A method for risk assessment of regional compound disaster chain under influence of tropical cyclone

Abstract

The present application relates to a kind of tropical cyclone influence under the method for evaluating the risk of regional composite disaster chain, comprising: obtaining the historical multi-source observation data of tropical cyclone and pre-processing, construct multidimensional disaster element dataset;Based on multidimensional disaster element dataset analysis each disaster element's nonlinear variation characteristic parameter, and based on nonlinear variation characteristic parameter constructs multidimensional disaster element quantification model and comprehensive disaster intensity index;The historical change track of comprehensive disaster intensity index is modeled, and abnormal mode and space-time characteristic parameter are extracted, and a spatio-temporal evolution prediction model of composite disaster chain is constructed;Real-time multi-source observation data of tropical cyclone is acquired, input into spatio-temporal evolution prediction model of composite disaster chain to predict the spatio-temporal evolution trend of composite disaster chain, and the spatio-temporal evolution trend of composite disaster chain is quantitatively analyzed, and the risk level of different regions is determined.The present application realizes the quantitative characterization and evaluation of the spatio-temporal evolution process of tropical cyclone composite disaster chain by integrating multi-source heterogeneous data.

Description

Technical Field

[0001] This invention relates to the field of disaster assessment technology, and in particular to a method for assessing the risk of regional complex disaster chains under the influence of tropical cyclones. Background Technology

[0002] Tropical cyclones, as one of the most destructive natural disasters globally, cause enormous loss of life and property to coastal areas every year. With the intensification of global climate change and the acceleration of coastal urbanization, the phenomenon of complex disaster chains triggered by tropical cyclones is becoming increasingly prominent, making risk assessment a core research direction in disaster prevention and mitigation. Current risk assessment methods mainly analyze single disaster types, lacking a deep understanding of the interaction mechanisms between various secondary disasters triggered by tropical cyclones. Existing assessment systems often treat rainstorms, floods, and typhoons separately, failing to accurately reflect the overall risk characteristics of complex disaster chains, leading to significant discrepancies between assessment results and actual disaster losses.

[0003] A key challenge in assessing the risk of complex disaster chains caused by tropical cyclones lies in the comprehensive quantification of multidimensional disaster elements. Tropical cyclone systems encompass multiple physical elements, including cyclone intensity, rainfall distribution, and wind field structure. These elements exhibit highly complex nonlinear variations across spatiotemporal scales, making it difficult for traditional single-indicator assessment methods to fully characterize their dynamic evolution. The difficulty in quantifying multidimensional elements further leads to technical bottlenecks in multi-source data fusion. Information from different sources, such as remote sensing data, ground observation data, and numerical simulation results, differs significantly in spatiotemporal resolution, accuracy standards, and data formats. The lack of an effective data integration framework makes it impossible to construct a unified assessment benchmark. The existence of data fusion problems directly restricts the accurate identification of the spatiotemporal evolution characteristics of disaster chains. Due to the inability to establish quantitative correlations between various secondary disasters, existing methods cannot accurately assess the risk levels of each link in a complex disaster chain. Therefore, this invention proposes a method for assessing the risk of regional complex disaster chains under the influence of tropical cyclones. Summary of the Invention

[0004] The purpose of this invention is to provide a method for risk assessment of regional complex disaster chains under the influence of tropical cyclones. It constructs a comprehensive evaluation system covering multi-dimensional indicators such as cyclone intensity classification, rainfall distribution, wind field simulation, and flood inundation range, and achieves quantitative characterization of the spatiotemporal evolution process of tropical cyclone complex disaster chains by effectively integrating multi-source heterogeneous data.

[0005] To achieve the above objectives, the present invention provides the following solution:

[0006] A method for assessing the risk of regional complex disaster chains under the influence of tropical cyclones includes:

[0007] Historical multi-source observation data of tropical cyclones were acquired and preprocessed to construct a multi-dimensional disaster element dataset.

[0008] Based on the multidimensional disaster element dataset, the nonlinear change characteristic parameters of each disaster element are analyzed, and a multidimensional disaster element quantitative model and a comprehensive disaster intensity index are constructed based on the nonlinear change characteristic parameters.

