Sea level rise influence risk assessment method, medium and system

Abstract

The invention provides a sea level rise influence risk assessment method, medium and system, and belongs to the technical field of risk prediction and assessment, and the method comprises the steps: firstly, analyzing historical sea level data through wavelet decomposition and an ARIMA model, and predicting a future change trend; meanwhile, a digital elevation model is established by using Kriging interpolation, and dynamic change of the terrain is described by combining real-time monitoring data. Secondly, identifying key infrastructures and influence factors by applying an analytic hierarchy process and principal component analysis, and extracting key features through singular value decomposition; and based on a singular value decomposition result, constructing a sensitivity and vulnerability matrix, calculating the exposure degree in combination with Monte Carlo simulation, and comprehensively evaluating the regional vulnerability. Furthermore, asset loss is predicted by applying a random forest algorithm, economic loss is calculated, and an ecological vulnerability index matrix is established to evaluate ecological influence. And finally, realizing quantitative risk assessment by adopting fuzzy comprehensive evaluation of an entropy weight method and a risk prediction model of support vector regression.

Description

Technical Field

[0001] The present invention belongs to the technical field of risk prediction and assessment, and in particular, relates to a method, medium and system for assessing the risk of sea level rise. Background Art

[0002] Under the current technological background, the main application scenarios and problems of sea level rise risk assessment are as follows: In the field of urban planning, traditional sea level rise risk assessment methods mainly rely on static geographic information systems and simple numerical models. These methods often ignore dynamic factors such as surface subsidence and tidal changes, resulting in insufficient accuracy of assessment results. In terms of infrastructure protection, existing assessment methods generally use linear regression or simple statistical models, which cannot effectively capture the nonlinear characteristics and multi-dimensional impacts of sea level rise. In the field of ecosystem protection, existing technologies often focus on the monitoring of a single indicator and lack a comprehensive assessment of ecosystem service value and biodiversity. In terms of economic loss assessment, traditional methods mainly rely on simple loss functions and deterministic models, which cannot effectively deal with uncertainty factors and multi-scenario analysis.

[0003] In order to improve the accuracy of risk assessment of the impact of sea level rise, existing technologies often use large-scale neural network models or large-scale data for processing, which leads to high computational complexity and high consumption of computing resources. Summary of the invention

[0004] In view of this, the present invention provides a method, medium and system for assessing the risk of sea level rise, which can solve the technical problems of high computational complexity and high consumption of computing resources in the prior art.

[0005] The present invention is implemented as follows: A first aspect of the present invention provides a method for assessing the risk of sea level rise, comprising the following steps: S10. Collect historical data on sea level rise, use wavelet decomposition algorithm to separate basic data and change data, and use ARIMA model to generate future forecast data; S20. Obtain the elevation data of the coastline, apply the Kriging interpolation method to establish a digital elevation model, and use monitoring equipment to obtain tide level and surface subsidence data in real time to achieve dynamic representation of the terrain; S30. Use the analytic hierarchy process to identify critical infrastructure and vulnerable assets, use principal component analysis to reduce the dimensionality of influencing factors, and extract key features through singular value decomposition to lay the foundation for risk assessment; S40. Based on the results of singular value decomposition, the sensitivity and vulnerability matrix is ​​constructed, and the affected area under different scenarios is calculated by Monte Carlo simulation method to obtain the exposure matrix and comprehensively evaluate the regional vulnerability. S50. Construct an asset loss matrix, use the random forest algorithm to predict asset losses, calculate the economic loss expectation matrix, and quantify the potential economic impact caused by sea level rise; S60. Construct an ecosystem service value matrix, calculate a biodiversity index matrix, comprehensively consider exposure, sensitivity and adaptive capacity, construct an ecological vulnerability index matrix, and assess the degree of impact on the ecosystem; S70, construct a fuzzy relationship matrix, calculate the weights using the entropy weight method, conduct fuzzy comprehensive evaluation, obtain comprehensive evaluation results, and provide multi-dimensional reference for the final risk assessment; S80, construct a risk index matrix, use singular value decomposition to extract the main risk feature vectors, use support vector regression to establish a risk prediction model, and achieve risk quantification and prediction; S90. Use the particle swarm optimization algorithm to optimize the risk prediction model parameters and generate the final risk assessment results.

[0006] Specifically, the step S10 includes: Step 101: Collect historical data on sea level rise in the assessment area, including relevant parameters such as tide level, sea water temperature, and salinity; Step 102: Decompose the historical data into basic data and change data using a wavelet decomposition algorithm, wherein the basic data reflects long-term trends and the change data reflects short-term fluctuations; Step 103: Using the ARIMA model in time series analysis, combined with the basic data and the change data, generate future prediction data for sea level rise, and the prediction formula is: ; in is the predicted value, is a constant term, is the autoregressive parameter, is the moving average parameter, is white noise; The step S20 specifically includes: Step 201: Obtaining coastline elevation data within the assessment area, including land elevation and seabed topography data; Step 202: Apply Kriging interpolation method to establish a coastline digital elevation model. The interpolation formula is: ; in is the estimated value of the unknown point, is the observed value of the known point, is the weight coefficient; Step 203: using tide monitoring stations and differential GPS surface settlement monitoring equipment deployed in the assessment area, real-time tide data and surface settlement data are acquired for dynamic updating of the model; The step S30 specifically includes: Step 301: Use the analytic hierarchy process to identify key infrastructure and vulnerable assets in the assessment area, and establish an asset value database; Step 302: Use principal component analysis (PCA) to reduce the dimension of the factors affecting sea level rise and construct an influencing factor matrix. ; Step 303: Use the singular value decomposition (SVD) method to decompose the influencing factor matrix Decompose it and get ,in and are the left and right singular vector matrices, is a diagonal matrix of singular values; The step S40 specifically includes: Step 401: Construct a sea level rise sensitivity matrix based on the SVD decomposition results. ; Step 402: Construct a vulnerability matrix ; Step 403: Using the Monte Carlo simulation method, combined with the coastline digital elevation model, tide data and surface subsidence data, the area of ​​the affected area under different sea level rise scenarios is calculated to obtain the exposure matrix ; It should be noted that in eigenvalue decomposition and singular value decomposition (SVD), represents the square root matrix of eigenvalues ​​or singular values. Specifically: for eigenvalue decomposition, ,in is a diagonal matrix, and the diagonal elements are the eigenvalues. It is a matrix formed by placing the square roots of the eigenvalues ​​on the diagonal. For singular value decomposition, ,in is a diagonal matrix whose diagonal elements are singular values. It is a matrix formed by placing the square roots of the singular values ​​on the diagonal.

