Multi-disaster-type disaster risk grade coupling assessment method

Abstract

The invention discloses a multi-disaster disaster risk grade coupling evaluation method. The method comprises the following steps: establishing a multi-dimensional database covering multiple disaster types; key features related to the disaster risk are extracted from the multi-dimensional database; establishing a risk assessment model according to the key features, and determining a plurality of single-disaster risk levels; identifying association rules, causal relationships or space-time coupling modes among different disasters by using a big data mining algorithm; coupling weights among different disasters are determined, and a coupling model is constructed to convert a coupling relation into a quantitative risk superposition effect; based on the coupling model, a single-disaster risk and a coupling effect are integrated, a multi-disaster comprehensive risk level is generated, through multi-dimensional data integration, quantitative coupling modeling and a dynamic adjustment mechanism, advantages are formed in evaluation precision, social and economic adaptability and model adaptability, and the method is more suitable for scenes needing accurate risk grading and decision support.

Description

Technical Field

[0001] This invention belongs to the field of disaster information management technology, specifically relating to a method for coupled assessment of multi-hazard risk levels. Background Technology

[0002] Disaster risk assessment is a comprehensive, scientific, and accurate evaluation of the various types of disaster risks faced by a city, providing strong protection for the city's sustainable development and the safety of residents' lives and property.

[0003] Currently, disaster level assessment primarily involves collecting disaster-related natural and social data using big data mining algorithms, determining the weights of each indicator using a combination of the Analytic Hierarchy Process (AHP) and entropy weighting, and establishing a comprehensive risk assessment model to evaluate the overall disaster risk value. For example, Chinese patent application number 201611270445.7 discloses a method for assessing regional flood disaster risk at different scales. This method includes: collecting spatial data of the area to be analyzed and related attribute data as initial data; compressing the initial data and storing it in a created spatial database; standardizing the initial data and unifying units; using a coupled model of the AHP and entropy weighting to determine the combined weights of each factor; assessing the flood disaster risk value within the study area, classifying risk levels, and generating a flood disaster risk index map. This method uses various collected natural and social data as input, employs multiple algorithmic coupled models to determine weights, and finally generates a flood risk level evaluation index map, improving the accuracy of flood disaster risk assessment. It also has the advantages of objectivity and efficiency, and more intuitively displays the potential flood disaster risk to the study area under given conditions.

[0004] However, the aforementioned disaster risk assessment is specifically designed for flood disasters. Its assessment indicators, model construction, and risk level classification are all based on the characteristics of flood disasters, without considering data and indicators for multiple disaster types. This results in limitations in disaster risk assessment. Therefore, we need to propose a multi-hazard risk level coupling assessment method to solve the above-mentioned problems, enabling it to conduct multi-hazard risk level coupling assessments and improve the comprehensiveness of disaster risk assessment. Summary of the Invention

[0005] The purpose of this invention is to provide a method for coupled assessment of multiple disaster risk levels, which can perform coupled assessment of multiple disaster risk levels, improve the comprehensiveness of disaster risk assessment, and solve the problems mentioned in the background art.

[0006] To achieve the above objectives, the present invention adopts the following technical solution: A method for coupled assessment of multi-hazard risk levels includes the following steps: S1. Establish a multidimensional database covering multiple disaster types. Integrate multi-source data, including spatiotemporal data, historical disaster information, environmental parameters, and socioeconomic data, within the multidimensional database. Disaster types include earthquakes, floods, typhoons, and debris flows. S2. Extract key features related to disaster risk from a multidimensional database; S3. Establish a risk assessment model based on key characteristics and determine the risk levels of multiple single disaster types respectively; S4. Use big data mining algorithms to identify the correlation rules, causal relationships, or spatiotemporal coupling patterns among different disasters; S5. Based on the identified association rules, causal relationships, or spatiotemporal coupling patterns, determine the coupling weights between different disasters and construct a coupling model to transform the coupling relationship into a quantitative risk superposition effect; S6. Based on the coupling model, integrate single-hazard risk and coupling effect to generate a multi-hazard comprehensive risk level; S7. Verify the accuracy of the coupling model using historical disaster data, and dynamically adjust the parameters of the coupling model using real-time data.

