A high-resolution seismic risk information analysis method and system

Abstract

The application discloses a high-resolution seismic risk information analysis method and system, and relates to the technical field of seismic risk analysis. The method comprises the following steps: collecting and preprocessing multi-source data of a target region to form a standardized multi-source data set; extracting multi-dimensional features for representing seismic risk by using the standardized multi-source data set; identifying a feature subset having essential causal correlation with the occurrence of an earthquake from the multi-dimensional features based on causal inference; determining a background seismic activity level based on historical earthquake records, and obtaining a confidence weight reflecting local cognition by quantifying the sparsity of available data and the complexity of geological structure at each spatial grid point; and combining dynamic information contained in the feature subset having essential causal correlation with the confidence weight reflecting local cognition. The application realizes fine analysis of the whole process of seismic risk from data fusion, feature selection, uncertainty quantification to dynamic evaluation.

Description

A high-resolution earthquake risk information analysis method and system Technical Field

[0001] This invention relates to the field of seismic hazard analysis technology, and in particular to a high-resolution seismic risk information analysis method and system. Background Technology

[0002] Probabilistic seismic hazard analysis (PSHA) is a core technology in modern earthquake risk quantification assessment. With advancements in observation technology, current high-resolution PSHA methods can integrate multi-source data from seismic, geological, and geodetic sources. Utilizing three-dimensional physical simulation and refined site effect analysis, they can generate seismic hazard zoning maps with spatial resolutions down to kilometers or even hundreds of meters, providing crucial information for seismic fortification of cities and major engineering projects. The core of this method lies in calculating the hyperprobability of seismic motion parameters at different locations by statistically analyzing historical seismic activity and combining it with seismic motion prediction models.

[0003] Existing high-resolution PSHA technology has a common limitation in the assessment of earthquake probability: it relies on the ergodic assumption, that is, using a uniform statistical model to extrapolate to each spatial grid point. It fails to explicitly distinguish and quantify the part of the uncertainty of earthquake occurrence that stems from insufficient understanding. Existing methods are unable to express the essential differences in the credibility of risk assessment caused by the differences in the degree of data support and the complexity of geological conditions at different spatial locations. Summary of the Invention

[0004] In view of the aforementioned existing problems, the present invention is proposed.

[0005] Therefore, this invention provides a high-resolution earthquake risk information analysis method to solve the problem that existing technologies, due to their reliance on ergodic assumptions, cannot accurately reflect the spatial differences in the reliability of risk assessment results.

[0006] To solve the above-mentioned technical problems, the present invention provides the following technical solution:

[0007] In a first aspect, the present invention provides a high-resolution earthquake risk information analysis method, which includes collecting and preprocessing multi-source data of a target area to form a standardized multi-source dataset;

[0008] By utilizing standardized multi-source datasets, multidimensional features for characterizing earthquake risk are extracted;

[0009] Based on causal inference, a subset of features that are essentially causally related to the occurrence of earthquakes is identified from multidimensional features;

[0010] Based on historical earthquake records, the background seismic activity level is determined. By quantifying the sparsity of available data and the complexity of geological structure at each spatial grid point, a confidence weight reflecting local cognition is obtained.

[0011] By combining the dynamic information contained in the feature subset of essential causal relationships with the confidence weights that reflect local cognition, the background seismic activity level is temporally regulated to obtain the grid-point differentiated earthquake occurrence probability.

[0012] Based on the grid-based differential earthquake occurrence probability, combined with the ground motion attenuation law and site effect, a probabilistic earthquake hazard analysis is conducted to obtain a high-resolution ground motion hazard zoning.

[0013] By integrating high-resolution seismic hazard zoning with information on disaster-bearing bodies, a high-resolution seismic risk map is generated.

[0014] As a preferred embodiment of the high-resolution seismic risk information analysis method of the present invention, the method includes the following steps: collecting and preprocessing multi-source data of the target area to form a standardized multi-source dataset:

[0015] Collect seismic catalog data, active fault data, crustal deformation data, and geophysical field data for the target area, and perform format conversion and coordinate system integration.

[0016] The seismic catalog data, active fault data, crustal deformation data, and geophysical field data that have completed format conversion and coordinate system one will be interpolated to a unified spatial grid and normalized.

[0017] The earthquake catalog data, active fault data, crustal deformation data, and geophysical field data that have undergone normalization processing are integrated into a standardized multi-source dataset.

[0018] As a preferred embodiment of the high-resolution earthquake risk information analysis method described in this invention, the method involves: extracting multi-dimensional features for characterizing earthquake risk using a standardized multi-source dataset, including the following steps:

[0019] Seismic activity parameters are calculated based on earthquake catalog data from standardized multi-source datasets.

[0020] Based on active fault data and crustal deformation data from standardized multi-source datasets, tectonic activity parameters are calculated.

[0021] Based on geophysical field data and active fault data from standardized multi-source datasets, crustal stress state parameters are calculated.

[0022] By combining seismic activity parameters, tectonic activity parameters, and crustal stress state parameters, a multidimensional feature is formed to characterize seismic risk.

[0023] As a preferred embodiment of the high-resolution earthquake risk information analysis method of the present invention, the method involves: identifying a subset of features that are essentially causally related to earthquake occurrence from multidimensional features based on causal inference, including the following steps:

[0024] By analyzing the prior knowledge dependencies among multidimensional features, earthquake occurrence, and potential confounding variables, a causal graph structure of multidimensional features, earthquake occurrence, and potential confounding variables is constructed.

[0025] Based on the aforementioned causal graph structure, a causal inference algorithm is applied to calculate the average treatment effect value of each feature in the multidimensional features on the occurrence of earthquakes.

[0026] The average treatment effect value was statistically analyzed, and a subset of features was selected from the multidimensional features to determine the feature subset that has an essential causal relationship with the occurrence of earthquakes.

[0027] As a preferred embodiment of the high-resolution earthquake risk information analysis method of the present invention, the determination of the background seismic activity level based on historical earthquake records includes the following steps:

[0028] A completeness analysis of historical earthquake records is conducted to determine the minimum complete magnitude of historical earthquake records;

[0029] By filtering historical earthquake records based on the minimum complete magnitude, a complete historical earthquake record is obtained.

[0030] A smoothing algorithm is used to calculate the earthquake occurrence rate of complete historical earthquake records on a spatial grid, which is then determined as the background seismic activity level.

[0031] As a preferred embodiment of the high-resolution seismic risk information analysis method described in this invention, the following steps are included: quantifying the sparsity of available data and the complexity of geological structures at each spatial grid point to obtain confidence weights reflecting local understanding.

[0032] The number of available seismic stations and engineering boreholes within a preset range around each spatial grid point is counted to obtain the sparsity of available data at each spatial grid point. The number of intersections of active faults and the number of changes in lithology within a preset range around each spatial grid point are calculated to obtain the complexity of the geological structure at each spatial grid point.

[0033] The sparsity of available data at each spatial grid point and the complexity of geological structure at each spatial grid point are normalized and weighted and fused to obtain the data-geological comprehensive heterogeneity index for each spatial grid point.

[0034] The data of each spatial grid point is inversely converted with the geological comprehensive heterogeneity index to obtain the confidence weight that reflects local cognition.

[0035] As a preferred embodiment of the high-resolution earthquake risk information analysis method of the present invention, the method combines the dynamic information contained in the feature subset with the confidence weight reflecting local cognition to temporally regulate the background seismic activity level and obtain the grid-point differentiated earthquake occurrence probability, including the following steps:

[0036] Extract the numerical changes of a subset of features representing essential causal relationships within a preset time window, as the dynamic information contained in the subset of features representing essential causal relationships;

[0037] Using the dynamic information contained in the feature subset of essential causal relationships as a correction factor, the background seismic activity level is corrected for the first time to obtain the preliminary corrected earthquake occurrence rate.

