002] As important producers in aquatic ecosystems, submerged macrophytes possess significant carbon sequestration capabilities. They absorb inorganic carbon from the water through photosynthesis, converting the carbon into organic matter during their growth process, which is then fixed within the plant and deposited in the sediment, constituting a key carbon sink in inland water systems. Studies have shown that submerged macrophytes play an important ecological role in maintaining carbon balance in the water, regulating CO2 flux across the water-air interface, and inhibiting greenhouse gas emissions. However, current assessments of the carbon sequestration capacity of submerged macrophytes are still primarily based on small-scale plot surveys, sampling measurements, and laboratory simulations, resulting in problems such as fragmented data, narrow coverage, and discontinuous calculations.

[0003] At the same time, the rapid development of remote sensing technology has provided an opportunity for large-scale identification and dynamic monitoring of ecological elements. Remote sensing images such as hyperspectral, multispectral, radar, and lidar can efficiently capture the reflectance characteristics of the water surface and have the potential to identify the distribution of submerged vegetation. However, existing methods are mostly limited to identifying the distribution of aquatic plants or extracting their area, lacking a complete systematic process for carbon sequestration estimation. For submerged plants in particular, their carbon sequestration capacity is not only affected by their distribution area but is also closely related to dynamic parameters such as vegetation structure, biomass density, species composition, and litter decomposition. Therefore, relying solely on empirical estimates of remote sensing area is unlikely to meet the needs of refined carbon budget management.

[0004] Therefore, there is an urgent need for a method system that integrates remote sensing image recognition, plant morphological structure modeling and life cycle carbon budget accounting to break through the limitations of traditional point surveys and static parameter replacement and realize large-scale, high-efficiency and refined submerged plant carbon sink measurement. Summary of the Invention

[0005] The technical problem to be solved by the present invention is to overcome the shortcomings of the existing technology and provide a submerged plant carbon sink calculation method based on space remote sensing and morphological modeling. It aims to construct a calculation process with parameter adaptability and feedback optimization capabilities through remote sensing information recognition, typical plant morphology classification, structural parameter estimation and carbon flux deduction, so as to realize quantitative estimation and dynamic evaluation of the carbon sink capacity of submerged plants.

[0006] The present invention is achieved through the following technical solutions: A method for estimating carbon sequestration of submerged macrophytes based on space remote sensing and morphological modeling includes the following steps: S1. Collect remote sensing images of the target water area using a drone and preprocess the remote sensing images to generate georeferenced images; S2. Based on the preprocessed remote sensing image, perform submerged plant image recognition training to identify submerged plant areas and non-vegetated areas in the image; S3. Based on the identification results of submerged plants, classify and model the submerged plants and estimate their three-dimensional volume; S4. Calculate the biomass of each submerged macrophyte category based on the calculated three-dimensional volume of each category, and perform a weighted average based on the proportion of the distribution area occupied or the absolute value to form the total biomass of the region; S5. Calculate the annual net carbon sink capacity of the submerged plant ecosystem per unit time based on the total biomass of the region, combined with the plant carbon content coefficient and litter carbon release parameters; S6. Automatically collect and calculate carbon flux data on an hourly scale to generate the Q_obs time series. Compare the calculated net carbon sink C_net with the model-estimated carbon sink Q_model. If the deviation between the two exceeds ±15% in the same time window, the feedback correction mechanism is triggered.

[0007] According to the above technical solution, preferably, in step S1, preprocessing the remote sensing image includes: Use GNSS+RTK positioning data and UAV attitude information to complete orthorectification and mosaic reconstruction; Use OpenCV and GDAL libraries to perform band normalization and illumination uniform processing on the remote sensing images; Identify non-water interference targets in the remote sensing image and perform mask processing, and use an improved NDWI water index plus morphological filtering combination to achieve accurate water area extraction; The final output image is divided into blocks and coordinate mosaicked to generate a georeferenced image.