[0009] The historical change trajectory of the comprehensive disaster intensity index is modeled, and the spatiotemporal characteristic parameters of abnormal patterns are extracted to construct a spatiotemporal evolution prediction model for the composite disaster chain.

[0010] Real-time multi-source observation data of tropical cyclones are acquired and input into the spatiotemporal evolution prediction model of the composite disaster chain to predict the spatiotemporal evolution trend of the composite disaster chain. The spatiotemporal evolution trend of the composite disaster chain is quantitatively analyzed to determine the risk level of different regions and complete the risk assessment of the regional composite disaster chain under the influence of tropical cyclones.

[0011] Optionally, historical multi-source observation data of tropical cyclones are acquired and preprocessed to construct a multi-dimensional disaster element dataset, including:

[0012] Historical multi-source observational data of tropical cyclones were obtained from satellite remote sensing, ground meteorological stations, and ocean buoys, including cyclone intensity data, rainfall distribution data, and wind field structure data;

[0013] After performing spatiotemporal registration, interpolation, and correction on the historical multi-source observation data, the characteristic parameters of cyclone intensity, rainfall distribution, and wind field structure are extracted to obtain a set of characteristic parameters.

[0014] Based on the set of feature parameters, the random forest algorithm is used to comprehensively analyze the cyclone intensity data and wind field structure data to determine the core influence area and intensity level distribution of the cyclone.

[0015] By combining the core impact area and intensity level distribution of the cyclone with rainfall distribution data for spatial overlay analysis, a risk distribution layer is obtained and disaster element data is fused to construct the multidimensional disaster element dataset.

[0016] Optionally, the nonlinear variation characteristic parameters of each disaster element are analyzed based on the multidimensional disaster element dataset, including:

[0017] The multidimensional disaster element dataset is analyzed using a nonlinear change analysis method to obtain preliminary nonlinear change characteristic parameters and dynamic change descriptions of each disaster element;

[0018] Based on the dynamic change description, the rate of change of cyclone intensity is calculated, and abrupt events are identified through the rate of change to obtain the abrupt event identifier of cyclone intensity.

[0019] Based on the mutation event identifier, the spatial characteristics of rainfall distribution during the occurrence of the mutation event are analyzed in conjunction with the spatial pattern of rainfall distribution to obtain the nonlinear change parameters of rainfall distribution.

[0020] Based on the aforementioned nonlinear variation parameters, the change characteristics of the wind field structure under the influence of abrupt events are extracted by combining the evolution trajectory of the wind field structure, and the dynamic evolution mode of the wind field structure is obtained.

[0021] The degree of correlation between various disaster elements is determined by identifying the mutation event, nonlinear change parameters, and dynamic evolution patterns. Based on the degree of correlation, the preliminary nonlinear change characteristic parameters of each disaster element are integrated and processed to obtain the final nonlinear change characteristic parameters.

[0022] Optionally, a multi-dimensional disaster element quantification model and a comprehensive disaster intensity index are constructed based on the aforementioned nonlinear variation characteristic parameters, including:

[0023] The nonlinear variation characteristic parameters are standardized to obtain the disaster element matrix, i.e., the multidimensional disaster element quantification model;

[0024] The disaster element matrix is ​​weighted and fused to obtain the comprehensive disaster intensity value and determine the comprehensive disaster intensity level distribution. If the comprehensive disaster intensity level distribution exceeds the preset level threshold, the weight coefficients are adjusted and optimized, and the comprehensive disaster intensity value is recalculated until it meets the preset threshold range to obtain the comprehensive disaster intensity index.

[0025] Optionally, the historical change trajectory of the comprehensive disaster intensity index is modeled, and anomalous patterns and spatiotemporal characteristic parameters are extracted to construct a spatiotemporal evolution prediction model for the composite disaster chain, including:

[0026] Long-term records of comprehensive disaster intensity indicators are obtained, and time series analysis is used to extract features from the trajectory changes of comprehensive disaster intensity to determine key time nodes and spatial distribution patterns.

[0027] Based on key time nodes and spatial distribution patterns, the deviation value between the current comprehensive disaster intensity index and historical data for the same period is calculated, and the deviation distribution results are obtained.

