Sea island ecological restoration effect comprehensive evaluation method

Abstract

The invention discloses an island ecological restoration effect comprehensive evaluation method, and particularly relates to the technical field of ecological environment evaluation. Obtaining a restoration area ecological basic parameter set, and constructing an original ecological characteristic matrix; calculating a restoration response curve based on the time sequence change characteristics, and determining the restoration response weight of each ecological factor; calculating a restoration net effect through a normalized difference model of parameters of a contrast area, and eliminating the influence of natural fluctuation; inputting the net effect vector into an ecological system structure coupling model, and quantifying the coupling strength between ecological factors to obtain a restoration comprehensive contribution degree; further performing space-time interpolation analysis to form a restoration effect distribution matrix and calculating a restoration vulnerability index; generating a repair optimization suggestion set according to a threshold value distribution result, and outputting a comprehensive evaluation report; according to the method, dynamic quantification and systematized diagnosis of the ecological restoration process are realized, the spatio-temporal evolution characteristics and restoration effect differences of the island ecosystem can be accurately reflected, and a scientific basis is provided for ecological restoration optimization and intelligent decision making.

Description

Technical Field

[0001] This invention relates to the field of ecological environment assessment technology, specifically to a comprehensive evaluation method for the effectiveness of island ecological restoration. Background Technology

[0002] In recent years, with the continuous strengthening of marine ecological environment protection policies, ecological restoration projects have become an important means to address island ecosystems severely affected by human activities. Island ecological restoration typically involves a multi-stage, coordinated process, including vegetation restoration, shoreline stabilization, intertidal habitat optimization, and seagrass bed reconstruction.

[0003] However, island ecosystems are influenced by multiple coupled effects of tides, hydrodynamics, salinity gradients, and sediment changes, resulting in a highly nonlinear and locally heterogeneous spatiotemporal response to their restoration processes. Currently, commonly used methods for evaluating restoration effectiveness are mostly based on single indicators, such as vegetation cover, biodiversity indices, or water quality parameters. While these methods can reflect local improvements, they are difficult to objectively quantify the overall restoration effect.

[0004] Especially in areas with small coral islands and sandbars, where the terrain is relatively flat and the ecological structure is simple, the effectiveness of restoration is often affected by the cyclical interference of seasonal tidal erosion and sedimentation, frequently resulting in a short-term "false recovery" phenomenon, where ecological indicators rise temporarily but the long-term trend deteriorates. How to construct a comprehensive evaluation system that takes into account multiple factors and can be dynamically updated for these typical small-scale islands, in order to identify weak links in restoration, has become an urgent technical problem to be solved in current ecological restoration projects. Summary of the Invention

[0005] The purpose of this invention is to provide a comprehensive evaluation method for the effectiveness of island ecological restoration, in order to address the shortcomings of the prior art.

[0006] To achieve the above objectives, the present invention provides the following technical solution: a method for comprehensive evaluation of the effectiveness of island ecological restoration, comprising: Obtain a set of basic ecological parameters within the target island restoration area, including vegetation cover (V), benthic organism density (B), water nutrient concentration (N), and sediment stability (P), and construct an original ecological feature matrix. ; Based on the temporal variation characteristics of each parameter, the repair response curve is calculated. And through dynamic weighting functions Obtain the set of repair response weight coefficients for each subsystem It is the output function; Obtain the parameter set of the adjacent unrepaired control area outside the repair area. The net effect vector of repair was obtained through a normalized difference model. ; Will The input is fed into the ecosystem structure coupling model M, and the comprehensive contribution of restoration is obtained based on the interdependence among ecological factors. M is used to quantify the coupling strength between different ecological factors. Spatiotemporal interpolation analysis was performed on C to obtain the distribution matrix S of the island's ecological restoration effectiveness, and the restoration vulnerability index of each spatial unit was calculated. Based on the threshold distribution of F, a set of restoration and optimization suggestions is generated, and a comprehensive evaluation report on the effectiveness of island ecological restoration is output.

[0007] Preferably, obtaining the basic ecological parameter set within the target island restoration area includes: Aerial surveys were conducted in different zones of the target island restoration area to collect surface reflectance data at different times. The vegetation cover V of each zone was calculated using the normalized vegetation index. Real-time data on benthic organism quantity and activity frequency are collected, and benthic organism density B is calculated based on the average density over multiple time periods. Seawater samples were collected at different tidal times, and the contents of total nitrogen, total phosphorus and dissolved oxygen were measured. The weighted summation was then used to calculate the comprehensive nutrient concentration N in the water. By combining multi-layer sediment sampling, the sediment particle size distribution and shear modulus change rate were determined, and the sediment stability P was calculated.

[0008] Preferably, the repair response curve is calculated based on the temporal variation characteristics of each parameter. ,include: Original ecological feature matrix The ecological parameters were interpolated over time to construct a unified time-scale sequence T. Using a combination algorithm of piecewise linear fitting and second-order polynomial curve analysis, the restoration response curves for vegetation cover V, benthic organism density B, water nutrient concentration N, and sediment stability P were calculated. .

[0009] Preferably, the dynamic weighting function is used. Obtain the set of repair response weight coefficients for each subsystem This includes: repairing response curves based on various parameters. Based on the characteristic parameters, construct a dynamic weighting function. The function takes the variance ratio of the characteristic parameters and the time delay factor as inputs, and determines the sensitivity coefficients of each ecological element through a normalized exponential decay model. ,in Characterizes the intensity of the response of the i-th ecological element to external disturbances; the sensitivity coefficient Standard deviation of original parameters In summary, the repair response weighting coefficients are calculated. The entropy weight normalization method is used to scale the weights.

[0010] Preferably, the parameter set of the adjacent unrepaired control area outside the repair area is obtained. ,include: Neighboring unrestored control areas with similar geographical location, topography, and hydrodynamic conditions to the restoration area were selected to establish an observation grid for the control areas and collect their basic ecological parameter sets. This includes vegetation coverage V′, benthic organism density B′, water nutrient concentration N′, and bottom sediment stability P′.