[0008] According to the above technical solution, preferably, step S2 includes: Based on the georeferenced image, the original bands are combined and operated on the basis of the multispectral image of the drone to generate multiple spectral index layers; With the U-Net convolutional neural network as the core architecture, the spectral index layer is used as the "multi-channel input". The encoder part of the U-Net model extracts the spatial texture features and spectral variation rules of the image, and the decoder part restores the extracted results to the original Figure 1 Consistent spatial resolution enables pixel-by-pixel classification; After the recognition is completed, a binary mask image is output, where "1" indicates the submerged plant area and "0" indicates the non-vegetation area; Combined with the known resolution of the UAV imagery, the total coverage area A of submerged plants was counted by pixel.

[0009] According to the above technical solution, preferably, in step S2, the spectral index layer includes the Normalized Aquatic Vegetation Index NDAVI, the Water Plant Index NDPI, and the Enhanced Vegetation Index EVI.

[0010] According to the above technical solution, preferably, in step S3, submerged plants are divided into three categories according to their structural morphology and growth mode, namely linear leaf type, tubular branch type, and creeping expansion type, and three-dimensional modeling is performed on each of them according to their geometric characteristics.

[0011] According to the above technical solution, preferably, in step S3, the modeling volume formula of the linear leaf class is: V=(4 / 3)×π×a×b×c, where a is the length of half a leaf, b is the width of half a leaf, and c is the thickness of half a leaf. The tubular branch class is defined as a combination of multiple cylinders for modeling, and the modeling volume formula of each cylinder is: V_i=π×(d_i / 2)2×h_i, where d_i is the diameter of the segment, h_i is the corresponding height, and the sum of multiple segments is the volume of the whole plant; the modeling volume formula of the creeping extension class is: V ， =A_local×t, where A_local is the average coverage area of ​​a single creeping plant and t is the average thickness.

[0012] According to the above technical solution, preferably, step S4 includes: Based on the three-dimensional volume Vi calculated for different types of submerged macrophytes, the dry matter density ρ corresponding to the selected plant species was called to calculate the dry weight of a single plant: Bio = Vi × ρ, where Vi is the three-dimensional volume of the linear leaf type, tubular branching type, or creeping extension type; Combined with the planting density D obtained by image density recognition, the dry weight per unit area is calculated as Bio_a=D×Bio; Combined with the total coverage area A of submerged plants, the total biomass of this type of submerged plants in the entire area is Bio_t=A×Bio_a=A×D×Bio; Modeling and calculation are performed for different types of submerged macrophytes to obtain the total biomass Bio_t_i of each type of submerged macrophyte. The weighted average is summarized according to the proportion of distribution area or absolute value to form the total regional biomass Bio_T=Σ(Bio_t_i).

[0013] According to the above technical solution, preferably, step S5 includes: The total annual carbon sequestration amount C_t=Bio_t×C_s, where C_s is the average proportion of carbon per unit dry weight, and the default value is 0.45; Annual carbon emissions C_d=Bio_t×R_d, where R_d is set based on regional environmental factors; Calculate the annual net carbon sink capacity of the submerged plant ecosystem per unit time, that is, the net carbon sink C_net=C_t-C_d.

[0014] The beneficial effects of the present invention are: Building on traditional remote sensing identification and ecological carbon sink research, this paper proposes a comprehensive, systematic approach that integrates space-based remote sensing, AI identification, plant structure modeling, and lifecycle carbon budget accounting. This approach offers significant technological innovation and increased application potential. Its core innovation lies in a systematic shift away from the traditional crude calculation method of "area estimation followed by empirical formulas." The approach pioneers the use of three-dimensional geometric modeling to parametrically describe the morphological structures of submerged plants (such as leaves and stems). This approach achieves a comprehensive modeling chain, from two-dimensional remote sensing identification to three-dimensional volume estimation and then to unit biomass conversion, significantly improving the accuracy and explanatory power of carbon sink measurements.

[0015] The present invention constructs a life cycle carbon budget model covering photosynthetic carbon fixation and litter carbon emission, systematically evaluating the net carbon sink capacity of submerged plants throughout their entire growth cycle, taking into account both carbon fixation contributions and carbon release risks, and making up for the current bias in the rough valuation based only on "positive carbon fixation", providing more balanced data support for water carbon flux research and greenhouse gas budgets.