[0038] The confidence weight, which reflects local perception, is used as a modulation coefficient and applied to the initially revised earthquake incidence rate.

[0039] The earthquake occurrence rate, after being preliminarily corrected by confidence weights reflecting local cognition, is used to determine the grid-specific earthquake occurrence probability for each spatial grid point.

[0040] As a preferred embodiment of the high-resolution earthquake risk information analysis method described in this invention, the method includes the following steps: based on the grid-point differentiated earthquake occurrence probability, combined with the ground motion attenuation law and site effect, a probabilistic earthquake hazard analysis is performed to obtain a high-resolution ground motion hazard zoning.

[0041] Based on differentiating the earthquake occurrence probability at grid points, the exceedance probability of the ground motion intensity generated by the potential source at each field point is calculated.

[0042] The ground motion attenuation law is applied to the exceedance probability of ground motion intensity generated by potential sources at each field point, and the field effect is used as an amplification factor to correct the exceedance probability of ground motion intensity generated by potential sources at each field point.

[0043] By performing probability integration on the exceedance probability of ground motion intensity generated by potential seismic sources at each field point, the exceedance probability of ground motion parameters at each field point within a given period is calculated.

[0044] Spatially distribute the exceedance probability of ground motion parameters at each field point within a given time period to obtain a high-resolution ground motion hazard zoning.

[0045] As a preferred embodiment of the high-resolution earthquake risk information analysis method of the present invention, the high-resolution seismic hazard zoning and disaster-bearing body information are fused to generate a high-resolution earthquake risk map, including the following steps:

[0046] Spatial registration and attribute association are performed between the spatial probability distribution of ground motion intensity characterized by high-resolution ground motion hazard zoning and the physical exposure and vulnerability characteristics contained in the disaster-bearing body information.

[0047] By integrating the strength of seismic hazard, the value of disaster-bearing bodies, and their vulnerability in each spatial unit, a risk index is generated for each spatial unit.

[0048] The risk indicators of each spatial unit are spatially visualized to generate a high-resolution earthquake risk map.

[0049] Secondly, the present invention provides a high-resolution earthquake risk information analysis system, including a data processing module for collecting and preprocessing multi-source data of a target area to form a standardized multi-source dataset.

[0050] The feature extraction module uses standardized multi-source datasets to extract multi-dimensional features to characterize earthquake risk, and identifies a subset of features that are essentially causally related to earthquake occurrence from the multi-dimensional features.

[0051] The cognitive weight quantification module, based on historical earthquake records, determines the background seismic activity level and obtains confidence weights that reflect local cognition by quantifying the sparsity of available data and the complexity of geological structures at each spatial grid point.

[0052] The synthesis module combines the dynamic information contained in the feature subset of essential causal relationships with the confidence weights that reflect local cognition to temporally regulate the background seismic activity level and obtain the grid-point differentiated earthquake occurrence probability.

[0053] The analysis module, based on the grid-point differentiated earthquake occurrence probability, combined with the ground motion attenuation law and site effect, performs probabilistic earthquake hazard analysis and obtains high-resolution ground motion hazard zoning.

[0054] The risk mapping module integrates high-resolution seismic hazard zoning with disaster-bearing body information to generate a high-resolution seismic risk map.

[0055] The beneficial effects of this invention are as follows: By collecting and preprocessing multi-source data to form a standardized dataset, extracting multi-dimensional risk features, and then selecting a subset of features that are essentially causally related to earthquake occurrence based on causal inference, and using historical earthquake records to determine the background activity level; by quantifying the data sparsity and geological complexity of each spatial grid point to obtain the local cognitive confidence weight, combining the dynamic information contained in the causal relationship features with the cognitive weight, and temporally regulating the background activity level to obtain the grid point-differentiated earthquake occurrence probability; combining the seismic motion attenuation law and site effect for probability analysis to generate a high-resolution seismic motion hazard zoning, and finally fusing it with the disaster-bearing body information to generate a high-resolution earthquake risk map, realizing a refined analysis of earthquake risk throughout the entire process from data fusion, feature selection, uncertainty quantification to dynamic assessment. Attached Figure Description

[0056] To more clearly illustrate the technical solutions of the embodiments of the present invention, the drawings used in the following description of the embodiments will be briefly introduced. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0057] Figure 1 is a flowchart of the high-resolution earthquake risk information analysis method.

[0058] Figure 2 is a schematic diagram of a high-resolution earthquake risk information analysis system. Detailed Implementation

[0059] To make the above-mentioned objects, features and advantages of the present invention more apparent and understandable, the specific embodiments of the present invention will be described in detail below with reference to the accompanying drawings.

[0060] Many specific details are set forth in the following description in order to provide a full understanding of the invention. However, the invention may also be practiced in other ways different from those described herein, and those skilled in the art can make similar extensions without departing from the spirit of the invention. Therefore, the invention is not limited to the specific embodiments disclosed below.

[0061] Secondly, the term "one embodiment" or "embodiment" as used herein refers to a specific feature, structure, or characteristic that may be included in at least one implementation of the present invention. The phrase "in one embodiment" appearing in different places in this specification does not necessarily refer to the same embodiment, nor is it a single or selective embodiment that is mutually exclusive with other embodiments.

[0062] Referring to Figures 1 and 2, an embodiment of the present invention is provided, which offers a high-resolution seismic risk information analysis method, comprising the following steps:

[0063] S1. Collect and preprocess multi-source data from the target area to form a standardized multi-source dataset.

[0064] S1.1 Collect seismic catalog data, active fault data, crustal deformation data, and geophysical field data of the target area, and perform format conversion and coordinate system conversion.

[0065] Furthermore, seismic catalog data, active fault data, crustal deformation data, and geophysical field data for the target area are obtained from different institutions, typically from seismic monitoring networks, geological survey institutions, satellite geodetic projects, and geophysical exploration databases. Due to differences in storage formats, time bases, and spatial reference systems, format conversion and coordinate system unification operations are performed. Format conversion involves transforming the proprietary formats of different data files into a universal one, while coordinate system unification requires unifying the spatial information of all data under the same geodetic coordinate system and map projection system. This transforms the integration problem of multi-source heterogeneous data into a standardized pipeline operation that can be processed in batches.

[0066] S1.2. The seismic catalog data, active fault data, crustal deformation data and geophysical field data that have completed format conversion and coordinate system one are interpolated to a unified spatial grid and normalized.

[0067] Furthermore, based on the data that has undergone format conversion and coordinate system unification, a unified spatial grid covering the entire target area is defined. The size of the spatial grid determines the spatial resolution of the final multi-source dataset. Using spatial interpolation algorithms, discrete point-based seismic catalog data, active fault data, crustal deformation data, and geophysical field data are interpolated to the center of each grid point in the unified spatial grid, so that each type of data forms a continuously distributed raster image in space. After completing the spatial interpolation, the various types of data at each grid point are immediately normalized. The deviation normalization method is used to scale data of different dimensions and orders of magnitude to the same numerical range. Data with different spatial resolutions and representations are resampled to the same scale, thereby realizing the direct correlation and operation of data with different physical meanings within the same grid cell. This overcomes the inherent difficulty of traditional methods in performing point-by-point quantitative correlation analysis due to the mismatch of data spatial scales, and generates a multidimensional data cube with strictly aligned spatial locations and comparable numerical ranges.

[0068] S1.3 Integrate the normalized seismic catalog data, active fault data, crustal deformation data, and geophysical field data into a standardized multi-source dataset.