[0011] Preferred, for Time standardization was performed to align its time series with the parameter set of the repair region. Maintain synchronization using a time registration function. Matching observation points from different time phases; Establish a normalized difference model The model uses the parameter set of the repair area. Parameter set of the control region Based on the difference, and combined with the set of repair response weight coefficients Weighted corrections were performed, and the normalized net effect of each ecological parameter was calculated. Where ε is a stability coefficient to prevent the denominator from being zero; the net effect results of each ecological parameter are integrated to form a net restoration effect vector. .

[0012] Preferably, the step of inputting ΔE into the ecosystem structure coupling model M and obtaining the comprehensive contribution of restoration based on the interdependence among ecological factors includes: Before inputting the net restoration effect vector ΔE into the ecosystem structure coupling model M, the components are standardized, and principal component analysis is used to extract the correlation features between the components of ΔE to obtain the initial feature matrix. ; based on A network graph of interdependence among ecological factors is constructed. The Pearson correlation coefficient and mutual information entropy are used to jointly determine the coupling strength between pairs of factors, and the coupling coefficient matrix is ​​calculated. ,in Characterizes the degree of structural coupling between the i-th ecological factor and the j-th ecological factor; when When the value exceeds the preset threshold τ, it is determined that there is a significant coupling relationship between the two factors; Using the coupling coefficient matrix C as input to model M, a structural coupling algorithm based on graph convolutional networks is employed to calculate the global coupling contribution of each ecological factor. ; The coupling contribution of each ecological factor Its corresponding net repair effect Calculate the overall contribution of restoration by weighting. .

[0013] Preferably, the step of performing spatiotemporal interpolation analysis on C to obtain the island ecological restoration effectiveness distribution matrix S includes: The overall contribution of each monitoring point to restoration A spatiotemporal dataset was constructed by combining geographic coordinate information, and the Kriging interpolation method was used to analyze the data at different monitoring periods and spatial locations. Perform spatial interpolation to generate a preliminary distribution layer of the repair results; A spatiotemporal interpolation analysis model was established based on the interpolation results. By incorporating the time dimension as a continuous variable into the model, and using a three-dimensional spline interpolation algorithm to calculate the predicted restoration contribution at any time and location, a distribution matrix S of island ecological restoration effectiveness is formed, where the matrix elements... This represents the repair effectiveness value of the i-th spatial unit at time j.

[0014] Preferably, the calculation of the repair vulnerability index of each spatial unit includes: calculating the repair stability index of each spatial unit based on the temporal variation characteristics of matrix S. And the volatility coefficient method is used to measure its time volatility; when When the stability value exceeds the set stability threshold θ, the unit is classified as a high-stability region; otherwise, it is marked as a vulnerable region. This comprehensive approach aims to improve stability. Construct a repair vulnerability index model using spatial gradient information. ,in The spatial gradient representing the repair effectiveness is normalized to obtain the repair vulnerability index distribution map of each unit.

[0015] Preferably, generating a set of repair and optimization suggestions based on the threshold distribution results of F includes: Based on the vulnerability index Based on the spatial distribution results, an ecological restoration risk level matrix F is constructed, where F consists of the vulnerability level corresponding to each spatial unit, and is based on... The threshold range is automatically divided into three categories: high-risk area, medium-risk area and low-risk area; Cluster analysis was performed on the risk level matrix F, and a partitioning optimization method based on the K-means and DBSCAN composite algorithm was used to identify spatial clusters with similar ecological degradation characteristics, forming a restoration priority distribution map, in which the average vulnerability value of the cluster center was used as the restoration priority weight. Based on the restoration priority distribution results, combined with the ecological element characteristic parameter set With repair response weight coefficient set It calls a remediation strategy matching model based on a combination of rule base and machine learning to automatically generate a set of remediation optimization suggestions, including the type of remediation measure, intervention intensity, time series and expected ecological response.

[0016] The technical effects and advantages provided by the present invention in the above technical solution are as follows: 1. This invention constructs a comprehensive evaluation system covering the entire chain, from the collection of basic ecological parameters, modeling of restoration response, extraction of net effects, calculation of coupled contributions, spatiotemporal interpolation analysis, to intelligent restoration optimization output. This system enables dynamic quantification and systematic diagnosis of the island ecological restoration process. Compared with traditional assessment methods that rely solely on single indicators or static observations, this invention can simultaneously capture the mutual coupling relationships and spatiotemporal evolution characteristics among multiple ecosystem elements, significantly improving the scientific rigor and accuracy of restoration effectiveness assessment.

[0017] 2. This invention can automatically identify key constraints and vulnerable areas in the restoration process of island ecosystems, forming a targeted set of restoration optimization suggestions, thus achieving a leap from "ecological monitoring" to "intelligent decision-making." The proposed ecosystem structure coupling model and spatiotemporal interpolation algorithm transform the assessment of restoration effects from static state judgment to dynamic trend prediction, significantly improving the interpretability and foresight of the assessment results. By combining a rule base with a machine learning-based restoration strategy matching model, this invention can not only output a quantitative comprehensive report on restoration effectiveness but also automatically recommend optimal restoration paths for different spatial units, reducing the uncertainty of human decision-making and improving the efficiency of restoration resource allocation. Attached Figure Description

[0018] To more clearly illustrate the technical solutions in the embodiments of this application or the prior art, the drawings used in the embodiments will be briefly introduced below. Obviously, the drawings described below are only some embodiments recorded in this invention. For those skilled in the art, other drawings can be obtained based on these drawings.

