Method and system for evaluating ecological restoration effect of seaweed field based on multi-source data fusion

Abstract

The present application relates to a seaweed field ecological restoration effect evaluation method and system based on multi-source data fusion, belonging to the technical field of seaweed field ecological restoration effect analysis, the present application can improve the overall precision of seaweed field vegetation coverage classification in turbid nearshore water body by calculating the carbon sink amount of the biological inversion model and the contrast between the restoration area and the reference area, and can effectively solve the failure problem of traditional underwater SfM method in turbid water body. Secondly, by constructing a Lagrangian particle tracking model of seaweed detritus, the trajectory and settlement distribution of detritus transported from the restoration area to the outside are simulated, the contribution of the detritus transport-deposition process to the carbon sink is quantified, and the carbon sink benefit of the seaweed field restoration project is provided with reliable data into the carbon trading system, thereby improving the carbon sink accounting precision. Finally, the feature importance ranking of the evaluation model is consistent with the ecological mechanism, the comprehensive recovery rate calculation has high repeatability, and the adoption rate of the optimization suggestion of the restoration measure is improved.

Description

Technical Field

[0001] This invention relates to the field of ecological restoration effect analysis technology, and in particular to a method and system for evaluating the ecological restoration effect of seaweed farms based on multi-source data fusion. Background Technology

[0002] Seaweed farms, as typical coastal marine ecosystems, possess core ecological functions such as ecological environment improvement, biodiversity maintenance, and blue carbon sink expansion. However, affected by factors such as coastal urbanization, eutrophication, seawater warming, and biological disturbance, seaweed farm ecosystems worldwide are experiencing large-scale degradation. In recent years, seaweed farm ecological restoration projects have been carried out both domestically and internationally, but the assessment of restoration effectiveness still mainly relies on artificial quadrat surveys and laboratory analysis, which suffers from prominent problems such as low monitoring efficiency, insufficient spatiotemporal coverage, single assessment dimensions, and poor accuracy in quantifying carbon sink contributions. Summary of the Invention

[0003] This invention overcomes the shortcomings of existing technologies and provides a method and system for evaluating the ecological restoration effect of seaweed farms based on multi-source data fusion.

[0004] To achieve the above objectives, the technical solution adopted by the present invention is as follows: The first aspect of this invention provides a method for evaluating the ecological restoration effect of seaweed farms based on multi-source data fusion, comprising the following steps: Multispectral data of the target area is collected and processed to calculate the red-green normalized vegetation index. K-means clustering and spatial consistency optimization are used to eliminate salt-and-pepper noise. A biomass inversion model was constructed, and the accuracy of the biomass inversion model was verified and spatiotemporal difference processing was performed to generate inversion data. Satellite remote sensing data and ground quadrat data were collected. The inversion data, satellite remote sensing data and ground quadrat data were fused using a Bayesian hierarchical model. The carbon sink volume of the restoration area and the reference area were compared to calculate the carbon sink enhancement effect. In conjunction with the enhancement effect, XGBoost gradient boosting tree is used as the basic evaluation model, with the overall recovery rate as the target variable. SHAP value analysis is introduced during the training process to quantify the contribution of each feature to the evaluation result and calculate the overall recovery rate. Early warning prompts are made based on the overall recovery rate.

[0005] Furthermore, in the evaluation method for the ecological restoration effect of seaweed farms based on multi-source data fusion, multispectral data of the target area is collected and processed to calculate the red-green normalized vegetation index. K-means clustering and spatial consistency optimization are used to eliminate salt-and-pepper noise. Specifically: A drone equipped with a five-band multispectral sensor was used to conduct aerial photography of the restoration area during low tide, collecting multispectral data and utilizing the water absorption characteristics in the near-infrared band. For each pixel, the deviation between the NIR band reflectance and the local mean is calculated. When the deviation exceeds the preset deviation threshold, the pixel is identified as a flare-contaminated pixel and corrected using neighborhood interpolation. The red-edge band reflectance and green band reflectance are obtained from the multispectral data, and the red-edge-green normalized vegetation index is calculated based on the red-edge band reflectance and green band reflectance. Using the red-green normalized vegetation index as the main feature, and combining the reflectance of the blue, green and red bands, a four-dimensional feature space is constructed. K-means clustering algorithm is used for unsupervised classification. For each pixel, the distribution of cluster labels in the 8-neighborhood is statistically analyzed. If the label of a pixel is inconsistent with the majority label of the neighborhood, the label of the pixel is corrected to the majority label of the neighborhood.

[0006] Furthermore, in the evaluation method for the ecological restoration effect of seaweed farms based on multi-source data fusion, a biomass inversion model is constructed, and the accuracy of the biomass inversion model is verified and spatiotemporal difference processing is performed to generate inversion data, specifically: Ground verification plots were set up within the remediation area to collect information on the wet weight and species of seaweed within the plots, and a biomass inversion model was constructed. The biomass inversion model satisfies the following relationship: , Where B is the biomass per unit area, and a, b, and c are the model parameters fitted using the least squares method. The red-edge-green normalized vegetation index; The model accuracy was evaluated using leave-one-out cross-validation. For uncovered areas, the spatial distribution map of biomass across the entire region was generated using permanent quadrat data deployed in the remediation area and data from adjacent cruises, and the biomass per unit area was calculated using the biomass inversion model to generate inversion data.