[0016] This method enables automated processing and multi-temporal analysis of remote sensing imagery. Combined with AI deep learning models, it can rapidly identify and dynamically update the spatial distribution of submerged macrophytes, adapting to ecological response assessments in the context of seasonal variations, climate impacts, and engineering interventions. This mechanism significantly expands the application scope of remote sensing data and enhances the system's real-time and intelligent capabilities in carbon monitoring scenarios.

[0017] This method boasts excellent adaptability and versatility, not limited to a specific type of water body. It can be applied to a variety of ecological units, including lakes, reservoirs, rivers, and wetlands. Model parameters can be flexibly adjusted based on environmental factors such as submerged plant species, transparency, water depth, and temperature, demonstrating a high degree of parametric design and scalability. This adaptive capability provides a standardized and localized approach to ecological governance and carbon sequestration management in different regions. BRIEF DESCRIPTION OF THE DRAWINGS

[0018] Figure 1 It is a flow chart of the submerged plant carbon sink calculation method disclosed in the present invention.

[0019] Figure 2 This is a schematic diagram of modeling the linear leaf-like model in step S3 of the present invention.

[0020] Figure 3 This is a schematic diagram of the branch class model in step S3 of the present invention.

[0021] Figure 4 This is a schematic diagram of modeling the creeping extension class model in step S3 of the present invention. DETAILED DESCRIPTION

[0022] In order to enable those skilled in the art to better understand the technical solution of the present invention, the present invention is further described in detail below with reference to the accompanying drawings and the best embodiment. Based on the embodiments of the invention, all other embodiments obtained by those skilled in the art without making any creative work shall fall within the scope of protection of the invention.

[0023] The method of the present invention constructs a carbon sink calculation path method with "remote sensing analysis - distribution identification - morphological modeling - biomass conversion - carbon flux deduction - measured feedback - results output" as the technical main line. It has the advantages of clear structure, closed-loop parameters, and strong application flexibility. It is particularly suitable for large-scale water bodies such as lakes, reservoirs, and rivers. Without relying on large-scale fixed facilities, it can be flexibly deployed and quickly estimated. It is suitable for research applications such as ecological monitoring and carbon sink management, and also has practical potential for docking with ecological compensation and carbon asset verification systems.

[0024] Example 1: As shown in the figure, the present invention includes the following steps: S1. Collect remote sensing images of the target water area using a drone, and preprocess the remote sensing images to generate geo-referenced images.

[0025] In this example, drones are used as remote sensing platforms. Fixed-wing or rotary-wing drones equipped with multispectral or visible light + near-infrared sensors are used to capture low-altitude, multi-angle, and multi-temporal images of target water areas. The drones typically fly at an altitude between 60 and 120 meters. The flight path is automatically generated based on the water boundary, supporting precise route programming and overlap settings to ensure sufficient spatial coverage and resolution.

[0026] After image acquisition is completed, the system automatically downloads the original image through a dedicated ground station or remote control data link and simultaneously imports it into the image preprocessing module. This module includes the following processing steps: Use GNSS+RTK positioning data and UAV attitude information to complete orthorectification and mosaic reconstruction (structured light matching + sparse point cloud reconstruction); Use OpenCV and GDAL libraries to normalize the image bands and unify the lighting; Automatically identify non-water interference targets (such as ships, reflections, and shoreline vegetation) in images and perform mask processing, and use the improved NDWI water index plus morphological filtering to achieve precise water area extraction; The final output image is divided into blocks and coordinate mosaicked to generate georeferenced image products for vegetation identification and feature extraction in subsequent modules.

[0027] This step has the integrated capabilities of flight planning, image acquisition and image standardization, and can quickly complete data acquisition and standardized processing in the lake area, greatly improving the system's timeliness, spatial accuracy and applicability.

[0028] S2. Based on the preprocessed remote sensing images, perform submerged plant image recognition training to identify submerged plant areas and non-vegetation areas in the images.