[0019] Figure 1 This is a flowchart of the method of the present invention. Detailed Implementation

[0020] To make the objectives, technical solutions, and advantages of the embodiments of the present invention clearer, the technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0021] For examples, please refer to Figure 1As shown in this embodiment, a comprehensive evaluation method for the effectiveness of island ecological restoration includes: Obtain a set of basic ecological parameters within the target island restoration area, including vegetation cover (V), benthic organism density (B), water nutrient concentration (N), and sediment stability (P), and construct an original ecological feature matrix. ; Based on the temporal variation characteristics of each parameter, the repair response curve is calculated. And through dynamic weighting functions Obtain the set of repair response weight coefficients for each subsystem It is the output function; Obtain the parameter set of the adjacent unrepaired control area outside the repair area. And the net repair effect vector ΔE is obtained through the normalized difference model; ΔE is input into the ecosystem structure coupling model M, and the comprehensive contribution of restoration is obtained based on the interdependence among ecological factors, where M is used to quantify the coupling strength between different ecological factors. Spatiotemporal interpolation analysis was performed on C to obtain the distribution matrix S of the island's ecological restoration effectiveness, and the restoration vulnerability index of each spatial unit was calculated. Based on the threshold distribution of F, a set of restoration and optimization suggestions is generated, and a comprehensive evaluation report on the effectiveness of island ecological restoration is output.

[0022] In this invention, vegetation cover V is obtained using a multispectral aerial survey method conducted by a UAV. The UAV, equipped with a multispectral sensor, performs zoned aerial photography under clear, low-wind conditions to acquire surface reflectance data in the blue, green, red, and near-infrared bands. Spectral calibration is performed by setting up a fixed ground reflector before data acquisition to eliminate measurement errors caused by differences between devices.

[0023] After image registration, the obtained image data is calculated using the Normalized Difference Vegetation Index (NDVI), which utilizes the difference in reflectance between red and near-infrared bands to obtain the vegetation index value for each pixel. To eliminate the influence of strong sunlight and cloud shadows on the NDVI results in island areas, this invention introduces an adaptive illumination correction algorithm. This algorithm automatically identifies abnormal illumination areas based on the brightness variation trend of the time series and performs weighted smoothing adjustment on their pixel values. The corrected NDVI results are spatially aggregated to obtain the average vegetation cover V for each zone.

[0024] In the process of island ecological restoration, benthic communities are an important indicator reflecting the level of ecological health. This invention deploys multiple in-situ biological monitoring devices in the intertidal zone and shallow sea areas. The devices include underwater camera units, lighting modules, and dissolved oxygen and temperature sensing modules. The devices record benthic biological image data at 10-minute sampling intervals and upload it to a data processing terminal in real time.

[0025] During the data analysis phase, the system utilizes a machine vision recognition algorithm based on deep convolutional neural networks to automatically detect benthic organisms, identify their morphological characteristics, and classify their taxa. The algorithm achieves identification by comparing known species feature templates in the sample database and records the number of organisms at each monitoring point. The final benthic organism density B is calculated as the average number of individuals per unit area at each monitoring point.

[0026] To reduce sampling bias caused by tidal changes, the system uses a time-weighted average method to calculate the effective observations at different tide levels, thereby improving the stability and comparability of benthic organism density data.

[0027] The concentration of nutrients (N) in the water was determined using a portable multi-parameter water quality analyzer, with samples taken at different tidal times. Sample analysis included three parameters: total nitrogen (TN), total phosphorus (TP), and dissolved oxygen (DO). Traditional methods use fixed weights for each indicator, which fails to reflect the impact of tidal energy differences on the water exchange rate. Therefore, this invention establishes a dynamic weight calculation model that adjusts the weights of each indicator based on tidal current intensity by real-time monitoring of hydrodynamic parameters (including flow velocity and tidal range).

[0028] Specifically, the model defines the weights of total nitrogen, total phosphorus, and dissolved oxygen as follows: The system calculates a correction coefficient β based on the combination of the average tidal velocity v and tidal range h during the sampling period. When the tidal intensity increases, the weight of nutrient components is appropriately increased, while the influence of the dissolved oxygen inverse weight term is reduced. The resulting comprehensive nutrient concentration N in the water body characterizes the dynamic change pattern of eutrophication in island waters.

[0029] The data processing module smooths the continuously monitored data over time, removes abnormal fluctuations, and generates time-series change curves for each sampling point for subsequent repair response analysis.

[0030] Substrate stability (P) is an important indicator reflecting the erosion resistance of the foundation structure in the remediation area. This invention periodically measures the shear wave propagation velocity of shallow sediments by deploying multiple substrate shear wave velocity test probes at various points within the remediation area. Higher shear wave velocities indicate a denser substrate structure and stronger resistance to disturbance.

[0031] Simultaneously, the system employs mechanical sampling to obtain sediment samples at different depths, determining particle size distribution and water content, and calculating the rate of change of sediment shear modulus. By comparing the rate of change of shear modulus at different depths, the stability of the sediment structure can be quantified. The final sediment stability P is determined by the weighted average of shear wave velocity and the rate of change of shear modulus.

[0032] To facilitate data consistency, the sediment stability results were standardized and converted into dimensionless parameters with values ​​ranging from 0 to 1. A P value close to 1 indicates extremely stable sediment; a P value less than 0.4 indicates loose sediment that is easily eroded by tidal currents.

[0033] After acquiring the above four types of parameters, the system registers the data according to the principle of spatial zoning consistency, mapping vegetation coverage V, benthic organism density B, water nutrient concentration N, and sediment stability P to the same geographic grid unit.

[0034] For the same grid cell, if multiple temporal observation data exist, the system uses a time-weighted average method to obtain a representative value. This results in the original ecological characteristic matrix. It consists of multiple four-dimensional vectors, each containing four index values: V, B, N, and P for the corresponding region. As an input variable for subsequent restoration response calculations, it can systematically describe the initial ecological state of the restoration area.