[0007] Preferably, the multidimensional database adopts a spatiotemporal database or a distributed database that supports spatiotemporal indexing and multi-source data integration. The data format is unified through an ETL process, and the multi-source data is mapped to a unified field. The spatiotemporal data includes remote sensing imagery and GPS positioning data accessed through a GIS system; historical disaster data includes disaster records collected from government emergency management departments and meteorological stations; environmental parameters include topographic data, soil type, and river network distribution; and socioeconomic data includes population density, GDP distribution, and infrastructure data.

[0008] The preferred key feature extraction process is as follows: A1. Perform data cleaning and standardization on multi-source data; A2. Based on the disaster risk formation mechanism, the characteristics are divided into spatiotemporal characteristics, physical characteristics, disaster-bearing body characteristics, and environmental characteristics, and a classification and labeling system is established; A3. Extracting physical features based on domain knowledge; A4. Principal component analysis was used to reduce dimensionality and extract comprehensive features. A5. Combine physical features with comprehensive features to form a combined feature set.

[0009] Preferably, the risk assessment model includes an earthquake assessment model, a flood assessment model, a typhoon assessment model, and a debris flow assessment model. The combined feature set is input into the corresponding risk assessment model according to the type of disaster. The risk assessment model outputs the corresponding single-disaster risk level, namely earthquake risk level, flood risk level, typhoon risk level, and debris flow risk.

[0010] Preferably, the association rule identification employs a priori algorithm, specifically including the following steps: B1. Statistically analyze the frequency of occurrence of all single disaster events; B2. Set a support threshold to retain high-frequency single disasters; B3. Perform a join operation on the frequent p-itemsets to generate candidate p+1 itemsets, and delete all candidates that contain non-frequent p-item subsets to obtain candidate frequent p+1 itemsets. B4. Calculate the confidence score for frequent itemsets and filter valid association rules based on the set confidence score threshold.

[0011] Preferably, the causal relationships between different disasters are identified using the Peter-Clarke algorithm. The causal relationship identification process is as follows: C1. Define the disaster characteristic variables, treat all disaster characteristics as nodes, and initially connect all nodes in pairs to form a completely undirected graph; C2. Remove unrelated edges by passing the unconditional independence test; C3. Introduce a set of k intermediate variables, and perform conditional independence tests on each pair of adjacent nodes under a set of possible conditions of size k. C4. Use collider structures for edge orientation to determine the causal direction; C5. Apply the orientation rules to orient the remaining edges.

[0012] Preferably, the spatiotemporal coupling pattern recognition process is as follows: Each disaster event is represented as a spatiotemporal point with spatial location and timestamp. The spatiotemporal neighborhood radius is defined, that is, the spatial neighborhood radius is 10 kilometers and the temporal neighborhood radius is 24 hours. If the number of spatiotemporal points (including the point itself) contained within the spatiotemporal neighborhood radius of a spatiotemporal point is not less than the set minimum number of samples (MinPts), then the point is marked as a core point. Points that are directly density-reachable from the core point are assigned to the same cluster. Density-connected core points and their density-reachable spatiotemporal points are clustered into a cluster. Each cluster represents a spatiotemporal coupling pattern.

[0013] Preferably, when determining the coupling weights, based on historical data and the aforementioned association rules, causal relationships, and spatiotemporal coupling patterns, the relative importance of each disaster in triggering cascading disasters or amplifying overall risk is analyzed, and a hierarchical structure model is constructed. The top layer is the target layer, the middle layer is the criterion layer, and the bottom layer is the alternative layer. Based on the analysis results of the relative importance of each disaster in triggering cascading disasters or amplifying overall risk, experts or historical statistics compare the relative importance of each disaster pairwise under the criterion layer to form a judgment matrix. Then, the maximum eigenvalue and eigenvector of each judgment matrix are calculated, and after normalization, the weight of each disaster on the overall risk is obtained. Finally, a consistency check is performed. If it fails, the judgment matrix is ​​readjusted; if it passes, the weight of each disaster on the overall risk is the coupling weight.

[0014] Preferably, the constructed coupling model is as follows: , in, , , , These are the quantitative values ​​for the single-hazard risk level of earthquakes, floods, typhoons, and mudslides, respectively. Let be the coupling weight of disaster i to disaster j, and let A and B be adjustment factors, where A+B=1. This is the regional vulnerability amplification factor.