[0028] If the deviation distribution result exceeds the preset statistical threshold, the comprehensive disaster intensity index at the current moment is determined to be an abnormal mode, and the spatiotemporal characteristic parameters of the abnormal mode are obtained.

[0029] By combining the aforementioned spatiotemporal characteristic parameters and long-term recorded data, a spatiotemporal evolution prediction model for the complex disaster chain is constructed.

[0030] Optionally, real-time multi-source observation data of tropical cyclones are acquired and input into the spatiotemporal evolution prediction model of the complex disaster chain to predict the spatiotemporal evolution trend of the complex disaster chain, including:

[0031] Acquire and integrate real-time multi-source observation data of tropical cyclones to obtain a fused dataset;

[0032] The fused dataset is structured to determine the spatiotemporal distribution pattern of disaster chain-related features, and abnormal features are judged based on the spatiotemporal distribution pattern. The weight distribution of key influencing factors in the abnormal features is obtained using the random forest algorithm.

[0033] The parameters of the spatiotemporal evolution prediction model of the composite disaster chain are updated based on the weight distribution of the key influencing factors. The fused dataset is then input into the updated spatiotemporal evolution prediction model of the composite disaster chain to obtain the spatiotemporal evolution trend of the composite disaster chain.

[0034] Optionally, a quantitative analysis of the spatiotemporal evolution trend of the complex disaster chain is performed to determine the risk level of different regions, including:

[0035] The scope of impact is determined based on the spatiotemporal evolution trend of the aforementioned complex disaster chain;

[0036] Obtain the population density of each region and analyze the degree of overlap between the affected area and each region;

[0037] Based on the population density and degree of overlap, the risk level of each region is determined.

[0038] Optionally, the method further includes: predicting disaster losses in each region according to a preset risk loss assessment system, wherein the risk loss assessment system is constructed by the correlation between historical disaster loss data and risk levels.

[0039] The beneficial effects of this invention are as follows:

[0040] This invention acquires multi-source observation data and performs spatiotemporal registration and standardization. It then employs a nonlinear variation feature extraction method to analyze the evolution characteristics of disaster elements, constructs a multi-dimensional disaster element quantification model, and integrates it into a comprehensive disaster intensity index. A spatiotemporal evolution analysis algorithm is used to model and predict the disaster intensity index, and the model parameters are dynamically adjusted based on real-time monitoring data to predict the evolution trend of complex disaster chains. Furthermore, a risk assessment algorithm is used to quantify and analyze the prediction results, determining the risk level distribution in different regions. This invention can effectively assess the risk of complex disaster chains triggered by tropical cyclones, providing a scientific basis for disaster prevention and mitigation decision-making. Attached Figure Description

[0041] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the drawings used in the embodiments will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0042] Figure 1 This is a flowchart illustrating a method for assessing the risk of regional complex disaster chains under the influence of tropical cyclones, according to an embodiment of the present invention. Detailed Implementation

[0043] 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. 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.

[0044] To make the above-mentioned objects, features and advantages of the present invention more apparent and understandable, the present invention will be further described in detail below with reference to the accompanying drawings and specific embodiments.

[0045] This embodiment provides a method for assessing the risk of regional complex disaster chains under the influence of tropical cyclones, such as... Figure 1 As shown, it includes:

[0046] Historical multi-source observation data of tropical cyclones were acquired and preprocessed to construct a multi-dimensional disaster element dataset.

[0047] Based on the multidimensional disaster element dataset, the nonlinear change characteristic parameters of each disaster element are analyzed, and a multidimensional disaster element quantitative model and a comprehensive disaster intensity index are constructed based on the nonlinear change characteristic parameters.

[0048] The historical change trajectory of the comprehensive disaster intensity index is modeled, and the spatiotemporal characteristic parameters of abnormal patterns are extracted to construct a spatiotemporal evolution prediction model for the composite disaster chain.

[0049] Real-time multi-source observation data of tropical cyclones are acquired and input into the spatiotemporal evolution prediction model of the composite disaster chain to predict the spatiotemporal evolution trend of the composite disaster chain. The spatiotemporal evolution trend of the composite disaster chain is quantitatively analyzed to determine the risk level of different regions and complete the risk assessment of the regional composite disaster chain under the influence of tropical cyclones.