[0035] Different ecological parameters in island ecological monitoring (such as vegetation cover V, benthic organism density B, water nutrient concentration N, and sediment stability P) typically have different sampling frequencies and time intervals. Directly calculating response curves would lead to time axis misalignment. Therefore, this invention first constructs a unified timescale sequence T.

[0036] The time scale T is generated based on the minimum sampling interval method: the system automatically reads the sampling time series of each parameter, takes the shortest interval Δt as the uniform time step, and interpolates the observation points of other parameters to complete the time step. The interpolation method adopts the locally weighted linear interpolation (LWLI) algorithm, which is based on three adjacent observation points at each target time and uses distance weights to perform linear prediction to keep the time series smooth and undistorted.

[0037] After interpolation, to reduce the impact of random fluctuations on subsequent curve fitting, this invention further introduces a sliding window averaging method. The window width is determined based on the typical response cycle of the ecological process: when the monitoring cycle is less than 30 days, the window width is set to 3; when the cycle is greater than 30 days, the window width is set to 5. The time series processed by the sliding average can eliminate short-term noise and maintain the long-term trend, providing a stable input for curve fitting.

[0038] After obtaining the smoothed time series, this invention establishes restoration response curves for each ecological parameter. This method is used to characterize the changing trends of ecological indicators over time after restoration. Since the restoration process of island ecosystems has phased characteristics, this invention employs a combined algorithm of "piecewise linear + second-order polynomial" to simultaneously characterize short-term disturbance responses and long-term restoration trends.

[0039] In the initial stages of restoration, ecological parameters often change rapidly, exhibiting near-linear growth or decline. The system identifies abrupt changes in slope within the time series, divides the curve into several sub-intervals, and uses linear fitting within each sub-interval to obtain the phased recovery rate.

[0040] During the stable phase, the changing trends of ecological parameters typically exhibit a slow, non-linear characteristic of rising or falling. This invention utilizes a second-order polynomial form for fitting, that is, using time as the independent variable and ecological parameters as the dependent variable to determine the rising arc and stable interval of the curve.

[0041] By continuously splicing piecewise linear and second-order polynomial results at boundary points and smoothing the splice points using cubic spline interpolation, a continuously differentiable repair response curve is formed. .

[0042] In each repair response curve, the system automatically extracts three types of key feature parameters: The response rate α, which is the average slope of the curve in the initial stage, is used to reflect the recovery speed. The lag period τ is the time required for an ecological parameter to go from a disturbed state to its first significant improvement. Peak recovery rate γ is the maximum increase in the curve when it reaches a stable stage.

[0043] These three types of characteristic parameters together describe the dynamic characteristics of an ecosystem from disturbance to recovery.

[0044] The dynamic weighting function defined in this invention The aim is to determine the restoration response sensitivity coefficient of each ecological element based on the above characteristic parameters. This reflects the changes in its weight in comprehensive ecological restoration.

[0045] Output function The core idea is that when a certain ecological element undergoes drastic temporal changes, recovers quickly, or has a short lag period, that element is more sensitive to the overall ecosystem's response, and its weight should be increased accordingly. The input variables of the function include: Original ecological parameter matrix Standardized values ​​in; Repair response curve The feature parameters α, τ, and γ extracted from them.

[0046] First, calculate the variance ratio for each ecological parameter. First, it is used to reflect the fluctuation intensity of this parameter in the time series; second, the time delay factor is defined. That is, the lag period τ and the total monitoring time The ratio. Output function. The formal definition is: ;in, and are model constants, representing the response amplification factor and the delay attenuation factor, respectively. In this invention... The value is 1. The value is 0.5, obtained based on experimental calibration.

[0047] The function's logic is as follows: if a certain parameter changes drastically over time ( (large), its weight This will increase; if there is a significant lag in recovery ( (Large), then weight Exponential decay is used to mathematically represent the dynamics of the recovery sensitivity.

[0048] In obtaining the sensitivity coefficient Then, the present invention incorporates the standard deviation of the original parameters. Jointly determine the repair response weighting coefficients The idea is to simultaneously consider the time response capability and numerical stability of the parameters in order to achieve objective weighting.

[0049] Specifically, the system first calculates the initial weight value for each ecological parameter. .

[0050] Subsequently, to ensure the comparability between weights, entropy weight normalization was used for scaling correction.

[0051] The implementation steps of the entropy weight method are as follows: For all Standardize it so that its numerical range is limited to between 0 and 1; Calculate the information entropy value of each parameter. The smaller the entropy value, the greater the information difference. Information utility value is calculated by reverse calculation based on information entropy. ; The final normalized weights are obtained by weighted averaging. .

[0052] This results in a set of repair response weight coefficients. These correspond to four ecological elements: vegetation coverage, benthic organism density, water nutrient concentration, and substrate stability.

[0053] In this invention, the scientific selection of the control area directly affects the reliability of the remediation effectiveness calculation. Therefore, the system uses Geographic Information System (GIS) data and a marine dynamic model to screen the areas surrounding the remediation area, selecting the nearest unrestored area that is most similar to the remediation area in terms of topography, substrate type, hydrodynamic conditions, and habitat type as the control area.

[0054] In practice, a spatial similarity matching algorithm is used to compare the feature vectors of the remediation area and the candidate area. This algorithm normalizes indicators such as terrain slope, water depth, bottom sediment particle size, and flow velocity to values ​​between 0 and 1, and calculates the Euclidean distance between the two areas. When the similarity is greater than 0.85, it can be identified as a qualified control area.

[0055] After determining the control area, a spatial observation grid identical to that of the restoration area is established in this area, with each grid unit maintaining a consistent scale (e.g., 100 meters × 100 meters), and data are collected in the same area as the restoration area. Similar ecological parameters include vegetation cover V′, benthic organism density B′, water nutrient concentration N′, and sediment stability P′.