[0007] Furthermore, in the evaluation method for the ecological restoration effect of seaweed farms based on multi-source data fusion, satellite remote sensing data and ground quadrat data are collected. A Bayesian hierarchical model is used to fuse the inverted data, satellite remote sensing data, and ground quadrat data. The carbon sequestration amounts in the restored area and the reference area are compared to calculate the carbon sequestration enhancement effect. Specifically: Satellite remote sensing data and ground quadrat data were collected, and data from three scales were fused using a Bayesian hierarchical model. Carbon content was measured for different algal species, and a multiple regression model was established between carbon content and algal species, growth stage, and environmental conditions. The carbon content was calculated using a multiple regression model of the carbon content with algal species, growth stage, and environmental conditions. High-precision seabed topographic data of the restoration area was obtained using multibeam sonar. Combined with measured tidal current and wave data of the restoration area, a two-dimensional hydrodynamic model was constructed to simulate the flow field and material transport process of the restoration area. A Lagrange particle tracking model for seaweed debris was constructed to simulate the trajectory and sedimentation distribution of debris transported from the remediation area to the outside, and to calculate the decomposition rate and total carbon sink of the debris. By comparing the carbon sequestration levels in the remediation area and the reference area, and combining this with the decomposition rate of debris, the carbon sequestration enhancement effect is calculated. This effect is then output quarterly and annually and incorporated into the comprehensive evaluation index system for remediation effectiveness.

[0008] Furthermore, in the evaluation method for the ecological restoration effect of seaweed farms based on multi-source data fusion, combined with the sinking effect, XGBoost gradient boosting tree is used as the basic evaluation model, with the overall recovery rate as the target variable. SHAP value analysis is introduced during the training process to quantify the contribution of each feature to the evaluation results and calculate the overall recovery rate, specifically including: Biodiversity index and benthic community structure data were obtained through manual surveys. Algal coverage and recovery rate were statistically analyzed. The algal coverage and recovery rate, biomass spatial distribution, carbon sink and carbon sequestration effect, human biodiversity index and benthic community structure data were uniformly coded into an evaluation feature vector. XGBoost gradient boosting tree is used as the basic evaluation model, with the overall recovery rate as the target variable. SHAP value analysis is introduced during training to quantify the contribution of each feature to the evaluation results. By analyzing SHAP values, the global importance ranking of each evaluation indicator and the local interpretation of the evaluation results of a single repair area are output. The data-driven weights of each indicator are determined based on SHAP values, and the overall recovery rate is calculated. When the overall recovery rate is lower than expected, the main limiting factors are identified by the feature with the largest negative contribution in the SHAP value, and targeted adjustment suggestions are output.

[0009] Furthermore, in the evaluation method for the ecological restoration effect of seaweed farms based on multi-source data fusion, an early warning is issued based on the comprehensive restoration rate, specifically as follows: Set a comprehensive recovery rate threshold, and determine whether the comprehensive recovery rate is greater than the comprehensive recovery rate threshold; When the overall recovery rate is greater than the overall recovery rate threshold, it is marked as a normal seaweed farm ecological restoration area; when the overall recovery rate is not greater than the overall recovery rate threshold, it is marked as an abnormal seaweed farm ecological restoration area. The normal seaweed farm ecological restoration area and the abnormal seaweed farm ecological restoration area are displayed in a preset manner, and relevant planning suggestions are generated based on the abnormal seaweed farm ecological restoration area.

[0010] A second aspect of the present invention provides a system for evaluating the ecological restoration effect of seaweed farms based on multi-source data fusion, comprising a memory and a processor. The memory includes a program for evaluating the ecological restoration effect of seaweed farms based on multi-source data fusion. When the program for evaluating the ecological restoration effect of seaweed farms based on multi-source data fusion is executed by the processor, it implements the steps of the method for evaluating the ecological restoration effect of seaweed farms based on multi-source data fusion as described in any one of the present invention.