[0007] The step S50 specifically includes: Step 501: Construct asset loss matrix ,in Indicates The assets in losses under different sea level rise scenarios; Step 502: Use the random forest algorithm to predict asset losses and update the loss matrix ; Step 503: Calculate the economic loss expectation matrix ,in is the sea level rise scenario probability vector; The step S60 specifically includes: Step 601: Constructing an ecosystem service value matrix ,in Indicates The ecosystem in The value of services under different sea level rise scenarios; Step 602: Calculate the Shannon-Wiener biodiversity index matrix , the calculation formula is: ; in is the number of species, For the The proportion of individuals of each species; Step 603: Constructing an ecological vulnerability index matrix ,in represents the matrix Hadamard product, represents the matrix Hadamard division, For the adaptability matrix, is the resilience matrix; The step S70 specifically includes: Step 701: Construct a fuzzy relationship matrix ,in Indicates The evaluation factors The degree of membership of each comment; Step 702: Calculate the weight vector using the entropy weight method , the weight calculation formula is: ; in For the The information entropy of each evaluation factor; Step 703: Perform fuzzy synthesis operation to obtain comprehensive evaluation results ,in represents the fuzzy composition operator; The step S80 specifically includes: Step 801: Constructing a risk index matrix ,in , , is the balance factor, , , is the corresponding weight matrix; Step 802: Use the singular value decomposition (SVD) method to decompose the risk index matrix Decompose and extract the main risk feature vectors; Step 803: Use the support vector regression (SVR) method to establish a risk prediction model based on the extracted risk feature vector, and its objective function is: ; in is the weight vector, is the bias term, and is the slack variable, is the penalty parameter; The step S90 specifically includes: Step 901: Optimize the parameters of the risk prediction model using the particle swarm optimization algorithm (PSO). The particle update formula is: ; ; in is the particle velocity, is the particle position, is the inertia weight, and is the acceleration constant, and is a random number, is the optimal position of an individual, is the global optimal position; Step 902: Generate a final sea level rise impact risk assessment result based on the optimized risk prediction model; Optionally, the method further includes step 903, formulating corresponding adaptation strategies and mitigation measures according to the risk assessment results, and regularly updating the assessment model.

[0008] A second aspect of the present invention provides a computer-readable storage medium, wherein the computer-readable storage medium stores program instructions, and when the program instructions are run in a computer, they are used to execute the above-mentioned method for assessing the risk of the impact of sea level rise.

[0009] A third aspect of the present invention provides a sea level rise impact risk assessment system, which includes the above-mentioned computer-readable storage medium.

[0010] Compared with the prior art, the method, medium and system for assessing the risk of sea level rise provided by the present invention have the following beneficial effects: First, in terms of data processing efficiency, the accuracy and efficiency of data processing are significantly improved by introducing the wavelet decomposition algorithm to separate the basic data and the change data of historical data and combining it with the ARIMA model for prediction.

[0011] Secondly, in terms of dynamic terrain characterization, the digital elevation model was established using the Kriging interpolation method, and dynamic data was obtained through real-time monitoring equipment, which significantly improved the accuracy of terrain characterization. In practical applications, the average error of terrain characterization was reduced from ±0.5 meters of traditional methods to ±0.15 meters, laying a more reliable foundation for subsequent risk assessment.

[0012] In terms of key factor identification, the combination of hierarchical analysis, principal component analysis and singular value decomposition has achieved accurate identification and dimensionality reduction of influencing factors. This method reduces the computational complexity of feature extraction from O(n³) to O(n²logn) while maintaining more than 90% information integrity.

[0013] In terms of model optimization, the particle swarm optimization algorithm is introduced to optimize the parameters of the support vector regression model, which significantly improves the model prediction accuracy.

[0014] In summary, the present invention solves the technical problems of the prior art that the calculation complexity is large and the calculation resources are consumed a lot. BRIEF DESCRIPTION OF THE DRAWINGS

[0015] Figure 1 A flow chart of the method provided by the present invention; Figure 2 This is a forecast map of sea level change trends; Figure 3 To assess regional surface subsidence maps; Figure 4 This is a graph analyzing the loss rate of key facilities under different scenarios; Figure 5 To map ecosystem service values ​​and biodiversity assessments; Figure 6 To assess the regional risk level distribution map; Figure 7 Compare the calculation time for each stage. DETAILED DESCRIPTION

[0016] In order to make the purpose, technical solution and advantages of the embodiments of the present invention more clear, the technical solution in the embodiments of the present invention will be clearly and completely described below in conjunction with the accompanying drawings in the embodiments of the present invention.

[0017] like Figure 1 FIG. 1 is a flow chart of a method for assessing the risk of sea level rise provided by the first aspect of the present invention. The method comprises the following steps: S10. Collect historical data on sea level rise, use wavelet decomposition algorithm to separate basic data and change data, and use ARIMA model to generate future forecast data; S20. Obtain the elevation data of the coastline, apply the Kriging interpolation method to establish a digital elevation model, and use monitoring equipment to obtain tide level and surface subsidence data in real time to achieve dynamic representation of the terrain; S30. Use the analytic hierarchy process to identify critical infrastructure and vulnerable assets, use principal component analysis to reduce the dimensionality of influencing factors, and extract key features through singular value decomposition to lay the foundation for risk assessment; S40. Based on the results of singular value decomposition, the sensitivity and vulnerability matrix is ​​constructed, and the affected area under different scenarios is calculated by Monte Carlo simulation method to obtain the exposure matrix and comprehensively evaluate the regional vulnerability. S50. Construct an asset loss matrix, use the random forest algorithm to predict asset losses, calculate the economic loss expectation matrix, and quantify the potential economic impact caused by sea level rise; S60. Construct an ecosystem service value matrix, calculate a biodiversity index matrix, comprehensively consider exposure, sensitivity and adaptive capacity, construct an ecological vulnerability index matrix, and assess the degree of impact on the ecosystem; S70, construct a fuzzy relationship matrix, calculate the weights using the entropy weight method, conduct fuzzy comprehensive evaluation, obtain comprehensive evaluation results, and provide multi-dimensional reference for the final risk assessment; S80, construct a risk index matrix, use singular value decomposition to extract the main risk feature vectors, use support vector regression to establish a risk prediction model, and achieve risk quantification and prediction; S90. Use the particle swarm optimization algorithm to optimize the risk prediction model parameters and generate the final risk assessment results.

[0018] The specific implementation methods of the above steps are described in detail below: The specific implementation of step S10 is to collect and process historical data of sea level rise to generate future prediction data. First, collect historical data of sea level rise in the assessment area, including relevant parameters such as tide level, sea water temperature, salinity, etc. These data usually come from multiple sources such as coastal stations, satellite remote sensing and marine surveys. The data collection time span is preferably not less than 30 years to ensure the representativeness and reliability of the data. Next, the collected historical data is decomposed into basic data and change data using a wavelet decomposition algorithm. Wavelet decomposition can effectively decompose time series data into components of different scales, where basic data reflects long-term trends and change data reflects short-term fluctuations. This step helps to identify the periodicity and trend characteristics of sea level changes. Wavelet decomposition uses discrete wavelet transform (DWT) and selects appropriate wavelet basis functions (such as Daubechies wavelets) for multi-scale decomposition. The number of decomposition layers is preferably set to 3-5 layers to balance computational efficiency and analysis accuracy. Then, the autoregressive integrated moving average (ARIMA) model in time series analysis is used to combine the obtained basic data and change data to generate future prediction data for sea level rise. The ARIMA model can capture the autocorrelation, trend and seasonal characteristics of time series, and is suitable for complex time series predictions such as sea level changes. The order (p, d, q) of the ARIMA model can be determined by the autocorrelation function (ACF) and partial autocorrelation function (PACF) analysis. Usually, p and q range from 0 to 2, and d ranges from 0 to 1. The prediction formula of the model is: ,in is the predicted value, is a constant term, is the autoregressive parameter, is the moving average parameter, is white noise. Model parameters are estimated by maximum likelihood estimation or conditional least squares method. The prediction time range is preferably the next 50-100 years to meet the needs of long-term planning. The purpose of this step is to provide reliable sea level rise prediction data for subsequent risk assessment and provide a scientific basis for decision-making.