[0015] Preferably, the accuracy verification process for the coupled model is as follows: E1. Collect historical disaster data and perform data cleaning and dataset partitioning to form a historical verification set for verification. E2. Use the historical disaster data to train the coupled model, and use the trained coupled model to predict disaster losses on the validation set. E3. Calculate the mean square error between the predicted and actual values, as well as the prediction accuracy of the coupled model; E4. Compare the calculated mean square error and the prediction accuracy of the coupled model with the target threshold to determine whether the prediction accuracy of the coupled model meets the requirements of disaster risk assessment accuracy. If the prediction accuracy of the coupled model meets the requirements of disaster risk assessment accuracy, then adjust the coupling weight parameters based on the real-time data stream. If the prediction accuracy of the coupled model does not meet the requirements of disaster risk assessment accuracy, then the model structure or parameters need to be readjusted based on historical data, and steps E2-E4 need to be executed again.

[0016] The multi-hazard risk level coupling assessment method proposed in this invention has the following advantages compared with the prior art: 1. This invention integrates spatiotemporal data, historical disaster data, environmental parameters, and socioeconomic data through a multidimensional database, combining disaster risk with the vulnerability of disaster-bearing bodies. The assessment results are closer to the actual loss scenarios, providing a more accurate basis for resource allocation. It identifies disaster association rules through big data mining algorithms and introduces coupling weights when constructing the coupling model, transforming disaster causal relationships and spatiotemporal coupling patterns into quantitative risk superposition effects. The accuracy of the model is verified through historical disaster data, and the parameters of the coupling model are dynamically adjusted using real-time monitoring data. It clearly covers four typical disasters: earthquakes, floods, typhoons, and debris flows. It can design exclusive assessment indicators for the physical characteristics of each disaster. By incorporating socioeconomic data, it can directly assess the impact of disasters on people's livelihoods and industries, making the assessment results more in line with the needs of government emergency decision-making. Based on big data mining algorithms to identify disaster coupling patterns, it can automatically discover new coupling relationships not covered by traditional models, thereby enabling coupling assessments of multiple disaster risk levels and improving the comprehensiveness of disaster risk assessment. Attached Figure Description

[0017] Figure 1 A flowchart of a method for coupled assessment of multiple disaster risk levels according to an embodiment of the present invention is shown; Figure 2 A flowchart illustrating the key feature extraction process according to an embodiment of the present invention is shown; Figure 3 A flowchart illustrating the association rule identification process according to an embodiment of the present invention is shown; Figure 4 A flowchart illustrating the causal relationship identification process according to an embodiment of the present invention is shown. Detailed Implementation

[0018] 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, and not all embodiments. The specific embodiments described herein are merely used to explain the present invention and are not intended to limit the present invention. 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.

[0019] This invention provides, for example Figure 1-4 The multi-hazard risk level coupling assessment method shown includes the following steps: S1. Establish a multidimensional database covering multiple disaster types. Integrate multi-source data, including spatiotemporal data, historical disaster information, environmental parameters, and socioeconomic data, within the multidimensional database. Disaster types include earthquakes, floods, typhoons, and debris flows. The multidimensional database adopts a spatiotemporal database or distributed database that supports spatiotemporal indexing and multi-source data integration. It unifies the data format through the ETL process and maps the multi-source data to a unified field. The ETL process includes extraction, transformation and loading. Extraction is to obtain the raw data from the disaster data source, transformation is to map the data to the target field, and loading is to store the transformed data into the target database to ensure data consistency and availability.