[0056] Since there may be a time inconsistency between the observations of the restoration area and the control area, this invention constructs a time registration function. This function is used to synchronize time series data. It compares the sampled time series data from the repair region and the control region. and The time difference ΔT between the two is calculated, and time correction is performed using weighted time interpolation. The interpolation weight is inversely proportional to the time interval; that is, the closer the time is to the target time, the higher its weight. If the control area is missing data at a certain time, linear interpolation is performed to fill in the missing data based on the observations at adjacent times.

[0057] Furthermore, to reduce long-term trend bias, the system employs a time-shift correction mechanism: when seasonal periodic changes in the ecological response of the remediation area are detected (e.g., annual tidal or temperature fluctuation cycles), the system automatically adjusts the time series of the control area to ensure that the seasonal phases of the two are consistent. After correction, the parameter set of the remediation area... Parameter set of the control region Achieving a one-to-one correspondence in the time dimension provides a reliable foundation for subsequent difference calculations.

[0058] To eliminate the impact of natural fluctuations, this invention establishes a normalized difference model to calculate the net ecological improvement effect between the remediation area and the control area. This model uses a set of parameters from the remediation area. Parameter set of the control area and repair response weight coefficient set The input is ΔE, and the output is the net effect vector of the repair.

[0059] Traditional methods typically employ simple interpolation. While this method represents the restoration effect, it ignores the scale differences and weighting effects between parameters, easily leading to incomparability between different ecological factors. This invention introduces a weighted normalization mechanism to ensure a unified measurement standard for different ecological elements in the comprehensive evaluation.

[0060] This invention defines the net effect component of repair. for: .in: This represents the average value of the i-th ecological parameter in the restoration area; This represents the average value of similar parameters in the corresponding control area; ε represents the calculated repair response weight; ε is a stability coefficient used to prevent numerical divergence caused by a denominator that is zero or close to zero. Its value ranges from 0.01 to 0.05, and in this embodiment of the invention, it is 0.02.

[0061] During the calculation process, the system first performs calculations on all... and Normalization (from 0 to 1) is performed to eliminate dimensional differences among the parameters. Then, the differences are calculated by substituting each parameter into the formula above and multiplied by their respective weights. .

[0062] The model's computation process is implemented in vectorized form within the data analysis module, enabling it to operate efficiently on large-scale ecological monitoring data.

[0063] After obtaining the net effect components of all ecological parameters Subsequently, the present invention integrates the results to form a net repair effect vector. Each component corresponds to one of the four ecological subsystems: vegetation, benthic organisms, water quality, and substrate.

[0064] To improve the stability of ΔE, this invention introduces a threshold determination and anomaly correction mechanism. The absolute value of each component of ΔE is calculated, and when the absolute value of any component exceeds a set threshold δ, the result is determined to be abnormal.

[0065] The threshold δ is defined based on empirical statistical results, when When the absolute value is greater than 0.8, it indicates that the difference between the repaired area and the control area is too large, which may be due to the influence of extreme external events (such as storm surges, typhoons, or sudden pollution). At this time, the system activates the local refitting algorithm, that is, in... During the calculation, observations from abnormal time periods are removed, and only data from adjacent time windows are used for recalculation, making the results smoother and more stable.

[0066] The final output ΔE vector is the net effect vector of restoration, representing the actual ecological improvement value after the restoration activities have eliminated the impact of natural fluctuations.

[0067] To ensure the consistency and comparability of the model input data, this invention first repairs the net effect vector. Standardize the process.

[0068] In its implementation, the system employs the Z-score normalization method, which means that each component... Subtracting the sample mean and then dividing by the sample standard deviation yields a standardized variable with a mean of 0 and a variance of 1. This eliminates differences in the dimensions and ranges of the various ecological parameters, ensuring a balanced weight distribution in subsequent modeling.

[0069] Subsequently, the system extracts features from the standardized data using Principal Component Analysis (PCA). The PCA algorithm solves for eigenvalues ​​and eigenvectors from the covariance matrix, selecting principal components that explain more than 85% of the population variance to construct the initial feature matrix. The matrix This essentially reflects the main change patterns among ecological factors, providing a foundation for the subsequent establishment of factor-dependent networks.

[0070] In this invention, the interdependence among ecological factors includes not only linear correlations but also nonlinear and implicit interactions. Therefore, this invention employs a joint modeling approach using Pearson correlation coefficient (Pearson r) and mutual information entropy (MI) to construct the coupling strength of ecological factors.

[0071] The Pearson correlation coefficient is used to measure the linear relationship between two ecological factors, and its value ranges from -1 to 1. When r is close to 1, it indicates a positive correlation; when it is close to -1, it indicates a negative correlation.

[0072] Mutual information entropy is used to characterize the degree of information dependence between two variables, that is, the degree to which the uncertainty of one variable is reduced by the other variable, and its value ranges from 0 to 1.

[0073] This invention defines the composite coupling coefficient between ecological factors i and j. The weighted combination of the two: Where α is the balance coefficient between linear and nonlinear correlation, and its value ranges from 0 to 1. In this embodiment, it is taken as 0.6.

[0074] All calculated Constructing the coupling coefficient matrix The diagonal elements of the matrix are fixed at 1, representing the complete dependence of ecological factors on themselves.

[0075] To identify significant coupling relationships, this invention sets a coupling strength threshold τ to determine which factors are strongly coupled. The threshold τ is selected based on the principle of statistical significance; when... When the value is greater than 0.5, the system considers the coupling relationship to be significant.

[0076] For weakly correlated factor pairs that do not meet the conditions, direct connections are not established in the coupled network to avoid interference from weak coupling noise.

[0077] Based on the significant coupling relationships in matrix C, the system constructs an interdependent network graph G(V, E) of ecological factors, where the node set V represents the set of ecological factors {V, B, N, P}, and the edge set E represents factor pairs with a coupling strength exceeding τ. The weight of each edge is determined by... This indicates the strength of interactions between factors. This network diagram visually describes the coupling structure within the ecosystem, providing a topological framework for subsequent calculations of coupling strength at the system level.