[0069] Furthermore, earthquake catalog data, active fault data, crustal deformation data, and geophysical field data, which have already been normalized and are now within a unified spatial grid, are correlated one-to-one according to their spatial grid locations. Each grid point contains a set of normalized attribute values ​​from different data sources. These attribute values ​​are organized in a predetermined order, creating a multidimensional feature vector for each grid point. The set of feature vectors from all grid points constitutes a standardized multi-source dataset. This transforms the analysis from multiple independent spatial data layers to an integrated, structured grid attribute table, changing the traditional layer-based analysis unit to a grid feature vector-based analysis unit. This overcomes the limitations of traditional raster data analysis, which can only perform single-layer or simple algebraic operations, and forms a standard input format that can be used for advanced multivariate statistical analysis.

[0070] S2. Using standardized multi-source datasets, extract multi-dimensional features to characterize earthquake risk.

[0071] S2.1 Calculate seismic activity parameters based on earthquake catalog data from standardized multi-source datasets.

[0072] Furthermore, based on the earthquake catalog data in the standardized multi-source dataset, the spatiotemporal benchmarks and formats have been unified. When calculating seismic activity parameters, the completeness of the earthquake catalog is confirmed, that is, the minimum complete magnitude is determined to ensure that all earthquake events with a magnitude greater than or equal to this value have been fully recorded. The calculation of the seismic moment release rate in the seismic activity parameter expression involves summing the seismic moments of each earthquake event in the catalog. The seismic moments can be obtained from the magnitude through empirical formulas. After summing, dividing by the total time length covered by the earthquake catalog yields the seismic moment release per unit time, i.e., the average seismic moment release rate. The design intent of this parameter is to use a scalar with a clear physical meaning to comprehensively characterize the long-term seismic energy release level of a region.

[0073] Specifically, it abandons the traditional approach of simply relying on earthquake frequency or average magnitude, and instead uses the fundamental physical quantity of earthquake moment, which has a more direct engineering and physical meaning, for accumulation and averaging, thus reflecting the intensity of regional tectonic activity more essentially. It overcomes the shortcomings of traditional seismic activity parameters, such as unreasonable weighting of earthquakes of different magnitudes or vague physical meanings. Its advantage is that it allows for objective comparison of seismic activity levels in different regions and at different times on a unified physical benchmark, providing a more robust input for risk assessment.

[0074] The expression for seismic activity parameters is:

[0075] ;

[0076] in, For seismic activity parameters, The total time span covered by the earthquake catalog. The total number of earthquake events. For earthquake event indexing, For the earthquake catalog Seismic moments of a single earthquake event.

[0077] It should be noted that the expression for seismic activity parameters applies to all events in the earthquake catalog (earthquake event index). From 1 to the total number of events Seismic moment Summing the results yields the total seismic moment release, which is then divided by the total time span covered by the catalog. Thus, the average seismic moment release rate per unit time is obtained. Seismic Moment It is a core parameter describing the physical processes of earthquake sources. It directly measures the product of fault slip area, average slip, and medium shear modulus. It has a clear physical relationship with the elastic wave energy released during earthquake rupture and quantifies the average rate at which crustal tectonic deformation releases energy through seismic activity in the study area over a long time scale.

[0078] S2.2 Calculate tectonic activity parameters based on active fault data and crustal deformation data from standardized multi-source datasets.

[0079] Furthermore, active fault data and crustal deformation data from standardized multi-source datasets have been interpolated onto a unified spatial grid. The calculation of the active parameters relies on the horizontal velocity field of the surface obtained by geodesy. The shear strain rate in the expression is a component of the strain rate tensor. It is necessary to obtain the spatial gradient of the velocity field and clarify the velocity difference between adjacent grid points. The partial derivatives of the velocity components in the vertical direction can be approximated by numerical methods such as the central difference method, and then substituted into the formula to calculate the shear strain rate. This quantifies the deformation rate that the crust is currently accumulating and that may lead to shear failure.

[0080] Specifically, by extracting mechanical information related to earthquake gestation from current crustal deformation observations, the continuous deformation field is transformed into a strain rate field reflecting the intensity of tectonic activity, realizing a shift from a kinematic to a dynamic perspective. This overcomes the limitation of relying solely on fault slip rates at the geological timescale to infer current activity. Its advantage is that it can capture more immediate and dynamic tectonic activity signals related to current tectonic loading, providing quantitative indicators that reflect the current state of crustal stress accumulation for earthquake hazard assessment.

[0081] The expression for the activity parameter is constructed as follows:

[0082] ;

[0083] in, To construct activity parameters, For the horizontal movement velocity of the earth's crust Components in direction The differential, For along Infinitely small distances along the coordinate axes For along Infinitely small distances along the coordinate axes For the horizontal movement velocity of the earth's crust Components in direction The differential.

[0084] It should be noted that the tectonic activity parameter formula used is based on the theory of continuum mechanics, extracting strain rate information from the horizontal velocity field of the Earth's surface obtained from geodesy. The physical core of the formula is the calculation of shear strain rate, which quantifies the spatial position of the Earth's crust. The rate of shape distortion occurring in the plane, express Movement in direction causes The gradient of directional velocity change describes the difference between east-west flow and north-south flow. Conversely, it means Movement in direction causes The gradient of the directional velocity change, averaged with these two shear components, yields the strain rate tensor component describing pure shear deformation. The rationale for this parameter lies in its direct reflection of the instantaneous rate at which the crust distorts to accumulate elastic strain energy, and its direct physical correlation with the gestation and occurrence of earthquakes. The current formula realizes a shift in observational perspective from long-term average to instantaneous state, enabling sensitive capture of minute crustal deformation anomalies caused by fault locking or stress loading. This allows the assessment of tectonic activity to no longer rely solely on inferences from past fault activity, but rather on direct measurement of the actual deformation process of the current crust. This greatly enhances the real-time nature and physical interpretability of judging the strength of current tectonic activity, providing more direct and dynamic observational evidence for identifying high-risk areas where stress is accumulating.

[0085] S2.3 Calculate crustal stress state parameters based on geophysical field data and active fault data from standardized multi-source datasets.

[0086] Furthermore, based on geophysical field data and active fault data from standardized multi-source datasets, which include a large amount of focal mechanism solutions, the calculation of crustal stress state parameters is an inversion process. The objective function is to find a regional stress tensor that minimizes the overall difference between the predicted fault slip direction under stress field and the slip direction observed by a large number of focal mechanism solutions. The calculation in the expression involves matching the observed values ​​with the theoretical predictions for each focal mechanism solution and evaluating the overall goodness of fit through weighted summation. The optimal stress tensor parameter is solved by minimizing this objective function. The design intent of this parameter is to invert the direction and relative magnitude of the regional stress field that controls the fault slip trend.

[0087] Specifically, by using the directionality of earthquake rupture itself as an indicator of the stress field, and by assembling a large number of focal mechanism solutions to constrain the regional average stress state, discrete focal mechanism information is condensed into a stress tensor that reflects the regional tectonic dynamic environment. The limitations of inferring the stress state from a single fault or local geological phenomenon can be overcome by extracting quantitative information on the regional stress field that drives these earthquakes from seismic activity, providing key parameters for understanding the mechanical environment of earthquake occurrence.

[0088] The expression for the crustal stress state parameter is:

[0089] ;

[0090] in, These are crustal stress state parameters. This is an index for the focal mechanism solution. For the first The weight of each solution, For the first The observed fault normal vector of a focal mechanism solution This is the theoretical fault normal vector predicted by the fault slip direction under the current inverted stress tensor field.