[0020] Spatiotemporal data includes remote sensing images and GPS positioning data accessed through GIS systems, such as the location of the earthquake epicenter and the extent of flooding. Historical disaster data includes disaster records collected from government emergency management departments and weather stations, such as earthquake magnitude and typhoon path; Environmental parameters include topographic data, soil type, and river network distribution; Socioeconomic data includes population density, GDP distribution, and infrastructure data, such as the location of hospitals and schools; S2. Extract key features related to disaster risk from a multidimensional database; The key feature extraction process is as follows: A1. Perform data cleaning and standardization on multi-source data; During data cleaning, K-nearest neighbor interpolation was used for spatiotemporal data, and mean filling was used for socioeconomic data. Outliers in physical characteristics were identified using box plots. The criteria filter data that exceeds three standard deviations. A2. Based on the disaster risk formation mechanism, the characteristics are divided into spatiotemporal characteristics, physical characteristics, disaster-bearing body characteristics, and environmental characteristics, and a classification and labeling system is established; Spatiotemporal characteristics include the longitude, latitude, and time window of the disaster occurrence, such as the longitude and latitude of the disaster point, the distance from the fault / river channel / coastline, and the time window such as the month of the disaster and seasonal periodicity. Physical characteristics include earthquake magnitude, focal depth, peak ground acceleration, flood inundation depth, flow velocity, return period, typhoon wind speed, wind circle radius, central air pressure, debris flow volume, and watershed slope; Characteristics of disaster-bearing bodies include population density, building seismic resistance level, and economic output density; Environmental characteristics include topographic slope, vegetation cover, and soil moisture content; A3. Extract physical features based on domain knowledge; according to disaster theory, experts define core physical indicators to extract flood risk, earthquake risk and debris flow risk. Flood risk extracts catchment area, river channel curvature coefficient and levee height; earthquake risk extracts fault type, site soil type and building site category; debris flow risk extracts watershed elevation difference, loose material reserves and rainfall intensity threshold. A4. Principal component analysis (PCA) is used for dimensionality reduction and comprehensive feature extraction. The formula for PCA is: , in, For the first Principal components, For the original features of multi-source data, To determine the contribution of each original feature to the nth principal component; A5, combine physical features with comprehensive features to form a combined feature set; S3. Establish a risk assessment model based on key characteristics and determine the risk levels of multiple single disaster types respectively; The risk assessment models include earthquake assessment models, flood assessment models, typhoon assessment models, and debris flow assessment models. Based on the type of disaster, a set of combined features is input into the corresponding risk assessment model. The risk assessment model outputs the corresponding single-hazard risk level, namely, earthquake risk level, flood risk level, typhoon risk level, and debris flow risk. The formula for the earthquake assessment model is as follows: Where, φ is the seismic risk attenuation rate, C is the regional coefficient, M is the earthquake magnitude, R is the distance from the epicenter, a is the magnitude influence parameter, used to describe the nonlinear influence of magnitude on earthquake intensity, and b is the distance attenuation parameter, used to control the attenuation rate of ground motion intensity with increasing distance; the formula expression for the flood assessment model is: Where L is the proportion of economic losses caused by the disaster to the total value of the disaster-bearing body, and D is the vertical height of the disaster-bearing body submerged by the flood. To reflect the foundation loss rate per unit inundation depth, Let be a depth-related nonlinear influence parameter, describing the rate at which the loss rate changes with flooding depth. The loss rate increases rapidly with increasing depth. The loss rate is linearly related to depth. The rate of loss is increasing at a slower pace; the formula for the typhoon assessment model is: ,in, The wind speed at a distance r from the center of the typhoon is... is the maximum wind speed at the center of the typhoon, and k is the wind speed attenuation coefficient, which is used to control the rate at which the wind speed decreases with increasing distance. The formula for the debris flow assessment model is as follows: In this context, 𝐸𝐼 is the debris flow susceptibility index, with a higher value indicating a greater likelihood of debris flow occurrence; 𝑤𝐸 is the weight of the Eth influencing factor, reflecting the relative importance of this factor to debris flow formation; and 𝑓𝐸 is the standardized value of the Eth influencing factor, reflecting the degree to which the actual state of this factor contributes to debris flow formation.

[0021] S4. Use big data mining algorithms to identify the correlation rules, causal relationships, or spatiotemporal coupling patterns among different disasters; Association rule identification employs a priori algorithms, specifically including the following steps: B1. Statistically analyze the frequency of occurrence of all single disaster events; B2. Set a support threshold to retain high-frequency single disasters; B3. Perform a join operation on the frequent p-itemsets to generate candidate p+1 itemsets, and delete all candidates that contain non-frequent p-item subsets to obtain candidate frequent p+1 itemsets. B4. Calculate the confidence score for frequent itemsets and filter valid association rules based on the set confidence score threshold; The formula for calculating valid rules is as follows:

[0022] Here, A and B represent two disaster types, such as A being a typhoon and B being a rainstorm. Let be the probability of disaster type A and disaster type B occurring simultaneously; let be the number of records in the historical disaster data that simultaneously meet the conditions for disaster type A and disaster type B; and let be the total sample size of the historical disaster data used for analysis. The probability of disaster type A occurring. The correlation between disaster type A and disaster type B. The frequency of occurrence of disaster type A alone. The frequency of occurrence of disaster type B alone; When Lift(A→B)>1, it means that disaster type A is positively correlated with disaster type B; When Lift(A→B)=1, it means that disaster type A and disaster type B are independent and unrelated. When Lift(A→B) < 1, it indicates that disaster type A is negatively correlated with disaster type B. For example, suppose there are 1000 historical disaster records. Among them, there are 200 records of typhoons (A) with Support(A) = 20%, 300 records of rainstorms (B) with Support(B) = 30%, and 150 records of both typhoons and rainstorms with Support(A∪B) = 15%. Then the support (A→B) = 15% / 100% = 15%, which means that 15% of the disasters include both typhoons and rainstorms. Confidence(A→B)=15% / 20%=75%, meaning that there is a 75% probability that a typhoon will be accompanied by heavy rain. Lift(A→B)=75% / 30%=2.5>1, indicating a strong positive correlation between typhoons and heavy rain; The causal relationships between different disasters are identified using the Peter-Clarke algorithm. The causal relationship identification process is as follows: C1. Define the disaster characteristic variables, namely H = earthquake magnitude, I = landslide area, C = rainfall, and D = vegetation coverage. Treat all disaster characteristics as nodes, and initially connect all nodes in pairs to form a completely undirected graph. C2. Remove unrelated edges by performing an unconditional independence test. Set the nodes to (X,Y). For each pair of nodes (X,Y), test whether they are unconditionally independent (i.e., without considering any intermediate variables). If they are independent, delete the edge between X and Y. If they are not independent, keep the edge between X and Y. For example, test whether earthquake magnitude (H) and vegetation cover (D) are independent; if they are independent, delete edge HD. C3. Introduce a set of k intermediate variables. For each pair of adjacent nodes, perform conditional independence tests on the possible condition set of size k. That is, for the remaining edge (X,Y), introduce a set of k intermediate variables S in turn to test whether X and Y are conditionally independent given S. If there exists an S that makes X and Y conditionally independent, delete the edge XY and record the set S (for subsequent orientation). k starts from 1 and increments until all possible combinations of intermediate variables have been tested. For example, test whether the earthquake magnitude (H) and landslide area (I) are independent for a given amount of rainfall (C), and if they are independent, delete edge HI; C4. Use a collider structure for edge orientation. The collider structure is a v-structure. Determine the causal direction. If there are three nodes XYZ, and X and Z are not directly connected, but both X and Z are connected to Y, and there is no set S such that X and Z are independent given S, then Y is a collision node of the v-structure, and the edge orientation is X→Y←Z. For example, if H and C are not directly connected in the earthquake magnitude (H) - landslide area (I) - rainfall (C), but H→I←C, then the landslide area (I) is a result of both the earthquake and the rainstorm (v-structure); C5, apply the orientation rules to orient the direction of the remaining edges. Using the conditional independence set S recorded in step C2, if X and Y are independent given S, and S contains the neighbor nodes of Y, then the edge orientation is X→Y or Y→X. This needs to be combined with prior knowledge such as time order. Time order constraint: In the causal relationship of disasters, the cause usually precedes the effect. For example, if the rainstorm occurs before the flood, the direction of the edge can be fixed accordingly, such as (C→I); The process of spatiotemporal coupled pattern recognition is as follows: Each disaster event is represented as a spatiotemporal point with spatial location and timestamp. The spatiotemporal neighborhood radius is defined as follows: the spatial neighborhood radius is 10 kilometers and the temporal neighborhood radius is 24 hours. If the number of spatiotemporal points (including the point itself) contained within the spatiotemporal neighborhood radius of a spatiotemporal point is not less than the set minimum number of samples (MinPts), then the point is marked as a core point. Points that are directly density reachable from the core point are assigned to the same cluster. Density-connected core points and their density-reachable spatiotemporal points are clustered into a cluster. Each cluster represents a spatiotemporal coupled pattern, such as continuous rainstorms and floods occurring in the same area within 24 hours. S5. Based on the identified association rules, causal relationships, or spatiotemporal coupling patterns, determine the coupling weights between different disasters, construct a coupling model to transform the coupling relationship into a quantitative risk superposition effect, and avoid assessment bias caused by simple superposition. When determining the coupling weights, based on historical data and the aforementioned association rules, causal relationships, and spatiotemporal coupling patterns, the relative importance of each disaster in triggering cascading disasters or amplifying overall risk is analyzed, constructing a hierarchical model. The top layer is the target layer (multi-hazard integrated risk), the middle layer is the criterion layer (inter-hazard interaction intensity, historical loss ratio), and the bottom layer is the scenario layer (earthquake, flood, typhoon, debris flow). Based on the analysis results of the relative importance of each disaster in triggering cascading disasters or amplifying overall risk, experts or historical statistics compare the relative importance of each disaster pairwise under the criterion layer to form a judgment matrix. Then, the maximum eigenvalue and eigenvector of each judgment matrix are calculated, normalized, and the weight of each disaster on the integrated risk is obtained. Finally, a consistency test is performed (if the combination confidence is less than 0.1, the verification passes). If it fails, the judgment matrix is ​​readjusted; if it passes, the weight of each disaster on the integrated risk is the coupling weight. The constructed coupling model is as follows: , in, , , , These are the quantitative values ​​for the single-hazard risk level of earthquakes, floods, typhoons, and mudslides, respectively. Let be the coupling weight of disaster i to disaster j, and let A and B be adjustment factors, where A+B=1. The parameters A, B, and C of the coupled model are used to amplify the regional vulnerability. The initial value is determined based on historical case regression fitting and is dynamically adjusted during the validation and optimization of the coupled model; S6. Based on the coupling model, integrate single-hazard risk and coupling effect to generate a multi-hazard comprehensive risk level; The integrated formula for single-hazard risk and coupling effects is:

[0023] in, For the comprehensive risk value, For single disaster type weighting, This is a coupling effect. These are the coupling weight coefficients. For the kth type of single-hazard risk; Normalize to the [0,1] interval and map to risk level, such as 0-0.2 for low risk and 0.8-1 for extremely high risk; S7. Verify the accuracy of the coupled model through historical disaster data and dynamically adjust the parameters of the coupled model using real-time data.

[0024] The accuracy verification process for the coupled model is as follows: E1. Collect historical disaster data and perform data cleaning and dataset partitioning to form a historical verification set for verification. E2. Use the historical disaster data to train the coupled model, and use the trained coupled model to predict disaster losses on the validation set. E3. Calculate the mean square error between the predicted and actual values, as well as the prediction accuracy of the coupled model; The formula for calculating the mean square error is: Where, 𝑀𝑆𝐸 is the mean squared error, l is the total number of historical disaster events, 𝑦𝑗 is the actual disaster loss of the j-th disaster event, and 𝑦𝑗 is the predicted disaster loss of the j-th disaster event by the coupled model; the formula for calculating the prediction accuracy of the coupled model is: Where P_j is the proportion of samples predicted as disaster type j that actually belong to that type, and R_j is the prediction accuracy among samples that actually belong to disaster type j. E4. Compare the calculated mean square error and the prediction accuracy of the coupled model with the target threshold to determine whether the prediction accuracy of the coupled model meets the requirements of disaster risk assessment accuracy. If the prediction accuracy of the coupled model meets the requirements of disaster risk assessment accuracy, then adjust the coupling weight parameters based on the real-time data stream. If the prediction accuracy of the coupled model does not meet the requirements of disaster risk assessment accuracy, then the model structure or parameters need to be readjusted based on historical data, and steps E2-E4 need to be executed again. The accuracy requirements for disaster risk assessment are determined with reference to national / industry norms and standards, such as disaster assessment-related standards issued by the Ministry of Emergency Management, the Ministry of Water Resources, and the Ministry of Natural Resources (e.g., the "Technical Specification for Natural Disaster Risk Assessment" and the "Technical Guidelines for the Compilation of Flood Risk Maps"), which clearly stipulate the accuracy requirements for different types of disasters (earthquakes, floods, typhoons, etc.). The target threshold is set by analyzing the distribution of prediction errors of similar models in historical disaster events (such as the mean and quantiles of the mean squared error). For example, if the median mean squared error of the historical coupled model is 50, the threshold can be set to 50 or adjusted based on this, such as relaxing it to 80 to adapt to new data. When the coupled model parameters are dynamically adjusted, data flow mining algorithms are used to receive new disaster data in real time and update the model parameters. When a new disaster combination occurs, the association rule base is updated by adjusting the parameters of the prior algorithm. By integrating spatiotemporal data, historical disaster data, environmental parameters, and socioeconomic data into a multidimensional database, disaster risk is combined with the vulnerability of disaster-bearing bodies. The assessment results are closer to actual loss scenarios, providing a more accurate basis for resource allocation. Big data mining algorithms identify disaster correlation rules, and coupling weights are introduced when constructing the coupled model. This transforms the causal relationships and spatiotemporal coupling patterns of disasters into quantitative risk superposition effects. For example, the coupling weight of typhoons and rainstorms is higher than that of earthquakes and debris flows, avoiding assessment bias and improving the scientific rigor of the coupled model. The accuracy of the model is verified through historical disaster data, and the parameters of the coupled model are dynamically adjusted using real-time monitoring data. For example, the correlation rules are updated based on newly occurring disaster cases to adapt to dynamic factors such as climate change and urban development. The system offers higher long-term prediction accuracy, clearly covering four typical disasters: earthquakes, floods, typhoons, and debris flows. It can design exclusive assessment indicators for the physical characteristics of each disaster, such as the inundation depth of floods and the wind speed radius of typhoons. It is more accurate in mapping the relationship between disaster mechanisms and risk levels. By incorporating socio-economic data, it can directly assess the impact of disasters on people's livelihoods and industries, such as the economic losses caused by factory shutdowns due to typhoons. The assessment results are more in line with the needs of government emergency decision-making, such as determining rescue priorities and post-disaster reconstruction budgets. Based on big data mining algorithms, it can identify disaster coupling patterns and automatically discover new coupling relationships not covered by traditional models, such as the potential correlation between extreme high temperatures and urban flooding. It adapts to the dynamic evolution of complex disaster scenarios. By integrating multi-dimensional data, using quantitative coupling modeling and dynamic adjustment mechanisms, it gains advantages in assessment accuracy, socioeconomic adaptability, and model self-adaptability, making it more suitable for scenarios requiring precise risk classification and decision support, such as urban emergency management and disaster prevention planning.