[0029] Because submerged plants reside in the lower layers of water year-round, their light reflection characteristics are weak, especially in the visible light band. Conventional image processing methods have difficulty distinguishing between water bodies and areas covered by submerged plants. To address this issue, this step constructs a deep learning-based image recognition module to train submerged plant image recognition and distinguish between submerged plant areas and non-vegetated areas in images.

[0030] First, based on the georeferenced imagery and the drone's multispectral imagery, the original red, green, blue, and near-infrared bands are combined and calculated to generate spectral index layers such as the NDAVI (Normalized Difference Aquatic Vegetation Index), NDPI (Water Plant Index), and EVI (Enhanced Vegetation Index). These indices amplify the reflectance differences between vegetation and the background water, improving subsequent recognition accuracy.

[0031] Subsequently, the system uses the U-Net convolutional neural network as the core architecture, and uses the above-mentioned multiple spectral index layers as "multi-channel inputs", similar to converting the original image into a stereo image containing multiple feature dimensions. The encoder part of the U-Net model can extract the spatial texture features and spectral variation patterns of the image, and the decoder part restores the extracted results to the original image. Figure 1 Consistent spatial resolution enables pixel-by-pixel classification.

[0032] To train the model, the system collects historical underwater video footage and on-site salvage locations, and manually annotates them to create a high-quality training sample set. The model uses Intersection over Union (IoU) as its primary loss function, focusing on improving the accuracy of boundary region recognition.

[0033] After identification, the system outputs a binary mask, where "1" indicates areas identified as submerged plants and "0" indicates areas without vegetation. Based on the known resolution of the drone imagery (e.g., 0.05 meters per pixel), the system calculates the total submerged plant cover area A (in square meters) per pixel. It then automatically extracts spatial boundaries through vectorization and outputs them in standard GIS formats such as Shapefile for subsequent modeling.

[0034] S3. Based on the identification results of submerged plants, classification modeling and three-dimensional volume estimation of submerged plants are performed.

[0035] In this example, submerged plants are divided into three categories according to their structural morphology and growth pattern: linear leaves, tubular branches, and creeping extensions. Three-dimensional models are then created based on their geometric characteristics for 3D volume estimation.

[0036] Linear-leaved plants (such as Vallisneria and Myriophyllum spicate) have slender, soft leaves arranged in a ribbon-like pattern. The configuration of a single plant can be approximated as a triaxial ellipsoid. The modeling volume formula is: V = (4 / 3) × π × a × b × c, where a is the half-leaf length (cm), b is the half-leaf width (cm), and c is the half-leaf thickness (cm). It should be noted that measurements are generally made using the maximum dimension of the unfolded leaf, often using the sample average.

[0037] Tubular-branching plants (such as Potamogeton pectinata and Hydrilla verticillata): These plants have short, thick stems and well-developed branches, exhibiting a typical layered cylindrical structure. They can be modeled as a combination of multiple cylinders. The volume of each cylinder is calculated as: V_i = π × (d_i / 2)² × h_i, where d_i is the diameter of the segment (cm) and h_i is the corresponding height (cm). The sum of the multiple segments is the volume of the entire plant.

[0038] Creeping and spreading plants (such as hornwort and foxtail algae): These plants grow close to the bottom, are low and spreading, and it is difficult to distinguish the boundaries of the individual plants. They are suitable for surface modeling. The modeling volume formula is: V ， =A_local×t, where A_local is the average coverage area of ​​a single creeping plant (m 2 ), t is the average thickness (cm).

[0039] There are two ways to input parameters: one is manual measurement: technicians use a typical sample plot (such as 1m 2 ) to measure various parameters (a, b, c, d, h, t); second, the system default experience parameter library: derived from the "China Typical Aquatic Plant Database", "Wetland Ecological Vegetation Atlas" and other documents, including the mean, standard deviation and regional variation coefficient, suitable for large-scale rapid estimation scenarios.

[0040] Through the above classification and modeling methods, the system can flexibly call the model structure and parameter combination according to the plant species and morphological characteristics, and quickly, efficiently and batch complete the three-dimensional volume and biomass modeling of submerged plants in a large area.

[0041] S4. Calculate the biomass of each submerged plant category based on the calculated three-dimensional volume of each category, and perform weighted average aggregation based on the proportion of the distribution area occupied or the absolute value to form the total regional biomass.