[0091] It should be noted that the crustal stress state parameter formula used is based on an inversion process according to physical principles. The formula does not directly calculate a scalar parameter, but rather defines an objective function. It solves for the optimal regional stress tensor by minimizing the difference between observations and predictions. In the formula, It is the first The fault normal vector observed by the focal mechanism solution represents the true geometric orientation of the rupture surface of the sub-earthquake. This refers to the normal vector most prone to slip on the same fault, theoretically predicted based on the Coulomb fracture criterion that the fault slip direction is collinear with the shear stress under a hypothetical uniform stress tensor field. The dot product of these two vectors... This reflects the degree of consistency between the observed rupture surface geometry and the optimal rupture surface geometry predicted by the current stress field. The closer the square of the dot product is to 1, the higher the consistency. The physical meaning of the objective function is to find an optimal regional stress tensor that minimizes the overall difference between the theoretically predicted fault slip direction and the fault slip direction observed in a large number of actual earthquakes under stress field driving. The weighted summation term... This allows for assigning different confidence levels to focal mechanism solutions of varying quality or importance, based on a fundamental physical principle: earthquakes are products of crustal stress fields, and their rupture mechanisms (focal mechanism solutions) record information about local stress states. By inverting a large number of focal mechanism solutions, the regional mean stress fields driving these earthquakes can be revealed.

[0092] S2.4 Combine seismic activity parameters, tectonic activity parameters, and crustal stress state parameters to form a multidimensional feature for characterizing seismic risk.

[0093] Furthermore, seismic activity parameters, tectonic activity parameters, and crustal stress state parameters are combined, with each grid point corresponding to a set of parameter values. The combination operation is performed independently on each spatial grid point, arranging the scalar values ​​representing seismic activity, the scalar values ​​representing tectonic deformation rate, and the relevant parameter values ​​representing stress state (such as the eigenvalues ​​or principal directions of the stress tensor) in a specific order to form a feature vector representing the comprehensive risk status of the grid point. The set of feature vectors of all grid points forms a multidimensional feature for characterizing seismic risk, thus constructing a comprehensive feature space that fully describes the earthquake-causing process (energy accumulation, tectonic background, and dynamic environment).

[0094] Specifically, parameters with different physical meanings and dimensions are integrated at the grid level after being standardized in the early stage, forming a high-dimensional data representation that can be used by machine learning algorithms to identify complex patterns. The in-depth analysis provides a high-dimensional and structured characteristic data foundation that can reflect the current state of seismic activity, tectonic background and crustal stress environment, which is a problem of insufficient information utilization caused by the use of only a single type of parameter or simple superposition of a few parameters in traditional risk assessment.

[0095] S3. Based on causal inference, identify a subset of features from multidimensional features that have an essential causal relationship with the occurrence of earthquakes.

[0096] S3.1 By analyzing the prior knowledge dependencies among multidimensional features, earthquake occurrence, and potential confounding variables, a causal graph structure of multidimensional features, earthquake occurrence, and potential confounding variables is constructed.

[0097] Furthermore, it integrates prior knowledge from geophysics, geology, and seismology, derived from long-term research practice rather than simply data-driven approaches; it identifies all risk factors that may affect earthquake occurrence or multidimensional characteristics, including but not limited to deep tectonic backgrounds such as the depth of the Moho discontinuity, crustal thickness, the direction and magnitude of regional stress fields, rock mechanical properties, pore fluid pressure, and the geometry and activity of major active fault zones. Based on an understanding of Earth system dynamics, it clarifies the causal relationship between factors. For example, it determines that the deep tectonic background controls the regional stress field pattern, the regional stress field affects the sliding behavior of active fault zones, and rock mechanical properties and pore fluid pressure jointly modulate the stress response of fault zones. The causal relationship is represented by nodes and directed edges. Nodes represent multidimensional characteristics, earthquake occurrence, and potential confounding variables, while directed edges represent the assumed causal influence direction, thus forming a directed acyclic graph structure.

[0098] Specifically, it transforms qualitative and conceptual causal analysis in geology into a structured causal dependency network containing explicit assumptions, making the hypotheses about the causes of earthquakes verifiable. It overcomes the inherent flaw of traditional statistical methods that may lead to spurious correlations due to ignoring confounding variables, ensuring that feature selection is based on causal relationships rather than superficial associations, and improving the physical interpretability and robustness of risk assessment.

[0099] S3.2 Based on the causal graph structure, the causal inference algorithm is applied to calculate the average treatment effect value of each feature in the multidimensional features on the occurrence of earthquakes.

[0100] Furthermore, based on the completed causal graph structure, an inverse probability weighted algorithm is applied to calculate the average treatment effect value, and the set of potential confounding variables identified in the causal graph is included. This is concretized into observable data, such as crustal thickness, rock type, and the direction of the background stress field; for each spatial grid point Use logistic regression model to estimate their propensity score That is, the value of the confusion variable for a given grid point. When processing variables The conditional probability that takes the value 1, the processing variable It can be defined as whether a certain seismic risk characteristic exceeds a specific threshold. The observation data of each spatial grid point is substituted into the ATE formula for calculation. Some processing groups are weighted, while The control group is partially weighted; the purpose of this weighting is to construct a pseudo-population in which variables are processed. Assignment and confusion variables This is irrelevant, thus simulating the environment of a randomized experiment. The weighted results of all spatial grid points are averaged to obtain the average treatment effect value ATE of a certain feature on the occurrence of earthquakes.

[0101] Specifically, the mathematical framework of causal inference from observational data is introduced into the field of earthquake risk assessment. By using a mathematical tool such as propensity score weighting, the influence of confounding variables is removed from observational data in the geosciences where real experiments cannot be conducted. This allows for a closer approximation of the causal relationship between features and earthquakes. It provides a comparable quantitative indicator with causal explanation for the importance assessment of each feature, so that feature selection no longer relies on statistical correlation but is based on the logical foundation of counterfactual causality, thereby improving the reliability of risk assessment conclusions.

[0102] The expression for the average treatment effect size is:

[0103] ;

[0104] in, This is the average treatment effect value. This represents the total number of spatial grid points. For spatial grid indexing, In the first Variables are processed at each spatial grid point. The observed values, For the first The result variable at each spatial grid point The observed values, For the first The values ​​of the set of confounding variables observed at each spatial grid point. For the first Tendency score for each spatial grid point.

[0105] S3.3 Statistically analyze the average treatment effect value, select feature subsets from the multidimensional features, and determine the feature subsets that have an essential causal relationship with the occurrence of earthquakes.

[0106] Furthermore, based on the average treatment effect value of each multidimensional feature, the multidimensional features are screened according to the statistical significance criteria for causal inference. This is usually done by calculating the standard error of the average treatment effect value and constructing confidence intervals. The uncertainty of the average treatment effect value estimation is quantified by using the bootstrap method or a theoretical formula based on the estimation equation. A level threshold is set. If the confidence interval of the average treatment effect value of a certain multidimensional feature does not contain zero, the multidimensional feature is considered to have a significant causal effect on the occurrence of earthquakes. All multidimensional features that pass the test are gathered together to form a feature subset. This feature subset is formally determined as the feature subset that has an essential causal relationship with the occurrence of earthquakes.

[0107] Specifically, the statistical decision-making mechanism in causal inference is introduced into the earthquake risk feature screening process. It not only relies on the size of the average treatment effect point estimate, but also considers its estimation accuracy. Through rigorous statistical testing, spurious correlation features that have large point estimates but high uncertainty and may be caused by random factors are excluded. This ensures that each feature in the final selected feature subset not only has a theoretical causal explanation, but also has statistically significant evidence to support it, thereby improving the reliability and interpretability of the constructed earthquake risk model.

[0108] S4. Determine the background seismic activity level based on historical earthquake records, including the following steps:

[0109] S4.1 Conduct a completeness analysis of historical earthquake records to determine the minimum complete magnitude of historical earthquake records.