[0025] Finally, it should be noted that the above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. Although the present invention has been described in detail with reference to the foregoing embodiments, those skilled in the art can still modify the technical solutions described in the foregoing embodiments or make equivalent substitutions for some of the technical features. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the protection scope of the present invention.

Claims

1. A method for coupled assessment of multi-hazard risk levels, characterized in that: Includes the following steps: S1. Establish a multidimensional database covering multiple disaster types. Integrate multi-source data, including spatiotemporal data, historical disaster information, environmental parameters, and socioeconomic data, within the multidimensional database. Disaster types include earthquakes, floods, typhoons, and debris flows. S2. Extract key features related to disaster risk from a multidimensional database; S3. Establish a risk assessment model based on key characteristics and determine the risk levels of multiple single disaster types respectively; S4. Use big data mining algorithms to identify the correlation rules, causal relationships, or spatiotemporal coupling patterns among different disasters; S5. Based on the identified association rules, causal relationships, or spatiotemporal coupling patterns, determine the coupling weights between different disasters and construct a coupling model to transform the coupling relationship into a quantitative risk superposition effect; S6. Based on the coupling model, integrate single-hazard risk and coupling effect to generate a multi-hazard comprehensive risk level; S7. Verify the accuracy of the coupling model using historical disaster data, and dynamically adjust the parameters of the coupling model using real-time data.

2. The method for coupled assessment of multi-hazard risk levels according to claim 1, characterized in that: The multidimensional database adopts a spatiotemporal database or distributed database that supports spatiotemporal indexing and multi-source data integration. The data format is unified through the ETL process, and the multi-source data is mapped to a unified field. The spatiotemporal data includes remote sensing imagery and GPS positioning data accessed through the GIS system; historical disaster data includes disaster records collected from government emergency management departments and meteorological stations; environmental parameters include topographic data, soil type and river network distribution; and socioeconomic data includes population density, GDP distribution and infrastructure data.

3. The method for coupled assessment of multi-hazard risk levels according to claim 2, characterized in that: The key feature extraction process is as follows: A1. Perform data cleaning and standardization on multi-source data; A2. Based on the disaster risk formation mechanism, the characteristics are divided into spatiotemporal characteristics, physical characteristics, disaster-bearing body characteristics, and environmental characteristics, and a classification and labeling system is established; A3. Extracting physical features based on domain knowledge; A4. Principal component analysis was used to reduce dimensionality and extract comprehensive features. A5. Combine physical features with comprehensive features to form a combined feature set.

4. The method for coupled assessment of multi-hazard risk levels according to claim 3, characterized in that: The risk assessment model includes an earthquake assessment model, a flood assessment model, a typhoon assessment model, and a debris flow assessment model. Based on the type of disaster, the combined feature set is input into the corresponding risk assessment model, and the risk assessment model outputs the corresponding single-disaster risk level, namely earthquake risk level, flood risk level, typhoon risk level, and debris flow risk.