[0019] The specific implementation method of step S20 is to obtain and process the coastline elevation data and establish a dynamic digital elevation model. First, obtain the coastline elevation data in the assessment area, including land elevation and seabed terrain data. These data can be obtained in a variety of ways, such as aerial lidar measurement, satellite remote sensing, sonar depth sounding, etc. The data resolution is preferably not less than 30 meters to ensure the accuracy of the model. Then, the Kriging interpolation method is applied to establish a coastline digital elevation model. The Kriging interpolation method is an optimal linear unbiased estimation method that can take into account spatial autocorrelation and is suitable for the interpolation of terrain data. The interpolation formula is: ,in is the estimated value of the unknown point, is the observed value of the known point, is the weight coefficient. The determination of the weight coefficient is based on the analysis of the variogram, and the commonly used variogram models include the spherical model, the exponential model and the Gaussian model. The interpolation grid size is preferably set to 10-30 meters to balance the calculation efficiency and the model accuracy. Then, the tide monitoring station and the differential GPS surface settlement monitoring equipment are arranged in the assessment area to obtain the tide data and the surface settlement data in real time. The density of the tide monitoring station is preferably set up one per 50-100 kilometers of coastline, and the density of the differential GPS monitoring point is preferably set up one per 10-20 square kilometers. These real-time data are used for the dynamic update of the model to reflect the changes in the coastline topography. The data update frequency is preferably once a month to capture seasonal changes. Finally, the real-time monitoring data is combined with the digital elevation model, and the sliding time window method is used to update the model. The time window size is preferably set to 3-6 months to balance the model stability and dynamic response capability. The purpose of this step is to establish a dynamic digital elevation model that can reflect the changes in the coastline topography in real time, and provide accurate terrain basic data for subsequent risk assessment.

[0020] The specific implementation method of step S30 is to identify key infrastructure and vulnerable assets, and perform dimensionality reduction and feature extraction on the influencing factors. First, the analytic hierarchy process (AHP) is used to identify key infrastructure and vulnerable assets in the assessment area, and an asset value database is established. The AHP method decomposes complex problems into multiple levels and elements by constructing a hierarchical model, and then determines the relative importance of each element through expert scoring and consistency testing. Evaluation indicators may include economic value, social impact, environmental sensitivity, etc. The consistency ratio (CR) threshold is preferably set to 0.1. If it exceeds this value, expert evaluation is required again. The asset value database should contain information such as asset type, geographical location, economic value, and service life. Then, the principal component analysis (PCA) is used to reduce the dimensionality of the factors affecting sea level rise and construct the influencing factor matrix F. PCA can reduce the data dimension while retaining the main information, thereby improving the efficiency of subsequent analysis. The influencing factors may include the rate of sea level rise, the rate of surface subsidence, the rate of coastal erosion, the frequency of storm surges, etc. Before performing PCA, the original data needs to be standardized to eliminate the dimensional effect. The selection of principal components is based on the cumulative variance contribution rate, and the threshold is preferably set to 85%, that is, the first few principal components with a cumulative variance contribution rate of 85% are selected. Then, the singular value decomposition (SVD) method is used to decompose the influencing factor matrix F to obtain , where U and V are the left and right singular vector matrices, is a singular value diagonal matrix. SVD can effectively extract the main features of the matrix, which plays an important role in data compression and feature extraction. In practical applications, only the first k largest singular values ​​and their corresponding singular vectors can be retained. The selection of k can be based on the cumulative energy ratio of the singular values, and the threshold is preferably set to 90%. The purpose of this step is to identify key risk bearers, reduce the dimension and extract features of complex influencing factors, and lay the foundation for subsequent risk assessment. Through the combined application of AHP, PCA and SVD, both expert knowledge and data-driven methods are taken into account, and the key features of the impact of sea level rise can be effectively extracted.

[0021] The specific implementation of step S40 is to construct a sensitivity and vulnerability matrix, and combine it with Monte Carlo simulation to calculate exposure and comprehensively evaluate regional vulnerability. First, based on the SVD decomposition results, a sea level rise sensitivity matrix is ​​constructed. The sensitivity matrix reflects the sensitivity of different regions to sea level rise, where the U matrix contains the main spatial pattern information. Then, the strength of these patterns is adjusted. Next, the vulnerability matrix is ​​constructed The vulnerability matrix reflects the contribution of different influencing factors to the risk of sea level rise, where the V matrix contains the main factor pattern information. It also plays a regulatory role. In practical applications, only the top m most significant patterns can be retained according to the size of the singular values. The selection of m can be based on the cumulative variance explanation rate, and the threshold is preferably set to 95%. Then, the Monte Carlo simulation method is used to combine the coastline digital elevation model, tidal data and surface subsidence data to calculate the area of ​​the affected area under different sea level rise scenarios to obtain the exposure matrix E. Monte Carlo simulation can take into account the uncertainty of parameters and provide more reliable risk assessment results. The number of simulations is preferably set to 10,000 times to ensure the stability of the results. In each simulation, the digital elevation model is updated according to the predicted sea level rise, tidal changes and surface subsidence rate, and then the area of ​​the flooded or affected area is calculated. The sea level rise scenario can be set based on the IPCC report, usually including low, medium and high scenarios, corresponding to the RCP2.6, RCP4.5 and RCP8.5 emission scenarios, respectively. The elements of the exposure matrix E It represents the proportion of the affected area of ​​the jth region under the i-th sea level rise scenario. Finally, the sensitivity matrix S, the vulnerability matrix V and the exposure matrix E are combined to construct a comprehensive vulnerability index matrix ,in represents the Hadamard product of the matrix (i.e., the multiplication of corresponding elements). Each element of the comprehensive vulnerability index matrix represents the comprehensive vulnerability score of the jth region under the i-th sea level rise scenario. For ease of interpretation and comparison, The elements of the matrix are normalized to the range of 0-1. The purpose of this step is to comprehensively assess the regional vulnerability, taking into account the three key factors of sensitivity, vulnerability and exposure, and providing an important basis for subsequent risk assessment. The combination of matrix operations and Monte Carlo simulations not only ensures computational efficiency but also takes into account uncertainty, and can provide decision makers with more comprehensive and reliable vulnerability assessment results.

[0022] The specific implementation of step S50 is to construct an asset loss matrix, and use the random forest algorithm to predict asset losses and calculate the expected economic losses. First, construct an asset loss matrix L, where represents the loss of the ith asset under the jth sea level rise scenario. Asset types may include buildings, infrastructure, agricultural land, etc. The sea level rise scenario is consistent with the setting in step S40. The initial loss estimate may be based on a functional relationship between the asset value and the flooding degree, such as a linear relationship or an exponential relationship. For example, for buildings, it can be assumed that the loss is exponentially related to the flooding depth: , where k is the loss coefficient, which can be obtained by fitting historical data. Next, the random forest algorithm is used to predict asset losses and update the loss matrix L. Random forest is an integrated learning method that can effectively handle nonlinear relationships and feature interactions by constructing multiple decision trees and taking their average prediction results. Input features may include asset type, geographical location, sea level rise, terrain features, etc. The number of trees in the random forest is preferably set to 500-1000, and the minimum number of samples for each node is preferably set to 1% of the total sample size. In order to improve the generalization ability of the model, k-fold cross validation is used, and k is preferably set to 5 or 10. Model performance can be evaluated by root mean square error (RMSE) and determination coefficient (R²). The smaller the RMSE and the closer the R² is to 1, the better the model performance. Then, calculate the expected economic loss matrix , where P is the probability vector of sea level rise scenarios. The scenario probabilities can be determined based on IPCC reports or expert judgment. For example, it can be assumed that the probabilities of low, medium, and high scenarios are 0.2, 0.5, and 0.3, respectively. Each element of the economic loss expectation matrix Represents the expected loss of the ith asset. In order to more comprehensively evaluate the economic impact, indirect economic losses, such as production disruptions, supply chain disruptions, etc., can also be considered. This can be achieved by introducing an economic impact multiplier, the size of which can be determined based on the type of industry and regional economic characteristics, usually between 1.2-2.0. Finally, the direct losses and indirect losses are added together to obtain the total expected economic loss. The purpose of this step is to quantify the potential economic impact caused by sea level rise and provide decision makers with specific economic loss estimates. Through the application of the random forest algorithm, complex nonlinear relationships can be captured and the accuracy of loss prediction can be improved. Combining scenario analysis and expected value calculation, it not only takes into account the impact of different sea level rise scenarios, but also provides a comprehensive economic loss assessment result.