[0110] Furthermore, standard methods for completeness analysis in seismology are applied to process historical earthquake records. The process is based on a fundamental observation: as earthquake magnitude decreases, the number of detectable events increases to a certain point, after which it begins to decline due to limitations in detection capabilities. The magnitude corresponding to the inflection point is the minimum complete magnitude. The analysis typically uses a magnitude-frequency relationship curve to plot the cumulative earthquake frequency as a function of magnitude. The inflection point where the curve begins to deviate from the linear Gutenberg-Richard relationship is identified as the minimum complete magnitude. Another commonly used method is the maximum curvature method, which finds the magnitude corresponding to the peak frequency in the magnitude distribution frequency histogram. Completeness analysis also needs to consider spatiotemporal variations, such as conducting analysis at different time periods or in different sub-regions to reflect the evolution of the monitoring network's capabilities.

[0111] Specifically, data quality control is shifted from subjective experience-based judgment to objective quantitative analysis. It acknowledges that the earthquake catalog is not complete for all magnitudes and provides a repeatable standard based on the statistical characteristics of the data itself to define the range of usable data. This overcomes the problem that directly using the entire original catalog may result in missing low-magnitude data due to insufficient detection capabilities, thus distorting the estimation of seismic activity parameters. It ensures that all subsequent statistical results based on earthquake frequency and energy calculations are built on a complete and consistent data sample, providing a reliable basic input for risk assessment.

[0112] S4.2. Based on the minimum complete magnitude, the historical earthquake records are filtered to obtain complete historical earthquake records.

[0113] Furthermore, an objective threshold of minimum complete magnitude is determined through completeness analysis. The implementation process involves scanning historical earthquake records one by one, comparing the magnitude value of each record with the minimum complete magnitude, retaining all records with magnitudes greater than or equal to the minimum complete magnitude, and removing all records with magnitudes less than the minimum complete magnitude. This ensures that the filtered historical earthquake records fully include all earthquake events above the minimum complete magnitude, without any systematic omissions due to limitations in monitoring capabilities.

[0114] S4.3. Using a smoothing algorithm, the earthquake occurrence rate of complete historical earthquake records is calculated on a spatial grid and determined as the background seismic activity level.

[0115] Furthermore, a Gaussian smoothing algorithm is applied to convert the spatial distribution information of complete historical seismic records into a continuous background seismic activity horizontal field. A regular spatial grid covering the study area is defined, and for each grid point in the grid, the spatial location... The algorithm iterates through every earthquake event in the complete historical earthquake record, calculates the Euclidean distance from the epicenter of each event to a grid point, and treats each earthquake event as a point source. Its contribution to the earthquake occurrence rate at the grid point is quantified by a Gaussian kernel function centered at the event epicenter. The Gaussian kernel function has the following form: ),in For distance, To control the standard deviation of the smoothing range, the contribution values ​​of all seismic events calculated using the corresponding Gaussian kernel function are plotted on the grid. The summation is performed at each point, and this summation value reflects the sum of the values ​​at each point. The cumulative intensity of seismic activity within a certain neighborhood centered on the earthquake is calculated by dividing this sum by the total time span. Area of ​​grid points The product of these factors is normalized to obtain the annual average earthquake occurrence rate per unit time and unit area at the grid points. All grid points The set constitutes a background seismic activity level model.

[0116] Specifically, discrete and sparse point-like seismic event data are transformed into a continuous and complete rate field through a smoothing process that considers spatial attenuation. This approach acknowledges the spatial correlation of seismic activity, meaning that the future seismic risk of a location depends not only on whether an earthquake has occurred but also significantly on the intensity of seismic activity within a certain surrounding area. It overcomes the spatial discontinuities and oversensitivity to event location errors caused by simply assigning seismic event counts to grid cells. By introducing a smoothing kernel that attenuates with distance, it effectively combines local seismic activity information with the regional tectonic background, generating a background seismic activity level model that reflects both the characteristics of earthquake clusters and spatial continuity.

[0117] It should be noted that the earthquake occurrence rate formula uses a Gaussian smoothing kernel to spatially diffuse discrete earthquake events, estimating the annual average earthquake occurrence rate at any point in space. Each earthquake event is considered a risk source with a spatially influential range, and its influence decreases Gaussianly with increasing distance. In the formula, ... It is the Gaussian kernel function, which quantizes the first... The weight of the impact of each earthquake event on the target point, distance The closer, the greater the impact; These are normalization coefficients that ensure the Gaussian kernel integrals to 1 in the two-dimensional plane, thus conserving the total energy of each seismic event; for all events in the catalog... The contributions of each seismic event are summed to indicate that the risk at the target point is the superposition result of the combined effects of all historical seismic events in the surrounding area. The total impact is normalized to the occurrence rate per unit time and unit area, thus obtaining the annual average earthquake occurrence rate. The previous approach acknowledged the spatial correlation of seismic activity, meaning that the seismic risk at a location depends not only on whether an earthquake has occurred, consistent with the observational fact that seismic activity is clustered along tectonic zones. However, it essentially assumed that the impact of earthquakes is localized and discontinuous, failing to reflect the spatial correlation of seismic activity and being extremely sensitive to earthquake location errors. The current formula, by introducing a Gaussian smoothing kernel with a clear physical meaning (spatial influence attenuation), transforms discrete point information into continuous field information. It uses a mathematically rigorous and physically interpretable method (similar to the concepts of thermal diffusion or physical fields) to characterize the propagation and superposition of seismic risk, generating a smooth and reasonable model of background seismic activity.

[0118] The expression for earthquake occurrence rate is:

[0119] ;

[0120] in, In spatial location The average annual earthquake incidence rate at that location The area of ​​a spatial grid cell. The total number of earthquake events with a magnitude greater than or equal to the minimum complete magnitude in a complete earthquake catalog. For the first The distance between the epicenters of an earthquake event The standard deviation of the Gaussian smoothing kernel. Index for earthquake events.

[0121] S5. By quantifying the sparsity of available data and the complexity of geological structure at each spatial grid point, a confidence weight reflecting local cognition is obtained.

[0122] S5.1 Count the number of available seismic stations and engineering boreholes within a preset range around each spatial grid point to obtain the sparsity of available data at each spatial grid point. Calculate the number of intersections of active faults and the number of changes in lithology within a preset range around each spatial grid point to obtain the complexity of the geological structure at each spatial grid point.

[0123] Furthermore, the quantification of data sparsity and geological complexity is addressed separately. For data sparsity, each spatial grid point is traversed, and an effective influence radius is preset with the grid point as the center. The total number of all available seismic stations and engineering boreholes falling within this circular area is counted. This number is directly used as a measure of the data sparsity of the grid point; the fewer the number, the sparser the data support. For geological complexity, the same process is implemented with the grid point as the center. Within another preset geological influence radius, the number of intersections of active faults and the frequency of occurrence of boundaries of different lithological units are counted. The sum of these two counts or the maximum value is used as a measure of the geological complexity of the grid point. The more fault intersections and the more frequent the lithological changes, the more complex the geological structure.

[0124] The expression for data sparsity is:

[0125] ;

[0126] in, for Data sparsity of each spatial grid point The normalization coefficient is... The total number of all available data points within the study area. For indexing data points, For the first The spatial lattice point and the first The distance between data points To effectively influence the radius.

[0127] S5.2 Normalize and weight the sparsity of available data at each spatial grid point and the complexity of geological structure at each spatial grid point to obtain the data-geological comprehensive heterogeneity index for each spatial grid point.

[0128] Furthermore, the sparsity of available data at each spatial grid point and the heterogeneity index of geological structure complexity at each spatial grid point are integrated into a comprehensive index. The data sparsity and geological complexity are normalized separately. For example, the min-max normalization method is used to linearly transform the value of each index to between zero and one, eliminating the influence of differences in the dimensions and numerical ranges of different indicators, making them comparable. Weights are assigned to the normalized data sparsity index and geological complexity index. The weight assignment can be determined based on the objective method of principal component analysis, reflecting the relative importance of the two indicators to the overall cognitive uncertainty. The weighted values ​​of the two indicators are then fused. The fusion method can be linear weighted summation, that is, the comprehensive heterogeneity index is equal to weight one multiplied by normalized data sparsity plus weight two multiplied by normalized geological complexity.