[0078] In this invention, the ecosystem structure coupling model M is constructed based on the Graph Convolutional Network (GCN) architecture to capture multi-level interactions between ecological factors.

[0079] Input layer: with feature matrix As the node feature input, the coupling coefficient matrix C is used as the adjacency matrix input; Propagation layer: This layer enables the transfer and fusion of information between nodes through graph convolution operations. The feature update of each node is obtained by a weighted sum of its own features and the features of its neighboring nodes, with the weights determined by the coupling coefficients in the C matrix. Activation and Normalization: The ReLU activation function is used to enhance the nonlinear expressive power of the model, and layer normalization is performed after each propagation to prevent gradient vanishing; Output layer: Outputs the coupling contribution of each node (i.e., ecological factor). This indicates the degree of influence of the factor on the overall system structure recovery.

[0080] The model's parameters are trained using a backpropagation algorithm to minimize the error between the model's output system coupling structure and the input ΔE data. Training samples can consist of observation data from multiple repair projects or multiple time periods to enhance the model's generalization ability.

[0081] In obtaining the coupling contribution of each ecological factor Subsequently, this invention defines the overall contribution of repair. for: The net restoration effect of each ecological factor. Its system coupling contribution The results are combined proportionally to form a comprehensive indicator. The value of is normalized by Min–Max to limit it to between 0 and 1, which facilitates comparison between different repair areas.

[0082] when A value close to 1 indicates that the restoration project has a significant promoting effect on the overall structure of the ecosystem; when... A value less than 0.3 indicates limited restoration effectiveness or that the ecosystem coupling structure has not yet formed a stable feedback loop.

[0083] To reconstruct a continuous spatial repair distribution based on discrete monitoring points, this invention employs the Kriging spatial interpolation algorithm.

[0084] This algorithm is based on the theory of regionalized variables and considers the overall contribution of repair. Treating it as a spatial random variable, we assume it satisfies the second-order stationarity condition, meaning the difference between any two points depends only on the distance. This is achieved by calculating the variogram. Describes spatial autocorrelation, where h represents the distance between two points.

[0085] The system first calculates the experimental variogram based on actual monitoring data, and then fits a theoretical model. The fitting model can be a spherical model, an exponential model, or a Gaussian model; this invention uses an exponential variogram to smoothly describe spatial continuity. Based on the fitting results, the system obtains the Kriging interpolation weight set. The predicted restoration contribution at any location (x, y) is calculated by weighted summation. That is, the predicted restoration contribution at the target location is equal to the weighted average of the restoration values ​​of all neighboring monitoring points multiplied by their respective spatial weights.

[0086] To reflect the dynamic changes in the restoration effectiveness, this invention further introduces a time dimension based on the Kriging space interpolation results, constructing a three-dimensional spline spatiotemporal interpolation model. .

[0087] This model uses time as a continuous variable and employs a three-dimensional B-spline interpolation algorithm to jointly model space and time, enabling the prediction of the repair contribution at any location at any time.

[0088] The specific implementation steps of the model include: The spatial interpolation results are stacked in time series to form a three-dimensional grid data structure; Spline control points are set along the time axis, and the spacing between the control points is automatically generated based on the monitoring frequency. Piecewise cubic spline functions are used to fit time series data to ensure that the interpolation results are smooth and continuous over time. Spline interpolation is performed simultaneously on the spatial and temporal layers to generate a four-dimensional interpolation result matrix, outputting continuous predicted values ​​of repair contribution.

[0089] The final distribution matrix of island ecological restoration effectiveness ,in This represents the repair effectiveness value of the i-th spatial unit at time j. The dimensions of the matrix correspond to the number of spatial grids and the number of time steps, achieving bidirectional spatiotemporal expansion.

[0090] After obtaining the repair effectiveness distribution matrix S, this invention calculates the repair stability index of each spatial unit. This is used to measure the temporal consistency of restoration effectiveness in different regions.

[0091] Stability Indicators Defined by the coefficient of variation (CV), it is the ratio of the standard deviation of a time series to the mean of the time series. Specifically: It equals the standard deviation of the repair effectiveness of the unit over the entire time period divided by its mean. When A smaller value (e.g., less than 0.3) indicates a stable repair process with minimal fluctuations; when... A large value (e.g., greater than 0.5) indicates that the repair effect in this area is significantly affected by external interference and has poor stability.

[0092] To ensure statistical significance, this invention excludes units with a time series length less than 5 from the calculation process. Calculations are performed to avoid misjudgments due to insufficient samples.

[0093] To further identify spatial regions where the repair effect is susceptible to disturbance, this invention establishes a repair vulnerability index model based on the repair stability λᵢ. .

[0094] The model consists of three parts: Average Repair Effectiveness : Indicates the average repair level of this unit; Time stability term This reflects the consistency of time during the repair process of this unit; Spatial gradient term : Indicates the spatial rate of change of the repair effectiveness of this unit and its surrounding units.

[0095] Spatial gradient The calculation employs the central difference method, which determines the gradient direction and magnitude by analyzing the differences in repair values ​​between adjacent cells. A larger gradient indicates significant differences in repair within the region, making it more susceptible to spatial imbalances.

[0096] Repair Vulnerability Index The calculation is implemented using a weighted linear combination method, specifically: Equal to a times Add b times Subtract c times Where a, b, and c are empirical weighting coefficients, satisfying In this embodiment of the invention, a, b, and c are taken as 0.4, 0.4, and 0.2, respectively, to take into account the effects of time fluctuations and spatial gradients.

[0097] Calculated After the values ​​are normalized to 0 to 1, a vulnerability index distribution map is generated. The higher the value, the greater the vulnerability of the region's ecological restoration, and the greater the need for subsequent maintenance and ecological consolidation.