[0023] The specific implementation of step S60 is to construct an ecosystem service value matrix, calculate a biodiversity index matrix, and comprehensively evaluate the impact on the ecosystem. First, construct an ecosystem service value matrix ESV, where Represents the service value of the ith ecosystem under the jth sea level rise scenario. Ecosystem types may include coastal ecosystems such as mangroves, salt marshes, coral reefs, and seagrass beds. Ecosystem services may include regulating services (such as carbon fixation and coastal protection), provisioning services (such as fishery resources), cultural services (such as ecotourism), etc. The value of services can be quantified by the benefit transfer method or the direct market valuation method. For services that are difficult to monetize, a relative value index can be used. For example, for carbon fixation services, the value can be calculated based on carbon sinks and carbon prices; for coastal protection services, it can be estimated based on the substitution cost method. Under different sea level rise scenarios, changes in the value of ecosystem services can be estimated by changes in ecosystem area and changes in service value per unit area. Next, the Shannon-Wiener biodiversity index matrix H is calculated, and the calculation formula is: , where S is the number of species, is the proportion of individuals of the i-th species. The calculation of the biodiversity index requires detailed species survey data, which can be obtained through field surveys, remote sensing analysis, and literature reviews. For areas where it is difficult to conduct a comprehensive survey, indicator species or key functional groups can be selected for assessment. Under different sea level rise scenarios, species composition and abundance may change, and predictions need to be made based on the species' response characteristics to sea level rise. The prediction method can use an ecological niche model or a qualitative analysis based on expert knowledge. Then, the ecological vulnerability index matrix is ​​constructed. ,in represents the matrix Hadamard product, represents the matrix Hadamard division, E is the exposure matrix (obtained from step S40), S is the sensitivity matrix, A is the adaptability matrix, and R is the resilience matrix. The sensitivity matrix S can be constructed based on the tolerance range of species or ecosystems to sea level rise. For example, mangroves are more tolerant to sea level rise than salt marshes. The adaptability matrix A can take into account factors such as the migration ability of species and the spatial connectivity of ecosystems. The resilience matrix R can be constructed based on factors such as the recovery rate of ecosystems and the status of seed banks. The element values ​​of these matrices can be determined by expert scoring or statistical analysis based on historical data, and the value range is usually 0-1. Each element of the ecological vulnerability index matrix It represents the ecological vulnerability score of the jth ecosystem or region under the i-th sea level rise scenario. In order to comprehensively assess the impact of the ecosystem, the ecosystem service value, biodiversity index and ecological vulnerability index can be combined to construct a comprehensive ecological impact index matrix CEI: ,in , , is the weight coefficient, which can be determined by expert scoring or hierarchical analysis method and meets The weight setting should reflect the importance that decision makers attach to different ecological indicators. For example, , , . Finally, based on the comprehensive ecological impact index matrix CEI, ecological hotspots and key vulnerable ecosystems can be identified to provide a basis for the formulation of ecological protection and adaptation strategies. The purpose of this step is to comprehensively assess the impact of sea level rise on ecosystems, taking into account three key aspects: ecosystem service value, biodiversity, and ecological vulnerability. Through matrix operations and index synthesis, the comprehensiveness of the assessment is guaranteed, and intuitive quantitative results are provided, which helps decision makers understand and compare the ecological impacts in different regions and under different scenarios.

[0024] The specific implementation of step S70 is to construct a fuzzy relationship matrix, calculate the weights using the entropy weight method, perform fuzzy comprehensive evaluation, and obtain a comprehensive evaluation result. First, construct a fuzzy relationship matrix ,in Indicates the membership of the i-th evaluation factor to the j-th comment. Evaluation factors may include economic losses, ecological impacts, social impacts, etc., which can be extracted from the results of the previous steps. The comment set can be set to five levels: "very low", "low", "medium", "high", and "very high". The membership can be determined by expert scoring or membership function based on data distribution. For example, for the economic loss factor, the membership function can be set based on the loss amount: , where a and b are thresholds, which can be determined based on historical data or expert judgment. For each evaluation factor, it is necessary to construct a membership function corresponding to all comments. Then, the weight vector W is calculated using the entropy weight method. The entropy weight method is an objective weighting method that can determine the weight according to the amount of information in the indicator. First, calculate the information entropy of the jth indicator: ,in , . Then calculate the information utility value: . Finally, calculate the weight: The advantage of the entropy weight method is that it can reduce the influence of subjective factors, but it should also be noted that if the variability of some important indicators is small, they may get lower weights. In this case, you can consider combining subjective weighting methods, such as the Delphi method, to adjust the weights. Then, perform fuzzy synthesis operations to obtain comprehensive evaluation results. ,in Represents the fuzzy synthesis operator. The fuzzy synthesis operator can be selected as weighted average type (M(·,+)) or main factor determination type (M(·,∧)). The calculation formula of weighted average type is: , the calculation formula of the main factor determination type is: The operator to be selected depends on the purpose of the evaluation. The weighted average type can comprehensively consider the influence of all factors, while the main factor determination type emphasizes the main influencing factors. After obtaining the B vector, the maximum membership method can be used to determine the final evaluation level, that is, the comment with the largest membership degree is selected as the comprehensive evaluation result. In addition, the fuzzy comprehensive evaluation index can also be calculated: ,in is the score corresponding to the jth comment (for example, the scores corresponding to "very low" to "very high" can be set to 1, 3, 5, 7, 9). The index provides a continuous score, which helps to make quantitative comparisons between different schemes or regions. Finally, based on the fuzzy comprehensive evaluation results, a risk level distribution map can be drawn to identify high-risk areas and analyze the main factors leading to high risks. The purpose of this step is to combine the various quantitative and qualitative indicators obtained in the previous steps to obtain an overall risk assessment result. Through the application of fuzzy theory, the uncertainty and ambiguity in the evaluation process can be effectively handled to provide more practical evaluation results. The advantage of the fuzzy comprehensive evaluation method is that it can comprehensively consider multiple evaluation indicators and handle the fuzzy relationship between indicators through the concept of membership, so as to obtain a comprehensive evaluation result.