[0129] Specifically, by integrating the multidimensional sources of cognitive uncertainty into a unified and operable quantitative indicator, we can identify data uncertainty and model uncertainty as the two core components of cognitive uncertainty. These two uncertainties may affect the reliability of the final assessment to varying degrees. By using weighted fusion rather than simple superposition, we can more reasonably simulate the joint contribution of the two to the total uncertainty. This overcomes the limitation of considering only one aspect of data or geological uncertainty and failing to comprehensively assess the cognitive state. It also avoids the crude treatment of simply treating uncertainties of different natures as equal. This results in a spatialized and comprehensive representation that can reflect the sufficiency of data support and the stability of the geological model.

[0130] S5.3. The data of each spatial grid point is inversely converted with the geological comprehensive heterogeneity index to obtain the confidence weight that reflects local cognition.

[0131] Furthermore, for each spatial grid point, the data-geological heterogeneity index is calculated by taking the reciprocal of the index. This means the confidence weight is equal to one divided by the data-geological heterogeneity index of the grid point. This inverse relationship implies that a higher data-geological heterogeneity index represents greater cognitive uncertainty at the grid point, resulting in a lower calculated confidence weight. Conversely, a lower index represents less cognitive uncertainty, resulting in a higher confidence weight. To ensure numerical stability and boundedness, a small normal number can be applied to the data-geological heterogeneity index as a smoothing factor before taking the reciprocal, preventing division by zero or near-zero values.

[0132] S6. By combining the dynamic information contained in the feature subset of essential causal relationships with the confidence weights that reflect local cognition, the background seismic activity level is temporally regulated to obtain the grid-point differentiated earthquake occurrence probability.

[0133] S6.1 Extract the numerical changes of the feature subset of essential causal relationships within a preset time window, as the dynamic information contained in the feature subset of essential causal relationships.

[0134] Furthermore, a preset time window is defined for each spatial grid point, such as a sliding time period. For each feature in the feature subset that is essentially causally related, such as calculating the mean, slope, variance, or rate of change of the feature value within the time window compared to an earlier time window, it reveals the degree of deviation of each causal feature from its long-term background state in the near term. This expands the perspective of risk assessment from static long-term average state to dynamic short-term change monitoring. Features that have been proven to be causally related to earthquakes, and short-term abnormal changes may contain key information about the pre-earthquake adjustment of crustal stress state. Dynamic signals are more indicative of earthquake triggering than static horizontal values.

[0135] S6.2. Using the dynamic information contained in the feature subset of essential causal relationships as a correction factor, the background seismic activity level is corrected for the first time to obtain the preliminary corrected earthquake occurrence rate.

[0136] Furthermore, the dynamic information contained in the feature subset of essential causal relationships is transformed into a correction factor to correct the background seismic activity level. A mathematical relationship is established to associate the background earthquake occurrence rate of each spatial grid point with the dynamic information reflecting the recent crustal state changes of the point. If the dynamic information of a certain grid point indicates an increase in the probability of earthquake occurrence, the correction factor is greater than one, thereby increasing the background occurrence rate of the point; otherwise, it is reduced.

[0137] Specifically, by fusing static long-term probabilities with dynamic short-term anomalous signals, the background seismic activity level based on long-term historical averages cannot reflect the current stress loading state, while the dynamic changes in causal characteristics can provide this short-term information. The generated earthquake occurrence rate not only includes the long-term background but also incorporates physically meaningful short-term modulation signals, making the probability estimation more timely.

[0138] S6.3. The confidence weight, which reflects local cognition, is used as a modulation coefficient and applied to the preliminary correction of the earthquake occurrence rate.

[0139] Furthermore, the preliminary corrected earthquake occurrence rate for each spatial grid point is multiplied by the corresponding confidence weight, which is less than or equal to one. Based on the magnitude of cognitive uncertainty at each grid point, the preliminary corrected earthquake occurrence rate is discounted, explicitly introducing cognitive uncertainty into the probability calculation. This acknowledges that the reliability of probability estimates varies at different locations and uses quantified weights to reflect this difference. As a result, the final probability value includes both random uncertainty and cognitive uncertainty. The probability value for high-risk but cognitively ambiguous areas is appropriately lowered, thus more realistically reflecting the limitations of knowledge.

[0140] S6.4. The earthquake occurrence rate, after being preliminarily corrected by the confidence weight modulated to reflect local cognition, is used to determine the grid-specific earthquake occurrence probability for each spatial grid point.

[0141] Furthermore, the modulated earthquake occurrence rate is confirmed as the final probability estimate for each spatial grid point. The modulated value is used as the output of the grid point differentiated earthquake occurrence probability. A series of adjustments are completed from background value to dynamic correction and then to cognitive weighting, thus completing the final integration of the probability model. Through step-by-step modulation, dynamic physical information and static cognitive information are systematically and hierarchically integrated into a unified probability framework. The resulting grid point differentiated earthquake occurrence probability is a result of multi-dimensional information fusion, which is supported by physical mechanisms and includes reliability measures, providing high-quality and interpretable input for subsequent high-resolution risk assessment.

[0142] S7. Based on the grid-point differentiated earthquake occurrence probability, combined with the ground motion attenuation law and site effect, a probabilistic earthquake hazard analysis is carried out to obtain a high-resolution ground motion hazard zoning.

[0143] S7.1. Based on the differential earthquake occurrence probability of grid points, calculate the exceedance probability of ground motion intensity generated by potential sources at each field point.

[0144] Furthermore, the core calculation process of probabilistic seismic hazard analysis takes as input the grid-point differentiated earthquake occurrence probability, i.e., the annual average occurrence rate of each potential source region. Based on the law of total probability, for each field point, considering all potential source regions, the integral is performed on all possible earthquake magnitudes and all possible source-to-field distances within each source region. The integral kernel function consists of three parts: the source occurrence rate, the magnitude probability density function, the distance conditional probability density function, and the conditional exceedance probability given by the seismic motion prediction equation.

[0145] Specifically, by combining the spatially differentiated source activity model with the ground motion prediction model through a rigorous probability integral framework, the uncertainty of the source is transformed into the uncertainty of the ground motion at the field point, resulting in an analytical solution for the ground motion hazard at the field point that fully considers the spatiotemporal uncertainties of earthquake occurrence.

[0146] It should be noted that, in order to calculate the total probability that the ground motion intensity at a field point exceeds a certain threshold, all possible earthquake scenarios that could cause strong ground motion at the field point (i.e., different sources, different magnitudes, and different locations) are considered. The probability of each scenario occurring is multiplied by the conditional probability that the scenario causes the ground motion to exceed the threshold. Finally, the contributions of all possible scenarios are integrated and summed.

[0147] The expression for the exceedance probability of vibration intensity is:

[0148] ;

[0149] in, This represents the probability of earthquake intensity exceeding the limit. These are the parameters for seismic intensity. The ground motion intensity threshold, To determine the total number of potential earthquake sources within the study area, An index for potential seismic source regions. For the first The average annual earthquake incidence rate of each potential seismic source The magnitude of the earthquake. For the first Given a magnitude at a potential earthquake source Distance from the epicenter The conditional probability density function, In the first The probability density function of earthquake magnitude at a potential hypocenter. This represents the distance from the epicenter to the site. For earthquake magnitude determination and distance Ground motion intensity under the condition Exceeding the earthquake intensity threshold The probability of.

[0150] S7.2 Apply the ground motion attenuation law to the exceedance probability of ground motion intensity generated by the potential source at each field point, and use the site effect as an amplification factor to correct the exceedance probability of ground motion intensity generated by the potential source at each field point.