[0098] To facilitate application and decision-making, this invention is based on the vulnerability index for repair. The distribution characteristics are used to set a three-level classification threshold: when When the value is less than 0.3, it is considered to be in the "stable recovery zone"; when When the value is between 0.3 and 0.6, it is considered a "recoverable zone"; when When the value is greater than 0.6, it is identified as a "highly vulnerable area".

[0099] The classification thresholds are determined through statistical analysis and empirical calibration, and can be dynamically adjusted according to different island ecosystem types. The system outputs a vulnerability distribution map and a classification table, providing a basis for subsequent restoration and maintenance and ecological risk early warning.

[0100] Arrange spatial units according to geographic index to form a two-dimensional matrix. ,in This represents the risk level value of the cell in the i-th row and j-th column. The matrix dimension is consistent with the repair effectiveness distribution matrix S, achieving a one-to-one correspondence between spatial locations.

[0101] To reduce the impact of local anomalies on risk classification, this invention employs a 3×3 sliding window median filtering algorithm to smooth matrix F. This method effectively suppresses isolated high-value points and improves the spatial continuity and interpretability of the risk distribution.

[0102] After constructing the risk level matrix F, this invention further identifies regional clusters with similar degradation or recovery characteristics through cluster analysis.

[0103] This invention employs a hybrid clustering model that combines K-means with DBSCAN (density-based spatial clustering algorithm).

[0104] The K-means algorithm is used to perform preliminary clustering of the overall structure of risk levels in the F matrix, and to determine the cluster centers and the number of categories K. The DBSCAN algorithm performs density correction based on the results of K-means and automatically identifies noise points and the boundaries of high-density clusters.

[0105] The initial value of K is set to 3, corresponding to the three risk areas of high, medium and low. DBSCAN sets the minimum number of neighborhood samples to 5, and the neighborhood radius ε is adaptively calculated based on the spatial resolution (usually 1.5 times the average grid spacing).

[0106] For each cluster in the clustering results, the system calculates its average vulnerability value. And sort them according to size, assigning repair priority weights. Clusters with high vulnerability have higher priority weights and are used to guide the generation of subsequent remediation recommendations.

[0107] To automatically generate optimized remediation suggestions for different types of risk areas, this invention proposes a remediation strategy matching model based on a rule base and machine learning. The model takes into account a risk level matrix F and a set of basic ecological parameters. and repair response weight coefficient set It outputs a repair and optimization strategy for each spatial unit.

[0108] The rule base contains three categories of remediation strategy templates, corresponding to vegetation restoration, water body regulation, and sediment improvement, respectively. Each template consists of a conditional expression and a corresponding remediation measure. For example: like For high-risk areas with vegetation coverage of less than 30%, it is recommended to use "replanting native vegetation + soil covering and stabilization". like If the area is classified as medium-risk and the nutrient concentration in the water is higher than the threshold of 2 mg / L, it is recommended to implement "ecological floating islands + nitrogen and phosphorus control".

[0109] Building upon rule-based matching, the model further utilizes a regression prediction module based on the random forest algorithm to quantitatively assess the expected ecological responses of different restoration measures. The random forest is trained using historical restoration case samples to predict the improvement of restoration measures on ecological indicators such as vegetation restoration rate and substrate stability.

[0110] The model integrates the rule base output with machine learning predictions, calculates a comprehensive benefit score using a weighted decision function, and selects the remediation strategy with the highest score. The output results form a set of remediation optimization suggestions. ,in It includes the type of remediation measures, the intensity of intervention, the timing of intervention, and the expected response value.

[0111] To achieve intuitive presentation of remediation assessment results and support decision-making, this invention designs an automated comprehensive report generation module. This module integrates a set of remediation optimization suggestions, a remediation effectiveness distribution matrix, comprehensive remediation contribution, and vulnerability index to generate a structured report document.

[0112] The report includes: Visualization layer of restoration effectiveness: The matrix S is converted into a color distribution map using the Geographic Information System (GIS) module to show the restoration contribution intensity of different areas; System stability analysis table: Lists the stability indicators, vulnerability indices, and risk levels of each region; Improve the risk zoning results: Display the area proportions and geographical distribution characteristics of high, medium and low risk areas in the form of tables and maps; Repair and optimization suggestion list: List the optimization suggestion set G in order of repair priority, and specify the target area, measure type and implementation period for each suggestion; Overall effectiveness evaluation conclusion: Automatically generates a comprehensive score and trend analysis of the overall restoration effect to assess the overall effectiveness of the restoration project.

[0113] The above description is merely a specific embodiment of this application, but the scope of protection of this application is not limited thereto. Any changes or substitutions that can be easily conceived by those skilled in the art within the scope of the technology disclosed in this application should be included within the scope of protection of this application.

Claims

1. A comprehensive evaluation method for the effectiveness of island ecological restoration, characterized in that: The method comprises the following steps: The basic ecological parameter set in the target island repair area is acquired, including vegetation coverage V, benthic organism density B, water body nutrient salt concentration N and bottom stability P, and an original ecological characteristic matrix is constructed ; according to the time sequence change characteristics of each parameter, the repair response curve thereof is calculated , and the repair response weight coefficient set of each subsystem is obtained through a dynamic weighting function ; and the output function is . Acquiring a parameter set of an adjacent non-repair control area outside the repair area and obtaining a repair net effect vector ΔE by a normalized difference model. inputting the ΔE into an ecosystem structure coupling model M to obtain a comprehensive contribution degree of repair according to the interdependence between ecological factors, wherein the M is used to quantify the coupling strength between different ecological factors; performing spatio-temporal interpolation analysis on the C to obtain an island ecological repair effectiveness distribution matrix S, and calculating a repair vulnerability index of each spatial unit; generating a repair optimization suggestion set according to the threshold distribution result of the F, and outputting an island ecological repair effectiveness comprehensive evaluation report.