[0042] The biomass estimation module completes the conversion from structural volume to organic dry weight based on 3D modeling, which is a key intermediate link in carbon sequestration capacity calculation. First, based on the 3D volume Vi (unit: cm) of different types of submerged plants (single plant) calculated in the previous stage, 3 ), call the dry matter density ρ (unit: g / cm) corresponding to the selected plant species 3 ), calculate the dry weight of a single plant Bio = Vi × ρ, where Vi is the three-dimensional volume of linear leaves, tubular branches, or creeping extensions, and ρ is the dry matter mass per unit volume after dehydration of the plant, which is derived from empirical databases or user measurements (obtained by sampling and drying methods). ρ varies among different plant species, for example, Vallisneria sinensis has a value of 0.43-0.48 g / cm 3 , Hydrilla verticillata 0.46~0.52 g / cm 3 , hornwort 0.40~0.46 g / cm 3 .

[0043] Secondly, the system combines the number of plants per square meter, that is, the planting density D (unit: plants / m 2 ), obtained through image density recognition, and further calculated the dry weight per unit area (g / m 2 ): Bio_a=D×Bio.

[0044] Combined with the total coverage area of ​​submerged plants A (unit: m) output by the remote sensing module in the early stage 2 ), and finally estimate the total biomass of a certain type of submerged plants in the entire area: Bio_t=A×Bio_a=A×D×Bio.

[0045] In actual operation, the system supports the coexistence of multiple submerged plants. In this case, the system will model each plant separately to obtain the total biomass Bio_t_i of each submerged plant, and perform weighted average summary according to the proportion of distribution area or absolute value to form the regional total biomass Bio_T=Σ(Bio_t_i).

[0046] This step achieves accurate estimation of basic data from plant morphology and volume to regional carbon storage, providing solid data support for subsequent carbon content estimation, carbon sink assessment and dynamic monitoring.

[0047] S5. Based on the total biomass of the region, combined with the plant carbon content coefficient and litter carbon release parameters, calculate the annual net carbon sink capacity of the submerged plant ecosystem per unit time.

[0048] Submerged macrophytes absorb inorganic carbon from water through photosynthesis and convert it into organic carbon, which is primarily stored in the plant's dry matter. The system multiplies the total biomass (Bio_t) by the carbon content factor (plant carbon content coefficient) C_s to calculate the total plant carbon sequestration (C_t). This is calculated as follows: Annual carbon sequestration (C_t) = Bio_t × C_s. C_s is the average proportion of carbon per unit dry weight (plant carbon content coefficient). The default value is 0.45, which can be adjusted based on species.

[0049] During the natural aging, withering, and decomposition processes of submerged macrophytes, some carbon is released back into the water or atmosphere in the form of CO2 and CH4. The system estimates the total annual carbon emissions (Cd) based on the regional biomass (Bio_t) and the carbon release parameter (R_d) from litterfall: C_d = Bio_t × R_d. R_d is set based on regional environmental factors. The default classification criteria are as follows: eutrophic waters (temperature > 25°C and heavy sediment deposition) have an R_d of 0.30; mesotrophic waters (normal temperature 15-25°C) have an R_d of 0.20; and oligotrophic waters (low temperature < 15°C and high transparency) have an R_d of 0.12.

[0050] Calculate the annual net carbon sink capacity of the submerged plant ecosystem per unit time. The net carbon sink represents the positive carbon balance value generated by the plant community per unit time, that is, the net carbon sink C_net=C_t-C_d.

[0051] S6. Automatically collect and calculate carbon flux data on an hourly scale to generate the Q_obs time series. Compare the calculated net carbon sink C_net with the model-estimated carbon sink Q_model. If the deviation between the two exceeds ±15% in the same time window, the feedback correction mechanism is triggered.