[0011] This invention addresses the shortcomings of the prior art and has the following beneficial effects: This invention collects and processes multispectral data from a target area to calculate the red-green normalized vegetation index (RDI). K-means clustering and spatial consistency optimization are used to eliminate salt-and-pepper noise, thereby constructing a biomass inversion model. The accuracy of the biomass inversion model is verified, and spatiotemporal difference processing is performed to generate inversion data. Satellite remote sensing data and ground quadrat data are collected, and a Bayesian hierarchical model is used to fuse the inversion data, satellite remote sensing data, and ground quadrat data. The carbon sink volume of the restoration area and the reference area are compared to calculate the carbon sink enhancement effect. Finally, combining the carbon sink enhancement effect, an XGBoost gradient boosting tree is used as the basic evaluation model, with the overall recovery rate as the target variable. SHAP value analysis is introduced during training to quantify the contribution of each feature to the evaluation results, and the overall recovery rate is calculated. Early warning prompts are provided based on the overall recovery rate. This invention, through its biomass inversion model, can improve the overall accuracy of seaweed cover classification in turbid nearshore waters and effectively solves the failure problem of traditional underwater SfM methods in turbid waters. Secondly, by constructing a Lagrange particle tracking model for seaweed debris, the trajectory and sedimentation distribution of debris transported from the remediation area can be simulated. This allows for the quantification of the contribution of debris transport-deposition processes to carbon sequestration, providing reliable data for incorporating the carbon sequestration benefits of seaweed farm remediation projects into the carbon trading system, thereby improving the accuracy of carbon sequestration accounting. Finally, the feature importance ranking of the evaluation model is consistent with the ecological mechanism, and the calculation of the overall restoration rate has high repeatability, increasing the adoption rate of remediation optimization recommendations. Attached Figure Description

[0012] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other embodiments can be obtained from these drawings without creative effort.

[0013] Figure 1 A flowchart of the overall process for evaluating the ecological restoration effect of seaweed farms based on multi-source data fusion is shown. Figure 2 A system block diagram of a seaweed farm ecological restoration effect evaluation system based on multi-source data fusion is shown. Detailed Implementation

[0014] To better understand the above-mentioned objectives, features, and advantages of the present invention, the present invention will be further described in detail below with reference to the accompanying drawings and specific embodiments. It should be noted that, unless otherwise specified, the embodiments and features described in these embodiments can be combined with each other.

[0015] Many specific details are set forth in the following description in order to provide a full understanding of the invention. However, the invention may also be practiced in other ways different from those described herein, and therefore the scope of protection of the invention is not limited to the specific embodiments disclosed below.

[0016] like Figure 1 As shown, the first aspect of this invention provides a method for evaluating the ecological restoration effect of seaweed farms based on multi-source data fusion, comprising the following steps: Multispectral data of the target area is collected and processed to calculate the red-green normalized vegetation index. K-means clustering and spatial consistency optimization are used to eliminate salt-and-pepper noise. A biomass inversion model was constructed, and the accuracy of the biomass inversion model was verified and spatiotemporal difference processing was performed to generate inversion data. Satellite remote sensing data and ground quadrat data were collected. The inversion data, satellite remote sensing data and ground quadrat data were fused using a Bayesian hierarchical model. The carbon sequestration of the restoration area and the reference area were compared to calculate the carbon sequestration enhancement effect. In conjunction with the sinking effect, XGBoost gradient boosting tree is used as the basic evaluation model, with the overall recovery rate as the target variable. SHAP value analysis is introduced during the training process to quantify the contribution of each feature to the evaluation results and calculate the overall recovery rate. Early warning prompts are given based on the overall recovery rate.

[0017] It should be noted that this invention, through a biological inversion model, can improve the overall accuracy of seaweed cover classification in turbid nearshore waters and effectively solve the failure problem of traditional underwater SfM methods in turbid waters. Secondly, by constructing a Lagrange particle tracking model for seaweed debris, simulating the trajectory and sedimentation distribution of debris transport from the remediation area, the contribution of debris transport-deposition processes to carbon sinks can be quantified. This provides reliable data for incorporating the carbon sink benefits of seaweed farm remediation projects into the carbon trading system, thereby improving the accuracy of carbon sink accounting. Finally, the feature importance ranking of the evaluation model is consistent with the ecological mechanism, and the calculation of the overall restoration rate has high repeatability, improving the adoption rate of remediation optimization recommendations.

[0018] Furthermore, in the evaluation method for the ecological restoration effect of seaweed farms based on multi-source data fusion, multispectral data of the target area is collected and processed to calculate the red-green normalized vegetation index. K-means clustering and spatial consistency optimization are used to eliminate salt-and-pepper noise. Specifically: A drone equipped with a five-band multispectral sensor was used to conduct aerial photography of the restoration area during low tide, collecting multispectral data and utilizing the water absorption characteristics in the near-infrared band. For each pixel, the deviation between the NIR band reflectance and the local mean is calculated. When the deviation exceeds the preset deviation threshold, the pixel is identified as a flare-contaminated pixel and corrected using neighborhood interpolation. The red-edge band reflectance and green band reflectance are obtained from multispectral data, and the red-edge-green normalized vegetation index is calculated based on the red-edge band reflectance and green band reflectance. Using the red-green normalized vegetation index as the main feature, and combining the reflectance of the blue, green and red bands, a four-dimensional feature space is constructed. K-means clustering algorithm is used for unsupervised classification. For each pixel, the distribution of cluster labels in the 8-neighborhood is statistically analyzed. If the label of a pixel is inconsistent with the majority label of the neighborhood, the label of the pixel is corrected to the majority label of the neighborhood.