2. The method according to claim 1, characterized in that: The method comprises the following steps: performing partition aerial survey on the target island repair area, collecting ground reflectivity data at different time phases, and calculating the vegetation coverage V of each partition by using a normalized vegetation index; collecting benthic organism quantity and activity frequency data in real time, and calculating the benthic organism density B according to multi-period average density; obtaining seawater samples at different tide time phases, measuring the total nitrogen, total phosphorus and dissolved oxygen content, and calculating the water body nutrient salt comprehensive concentration N by weighted summation; combining multi-layer sediment sampling, measuring the sediment particle size distribution and shear modulus change rate, and calculating the sediment stability P.

3. The method according to claim 1, characterized in that: The repair response curve of each parameter is calculated according to the time sequence variation characteristics of the parameter , comprising: Original ecological feature matrix The ecological parameters were interpolated over time to construct a unified time-scale sequence T. Using a combination algorithm of piecewise linear fitting and second-order polynomial curve analysis, the restoration response curves for vegetation cover V, benthic organism density B, water nutrient concentration N, and sediment stability P were calculated. .

4. The method according to claim 3, characterized in that: The dynamic weighting function is determined by the variance ratio of the characteristic parameters and the time delay factor The repair response weight coefficient set of each subsystem is obtained , comprising: based on the parameter repair response curve The characteristic parameters are used to construct a dynamic weighting function , the function takes the variance ratio of the characteristic parameters and the time delay factor as input, and determines the sensitivity coefficient of each ecological element by a normalized exponential decay model , wherein The response intensity of the i-th ecological element to external disturbance is represented; the sensitivity coefficient The original parameter standard deviation The repair response weight coefficient is calculated by synthesizing And the scale correction is carried out on each weight by using the entropy weight normalization method.

5. The method according to claim 1, characterized in that: the acquisition of the set of parameters outside the repair zone adjacent to the non-repaired control zone comprising: Neighboring unrestored control areas with similar geographical location, topography, and hydrodynamic conditions to the restoration area were selected to establish an observation grid for the control areas and collect their basic ecological parameter sets. This includes vegetation coverage V′, benthic organism density B′, water nutrient concentration N′, and bottom sediment stability P′.

6. The method according to claim 5, characterized in that: To perform time normalization, making the time series of the repair area parameter set synchronous with the time series of the reference area parameter set, by a time registration function match the different time points observations; A normalized difference model is established , which is based on the difference between the parameter set of the restoration area and the parameter set of the control area , and is weighted and corrected by combining the restoration response weight coefficient set . The normalized net effect of each ecological parameter is calculated , where ε is a stability coefficient to prevent the denominator from being zero; and the net effect results of each ecological parameter are integrated to form a restoration net effect vector .

7. The method according to claim 1, characterized in that: The method comprises the following steps: The method comprises the following steps: Before inputting the repair net effect vector ΔE into the ecosystem structure coupling model M, each component is normalized, the correlation characteristics between the ΔE components are extracted by principal component analysis, and an initial characteristic matrix is obtained ; Based on The interdependence network graph between ecological factors is constructed, the coupling strength of each two factors is determined by combining the Pearson correlation coefficient and mutual information entropy, and a coupling coefficient matrix is calculated , wherein The structural coupling degree of the i th ecological factor and the j th ecological factor is represented; when The significant coupling relationship exists between the two factors when the coupling coefficient is greater than a preset threshold τ. The coupling coefficient matrix C is taken as an input of the model M, and a structure coupling algorithm based on a graph convolution network is used to calculate the global coupling contribution degree of each ecological factor ; Coupling contribution of each ecological factor The corresponding repair net effect Superimposed by weight, calculate the repair comprehensive contribution .

8. The method according to claim 1, characterized in that: The method comprises the following steps: The method comprises the following steps: The repair comprehensive contribution degree of each monitoring point The spatio-temporal dataset is constructed in combination with geographic coordinate information, and a Kriging interpolation method is adopted to perform spatial interpolation on the monitoring time periods and spatial positions of The repair effectiveness preliminary distribution layer is generated. Establishing a space-time interpolation analysis model based on the interpolation result The time dimension is taken as a continuous variable in modeling, the repair contribution prediction value at any time and any position is calculated through a three-dimensional spline interpolation algorithm, and a sea island ecological repair effect distribution matrix S is formed, wherein matrix elements represent the repair effect value of the i-th space unit at the j-th time.

9. The method according to claim 8, characterized in that: The calculation of the repair vulnerability index of each spatial unit comprises: calculating the repair stability index of each spatial unit according to the time sequence variation characteristics of the matrix S , and the time fluctuation degree is measured by using the fluctuation coefficient method; when the stability threshold θ is higher than the set stability threshold θ, the determination unit is a high stability area, otherwise it is marked as a fragile area; the comprehensive repair stability and the spatial gradient information are used to construct a repair vulnerability index model , wherein the spatial gradient of the repair effect, and the repair vulnerability index distribution diagram of each unit is obtained through normalization processing.

10. The method according to claim 9, characterized in that: performing cluster analysis on the risk level matrix F, identifying spatial clusters with similar ecological degradation characteristics by using a partition optimization method based on a K-means and DBSCAN composite algorithm, and forming a repair priority distribution map, wherein the average vulnerability value of the cluster center is used as a repair priority weight. Based on the spatial distribution results of the restoration vulnerability index , a restoration risk level matrix F is constructed, where F is composed of the vulnerability levels of each spatial unit and is automatically divided into three categories of high-risk area, medium-risk area and low-risk area according to the threshold interval . ​ According to the repair priority distribution result, combined with the ecological element characteristic parameter set And the repair response weight coefficient set , call the repair strategy matching model based on the combination of rule base and machine learning, automatically generate a set of repair optimization suggestions, including repair measure type, intervention intensity, time sequence and expected ecological response.

Images