[0050] Specifically, this embodiment acquires multi-source observation data and performs spatiotemporal registration and standardization. It then employs a nonlinear variation feature extraction method to analyze the evolution characteristics of disaster elements, constructs a multi-dimensional disaster element quantification model, and integrates it into a comprehensive disaster intensity index. A spatiotemporal evolution analysis algorithm is used to model and predict the disaster intensity index, and the model parameters are dynamically adjusted based on real-time monitoring data to predict the evolution trend of complex disaster chains. Furthermore, a risk assessment algorithm is used to quantify and analyze the prediction results, determining the risk level distribution in different regions. This embodiment can effectively assess the risk of complex disaster chains triggered by tropical cyclones, providing a scientific basis for disaster prevention and mitigation decision-making.

[0051] Furthermore, historical multi-source observation data of tropical cyclones were acquired and preprocessed to construct a multi-dimensional disaster element dataset, including:

[0052] Historical multi-source observational data of tropical cyclones were obtained from satellite remote sensing, ground meteorological stations, and ocean buoys, including cyclone intensity data, rainfall distribution data, and wind field structure data;

[0053] After performing spatiotemporal registration, interpolation, and correction on the historical multi-source observation data, the characteristic parameters of cyclone intensity, rainfall distribution, and wind field structure are extracted to obtain a set of characteristic parameters.

[0054] Based on the set of feature parameters, the random forest algorithm is used to comprehensively analyze the cyclone intensity data and wind field structure data to determine the core influence area and intensity level distribution of the cyclone.

[0055] By combining the core impact area and intensity level distribution of the cyclone with rainfall distribution data for spatial overlay analysis, a risk distribution layer is obtained and disaster element data is fused to construct the multidimensional disaster element dataset.

[0056] Specifically, this embodiment first acquires data on the cloud structure and intensity distribution of the cyclone using satellite remote sensing imagery. For example, using infrared channel data from the Fengyun-4 satellite, areas with cloud top temperatures below -60°C are extracted as the cyclone core region. Reflectivity is calculated using the visible light channel, and the cyclone intensity is estimated to be level 12 (wind speed approximately 33 m / s). Simultaneously, ground meteorological stations record hourly updated air pressure and wind speed data. For example, a station records an air pressure of 980 hPa and a wind speed of 25 m / s. A regional air pressure field is generated using interpolation algorithms (such as inverse distance weighting), with the error controlled within ±2 hPa. Ocean buoy monitoring information provides sea surface temperature and wave height data. For example, a buoy records a sea surface temperature of 28.5°C and a wave height of 3.2 meters. Time series analysis is used to predict the cyclone path shift trend, and the prediction accuracy is optimized to ±5 kilometers using a Kalman filter algorithm. Subsequently, for spatiotemporal registration, data with different resolutions and coordinate systems were standardized. For example, satellite imagery has a resolution of 1 km, ground station data is distributed as points, and ocean buoy data has a time interval of 30 minutes. Kriging interpolation was used to unify the spatial resolution to a 500-meter grid, and the time interval was standardized to once per hour. Projection transformation was performed using the WGS84 coordinate system, with errors controlled within ±0.5 meters to ensure data consistency. Finally, a standardized multidimensional disaster element dataset was generated, integrating cyclone intensity (wind speed 33 m / s), rainfall distribution (hourly rainfall 50 mm, based on satellite inversion and ground data fusion), and wind field structure (based on Doppler radar wind speed vector field, maximum wind speed radius approximately 50 km) into a NetCDF format file.

[0057] Furthermore, based on the aforementioned multidimensional disaster element dataset, the nonlinear variation characteristic parameters of each disaster element are analyzed, including:

[0058] The multidimensional disaster element dataset is analyzed using a nonlinear change analysis method to obtain preliminary nonlinear change characteristic parameters and dynamic change descriptions of each disaster element;

[0059] Based on the dynamic change description, the rate of change of cyclone intensity is calculated, and abrupt events are identified through the rate of change to obtain the abrupt event identifier of cyclone intensity.

[0060] Based on the mutation event identifier, the spatial characteristics of rainfall distribution during the occurrence of the mutation event are analyzed in conjunction with the spatial pattern of rainfall distribution to obtain the nonlinear change parameters of rainfall distribution.