[0019] It should be noted that a drone equipped with a five-band multispectral sensor (blue 450nm, green 560nm, red 650nm, red-edge 705nm, near-infrared 840nm) was used to conduct aerial photography of the remediation area during low tide. The flight altitude was set at 80-120m, ground resolution ≤3cm, forward overlap ≥80%, and lateral overlap ≥70%. High-resolution orthophotos (DOM) and digital surface models (DSM) were acquired simultaneously. Utilizing the water absorption characteristics in the near-infrared band (NIR, 840nm), the deviation between the NIR band reflectance and the local mean was calculated for each pixel. When the deviation exceeded a threshold, the pixel was identified as a flare-contaminated pixel and corrected using neighborhood interpolation. Red-edge-green normalized vegetation index (RE-GNDVI) was calculated. , in, For red-edge band reflectivity, This refers to the green band reflectance. This index utilizes the high sensitivity of the red-edge band to chlorophyll content to effectively enhance the spectral differences between submerged algae and water bodies, as well as sandy seabeds.

[0020] Using RE-GNDVI value as the primary feature and combining reflectance in the blue, green, and red bands, a four-dimensional feature space was constructed. Unsupervised classification was performed using the K-means clustering algorithm (K=4, corresponding to dense algae areas, sparse areas, sandy seabed, and water bodies, respectively). Furthermore, a spatial consistency constraint was introduced: for each pixel, the distribution of cluster labels within its 8-neighborhood was statistically analyzed. If a pixel's label differed from the majority label in its neighborhood, the pixel's label was corrected to match the majority label in its neighborhood to eliminate "salt-and-pepper noise."

[0021] Furthermore, in the evaluation method for the ecological restoration effect of seaweed farms based on multi-source data fusion, a biomass inversion model is constructed, and the accuracy of the biomass inversion model is verified and spatiotemporal difference processing is performed to generate inversion data, specifically: Ground validation plots were set up within the remediation area to collect information on the wet weight and species of seaweed within the plots. A biomass inversion model was then constructed, which satisfies the following relationship: , Where B is the biomass per unit area, and a, b, and c are the model parameters fitted using the least squares method. The red-edge-green normalized vegetation index; The model accuracy was evaluated using leave-one-out cross-validation. For uncovered areas, the spatial distribution map of biomass across the entire region was generated using permanent quadrat data deployed in the remediation area and data from adjacent cruises, and the biomass per unit area was calculated using the biomass inversion model to generate inversion data.

[0022] It should be noted that, based on the calculation results of the above formulas, leave-one-out cross-validation is used to evaluate the model accuracy, requiring R² ≥ 0.85. For uncovered areas, Kriging interpolation is used to generate a spatial distribution map of biomass across the entire region, utilizing permanent quadrat data deployed within the remediation area and data from adjacent cruises. This method achieves an overall accuracy of ≥ 94% for seaweed cover classification in turbid nearshore waters, and the biomass inversion accuracy is significantly improved compared to traditional quadrat methods, effectively solving the failure problem of traditional underwater SfM methods in turbid waters.

[0023] Furthermore, in the evaluation method for the ecological restoration effect of seaweed farms based on multi-source data fusion, satellite remote sensing data and ground quadrat data are collected. A Bayesian hierarchical model is used to fuse the inverted data, satellite remote sensing data, and ground quadrat data. The carbon sequestration amounts in the restored area and the reference area are compared to calculate the carbon sequestration enhancement effect. Specifically: Satellite remote sensing data and ground quadrat data were collected, and data from three scales were fused using a Bayesian hierarchical model. Carbon content was measured for different algal species, and a multiple regression model was established between carbon content and algal species, growth stage, and environmental conditions. The carbon content was calculated using a multiple regression model of carbon content with algal species, growth stage, and environmental conditions. High-precision seabed topographic data of the restoration area was obtained using multibeam sonar. Combined with measured tidal current and wave data of the restoration area, a two-dimensional hydrodynamic model was constructed to simulate the flow field and material transport process in the restoration area. A Lagrange particle tracking model for seaweed debris was constructed to simulate the trajectory and sedimentation distribution of debris transported from the remediation area to the outside, and to calculate the decomposition rate and total carbon sink of the debris. By comparing the carbon sequestration levels in the remediation area and the reference area, and combining this with the decomposition rate of debris, the carbon sequestration enhancement effect is calculated. This effect is then output quarterly and annually and incorporated into the comprehensive evaluation index system for remediation effectiveness.

[0024] It should be noted that biomass data at three scales were fused: (a) satellite remote sensing data (Sentinel-2 MSI, 10m resolution) to obtain the large-scale distribution area of ​​algae in the restoration area and surrounding sea; (b) inversion data for fine-scale biomass inversion in the restoration area; and (c) ground quadrat data for model calibration and validation. A Bayesian hierarchical model was used to fuse the data at the three scales. , Among them, weight , , The spatial resolution of each data source and the accuracy of historical verification are adaptively determined. For position Estimated seaweed biomass after fusion; It is satellite remote sensing in location Biomass obtained from inversion; It is a drone multispectral location Biomass obtained from inversion; It is a location The biomass at the location is determined by measured or spatially interpolated data from ground quadrats. The weights satisfy the normalization constraint. , Carbon content was measured for different algal species (using the CHNS mode of an elemental analyzer), and a multiple regression model was established between carbon content and algal species, growth stage, and environmental conditions. , in, It refers to the carbon content of seaweed. It is a multiple regression mapping function. It is a type of seaweed. It is the growing season; T represents the water temperature. It refers to the concentration of nutrients.