[0052] Specifically, to ensure that the carbon sink values ​​estimated by the model are closer to the actual carbon exchange capacity of the ecosystem, the "on-site water-air boundary carbon exchange dynamic observation submodule" is deployed, using the principle of non-intrusive open flux observation. This module integrates the following core equipment: High-frequency infrared gas analyzer (IRGA) for real-time monitoring of CO2 / CH4 concentration changes in the atmospheric boundary layer; micro-meteorological sensor array (wind speed, wind direction, air temperature, relative humidity); dissolved gas concentration sensor in water (electrode or optical); data synchronization acquisition system and dynamic flux calculation unit.

[0053] In this example, the “open boundary – high frequency sampling – gradient solution” method is used to derive the carbon flux. According to the improved Fick diffusion formula: Q = k × (Cw-Ca), where Q is the carbon flux (mg C / m 2 / h), representing the net amount of carbon released or absorbed by the water body into the air; k is the gas exchange coefficient, which is dynamically calculated by the system, taking into account environmental parameters such as wind speed disturbance, water surface temperature difference, and Reynolds number; Cw is the concentration of dissolved inorganic carbon (DIC) in the water body, in mg / L; Ca is the concentration of CO2 in the air, in ppm (converted by standard).

[0054] The system automatically collects and calculates carbon flux data on an hourly scale, generating a Q_obs time series. Simultaneously, the system infers the model-estimated carbon sink, Q_model (converted from C_net), based on remote sensing and modeling processes. When the two deviate by more than ±15% within the same time window, a feedback correction mechanism is triggered, fine-tuning the carbon content factor (plant carbon content coefficient) or litter carbon release parameters. By fitting residual trends, the system dynamically updates the parameter library for the corresponding species or environmental scenario, enhancing the adaptability and accuracy of the regional model.

[0055] Through this dynamic observation and feedback correction mechanism, "measurement-driven model adaptive optimization" can be achieved, continuously reducing the deviation between simulation results and actual ecological processes, and improving the authenticity, reliability and engineering applicability of submerged plant carbon sink estimates.

[0056] S7. After completing the full-process modeling and observation analysis of the carbon sequestration capacity of submerged macrophytes, the results are output in various forms. The system results support visualization in various ways such as graphics, zoning, and dynamic trending.

[0057] The system flexibly generates carbon sink information displays and data structures at different levels, including submerged plant distribution, carbon sink spatial gradients, and biomass composition, supporting a multi-angle and multi-scale representation of regional carbon storage patterns. Users can perform overlay analysis, trend analysis, or classification statistics based on the resulting layers, supporting ecological assessments, management optimization, and carbon trading. The system's results are well-connected, with configurable output interfaces for consistent integration with third-party platforms (such as carbon sink accounting systems, water ecological supervision platforms, and ecological compensation accounting modules).

[0058] Example 2: Based on the above Example 1, this example discloses a specific operation path of a submerged plant carbon sink measurement method based on space remote sensing and morphological modeling, demonstrating how this method can complete the refined deduction of carbon flux through data acquisition, image processing, structural modeling, biomass measurement and carbon sink estimation processes in actual applications.

[0059] First, during the remote sensing image acquisition phase, a rotary-wing drone equipped with a multispectral imaging system was used to conduct flight operations over the target water area. The flight altitude was controlled at 80 meters, with a forward overlap of 75% and a lateral overlap of 65%. The multi-channel image data (red, green, blue, and near-infrared) acquired by the drone was corrected using GNSS+RTK positioning data and IMU attitude parameters to complete orthorectification and image stitching. Subsequently, band normalization, brightness balancing, and image tiling were performed using the GDAL and OpenCV libraries. The NDWI index was used to extract water masks and remove interfering objects in non-water areas, ultimately producing a standardized remote sensing image.

[0060] Next, in the submerged plant spatial identification phase, the processed image is fed into a trained U-Net deep learning model. The input includes vegetation index layers such as NDAVI, NDPI, and EVI, and the output is a pixel-by-pixel submerged plant classification mask. The system counts the number of pixels in the mask and, based on the image resolution (e.g., 0.05 meters / pixel), calculates the total submerged plant distribution area A. The corresponding shapefile vector file is then exported for subsequent modeling.