[0061] Based on the aforementioned nonlinear variation parameters, the change characteristics of the wind field structure under the influence of abrupt events are extracted by combining the evolution trajectory of the wind field structure, and the dynamic evolution mode of the wind field structure is obtained.

[0062] The degree of correlation between various disaster elements is determined by identifying the mutation event, nonlinear change parameters, and dynamic evolution patterns. Based on the degree of correlation, the preliminary nonlinear change characteristic parameters of each disaster element are integrated and processed to obtain the final nonlinear change characteristic parameters.

[0063] Specifically, in this embodiment, the time series data of cyclone intensity is first processed using wavelet transform. Assuming the dataset contains cyclone intensity values ​​recorded hourly for a time span of 72 hours, with intensity values ​​ranging from 20 to 80 m / s, the signal is decomposed to level 3 using Discrete Wavelet Transform (DWT). High-frequency components are extracted using Daubechies wavelet basis functions, and wavelet coefficients at each time point are calculated. It is found that the rate of change of intensity reaches 5.2 m / s / hour at the 24th hour, exceeding the preset threshold of 3.0 m / s / hour. The system automatically marks this as an intensity abrupt change event and records the abrupt change time and the correlation coefficient of 0.85 as feature parameters. Next, the spatial pattern of rainfall distribution was analyzed. Principal Component Analysis (PCA) combined with nonlinear kernel functions (such as radial basis function kernels) was used to reduce the dimensionality of the rainfall grid data (resolution 0.1 degrees × 0.1 degrees, coverage 1000 km × 1000 km). The first two principal components were extracted, with contribution rates of 65% and 20%, respectively. By calculating the spatial gradient change rate of the principal components, the rainfall intensity gradient in the central region was found to be 12.5 mm / h / km, characterizing the nonlinear concentrated features of the rainfall distribution. A parameter table was generated to store the gradient values ​​and the distribution pattern coefficient 0.72. Subsequently, the evolution trajectory of the wind field structure was analyzed. The Lyapunov exponent calculation method in chaos theory was used. Wind field vector data (updated every 6 hours, wind speed range 5 to 50 m / s) was input, and a phase space was constructed using the time-delay embedding method. The delay time was set to 3 hours, and the embedding dimension was 5. The maximum Lyapunov exponent was calculated to be 0.15, indicating that the wind field evolution has strong nonlinear chaotic characteristics. The system automatically saved the exponent value and the trajectory fractal dimension of 1.8 as feature parameters. Finally, the aforementioned cyclone intensity abrupt change events, rainfall distribution characteristic parameters, and wind field chaos index are integrated into the disaster element database to form a set of nonlinear change characteristic parameters.

[0064] Furthermore, based on the aforementioned nonlinear variation characteristic parameters, a multidimensional disaster element quantification model and a comprehensive disaster intensity index are constructed, including:

[0065] The nonlinear variation characteristic parameters are standardized to obtain the disaster element matrix, i.e., the multidimensional disaster element quantification model;

[0066] The disaster element matrix is ​​weighted and fused to obtain the comprehensive disaster intensity value and determine the comprehensive disaster intensity level distribution. If the comprehensive disaster intensity level distribution exceeds the preset level threshold, the weight coefficients are adjusted and optimized, and the comprehensive disaster intensity value is recalculated until it meets the preset threshold range to obtain the comprehensive disaster intensity index.