[0025] The specific implementation of step S80 is to construct a risk index matrix, extract the main risk feature vectors using singular value decomposition, and establish a risk prediction model using support vector regression. First, construct a risk index matrix ,in , , is the balance factor, , , is the corresponding weight matrix, is the economic loss expectation matrix (obtained from step S50), is the ecological vulnerability index matrix (obtained from step S60), and B is the fuzzy comprehensive evaluation result (obtained from step S70). The setting of the balance factor reflects the relative importance that decision makers attach to economic, ecological and comprehensive impacts, which can be determined through expert consultation or policy analysis. For example, , , The weight matrix can be determined based on the importance and reliability of each indicator, for example, it can be constructed using the analytic hierarchy process (AHP). Each element of the risk index matrix RI represents the comprehensive risk index of the jth region or asset under the i-th sea level rise scenario. Then, the singular value decomposition (SVD) method is used to decompose the risk index matrix RI and extract the main risk feature vector. The SVD decomposition formula is: , where U and V are the left and right singular vector matrices, is a diagonal matrix of singular values. The singular values ​​are arranged from large to small, reflecting the importance of the corresponding eigenvectors. The first k largest singular values ​​and their corresponding singular vectors can be selected. The selection of k can be based on the cumulative variance explanation rate, and the threshold is preferably set to 90%. This can greatly reduce the data dimension while retaining the main risk feature information. Then, the support vector regression (SVR) method is used to establish a risk prediction model based on the extracted risk feature vectors. SVR is a powerful nonlinear regression method that can effectively handle high-dimensional data and complex nonlinear relationships. The objective function of SVR is: , where w is the weight vector and b is the bias term, and is a slack variable, and C is a penalty parameter. The constraints are: , , ,in is the kernel function, is the parameter of the insensitive loss function. The choice of kernel function has an important impact on the performance of the model. Commonly used kernel functions include linear kernel, polynomial kernel and radial basis function (RBF) kernel. In practice, RBF kernel usually performs better and its form is: ,in is the kernel parameter. The model's hyperparameters (such as C, , ) can be optimized by grid search combined with cross validation. To evaluate model performance, indicators such as root mean square error (RMSE), mean absolute error (MAE), and coefficient of determination (R²) can be used. In addition, leave-one-out cross validation (LOOCV) can be used to evaluate the generalization ability of the model. Finally, based on the trained SVR model, the risks under different sea level rise scenarios can be predicted. The prediction results can be used to generate risk maps, identify high-risk areas, and analyze the contribution of different factors to the risk. The purpose of this step is to establish a model that can accurately predict the risk of sea level rise and provide decision makers with a scientific risk assessment tool. Through the application of SVD, effective feature extraction and dimensionality reduction are achieved; while the SVR model can capture complex nonlinear relationships and provide accurate risk predictions. The advantage of this method is that it can handle high-dimensional data, adapt to risk assessment needs at different scales, and can update the model and predict risks based on new data.

[0026] The specific implementation method of step S90 is to optimize the parameters of the risk prediction model using the particle swarm optimization algorithm, generate the final risk assessment results, and formulate corresponding adaptation strategies and mitigation measures. First, the parameters of the risk prediction model are optimized using the particle swarm optimization algorithm (PSO). PSO is a heuristic optimization algorithm that can effectively handle complex nonlinear optimization problems. The particle update formula is: , ,in is the particle velocity, is the particle position, w is the inertia weight, and is the acceleration constant, and is a random number in the interval [0,1]. is the individual optimal position, and g is the global optimal position. A particle represents a set of candidate values ​​for model parameters. For example, for the SVR model, a particle can include a penalty parameter C, an insensitive loss function parameter and kernel function parameters The size of the particle swarm is usually set to 20-50, and the number of iterations can be set to 100-500, depending on the complexity of the problem and the computing resources. The inertia weight w can be set to a linear decrease strategy, with the initial value set to 0.9 and the final value set to 0.4. Acceleration constant and It is usually set to 2. The optimization objective function can be the prediction error of the model (such as RMSE) or the cross-validation score. In order to avoid overfitting, k-fold cross-validation (k is usually set to 5 or 10) can be used to evaluate the model performance. The advantage of the PSO algorithm is that it can quickly converge to the global optimal solution and is suitable for dealing with optimization problems in continuous parameter space. Then, based on the optimized risk prediction model, the final sea level rise impact risk assessment results are generated. Optionally, the risk assessment results may include one or more of the following aspects: (1) Spatial distribution of risk: generate risk maps to show the risk levels of different regions, which can be visualized using GIS technology; (2) Time series risk prediction: predict the risks at different time points in the future (such as 2030, 2050, and 2100) and analyze the evolution trend of risks; (3) Scenario analysis: based on different sea level rise scenarios (such as IPCC's RCP2.6, RCP4.5, and RCP8.5 scenarios), generate corresponding risk assessment results and analyze the risk differences under different scenarios; (4) Sensitivity analysis: by changing the model input parameters, analyze the impact of different factors on the risk and identify key risk drivers; (5) Uncertainty analysis: using the Monte Carlo simulation method to generate confidence intervals for risk prediction and evaluate the reliability of the prediction results. The purpose of this step is to improve the performance of the risk prediction model through optimization algorithms, generate comprehensive and reliable risk assessment results, and formulate scientific adaptation strategies and mitigation measures based on these results. Through the application of the PSO algorithm, model parameters can be effectively optimized and prediction accuracy can be improved; the multi-dimensional risk assessment results and dynamic update mechanism provide decision makers with comprehensive and timely decision support, which helps to formulate more effective strategies to cope with sea level rise.

[0027] In order to better understand and implement the present invention, a specific embodiment 1 of the method provided by the first aspect of the present invention is provided below. The specific implementation of each step in this embodiment 1 is described in detail as follows: The specific implementation of step S10 is as follows: First, collect historical data on sea level rise in the assessment area, including relevant parameters such as tide level, sea water temperature, salinity, etc. The default time span of data collection is no less than 30 years to ensure the representativeness and reliability of the data.

[0028] Next, the collected historical data is decomposed into basic data and change data using the wavelet decomposition algorithm. Wavelet decomposition uses discrete wavelet transform (DWT) and selects appropriate wavelet basis functions (such as Daubechies wavelet) for multi-scale decomposition. The decomposition process can be expressed as: , in is the original signal, For the Layer approximation coefficient (basic data), For the Layer detail coefficient (variable data). Decomposition layer number The default setting is 3-5 layers to balance computational efficiency and analysis accuracy.

[0029] Then, the autoregressive integrated moving average (ARIMA) model in time series analysis is used to combine the basic data and change data obtained to generate future prediction data for sea level rise. The expression of the ARIMA model is: , in is the lag operator, is an autoregressive polynomial, is the difference operator, is the moving average polynomial, is white noise. The order of the model is It can be determined by autocorrelation function (ACF) and partial autocorrelation function (PACF) analysis, usually and The value range is 0-2. The value range is 0-1.

[0030] The model parameters can be estimated using the maximum likelihood estimation method: , in is the likelihood function, is the sample size, is the white noise variance.

[0031] The forecast time range defaults to the next 50-100 years to meet long-term planning needs. The forecast results can be used in the subsequent risk assessment process.

[0032] The specific implementation of step S20 is as follows: First, obtain the coastline elevation data in the assessment area, including land elevation and seabed topography data. The data resolution is no less than 30 meters by default to ensure the accuracy of the model.

[0033] Next, the Kriging interpolation method is used to establish the coastline digital elevation model. The basic formula of Kriging interpolation is: , in is the estimated value of the unknown point, is the observed value of the known point, is the weight coefficient. The weight coefficient is determined based on the variance function analysis. The variance function model can be a spherical model, an exponential model, or a Gaussian model. Taking the spherical model as an example, its expression is: , in is the variation function value, is the distance between sampling points, is the nugget value, is the base value, The interpolation grid size is set to 10-30 meters by default to balance computational efficiency and model accuracy.

[0034] Then, tide monitoring stations and differential GPS surface subsidence monitoring equipment are deployed in the assessment area to obtain tide data and surface subsidence data in real time. The default density of tide monitoring stations is one for every 50-100 kilometers of coastline, and the default density of differential GPS monitoring points is one for every 10-20 square kilometers.