[0061] During the modeling phase, the system categorizes submerged macrophytes into linear leaf-shaped and tubular branch-shaped groups based on regional species characteristics, and calls the modeling modules for each: Vallisneria is modeled as an ellipsoid with parameters a=10cm, b=0.5cm, and c=0.1cm; Hydrilla verticillata is modeled as a cylindrical assembly consisting of a main stem (d=0.8cm, h=12cm) and two secondary branches (d=0.6cm, h=8cm; d=0.5cm, h=6cm). The system calculates each volume V_i and, based on the species dry matter density ρ (e.g., Vallisneria ρ=0.45g / cm 3 , black algae ρ = 0.48g / cm 3 ) Estimate the dry weight of a single plant Bio=V×ρ.

[0062] Then, in the biomass estimation stage, the average planting density of the region was obtained through sample surveys, which was D = 120 plants / m 2 Combined with the remote sensing identification area A, the total biomass of the region is calculated as Bio_total = A × D × Bio. For areas where multiple plant species coexist, the system will calculate Bio_total_i for each species and summarize them into Bio_T by weighted area.

[0063] Next, during the carbon sequestration calculation phase, we set the carbon content ratio (C_s) to 0.45 and the litter carbon release coefficient (R_d) to 0.20 (based on water transparency and temperature). The system calculates annual carbon sequestration using the formula C_t = Bio_total × C_s, annual carbon emissions using C_d = Bio_total × R_d, and net carbon sequestration capacity using C_net = C_t - C_d.

[0064] During the field observation feedback phase, open carbon flux observation points were set up in representative areas, equipped with infrared CO2 analyzers, micrometeorological sensors, and water gas concentration probes. The system records the measured flux Q_obs on an hourly basis and derives the model Q_model using the Fick equation. If the deviation exceeds 15%, the system will correct C_s and R_d, automatically updating the parameter curve.

[0065] Finally, the system exports outputs including GeoTIFF distribution maps, biomass maps, carbon sink intensity maps, and PDF graphic reports. It also supports exporting administrative unit statistics and exporting them to third-party platforms. The system automatically fits correction factors based on error trends each year, enabling adaptive, iterative updates to the model parameter library.

[0066] In summary, this application forms a standardized, modular, and universal submerged plant carbon sink calculation technology path from remote sensing acquisition, image processing, plant classification modeling, biomass estimation, annual carbon flux extrapolation to field measurement feedback correction. It has the advantages of clear structure, closed-loop parameters, and strong application flexibility, and is suitable for submerged plant carbon sink monitoring scenarios in different regions and multiple types of water bodies.

[0067] The above is only a preferred embodiment of the present invention. It should be pointed out that for ordinary technicians in this technical field, several improvements and modifications can be made without departing from the principles of the present invention. These improvements and modifications should also be regarded as within the scope of protection of the present invention.

Claims

1. A method for estimating carbon sinks of submerged plants based on space remote sensing and morphological modeling, characterized in that: The steps include: S1. Collect remote sensing images of the target water area using a drone and preprocess the remote sensing images to generate georeferenced images; S2. Based on the preprocessed remote sensing image, perform submerged plant image recognition training to identify submerged plant areas and non-vegetated areas in the image; S3. Based on the identification results of submerged plants, classify and model the submerged plants and estimate their three-dimensional volume; S4. Calculate the biomass of each submerged macrophyte category based on the calculated three-dimensional volume of each category, and perform a weighted average based on the proportion of the distribution area occupied or the absolute value to form the total biomass of the region; S5. Based on the total biomass of the region, combined with the plant carbon content coefficient and litter carbon release parameters, calculate the annual net carbon sink capacity of the submerged plant ecosystem per unit time.

2. The method for calculating carbon sinks of submerged plants based on space remote sensing and morphological modeling according to claim 1, characterized in that: In step S1, preprocessing the remote sensing image includes: Use GNSS+RTK positioning data and UAV attitude information to complete orthorectification and mosaic reconstruction; Use OpenCV and GDAL libraries to perform band normalization and illumination uniform processing on the remote sensing images; Identify non-water interference targets in the remote sensing image and perform mask processing, and use an improved NDWI water index plus morphological filtering combination to achieve accurate water area extraction; The final output image is divided into blocks and coordinate mosaicked to generate a georeferenced image.