[0025] Based on the hydrodynamic sub-model, a Lagrange particle tracking model for algal debris was constructed to simulate the trajectory and sedimentation distribution of debris transported from the remediation zone outwards. During transport, the debris undergoes decomposition, and the decomposition rate follows first-order kinetics: , Where m is the mass of the debris. Let θ be the baseline decomposition rate constant, θ be the temperature coefficient, and T be the water temperature. A certain proportion of the detrital carbon deposited on the seabed (burial efficiency ε) is permanently stored. Total carbon sequestration. for: , Where η is the proportion of removable carbon sinks (carbon removed through harvesting), and the integral term is the contribution of detrital sedimentary carbon sinks. It represents the dual integral of time and space.

[0026] By comparing the carbon sequestration levels in the remediation zone and the reference zone, the carbon sequestration enhancement effect can be calculated. , in, It has the effect of increasing foreign exchange reserves. It is the total carbon sequestration in the remediation area. This refers to the total carbon sequestration in the reference area.

[0027] The effects of foreign exchange reserves will be output on a quarterly and annual basis and incorporated into the comprehensive evaluation index system for the repair effect.

[0028] Furthermore, in the evaluation method for the ecological restoration effect of seaweed farms based on multi-source data fusion, combined with the sinking effect, XGBoost gradient boosting tree is used as the basic evaluation model, with the overall recovery rate as the target variable. SHAP value analysis is introduced during the training process to quantify the contribution of each feature to the evaluation results and calculate the overall recovery rate, specifically including: Biodiversity index and benthic community structure data were obtained through manual surveys. Algal coverage and recovery rate were statistically analyzed. Based on algal coverage and recovery rate, biomass spatial distribution, carbon sink and carbon sequestration effect, human biodiversity index and benthic community structure data were uniformly coded into evaluation feature vectors to form a feature set. XGBoost gradient boosting tree is used as the basic evaluation model, with the overall recovery rate as the target variable. SHAP value analysis is introduced during training to quantify the contribution of each feature to the evaluation results. By analyzing SHAP values, the global importance ranking of each evaluation indicator and the local interpretation of the evaluation results of a single repair area are output. The data-driven weights of each indicator are determined based on SHAP values, and the overall recovery rate is calculated. The data-driven weights of each indicator are determined based on the SHAP value, and the overall recovery rate is calculated. , in, Obtained by normalizing the SHAP value. Let be the single-index recovery rate of the j-th indicator. Referring to the grading criteria of GB / T45031-2024, the repair effect is divided into four levels: excellent (CRR≥85%), good (70%≤CRR<85%), qualified (50%≤CRR<70%), and unqualified (CRR<50%).

[0029] When the overall recovery rate is lower than expected, the main limiting factors are identified by the feature with the largest negative contribution in the SHAP value, and targeted adjustment suggestions are output.

[0030] It should be noted that XGBoost gradient boosting trees are used as the basic evaluation model, with the overall recovery rate as the objective variable. During training, SHAP (SHapley Additive exPlanations) value analysis is introduced to quantify the contribution of each feature to the evaluation results. , in, Let be the SHAP value of the j-th feature, F be the total feature set, and S be a subset of features. Through SHAP value analysis, the system outputs: (a) the global importance ranking of each evaluation indicator; and (b) a local interpretation of the evaluation results for a single recovery zone (which indicators promoted or inhibited the overall score). When the overall recovery rate is lower than expected, the system automatically identifies the main limiting factors (through the feature with the largest negative contribution in the SHAP value) and outputs targeted adjustment suggestions.

[0031] It should be noted that the feature importance ranking of the assessment model is consistent with the ecological mechanism, the calculation of the overall restoration rate has high repeatability, and the adoption rate of optimization suggestions for remediation measures is improved.

[0032] Furthermore, in the evaluation method for the ecological restoration effect of seaweed farms based on multi-source data fusion, early warning is issued based on the comprehensive restoration rate, specifically as follows: Set a comprehensive recovery rate threshold and determine whether the comprehensive recovery rate is greater than the comprehensive recovery rate threshold; When the overall recovery rate is greater than the overall recovery rate threshold, it is marked as a normal seaweed farm ecological restoration area; when the overall recovery rate is not greater than the overall recovery rate threshold, it is marked as an abnormal seaweed farm ecological restoration area. The ecological restoration areas of normal seaweed beds and abnormal seaweed beds are displayed in a preset manner, and relevant planning suggestions are generated based on the ecological restoration areas of abnormal seaweed beds.