5. The method for coupled assessment of multi-hazard risk levels according to claim 4, characterized in that: Association rule identification employs a priori algorithms, specifically including the following steps: B1. Statistically analyze the frequency of occurrence of all single disaster events; B2. Set a support threshold to retain high-frequency single disasters; B3. Perform a join operation on the frequent p-itemsets to generate candidate p+1 itemsets, and delete all candidates that contain non-frequent p-item subsets to obtain candidate frequent p+1 itemsets. B4. Calculate the confidence score for frequent itemsets and filter valid association rules based on the set confidence score threshold.

6. The method for coupled assessment of multi-hazard risk levels according to claim 5, characterized in that: The causal relationships between different disasters are identified using the Peter-Clarke algorithm. The causal relationship identification process is as follows: C1. Define the disaster characteristic variables, treat all disaster characteristics as nodes, and initially connect all nodes in pairs to form a completely undirected graph; C2. Remove unrelated edges by passing the unconditional independence test; C3. Introduce a set of k intermediate variables, and perform conditional independence tests on each pair of adjacent nodes under a set of possible conditions of size k. If they are independent, remove the edge. C4. Use collider structures for edge orientation to determine the causal direction; C5. Apply the orientation rules to orient the remaining edges.

7. The method for coupled assessment of multi-hazard risk levels according to claim 6, characterized in that: The process of spatiotemporal coupled pattern recognition is as follows: Each disaster event is represented as a spatiotemporal point with spatial location and timestamp. The spatiotemporal neighborhood radius is defined as follows: the spatial neighborhood radius is 10 kilometers and the temporal neighborhood radius is 24 hours. If the number of spatiotemporal points contained within the spatiotemporal neighborhood radius of a spatiotemporal point is not less than the set minimum number of samples, then the point is marked as a core point. Points that are directly density reachable from the core point are assigned to the same cluster. Density-connected core points and their density-reachable spatiotemporal points are clustered into a cluster. Each cluster represents a spatiotemporal coupled pattern.

8. The method for coupled assessment of multi-hazard risk levels according to claim 7, characterized in that: When determining the coupling weights, based on historical data and the aforementioned association rules, causal relationships, and spatiotemporal coupling patterns, the relative importance of each disaster in triggering cascading disasters or amplifying overall risk is analyzed, and a hierarchical model is constructed. The top layer is the target layer, the middle layer is the criterion layer, and the bottom layer is the alternative layer. Based on the analysis results of the relative importance of each disaster in triggering cascading disasters or amplifying overall risk, experts or historical statistics compare the relative importance of each disaster pairwise under the criterion layer to form a judgment matrix. Then, the maximum eigenvalue and eigenvector of each judgment matrix are calculated, and after normalization, the weight of each disaster on the overall risk is obtained. Finally, a consistency check is performed. If it fails, the judgment matrix is ​​readjusted; if it passes, the weight of each disaster on the overall risk is the coupling weight.

9. The method for coupled assessment of multi-hazard risk levels according to claim 8, characterized in that: The constructed coupling model is as follows: ， in, , , , These are the quantitative values ​​for the single-hazard risk level of earthquakes, floods, typhoons, and mudslides, respectively. Let be the coupling weight of disaster i to disaster j, and let A and B be adjustment factors, where A+B=1. This is the regional vulnerability amplification factor.

10. The method for coupled assessment of multi-hazard risk levels according to claim 9, characterized in that: The accuracy verification process for the coupled model is as follows: E1. Collect historical disaster data and perform data cleaning and dataset partitioning to form a historical verification set for verification. E2. Use the historical disaster data to train the coupled model, and use the trained coupled model to predict disaster losses on the validation set. E3. Calculate the mean square error between the predicted and actual values, as well as the prediction accuracy of the coupled model; E4. Compare the calculated mean square error and the prediction accuracy of the coupled model with the target threshold to determine whether the prediction accuracy of the coupled model meets the requirements of disaster risk assessment accuracy. If the prediction accuracy of the coupled model meets the requirements of disaster risk assessment accuracy, then adjust the coupling weight parameters based on the real-time data stream. If the prediction accuracy of the coupled model does not meet the requirements of disaster risk assessment accuracy, then the model structure or parameters need to be readjusted based on historical data, and steps E2-E4 need to be executed again.

Images