[0067] Specifically, this embodiment focuses on quantifying the cyclone intensity index. Based on historical meteorological data, the cyclone's central pressure is 960 hPa. A standardized formula is used to convert this pressure into an intensity index: Index = (Standard Pressure - Actual Pressure) / Standard Deviation. The standard pressure is 1013 hPa, and the standard deviation is 20 hPa. The calculated index is (1013 - 960) / 20 = 2.65, indicating a high cyclone intensity. Next, regarding the rainfall intensity distribution matrix, assuming the rainfall data in a certain area is a 5x5 grid matrix with a central point rainfall of 100 mm, gradually decreasing to 20 mm at the periphery, a smooth distribution matrix is ​​generated using a spatial interpolation algorithm (such as inverse distance weighting). The weight of each grid point is calculated, with the central point having a weight of 1, and the periphery weights decreasing inversely proportional to the square of the distance. The maximum value of the standardized rainfall intensity matrix is ​​0.85. Then, for the wind field velocity vector field, assuming a wind speed of 25 m / s at a certain point and a wind direction of 45 degrees northeast, the vector is decomposed into an x-axis component of 17.68 m / s and a y-axis component of 17.68 m / s. Combining data from multiple observation points within the region, a continuous wind field distribution map is generated using Kriging interpolation, and the average wind speed intensity is calculated to be 0.78. Finally, a weighted fusion algorithm is used to integrate the above indicators, setting the weight of the cyclone intensity index to 0.4, the weight of the rainfall intensity to 0.3, and the weight of the wind field velocity to 0.3. The comprehensive disaster intensity index is calculated as 2.65 × 0.4 + 0.85 × 0.3 + 0.78 × 0.3 = 1.06 + 0.255 + 0.234 = 1.549. The evaluation value is 1.549, indicating that the disaster intensity is at a medium-to-high level.

[0068] Furthermore, the historical trajectory of the comprehensive disaster intensity index is modeled, and anomalous patterns and spatiotemporal characteristic parameters are extracted to construct a spatiotemporal evolution prediction model for the composite disaster chain, including:

[0069] Long-term records of comprehensive disaster intensity indicators are obtained, and time series analysis is used to extract features from the trajectory changes of comprehensive disaster intensity to determine key time nodes and spatial distribution patterns.

[0070] Based on key time nodes and spatial distribution patterns, the deviation value between the current comprehensive disaster intensity index and historical data for the same period is calculated, and the deviation distribution results are obtained.

[0071] If the deviation distribution result exceeds the preset statistical threshold, the comprehensive disaster intensity index at the current moment is determined to be an abnormal mode, and the spatiotemporal characteristic parameters of the abnormal mode are obtained.

[0072] By combining the aforementioned spatiotemporal characteristic parameters and long-term recorded data, a spatiotemporal evolution prediction model for the complex disaster chain is constructed.

[0073] Specifically, this embodiment first processes historical data on disaster intensity indicators using a spatiotemporal evolution analysis algorithm. Disaster intensity indicator data from the past 10 years was collected, recorded monthly for each year, totaling 120 data points, with a data range between 0 and 100. For example, the disaster intensity indicator for January 2020 was 45.5, for January 2021 it was 48.2, and so on. The ARIMA (Autoregressive Integral Moving Average) model from time series analysis is used for modeling, with parameters set to (p=1, d=1, q=1). By fitting the historical data, the predicted value and residual for each time point are obtained, and the mean and standard deviation for the same historical period are calculated. For example, the historical mean for January is 46.8, and the standard deviation is 2.5. Next, a statistical threshold is set as the mean plus or minus two standard deviations, i.e., the threshold range is 41.8 to 51.8. If the disaster intensity indicator for January 2023 is 55.3, exceeding the threshold range, it is judged as an abnormal evolution pattern. Furthermore, by analyzing the temporal and spatial distribution of anomalies and combining Geographic Information System (GIS) technology, disaster intensity indicators are correlated with geographical locations. Assuming anomalies are concentrated in a certain area (e.g., coordinates between 120-122 degrees East longitude and 30-32 degrees North latitude), a disaster intensity heatmap is generated using spatial interpolation algorithms (e.g., Kriging interpolation). This reveals that the areas of intensity anomalies overlap with historical disaster chains (e.g., areas prone to flooding-landslides). Finally, based on the above analysis, a disaster chain spatiotemporal evolution prediction model is constructed. Using a Bayesian network algorithm, historical disaster chain data and current anomaly indicators are input, and model parameters are output. For example, the disaster chain trigger probability is 0.75, and the propagation speed parameter is 0.3 units / day. The model predicts that within the next month, the disaster chain may expand its impact area to the region between 123 degrees East longitude and 33 degrees North latitude.