[0033] It should be noted that the relevant planning recommendations include: adjusting algae species ratios, optimizing deployment density, substrate improvement schemes, and intervention measures for environmental stress factors. The adjusted schemes are fed back to the digital twin for effect simulation, forming a closed loop of "assessment—diagnosis—optimization—simulation—reassessment".

[0034] In addition, this method also includes: Within the remediation area, at least 15 sediment sampling points were arranged in a grid pattern, and surface sediment samples (0-10 cm) were obtained using a sediment core collector. Sediment samples were also collected from a reference point outside the remediation area as a control. For the main algal species to be restored in the restoration zone, species-specific TaqMan probes were designed, and absolute quantification was performed using digital PCR. Digital PCR divided the DNA template into tens of thousands of microreaction units, and the original template concentration was calculated by Poisson distribution. A quantitative conversion model between eDNA copy number and seaweed-derived organic carbon content was established. Lead-210 radionuclide dating was performed on sediment cores to determine the sediment accumulation rate. Combined with the seaweed-derived carbon content quantified by eDNA, the burial flux of seaweed-derived carbon was calculated. The total carbon sink amount is corrected based on the burial flux of the seaweed-derived carbon, and the carbon sink enhancement effect is updated.

[0035] It should be noted that, since the calculation of carbon sinks in clastic sediments mainly relies on hydrodynamic models and decomposition kinetic parameter estimations, and lacks direct empirical detection methods for algal-derived carbon in sediments, the "verifiability" of carbon sink calculations is insufficient. Therefore, a quantitative conversion model between eDNA copy number and algal-derived organic carbon content is established: , in, The organic carbon content of algae at depth z (g C / g dry sediment). Let be the eDNA copy number at depth z (copies / g dry sediment), α be the conversion factor between eDNA copy number and biomass, β(z) be the depth attenuation correction factor (considering DNA degradation in sediment), and γ be the background correction term. Lead-210 radionuclide dating was performed on the sediment core to determine the sediment accumulation rate (cm / yr). Combined with the seaweed-derived carbon content quantified by eDNA, the burial flux of seaweed-derived carbon was calculated. , in, Carbon burial flux (g C / m² / yr). The dry bulk density of the sediment is expressed in g / cm³. The sediment accumulation rate is expressed as cm / yr.

[0036] It should be noted that this invention enables species-level tracing and absolute quantification of seaweed-derived carbon in sediments, upgrading carbon sink accounting from "model estimation" to "empirical detection," thereby reducing the uncertainty of carbon sink estimation.

[0037] In addition, this method also includes: The restoration area and its surrounding sea area are divided into regular grids, with each grid cell serving as a node in the graph. The edge connection rules between nodes include: establishing edges between adjacent grids; and establishing weighted edges for grids with significant material exchange based on the material transport probability of the hydrodynamic model, thereby constructing an adjacency matrix. A graph attention network is used to learn the spatial correlation strength between nodes. Each node is updated to generate a GAT model, which is then embedded into a digital twin. The digital twin provides hydrodynamic field and material transport information as a graph structure prior for GAT, and the prediction results of GAT are fed back to the digital twin as spatial constraints on the ecological state. Based on the spatial correlation weights learned from GAT, a risk propagation network is constructed. When the assessment indicators of a certain grid cell show a significant decline, the risk propagation network is used to predict the path and speed of risk diffusion to surrounding grids. Based on the spatial correlation analysis results of GAT, the spatial layout of remediation measures is optimized. Assessment indicators include biomass and health status. It should be noted that a "risk propagation network" is constructed based on the spatial correlation weights learned by GAT. When an assessment metric (such as biomass or health status) of a grid cell shows a significant decline, the risk propagation network is used to predict the path and speed of risk diffusion to surrounding grid cells. , in, For node j in time next The risk index of time, Let be the risk index of node j at time t. This is the risk propagation rate coefficient. Let be the attention weight of node j to node i, adaptively calculated from the node features. This is a set of nodes. When it is predicted that a risk will spread to a critical ecological area (such as a core restoration area), an early warning is triggered, and isolation or intervention measures are recommended. Based on the spatial correlation analysis results of GAT, the spatial layout of restoration measures is optimized. For example, if GAT identifies a grid cell as a "spatial hub" (i.e., having a high influence on multiple surrounding grids), restoration measures are prioritized and strengthened in that area.

[0038] It should be noted that, in specific implementation, the UAV's five-band sensor can use blue 450nm, green 560nm, red 650nm, red-edge 705nm, and near-infrared 840nm; the flight altitude can be set to 80-120m, ground resolution ≤3cm, forward overlap ≥80%, and lateral overlap ≥70%. In K-means clustering, K can be 4, or adjusted according to the actual seaweed distribution type. The biomass inversion model parameters are fitted using the least squares method, and verification requirements are... Satellite remote sensing data can be Sentinel-2MSI or other multispectral satellite data with red-edge bands. Bayesian fusion weights can be adaptively determined by spatial resolution and historical verification accuracy. Parameters such as decomposition rate, temperature coefficient, and burial efficiency in carbon sink calculations can be calibrated according to different sea areas and algal species. XGBoost can be replaced with other gradient boosting tree models, and SHAP values ​​can be used for feature contribution interpretation. The number of digital PCR microreaction units in eDNA detection can be adjusted according to the instrument and sample. GAT can be replaced with other graph neural network models, and the digital twin can integrate hydrodynamic, ecological, and remediation simulation modules.