[0151] Furthermore, by applying site effect corrections to the exceedance probability of ground motion intensity, the ground motion attenuation law is already implicitly contained within... In the function, the data is usually derived from the records of the bedrock site. The correction for the site effect is to convert the bedrock ground motion to the ground surface through the site amplification factor. The calculated bedrock ground motion intensity is multiplied by an amplification factor related to the site type and ground motion frequency to obtain the ground surface ground motion intensity. This separates and processes the global ground motion attenuation law from the local site response.

[0152] S7.3. Perform probability integration on the exceedance probability of ground motion intensity generated by potential seismic sources at each field point, and calculate the exceedance probability of ground motion parameters at each field point within a given period.

[0153] Furthermore, the annual exceedance probability is transformed into the exceedance probability over a given period of time. Assuming that the occurrence of earthquake events follows a Poisson process in time, the annual average occurrence rate is calculated using the Poisson distribution formula. Convert to time Probability of exceeding within the year It introduces a time dimension, transforming the measurement of risk from annual occurrence rate to risk within the design reference period. It uses probability theory to transform short-term occurrence rate into long-term risk, and the resulting probability value has clear engineering significance and is directly aligned with the seismic fortification standards of building structures.

[0154] The expression for the exceedance probability of ground motion parameters is:

[0155] ;

[0156] in, For the future Within a year, the intensity of ground motion at the field point Exceeding the earthquake intensity threshold The probability of the event, Given a time period of years, Ground motion intensity at the field point Exceeding the seismic intensity threshold within one year The average incidence rate.

[0157] S7.4 Spatial distribution of the exceedance probability of ground motion parameters at each field point within a given time period to obtain high-resolution ground motion hazard zoning.

[0158] Furthermore, the exceedance probability values ​​of all ground motion parameters within a given timeframe in the study area are spatially interpolated or directly rendered according to their geographic coordinates to generate continuous contour maps or raster maps. The numerical calculation results are transformed into intuitive spatial decision support information products. Through spatial display, the geographical distribution pattern of seismic hazard is revealed, rather than just discrete point data. The generated high-resolution seismic hazard zoning map can clearly show the spatial pattern of high-risk and low-risk areas.

[0159] S8. Integrate high-resolution seismic hazard zoning with disaster-bearing body information to generate a high-resolution seismic risk map.

[0160] S8.1 Spatial registration and attribute association are performed between the spatial probability distribution of ground motion intensity characterized by the high-resolution ground motion hazard zoning and the physical exposure and vulnerability characteristics contained in the disaster-bearing body information.

[0161] Furthermore, the high-resolution seismic hazard zoning map is precisely georeferenced with the disaster-bearing body spatial distribution database to ensure that the location of each spatial unit corresponds one-to-one. For each spatial unit, the seismic intensity exceedance probability curve provided by its seismic hazard zoning is extracted. The quantity, physical value and corresponding vulnerability curves of various disaster-bearing bodies within the unit are obtained from the disaster-bearing body information database. This integrates the three risk components of physical hazard, exposure and vulnerability under a unified spatial framework, ensuring that the risk value of each spatial unit is derived based on the specific hazard level and disaster-bearing body condition of the unit.

[0162] S8.2. By integrating the strength of seismic hazard, the value of disaster-bearing bodies, and their vulnerability in each spatial unit, a risk index is generated for each spatial unit.

[0163] Furthermore, using a probabilistic risk integral method, for each spatial unit, the ground motion exceedance probability curve, the vulnerability curve of the disaster-bearing body, and the value information of the disaster-bearing body are convolved and calculated. The calculation covers all possible ground motion intensity levels, multiplies the exceedance probability, the damage probability of the disaster-bearing body at each intensity level, and the corresponding loss value, and sums them over all intensity levels to obtain the expected annual average loss or other risk indicators of the spatial unit.

[0164] S8.3. Visualize the risk indicators of each spatial unit to generate a high-resolution seismic risk map.

[0165] Furthermore, gradient colors are used to represent the level of risk, thus clearly displaying the spatial distribution pattern of risk. This transforms numerical risk calculation results into intuitive spatial decision maps, leveraging the power of spatial visualization to compress complex risk data into easily understandable image information, highlighting high-risk areas and spatial patterns.

[0166] This embodiment also provides a high-resolution earthquake risk information analysis system, including: a data processing module, which collects and preprocesses multi-source data of the target area to form a standardized multi-source dataset;

[0167] The feature extraction module uses standardized multi-source datasets to extract multi-dimensional features to characterize earthquake risk, and identifies a subset of features that are essentially causally related to earthquake occurrence from the multi-dimensional features.

[0168] The cognitive weight quantification module, based on historical earthquake records, determines the background seismic activity level and obtains confidence weights that reflect local cognition by quantifying the sparsity of available data and the complexity of geological structures at each spatial grid point.

[0169] The synthesis module combines the dynamic information contained in the feature subset of essential causal relationships with the confidence weights that reflect local cognition to temporally regulate the background seismic activity level and obtain the grid-point differentiated earthquake occurrence probability.

[0170] The analysis module, based on the grid-point differentiated earthquake occurrence probability, combined with the ground motion attenuation law and site effect, performs probabilistic earthquake hazard analysis and obtains high-resolution ground motion hazard zoning.

[0171] The risk mapping module integrates high-resolution seismic hazard zoning with disaster-bearing body information to generate a high-resolution seismic risk map.

[0172] This embodiment also provides a computer device applicable to the high-resolution earthquake risk information analysis method, including: a memory and a processor; the memory is used to store computer-executable instructions, and the processor is used to execute the computer-executable instructions to implement the high-resolution earthquake risk information analysis method proposed in the above embodiment.

[0173] The computer device can be a terminal, comprising a processor, memory, communication interface, display screen, and input devices connected via a system bus. The processor provides computing and control capabilities. The memory includes non-volatile storage media and internal memory. The non-volatile storage media stores the operating system and computer programs. The internal memory provides an environment for the operation of the operating system and computer programs stored in the non-volatile storage media. The communication interface is used for wired or wireless communication with external terminals; wireless communication can be achieved through Wi-Fi, carrier networks, NFC (Near Field Communication), or other technologies. The display screen can be an LCD screen or an e-ink screen. The input devices can be a touch layer covering the display screen, buttons, a trackball, or a touchpad on the computer device's casing, or an external keyboard, touchpad, or mouse.

[0174] This embodiment also provides a storage medium storing a computer program, which, when executed by a processor, implements the high-resolution seismic risk information analysis method proposed in the above embodiments. The storage medium can be implemented by any type of volatile or non-volatile storage device or a combination thereof, such as Static Random Access Memory (SRAM), Electrically Erasable Programmable Read-Only Memory (EEPROM), Erasable Programmable Read Only Memory (EPROM), Programmable Red-Only Memory (PROM), Read-Only Memory (ROM), magnetic storage, flash memory, magnetic disk, or optical disk.

[0175] In summary, this invention collects and preprocesses multi-source data to form a standardized dataset, extracts multi-dimensional risk features, and then selects a subset of features with essential causal relationships to earthquake occurrence based on causal inference. Historical earthquake records are used to determine the background activity level. By quantifying the data sparsity and geological complexity of each spatial grid point, local cognitive confidence weights are obtained. The dynamic information contained in the causal relationship features is combined with the cognitive weights to temporally regulate the background activity level, obtaining grid-specific earthquake occurrence probabilities. Probability analysis is performed by combining ground motion attenuation patterns and site effects to generate high-resolution ground motion hazard zoning. Finally, this is fused with disaster-bearing body information to generate a high-resolution earthquake risk map, achieving a refined analysis of earthquake risk throughout the entire process, from data fusion, feature selection, uncertainty quantification to dynamic assessment.

[0176] It should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and are not intended to limit it. Although the present invention has been described in detail with reference to preferred embodiments, those skilled in the art should understand that modifications or equivalent substitutions can be made to the technical solutions of the present invention without departing from the spirit and scope of the technical solutions of the present invention, and all such modifications or substitutions should be covered within the scope of the claims of the present invention.