[0035] Finally, the real-time monitoring data is combined with the digital elevation model, and the sliding time window method is used to update the model. The time window size is set to 3-6 months by default, and the update formula can be expressed as: , in is the updated digital elevation model. is the model at the previous moment, The model generated for the new observation data, is the smoothing coefficient, and its value range is [0,1].

[0036] The specific implementation of step S30 is as follows: First, the analytic hierarchy process (AHP) is used to identify key infrastructure and vulnerable assets in the assessment area and establish an asset value database. The basic steps of AHP include: building a hierarchy, constructing a judgment matrix, calculating a weight vector, and consistency testing. The judgment matrix is ​​constructed using a 1-9 scale, and the weight vector can be calculated using the eigenvalue method: , in is the judgment matrix, is the weight vector, is the maximum eigenvalue. The calculation formula of consistency ratio (CR) is: , , in is the consistency indicator, is the random consistency indicator, is the matrix order. The CR threshold is set to 0.1 by default.

[0037] Then, principal component analysis (PCA) was used to reduce the dimension of the factors affecting sea level rise and construct an influencing factor matrix. The basic steps of PCA include: data standardization, calculation of covariance matrix, calculation of eigenvalues ​​and eigenvectors, and selection of principal components. The standardization formula is: , in is the standardized value, is the original value, is the mean, The selection of principal components is based on the cumulative variance contribution rate, and the default threshold is 85%.

[0038] Then, the singular value decomposition (SVD) method is used to analyze the influencing factor matrix Decompose it and get: , in and are the left and right singular vector matrices, is a diagonal matrix of singular values. The selection of singular values ​​can be based on the cumulative energy ratio of the singular values, and the default threshold is set to 90%.

[0039] The specific implementation of step S40 is as follows: First, based on the SVD decomposition results, the sensitivity matrix of sea level rise is constructed. and vulnerability matrix .

[0040] Then, the Monte Carlo simulation method was used to calculate the area of ​​the affected area under different sea level rise scenarios by combining the coastline digital elevation model, tide data and surface subsidence data to obtain the exposure matrix The basic steps of Monte Carlo simulation include: defining probability distribution, generating random numbers, calculating model output, and statistically analyzing the results. The number of simulations is set to 10,000 by default to ensure the stability of the results.

[0041] The exposure calculation formula can be expressed as: , in For the sea ​​level rise scenario The exposure of the area is the affected area, is the total area of ​​the region.

[0042] Finally, construct the comprehensive vulnerability index matrix ,in Represents the Hadamard product of matrices (that is, the multiplication of corresponding elements).

[0043] The specific implementation of step S50 is as follows: First, construct the asset loss matrix ,in Indicates The assets in Initial loss estimates can be based on a functional relationship between asset value and flooding severity, such as an exponential relationship: , in is the loss coefficient, which can be obtained by fitting historical data.

[0044] Next, the random forest algorithm is used to predict asset losses and update the loss matrix The basic principle of random forest is to construct multiple decision trees and take their average prediction results. The splitting criterion of the decision tree can be the Gini index: , in For Node Middle The proportion of class samples, is the number of categories. The number of trees in the random forest is set to 500-1000 by default, and the minimum number of samples per node is set to 1% of the total sample size by default.

[0045] Then, calculate the economic loss expectation matrix ,in is the probability vector of sea level rise scenario. The economic impact multiplier can be introduced to consider indirect economic losses, and the final economic loss expectation can be expressed as: , in is the expected total economic loss, It is the economic impact multiplier matrix, and its value usually ranges from 1.2 to 2.0.

[0046] The specific implementation of step S60 is as follows: First, construct an ecosystem service value matrix ,in Indicates The ecosystem in The service value under each sea level rise scenario can be quantified using the benefit transfer method or the direct market valuation method.

[0047] Next, calculate the Shannon-Wiener biodiversity index matrix , the calculation formula is: , in is the number of species, For the The proportion of individuals of a species.

[0048] Then, the ecological vulnerability index matrix was constructed ,in represents the matrix Hadamard product, represents the matrix Hadamard division, For the adaptability matrix, is the resilience matrix. The element values ​​of these matrices can be determined by expert scoring or statistical analysis based on historical data, and the value range is usually 0-1.

[0049] Finally, construct a comprehensive ecological impact index matrix : , in , , is the weight coefficient and satisfies .

[0050] The specific implementation of step S70 is as follows: First, construct the fuzzy relationship matrix ,in Indicates The evaluation factors The membership degree of each comment can be determined by expert scoring method or membership function method based on data distribution. For example, the triangular membership function can be expressed as: , in , , are the parameters of the membership function.

[0051] Next, the entropy weight method is used to calculate the weight vector First, calculate the Information entropy of indicators: , in , . Then calculate the information utility value: .

[0052] Finally calculate the weight: .

[0053] Then, fuzzy synthesis operation is performed to obtain the comprehensive evaluation result. ,in Represents the fuzzy synthesis operator. You can choose the weighted average type (M(·,+)) or the main factor determination type (M(·,∧)) operator. The calculation formula of the weighted average type is: , The calculation formula of the main factor determination type is: .

[0054] Finally, the fuzzy comprehensive evaluation index can be calculated: , in For the The score corresponding to each comment.

[0055] The specific implementation of step S80 is as follows: First, construct the risk index matrix : , in , , is the balance factor, , , is the corresponding weight matrix.

[0056] Then, the singular value decomposition (SVD) method is used to transform the risk index matrix Decompose and extract the main risk feature vector. The SVD decomposition formula is: , in and are the left and right singular vector matrices, is a diagonal matrix of singular values. The largest singular values ​​and their corresponding singular vectors, The choice can be based on the cumulative variance explained, with a default threshold of 90%.

[0057] Then, the support vector regression (SVR) method is used to establish a risk prediction model based on the extracted risk feature vector. The objective function of SVR is: , in is the weight vector, is the bias term, and is the slack variable, is the penalty parameter. The constraints are: , , , in is the kernel function, is the parameter of the insensitive loss function. Commonly used kernel functions include linear kernel, polynomial kernel and radial basis function (RBF) kernel. The form of RBF kernel is: , in is the kernel parameter.

[0058] The model's hyperparameters (such as , , ) can be optimized by grid search combined with cross validation. The indicators for evaluating model performance can use root mean square error (RMSE), mean absolute error (MAE) and determination coefficient (R²). The calculation formula of RMSE is: , in is the actual value, is the predicted value, is the number of samples.

[0059] The specific implementation of step S90 is as follows: First, the particle swarm optimization algorithm (PSO) is used to optimize the parameters of the risk prediction model. The particle update formula of PSO is: , , in is the particle velocity, is the particle position, is the inertia weight, and is the acceleration constant, and is a random number in the interval [0,1]. is the optimal position of an individual, is the global optimal position. The size of the particle swarm is usually set to 20-50, and the number of iterations can be set to 100-500. Inertia weight A linear reduction strategy can be adopted: , in is the initial inertia weight (usually 0.9), is the final inertia weight (usually 0.4), is the maximum number of iterations, is the current iteration number.

[0060] Then, based on the optimized risk prediction model, the final sea level rise impact risk assessment results are generated, including risk spatial distribution, time series risk prediction, scenario analysis, sensitivity analysis and uncertainty analysis.

[0061] For uncertainty analysis, the Monte Carlo simulation method can be used to generate a confidence interval for risk prediction. The calculation formula for the confidence interval is: , in is the sample mean, The degrees of freedom are of Distribution Quantile, is the sample standard deviation, is the sample size.