[0039] like Figure 2 As shown, the second aspect of the present invention provides a system for evaluating the ecological restoration effect of seaweed farms based on multi-source data fusion, including a memory and a processor. The memory includes a program for evaluating the ecological restoration effect of seaweed farms based on multi-source data fusion. When the program for evaluating the ecological restoration effect of seaweed farms based on multi-source data fusion is executed by the processor, it implements any of the steps of the method for evaluating the ecological restoration effect of seaweed farms based on multi-source data fusion.

[0040] In the several embodiments provided in this application, it should be understood that the disclosed devices and methods can be implemented in other ways. The device embodiments described above are merely illustrative. For example, the division of units is only a logical functional division, and in actual implementation, there may be other division methods, such as: multiple units or components can be combined, or integrated into another system, or some features can be ignored or not executed. In addition, the coupling, direct coupling, or communication connection between the various components shown or discussed can be through some interfaces, and the indirect coupling or communication connection between devices or units can be electrical, mechanical, or other forms.

[0041] The units described above as separate components may or may not be physically separate. The components shown as units may or may not be physical units. They may be located in one place or distributed across multiple network units. Some or all of the units may be selected to achieve the purpose of this embodiment according to actual needs.

[0042] In addition, in the various embodiments of the present invention, each functional unit can be integrated into one processing unit, or each unit can be a separate unit, or two or more units can be integrated into one unit; the integrated unit can be implemented in hardware or in the form of hardware plus software functional units.

[0043] Those skilled in the art will understand that all or part of the steps of the above method embodiments can be implemented by hardware related to program instructions. The aforementioned program can be stored in a computer-readable storage medium. When the program is executed, it performs the steps of the above method embodiments. The aforementioned storage medium includes various media capable of storing program code, such as mobile storage devices, read-only memory (ROM), random access memory (RAM), magnetic disks, or optical disks.

[0044] Alternatively, if the integrated units of this invention are implemented as software functional modules and sold or used as independent products, they can also be stored in a computer-readable storage medium. Based on this understanding, the technical solutions of the embodiments of this invention, or the parts that contribute to the prior art, can be embodied in the form of a software product. This computer software product is stored in a storage medium and includes several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute all or part of the methods of the various embodiments of this invention. The aforementioned storage medium includes various media capable of storing program code, such as mobile storage devices, ROM, RAM, magnetic disks, or optical disks.

[0045] The above are merely specific embodiments of the present invention, but the scope of protection of the present invention is not limited thereto. Any variations or substitutions that can be easily conceived by those skilled in the art within the technical scope disclosed in the present invention should be included within the scope of protection of the present invention. Therefore, the scope of protection of the present invention should be determined by the scope of the claims.

Claims

1. A method for evaluating the ecological restoration effect of seaweed farms based on multi-source data fusion, characterized in that, Includes the following steps: Multispectral data of the target area is collected and processed to calculate the red-green normalized vegetation index. K-means clustering and spatial consistency optimization are used to eliminate salt-and-pepper noise. A biomass inversion model was constructed, and the accuracy of the biomass inversion model was verified and spatiotemporal difference processing was performed to generate inversion data. Satellite remote sensing data and ground quadrat data were collected. The inversion data, satellite remote sensing data and ground quadrat data were fused using a Bayesian hierarchical model. The carbon sequestration of the restoration area and the reference area were compared to calculate the carbon sequestration enhancement effect. In conjunction with the sinking effect, XGBoost gradient boosting tree is used as the basic evaluation model, with the overall recovery rate as the target variable. SHAP value analysis is introduced during the training process to quantify the contribution of each feature to the evaluation result and calculate the overall recovery rate. Early warning prompts are made based on the overall recovery rate.

2. The method for evaluating the ecological restoration effect of seaweed farms based on multi-source data fusion according to claim 1, characterized in that, Multispectral data of the target region is collected and processed to calculate the red-green normalized vegetation index. K-means clustering and spatial consistency optimization are then used to eliminate salt-and-pepper noise. Specifically: A drone equipped with a five-band multispectral sensor was used to conduct aerial photography of the restoration area during low tide, collecting multispectral data and utilizing the water absorption characteristics in the near-infrared band. For each pixel, the deviation between the NIR band reflectance and the local mean is calculated. When the deviation exceeds the preset deviation threshold, the pixel is identified as a flare-contaminated pixel and corrected using neighborhood interpolation. The red-edge band reflectance and green band reflectance are obtained from the multispectral data, and the red-edge-green normalized vegetation index is calculated based on the red-edge band reflectance and green band reflectance. Using the red-green normalized vegetation index as the main feature, and combining the reflectance of the blue, green and red bands, a four-dimensional feature space is constructed. K-means clustering algorithm is used for unsupervised classification. For each pixel, the distribution of cluster labels in the 8-neighborhood is statistically analyzed. If the label of a pixel is inconsistent with the majority label of the neighborhood, the label of the pixel is corrected to the majority label of the neighborhood.