Claims

1. A high-resolution seismic risk information analysis method, characterized in that: include, Collect and preprocess multi-source data of the target area to form a standardized multi-source dataset; use the standardized multi-source dataset to extract multi-dimensional features to characterize earthquake risk; Based on causal inference, a subset of features that are essentially causally related to the occurrence of earthquakes is identified from multidimensional features; Based on historical earthquake records, the background seismic activity level is determined. By quantifying the sparsity of available data and the complexity of geological structure at each spatial grid point, a confidence weight reflecting local cognition is obtained. The dynamic information contained in the feature subset of essential causal relationships is combined with the confidence weight reflecting local cognition to temporally regulate the background seismic activity level and obtain the grid point-differentiated earthquake occurrence probability. Based on the differentiated earthquake occurrence probability at grid points, combined with ground motion attenuation patterns and site effects, probabilistic earthquake hazard analysis is conducted to obtain high-resolution ground motion hazard zoning. This high-resolution hazard zoning is then fused with information on disaster-bearing bodies to generate a high-resolution earthquake risk map. By quantifying the sparsity of available data and the complexity of geological structures at each spatial grid point, confidence weights reflecting local perceptions are obtained. This includes the following steps: counting the number of available seismic stations and boreholes within a preset range around each spatial grid point to obtain the sparsity of available data at each spatial grid point; calculating the number of intersections of active faults and the number of changes in lithology within a preset range around each spatial grid point to obtain the complexity of geological structures at each spatial grid point; and normalizing and weighting the sparsity of available data and the complexity of geological structures at each spatial grid point to obtain a comprehensive data-geological heterogeneity index for each spatial grid point. The process involves: 1) Inversely converting the data of each spatial grid point with the geological comprehensive heterogeneity index to obtain a confidence weight reflecting local perception; 2) Combining the dynamic information contained in the feature subset of essential causal relationships with the confidence weight reflecting local perception to temporally modulate the background seismic activity level, obtaining the grid-specific earthquake occurrence probability, including the following steps: Extracting the numerical changes of the feature subset of essential causal relationships within a preset time window as the dynamic information contained in the feature subset of essential causal relationships; 3) Using the dynamic information contained in the feature subset of essential causal relationships as a correction factor to perform the first correction on the background seismic activity level, obtaining a preliminary corrected earthquake occurrence rate; 4) Using the confidence weight reflecting local perception as a modulation coefficient to apply to the preliminary corrected earthquake occurrence rate; 5) Determining the grid-specific earthquake occurrence probability for each spatial grid point using the preliminary corrected earthquake occurrence rate modulated by the confidence weight reflecting local perception.

2. The high-resolution seismic risk information analysis method as described in claim 1, characterized in that: The process of collecting and preprocessing multi-source data from the target area to form a standardized multi-source dataset includes the following steps: collecting seismic catalog data, active fault data, crustal deformation data, and geophysical field data from the target area, and performing format conversion and coordinate system one adjustment; interpolating the seismic catalog data, active fault data, crustal deformation data, and geophysical field data that have undergone format conversion and coordinate system one adjustment to a unified spatial grid, and performing normalization processing; and integrating the normalized seismic catalog data, active fault data, crustal deformation data, and geophysical field data into a standardized multi-source dataset.

3. The high-resolution seismic risk information analysis method as described in claim 2, characterized in that: Using standardized multi-source datasets, multidimensional features for characterizing earthquake risk are extracted, including the following steps: calculating seismic activity parameters based on earthquake catalog data in the standardized multi-source datasets; calculating tectonic activity parameters based on active fault data and crustal deformation data in the standardized multi-source datasets; calculating crustal stress state parameters based on geophysical field data and active fault data in the standardized multi-source datasets; and combining the seismic activity parameters, tectonic activity parameters, and crustal stress state parameters to form multidimensional features for characterizing earthquake risk.

4. The high-resolution seismic risk information analysis method as described in claim 3, characterized in that: Based on causal inference, the subset of features that are essentially causally related to earthquake occurrence is identified from multidimensional features, including the following steps: by analyzing the prior knowledge dependencies between multidimensional features, earthquake occurrence, and potential confounding variables, a causal graph structure of multidimensional features, earthquake occurrence, and potential confounding variables is constructed; based on the causal graph structure, a causal inference algorithm is applied to calculate the average treatment effect value of each feature in the multidimensional features on earthquake occurrence; the average treatment effect value is statistically analyzed, and a subset of features is selected from the multidimensional features to determine the subset of features that are essentially causally related to earthquake occurrence.

5. The high-resolution seismic risk information analysis method as described in claim 4, characterized in that: Determining the background seismic activity level based on historical earthquake records includes the following steps: performing a completeness analysis on the historical earthquake records to determine the minimum complete magnitude of the historical earthquake records; filtering the historical earthquake records based on the minimum complete magnitude to obtain complete historical earthquake records; and using a smoothing algorithm to calculate the earthquake occurrence rate of the complete historical earthquake records on a spatial grid to determine the background seismic activity level.

6. The high-resolution seismic risk information analysis method as described in claim 5, characterized in that: Based on the grid-point differentiated earthquake occurrence probability, combined with ground motion attenuation laws and site effects, probabilistic seismic hazard analysis is performed to obtain high-resolution ground motion hazard zoning. The steps include: calculating the ground motion intensity exceedance probability of potential sources at each field point based on the grid-point differentiated earthquake occurrence probability; applying the ground motion attenuation law to the ground motion intensity exceedance probability of potential sources at each field point, using the site effect as an amplification factor to correct the ground motion intensity exceedance probability of potential sources at each field point; performing probability integration on the ground motion intensity exceedance probability of potential sources at each field point to calculate the ground motion parameter exceedance probability of each field point within a given timeframe; and spatially distributing the ground motion parameter exceedance probability of each field point within a given timeframe to obtain high-resolution ground motion hazard zoning.

7. The high-resolution seismic risk information analysis method as described in claim 6, characterized in that: The process of fusing high-resolution seismic hazard zoning with disaster-bearing body information to generate a high-resolution seismic risk map includes the following steps: spatially registering and associating the spatial probability distribution of seismic intensity represented by the high-resolution seismic hazard zoning with the physical exposure and vulnerability characteristics implied by the disaster-bearing body information; generating risk indicators for each spatial unit by fusing the seismic hazard strength, disaster-bearing body value, and vulnerability degree within each spatial unit; and spatially visualizing the risk indicators of each spatial unit to generate a high-resolution seismic risk map.

8. A high-resolution earthquake risk information analysis system, based on the high-resolution earthquake risk information analysis method according to any one of claims 1 to 7, characterized in that: include, The data processing module collects and preprocesses multi-source data from the target area to form a standardized multi-source dataset. The feature extraction module uses standardized multi-source datasets to extract multi-dimensional features to characterize earthquake risk, and identifies a subset of features that are essentially causally related to earthquake occurrence from the multi-dimensional features. The cognitive weight quantification module, based on historical earthquake records, determines the background seismic activity level and obtains confidence weights reflecting local cognition by quantifying the sparsity of available data and the complexity of geological structures at each spatial grid point. The synthesis module combines the dynamic information contained in the feature subsets of essential causal relationships with the confidence weights reflecting local cognition to temporally regulate the background seismic activity level and obtain grid-point differentiated earthquake occurrence probabilities. The analysis module, based on the grid-point differentiated earthquake occurrence probabilities and combined with ground motion attenuation patterns and site effects, conducts probabilistic seismic hazard analysis to obtain high-resolution ground motion hazard zoning. The risk mapping module integrates the high-resolution ground motion hazard zoning with disaster-bearing body information to generate a high-resolution earthquake risk map.

Images