[0074] Furthermore, real-time multi-source observation data of tropical cyclones are acquired and input into the spatiotemporal evolution prediction model of the complex disaster chain to predict the spatiotemporal evolution trend of the complex disaster chain, including:

[0075] Acquire and integrate real-time multi-source observation data of tropical cyclones to obtain a fused dataset;

[0076] The fused dataset is structured to determine the spatiotemporal distribution pattern of disaster chain-related features, and abnormal features are judged based on the spatiotemporal distribution pattern. The weight distribution of key influencing factors in the abnormal features is obtained using the random forest algorithm.

[0077] The parameters of the spatiotemporal evolution prediction model of the composite disaster chain are updated based on the weight distribution of the key influencing factors. The fused dataset is then input into the updated spatiotemporal evolution prediction model of the composite disaster chain to obtain the spatiotemporal evolution trend of the composite disaster chain.

[0078] Specifically, this embodiment first obtains cyclone path information from satellite remote sensing data. Assuming the current cyclone center is located at 120.5 degrees east longitude and 25.3 degrees north latitude, moving at a speed of 15 kilometers per hour in a west-northwest direction, the cyclone path is smoothed using a Kalman filter algorithm. Combined with historical trajectory data, the predicted location for the next 6 hours is calculated to be 119.8 degrees east longitude and 26.1 degrees north latitude, with an error range controlled within ±0.2 degrees. Next, precipitation cloud image data is fused, and radar monitoring is used to obtain precipitation intensity distribution. Assuming the current core area rainfall is 50 millimeters per hour, a convolutional neural network model is used to extract cloud image features, predicting that the rainfall will increase to 60 millimeters per hour in the next 3 hours, expanding the coverage area to a radius of 100 kilometers. Simultaneously, wind field distribution information is collected. Assuming the current maximum wind speed is 30 meters per second and the wind field radius is 200 kilometers, based on the numerical weather prediction model (WRF) and combined with real-time wind speed data to update boundary conditions, the wind speed is predicted to increase to 35 meters per second in the next 12 hours, expanding the wind field radius to 250 kilometers. Subsequently, the aforementioned multi-source data was input into the spatiotemporal evolution prediction model. Using a Long Short-Term Memory (LSTM) network algorithm, combined with cyclone path, rainfall, and wind field parameters, the model weights were dynamically adjusted to predict the evolution trend of the complex disaster chain. The model concluded that the cyclone might trigger a storm surge within the next 24 hours, affecting coastal areas with an estimated storm surge height of 1.5 meters and an impact range of approximately 50 kilometers of coastline. Finally, a Bayesian update method was used to correct the model parameters hourly based on real-time monitoring data. For example, the cyclone movement speed error was adjusted from ±0.2 degrees to ±0.15 degrees to improve prediction accuracy and ensure the reliability of the disaster chain evolution trend prediction results.

[0079] Furthermore, a quantitative analysis of the spatiotemporal evolution trend of the aforementioned complex disaster chain is conducted to determine the risk levels of different regions, including:

[0080] The scope of impact is determined based on the spatiotemporal evolution trend of the aforementioned complex disaster chain;

[0081] Obtain the population density of each region and analyze the degree of overlap between the affected area and each region;

[0082] Based on the population density and degree of overlap, the risk level of each region is determined.

[0083] Furthermore, the method also includes: predicting disaster losses in each region according to a preset risk loss assessment system, wherein the risk loss assessment system is constructed by the correlation between historical disaster loss data and risk levels.

[0084] Specifically, this embodiment first obtains disaster impact range data through a disaster chain evolution prediction model. It assumes that a Bayesian network model based on historical data predicts a combined flood and landslide disaster probability of 0.75 in a certain region, with an impact range radius of 10 kilometers and center point coordinates of (X: 120.5, Y: 30.2). Next, a spatial overlay analysis algorithm is used to compare the predicted impact range with geographic information data of densely populated areas. Assuming the center point coordinates of the densely populated area are (X: 120.6, Y: 30.1), the calculated overlap ratio between the two areas is 0.65, while the preset danger threshold is 0.5. Since the overlap exceeds the threshold, the area is automatically identified as a potentially high-risk area. Subsequently, the risk level was further quantified using a risk assessment algorithm. Combining the disaster occurrence probability of 0.75, population density of 5000 people per square kilometer, and overlap area ratio of 0.65, the risk index was calculated as 0.75 × 0.65 × 5000 = 2437.5. Assuming the risk level classification criteria are: an index greater than 2000 is high risk, 1000 to 2000 is medium risk, and less than 1000 is low risk, this area was therefore classified as high risk. Finally, a risk level distribution map was generated based on Geographic Information System (GIS) technology. The risk index of all areas was spatially visualized. Assuming there are 10 sub-regions, 3 with a risk index exceeding 2000 were marked as red (high risk), 5 with a risk index between 1000 and 2000 were marked as yellow (medium risk), and 2 with a risk index below 1000 were marked as green (low risk). The distribution map was output and stored as a vector format file.