[0062] Through the above detailed mathematical models and algorithms, a comprehensive assessment and prediction of the risk of sea level rise can be achieved, providing decision makers with a scientific decision support tool. This method takes into account natural, social and economic factors, can adapt to the risk assessment needs of different scales and regions, and maintains the timeliness and accuracy of the model through a dynamic update mechanism.

[0063] A second aspect of the present invention provides a computer-readable storage medium, wherein the computer-readable storage medium stores program instructions, and when the program instructions are run in a computer, they are used to execute the above-mentioned method for assessing the risk of the impact of sea level rise.

[0064] A third aspect of the present invention provides a sea level rise impact risk assessment system, which includes the above-mentioned computer-readable storage medium.

[0065] Specifically, the principle of the present invention is: Based on multi-level and multi-dimensional system engineering theory, efficient and accurate risk assessment is achieved through algorithm optimization and model integration. The specific principles are explained as follows: At the data processing level, the principle of using wavelet decomposition algorithm to process historical data in this invention is that sea level change data has obvious multi-scale characteristics, including two main components: long-term trend and short-term fluctuation. Wavelet decomposition can analyze in both time and frequency domains, decomposing the original signal into different frequency components through wavelet basis functions of different scales, thereby effectively separating the basic trend and fluctuation characteristics. The mathematical basis of this decomposition method is: ; in, is the wavelet basis function, is the wavelet coefficient. This decomposition enables the subsequent ARIMA model to make more accurate predictions for trend terms and volatility terms respectively.

[0066] In terms of terrain characterization, the core principle of Kriging interpolation is based on variogram theory and minimum variance unbiased estimation. This method takes into account spatial autocorrelation and optimizes interpolation weights by building a spatial statistical model. The variance of the estimated value is minimal and there is no systematic bias. The theoretical basis of Kriging interpolation is: ; in, is the variance function, is the spatial distance, is the number of sample logarithms. This method can effectively handle the heterogeneity and anisotropy of terrain.

[0067] In terms of feature extraction and dimensionality reduction, the principle of using principal component analysis (PCA) combined with singular value decomposition (SVD) in this invention is that the factors affecting sea level rise have the characteristics of high dimension and high correlation. PCA converts the original features into linearly independent principal components through orthogonal transformation, while SVD provides an effective calculation method from an algebraic perspective. Its mathematical expression is: ; in, The singular values ​​in are arranged in descending order, and retaining the first k largest singular values ​​can achieve the optimal low-rank approximation, which not only reduces the data dimension but also retains the main information.

[0068] In terms of risk quantification, the multidimensional matrix system constructed by the present invention is based on the risk theory framework: Risk = Hazard × Vulnerability × Exposure. By performing Hadamard product operations on each matrix, an accurate combination of risk factors is achieved: ; in, Represents the matrix Hadamard product. This operation method ensures the physical meaning of risk assessment.

[0069] At the model optimization level, the principle of using the particle swarm optimization (PSO) algorithm to optimize the support vector regression (SVR) model is that the performance of the SVR model is highly dependent on the selection of its hyperparameters, and the PSO algorithm can quickly find the optimal solution in the high-dimensional parameter space by simulating the intelligent behavior of the group. Its iterative process follows: ; This optimization mechanism ensures that the model has high prediction accuracy while maintaining generalization ability.

[0070] In general, the technical solution of the present invention constructs a logically rigorous and hierarchical risk assessment system through system integration and optimized combination. Each module is closely connected through data flow and control flow to form a complete technical chain. This systematic approach not only ensures the scientificity and reliability of the assessment results, but also greatly improves the calculation efficiency.

[0071] In order to further better understand and implement the present invention, Example 2 of a specific application scenario of the present invention is provided below: This example selects a typical bay area in a coastal city as the evaluation object. The area is about 200 square kilometers and the coastline is about 45 kilometers long. It is a typical subtropical bay area. The area contains various functional areas such as important port facilities, chemical parks, residential areas and mangrove protection areas.

[0072] In step S10, the sea level observation data of the area from 2000 to 2023 are first collected. The data sources include the hourly observation records of the five tide stations set up in the area, and the data content includes parameters such as tide level, sea water temperature and salinity. The 23-year historical data are processed using the wavelet decomposition algorithm, and the db4 wavelet basis function is selected for 4-layer decomposition. The basic data obtained after decomposition show that the average annual sea level rise rate in the area is 4.2 mm per year. The change data show that due to the influence of the monsoon climate, the region has significant seasonal fluctuation characteristics with a fluctuation range of ±15 cm. Based on the decomposition results, the ARIMA (2,1,1) model is used for prediction. The prediction results show that by 2050, the sea level in the region will rise by 12.5 cm relative to 2023, as shown in Figure 2. Figure 2 The data processing at this stage only takes 30 minutes, which saves about 70% of the processing time compared to traditional methods.

[0073] In step S20, the elevation data of the coastline was obtained by UAV laser radar measurement with a resolution of 0.5 meters. The digital elevation model was constructed using the Kriging interpolation method, and the interpolation grid size was set to 5 meters × 5 meters. At the same time, 20 differential GPS monitoring points were deployed in the assessment area to monitor the surface subsidence in real time. The monitoring data showed that the average subsidence rate in the area was 2.1 mm per year, and the subsidence rate in the local area of ​​the chemical park reached 8.5 mm per year due to excessive groundwater exploitation. Figure 3 The construction process of the digital elevation model is completed automatically by computer, and it only takes 2 hours to generate an accurate terrain model covering the entire area.

[0074] In step S30, the key facilities in the area are identified and graded using the analytic hierarchy process. The evaluation results are shown in Table 1 below: Table 1 Facility evaluation results

[0075] The principal component analysis method was used to reduce the dimensionality of the influencing factors from the initial 28 to 8 main factors, with a cumulative contribution rate of 85.3%. The eigenvector was extracted through singular value decomposition, and the calculation process only took 15 minutes, greatly improving the calculation efficiency.

[0076] In step S40, the sensitivity and vulnerability matrix is ​​constructed based on the SVD decomposition results. The Monte Carlo method is used to perform 10,000 simulations, and three sea level rise scenarios are set: a low scenario (a rise of 10 cm), a medium scenario (a rise of 25 cm), and a high scenario (a rise of 50 cm). The simulation results show that under the medium scenario, about 12.3% of the area will be directly affected. The simulation calculation uses parallel computing technology, and all calculations can be completed in just 45 minutes on an 8-core processor.

[0077] In step S50, an asset loss matrix is ​​constructed. The random forest algorithm is used to predict the loss of various types of assets, and a random forest model containing 1,000 decision trees is established. The prediction results are as follows: Figure 4 As shown in Table 2 below: Table 2 Scenario loss table

[0078] In step S60, the ecosystem in the area is evaluated. There are 2.8 square kilometers of mangroves and 5.6 square kilometers of coastal wetlands in the area. Through field surveys and remote sensing image analysis, the service value of the main ecosystem types in the area is calculated. The results are as follows: Figure 5 As shown in Table 3 below, Figure 5 The vertical axis unit is area (km 2 ) / Service value (10,000 yuan / km 2 / Year).

[0079] Table 3 Ecological value table

[0080] In step S70, a fuzzy comprehensive evaluation method is used to evaluate the risk in multiple dimensions. An evaluation index system is established, including three dimensions: natural, social, and economic. The entropy weight method is used to calculate the weights, and the main index weights are shown in Table 4 below: Table 4 Multi-dimensional indicator weight table

[0081] In step S80, a risk index matrix is ​​constructed, and the balance factors α=0.4, β=0.35, and γ=0.25 are set. Grid computing technology is used for parallel processing, and the assessment area is divided into 2000 grid cells for risk calculation. The support vector regression model is trained using the cross-validation method, and the average prediction error of the model is 3.2%. The entire calculation process takes only 1.5 hours, which is 65% more efficient than the traditional method.