3. The method for evaluating the ecological restoration effect of seaweed farms based on multi-source data fusion according to claim 1, characterized in that, A biomass inversion model was constructed, and the accuracy of the model was verified and spatiotemporal difference processing was performed to generate inversion data, specifically as follows: Ground verification plots were set up within the remediation area to collect information on the wet weight and species of seaweed within the plots, and a biomass inversion model was constructed. The biomass inversion model satisfies the following relationship: ， Where B is the biomass per unit area, and a, b, and c are the model parameters fitted using the least squares method. The red-edge-green normalized vegetation index; The model accuracy was evaluated using leave-one-out cross-validation. For uncovered areas, the spatial distribution map of biomass across the entire region was generated using permanent quadrat data deployed in the remediation area and data from adjacent cruises, and the biomass per unit area was calculated using the biomass inversion model to generate inversion data.

4. The method for evaluating the ecological restoration effect of seaweed farms based on multi-source data fusion according to claim 1, characterized in that, Satellite remote sensing data and ground quadrat data were collected. A Bayesian hierarchical model was used to fuse the inverted data, satellite remote sensing data, and ground quadrat data. The carbon sequestration amounts in the remediation area and the reference area were compared to calculate the carbon sequestration enhancement effect. Specifically: Satellite remote sensing data and ground quadrat data were collected, and data from three scales were fused using a Bayesian hierarchical model. Carbon content was measured for different algal species, and a multiple regression model was established between carbon content and algal species, growth stage, and environmental conditions. The carbon content was calculated using a multiple regression model of the carbon content with algal species, growth stage, and environmental conditions. High-precision seabed topographic data of the restoration area was obtained using multibeam sonar. Combined with measured tidal current and wave data of the restoration area, a two-dimensional hydrodynamic model was constructed to simulate the flow field and material transport process of the restoration area. A Lagrange particle tracking model for seaweed debris was constructed to simulate the trajectory and sedimentation distribution of debris transported from the remediation area to the outside, and to calculate the decomposition rate and total carbon sink of the debris. By comparing the carbon sequestration levels in the remediation area and the reference area, and combining this with the decomposition rate of debris, the carbon sequestration enhancement effect is calculated. This effect is then output quarterly and annually and incorporated into the comprehensive evaluation index system for remediation effectiveness.

5. The method for evaluating the ecological restoration effect of seaweed farms based on multi-source data fusion according to claim 3, characterized in that, To combine the effects of sink enhancement, XGBoost gradient boosting tree is used as the basic evaluation model, with the overall recovery rate as the target variable. SHAP value analysis is introduced during training to quantify the contribution of each feature to the evaluation results and calculate the overall recovery rate, specifically including: Biodiversity index and benthic community structure data were obtained through manual surveys. Algal coverage and recovery rate were statistically analyzed. The algal coverage and recovery rate, biomass spatial distribution, carbon sink and carbon sequestration effect, human biodiversity index and benthic community structure data were uniformly coded into an evaluation feature vector. XGBoost gradient boosting tree is used as the basic evaluation model, with the overall recovery rate as the target variable. SHAP value analysis is introduced during training to quantify the contribution of each feature to the evaluation results. By analyzing SHAP values, the global importance ranking of each evaluation indicator and the local interpretation of the evaluation results of a single repair area are output. The data-driven weights of each indicator are determined based on SHAP values, and the overall recovery rate is calculated. When the overall recovery rate is lower than expected, the main limiting factors are identified by the feature with the largest negative contribution in the SHAP value, and targeted adjustment suggestions are output.

6. The method for evaluating the ecological restoration effect of seaweed farms based on multi-source data fusion according to claim 1, characterized in that, The warning will be issued based on the overall recovery rate, specifically as follows: Set a comprehensive recovery rate threshold, and determine whether the comprehensive recovery rate is greater than the comprehensive recovery rate threshold; When the overall recovery rate is greater than the overall recovery rate threshold, it is marked as a normal seaweed farm ecological restoration area; when the overall recovery rate is not greater than the overall recovery rate threshold, it is marked as an abnormal seaweed farm ecological restoration area. The normal seaweed farm ecological restoration area and the abnormal seaweed farm ecological restoration area are displayed in a preset manner, and relevant planning suggestions are generated based on the abnormal seaweed farm ecological restoration area.

7. A system for evaluating the ecological restoration effect of seaweed farms based on multi-source data fusion, characterized in that, The device includes a memory and a processor. The memory includes a program for evaluating the ecological restoration effect of seaweed farms based on multi-source data fusion. When the processor executes the program for evaluating the ecological restoration effect of seaweed farms based on multi-source data fusion, it implements the steps of the method for evaluating the ecological restoration effect of seaweed farms based on multi-source data fusion as described in any one of claims 1-6.

Images