3. The method for calculating carbon sinks of submerged plants based on space remote sensing and morphological modeling according to claim 1, characterized in that: Step S2 includes: Based on the georeferenced image, the original bands are combined and operated on the basis of the multispectral image of the drone to generate multiple spectral index layers; Using the U-Net convolutional neural network as its core architecture, the spectral index layer is used as a "multi-channel input." The encoder part of the U-Net model extracts the spatial texture features and spectral variation patterns of the image, while the decoder part restores the extracted results to the same spatial resolution as the original image, achieving pixel-by-pixel classification. After the recognition is completed, a binary mask image is output, where "1" indicates the submerged plant area and "0" indicates the non-vegetation area; Combined with the known resolution of the UAV imagery, the total coverage area A of submerged plants was counted by pixel.

4. The method for calculating carbon sinks of submerged plants based on space remote sensing and morphological modeling according to claim 3, characterized in that: In step S2, the spectral index layer includes the Normalized Difference Aquatic Vegetation Index (NDAVI), the Water Plant Index (NDPI), and the Enhanced Vegetation Index (EVI).

5. The method for calculating carbon sinks of submerged plants based on space remote sensing and morphological modeling according to claim 3, characterized in that: In step S3, submerged plants are divided into three categories according to their structural morphology and growth pattern, namely linear leaf type, tubular branch type, and creeping expansion type, and three-dimensional modeling is performed on each of them according to their geometric characteristics.

6. The method for calculating carbon sinks of submerged plants based on space remote sensing and morphological modeling according to claim 5, characterized in that: In step S3, the modeling volume formula of the linear leaf class is: V=(4 / 3)×π×a×b×c, where a is the half leaf length, b is the half leaf width, and c is the half leaf thickness. The tubular branch type is defined as a combination of multiple cylinders. The modeling volume formula of each cylinder is: V_i=π×(d_i / 2)2×h_i, where d_i is the diameter of the segment and h_i is the corresponding height. The sum of multiple segments is the volume of the entire plant. The modeling volume formula of the creeping extension class is: V ， =A_local×t, where A_local is the average coverage area of ​​a single creeping plant and t is the average thickness.

7. The method for calculating carbon sinks of submerged plants based on space remote sensing and morphological modeling according to claim 6, characterized in that: Step S4 includes: Based on the three-dimensional volume Vi calculated for different types of submerged macrophytes, the dry matter density ρ corresponding to the selected plant species was called to calculate the dry weight of a single plant: Bio = Vi × ρ, where Vi is the three-dimensional volume of the linear leaf type, tubular branching type, or creeping extension type; Combined with the planting density D obtained by image density recognition, the dry weight per unit area is calculated as Bio_a=D×Bio; Combined with the total coverage area A of submerged plants, the total biomass of this type of submerged plants in the entire area is Bio_t=A×Bio_a=A×D×Bio; Modeling and calculation are performed for different types of submerged macrophytes to obtain the total biomass Bio_t_i of each type of submerged macrophyte. The weighted average is summarized according to the proportion of distribution area or absolute value to form the total regional biomass Bio_T=Σ(Bio_t_i).

8. The method for calculating carbon sinks of submerged plants based on space remote sensing and morphological modeling according to claim 7, characterized in that: Step S5 includes: The total annual carbon sequestration amount C_t=Bio_t×C_s, where C_s is the average proportion of carbon per unit dry weight, and the default value is 0.45; Annual carbon emissions C_d=Bio_t×R_d, where R_d is set based on regional environmental factors; Calculate the annual net carbon sink capacity of the submerged plant ecosystem per unit time, that is, the net carbon sink C_net=C_t-C_d.

9. The method for calculating carbon sinks of submerged plants based on space remote sensing and morphological modeling according to claim 8, characterized in that: Also includes: S6. Automatically collect and calculate carbon flux data on an hourly scale to generate the Q_obs time series. Compare the calculated net carbon sink C_net with the model-estimated carbon sink Q_model. If the deviation between the two exceeds ±15% in the same time window, the feedback correction mechanism is triggered.

Images