[0082] In step S90, the risk prediction model is optimized using the particle swarm optimization algorithm. The particle swarm size is set to 200, the maximum number of iterations is 1000, and the acceleration constant c1=c2=2.0. The performance of the optimized model is significantly improved, and the prediction accuracy is increased from 86.5% to 94.8%. The final risk assessment results are as follows: Figure 6 As shown in Table 5 below: Table 5 Risk Assessment Table

[0083] The implementation process of this embodiment shows that this method significantly improves the calculation efficiency while ensuring the evaluation accuracy. Figure 7 As shown in the figure. It is mainly reflected in: (1) Parallel computing technology is used in the data processing stage, which increases the processing speed by 70%. (2) The terrain model construction process is automated, saving a lot of manual operation time. (3) Grid computing technology is used in the risk calculation stage, which increases the computing efficiency by 65%. (4) The improved particle swarm algorithm is used in the model optimization stage, which increases the convergence speed by 50%. The entire evaluation process from data collection to final result output only takes 5 working days, while traditional methods usually take more than 1 month. At the same time, the resource consumption of this method is also greatly reduced, and only a medium-configuration workstation is needed to complete all computing tasks.

[0084] The above description is only a specific implementation mode of the present invention, but the protection scope of the present invention is not limited thereto. Any technician familiar with the technical field can easily think of changes or substitutions within the technical scope disclosed by the present invention, which should be covered by the protection scope of the present invention.

Claims

1. A method for assessing the risk of sea level rise, characterized in that: The following steps are involved: S10. Collect historical data on sea level rise, use wavelet decomposition algorithm to separate basic data and change data, and use ARIMA model to generate future forecast data; S20. Obtain the elevation data of the coastline, apply the Kriging interpolation method to establish a digital elevation model, and use monitoring equipment to obtain tide level and surface subsidence data in real time to achieve dynamic representation of the terrain; S30. Use the analytic hierarchy process to identify critical infrastructure and vulnerable assets, use principal component analysis to reduce the dimensionality of influencing factors, and extract key features through singular value decomposition to lay the foundation for risk assessment; S40. Based on the results of singular value decomposition, the sensitivity and vulnerability matrix is ​​constructed, and the affected area under different scenarios is calculated by Monte Carlo simulation method to obtain the exposure matrix and comprehensively evaluate the regional vulnerability. S50. Construct an asset loss matrix, use the random forest algorithm to predict asset losses, calculate the economic loss expectation matrix, and quantify the potential economic impact caused by sea level rise; S60. Construct an ecosystem service value matrix, calculate a biodiversity index matrix, comprehensively consider exposure, sensitivity and adaptive capacity, construct an ecological vulnerability index matrix, and assess the degree of impact on the ecosystem; S70, construct a fuzzy relationship matrix, calculate the weights using the entropy weight method, conduct fuzzy comprehensive evaluation, obtain comprehensive evaluation results, and provide multi-dimensional reference for the final risk assessment; S80, construct a risk index matrix, use singular value decomposition to extract the main risk feature vectors, use support vector regression to establish a risk prediction model, and achieve risk quantification and prediction; S90. Use the particle swarm optimization algorithm to optimize the risk prediction model parameters and generate the final risk assessment results.

2. A method for assessing the risk of sea level rise according to claim 1, characterized in that: The step S10 specifically includes: Step 101, collecting historical data of sea level rise in the assessment area, including tide level, sea water temperature, and salinity; Step 102: Decompose the historical data into basic data and change data using a wavelet decomposition algorithm, wherein the basic data reflects long-term trends and the change data reflects short-term fluctuations; Step 103: using the ARIMA model in time series analysis, combining the basic data and the change data, to generate future prediction data for sea level rise; The step S20 specifically includes: Step 201: Obtaining coastline elevation data within the assessment area, including land elevation and seabed topography data; Step 202: Apply Kriging interpolation method to establish a coastline digital elevation model; Step 203: using tide level monitoring stations and differential GPS surface settlement monitoring equipment deployed in the assessment area to obtain tide level data and surface settlement data in real time; The step S30 specifically includes: Step 301: Use the analytic hierarchy process to identify key infrastructure and vulnerable assets in the assessment area, and establish an asset value database; Step 302: Use principal component analysis to reduce the dimension of the factors affecting sea level rise and construct an influencing factor matrix; Step 303: Decompose the influencing factor matrix using a singular value decomposition method to obtain a decomposition result.

3. A method for assessing the risk of sea level rise according to claim 2, characterized in that: The step S40 specifically includes: Step 401: construct a sea level rise impact sensitivity matrix based on the decomposition results; Step 402: construct a vulnerability matrix to reflect the vulnerability of the region to sea level rise; Step 403: Using the Monte Carlo simulation method, combined with the coastline digital elevation model, tide data and surface subsidence data, the area of ​​the affected area under different sea level rise scenarios is calculated to obtain an exposure matrix.

4. A method for assessing the risk of sea level rise according to claim 3, characterized in that: The step S50 specifically includes: Step 501: construct an asset loss matrix to represent the losses of different assets under various sea level rise scenarios; Step 502: Use the random forest algorithm to predict asset losses and update the loss matrix; Step 503: Calculate the economic loss expectation matrix, multiply the loss matrix by the sea level rise scenario probability vector, and quantify the potential economic impact.

5. A method for assessing the risk of sea level rise according to claim 4, characterized in that: The step S60 specifically includes: Step 601: construct an ecosystem service value matrix to represent the service value of different ecosystems under various sea level rise scenarios; Step 602, calculating the Shannon-Wiener biodiversity index matrix to assess the regional biodiversity status; Step 603: Construct an ecological vulnerability index matrix, comprehensively consider exposure, sensitivity, adaptability and resilience, and assess the degree of impact on the ecosystem.

6. A method for assessing the risk of sea level rise according to claim 5, characterized in that: The step S70 specifically includes: Step 701: construct a fuzzy relationship matrix to represent the degree of membership of each evaluation factor to different comments; Step 702, using the entropy weight method to calculate the weight vector, taking into account the information entropy of each indicator; Step 703: Perform fuzzy synthesis operations to obtain comprehensive evaluation results, providing multi-dimensional references for risk assessment.

7. A method for assessing the risk of sea level rise according to claim 6, characterized in that: The step S80 specifically includes: Step 801, constructing a risk index matrix, integrating economic losses, ecological vulnerability and fuzzy comprehensive evaluation results; Step 802: Decompose the risk index matrix using the singular value decomposition method to extract the main risk feature vector; Step 803: Use support vector regression method to establish a risk prediction model based on the extracted risk feature vector.

8. A method for assessing the risk of sea level rise according to claim 7, characterized in that: The step S90 specifically includes: Step 901: Optimizing the parameters of the risk prediction model using a particle swarm optimization algorithm; Step 902: Generate a final sea level rise impact risk assessment result based on the optimized risk prediction model.

9. A computer-readable storage medium, characterized in that: The computer-readable storage medium stores program instructions, and when the program instructions are run in a computer, they are used to execute the sea level rise impact risk assessment method described in any one of claims 1-8.

10. A sea level rise risk assessment system, characterized in that: A computer-readable storage medium comprising the computer-readable storage medium of claim 9.

Images