[0085] The embodiments described above are merely preferred embodiments of the present invention and are not intended to limit the scope of the present invention. Various modifications and improvements made to the technical solutions of the present invention by those skilled in the art without departing from the spirit of the present invention should fall within the protection scope defined by the claims of the present invention.

Claims

1. A method for assessing the risk of regional complex disaster chains under the influence of tropical cyclones, characterized in that, include: Historical multi-source observation data of tropical cyclones were acquired and preprocessed to construct a multi-dimensional disaster element dataset. The analysis of nonlinear change characteristic parameters of each disaster element based on the multidimensional disaster element dataset is as follows: Nonlinear change analysis methods are used to analyze the multidimensional disaster element dataset to obtain preliminary nonlinear change characteristic parameters and dynamic change descriptions of each disaster element; based on the dynamic change descriptions, the rate of change of cyclone intensity is calculated, and abrupt events are identified using the rate of change values ​​to obtain abrupt event identifiers for cyclone intensity; based on the abrupt event identifiers, the spatial characteristics of rainfall distribution during abrupt events are analyzed in conjunction with spatial patterns of rainfall distribution to obtain nonlinear change parameters of rainfall distribution; based on the nonlinear change parameters, the change characteristics of wind field structure under the influence of abrupt events are extracted in conjunction with the evolution trajectory of wind field structure to obtain dynamic evolution patterns of wind field structure; the degree of correlation influence of each disaster element is determined through the abrupt event identifiers, nonlinear change parameters, and dynamic evolution patterns, and the preliminary nonlinear change characteristic parameters of each disaster element are integrated based on the degree of correlation influence to obtain final nonlinear change characteristic parameters; Based on the aforementioned nonlinear variation characteristic parameters, a multidimensional disaster element quantification model and a comprehensive disaster intensity index are constructed: the nonlinear variation characteristic parameters are standardized to obtain a disaster element matrix, i.e., the multidimensional disaster element quantification model; the disaster element matrix is ​​weighted and fused to obtain a comprehensive disaster intensity value and determine the comprehensive disaster intensity level distribution; if the comprehensive disaster intensity level distribution exceeds a preset level threshold, the weight coefficients are adjusted and optimized, and the comprehensive disaster intensity value is recalculated until it meets the preset threshold range to obtain the comprehensive disaster intensity index; The historical change trajectory of the comprehensive disaster intensity index is modeled, and the spatiotemporal characteristic parameters of abnormal patterns are extracted to construct a spatiotemporal evolution prediction model for the composite disaster chain. Real-time multi-source observation data of tropical cyclones are acquired and input into the spatiotemporal evolution prediction model of the composite disaster chain to predict the spatiotemporal evolution trend of the composite disaster chain. The spatiotemporal evolution trend of the composite disaster chain is quantitatively analyzed to determine the risk level of different regions and complete the risk assessment of the regional composite disaster chain under the influence of tropical cyclones.

2. The method for assessing the risk of regional complex disaster chains under the influence of tropical cyclones according to claim 1, characterized in that, Historical multi-source observation data of tropical cyclones were acquired and preprocessed to construct a multi-dimensional disaster element dataset, including: Historical multi-source observational data of tropical cyclones were obtained from satellite remote sensing, ground meteorological stations, and ocean buoys, including cyclone intensity data, rainfall distribution data, and wind field structure data; After performing spatiotemporal registration, interpolation, and correction on the historical multi-source observation data, the characteristic parameters of cyclone intensity, rainfall distribution, and wind field structure are extracted to obtain a set of characteristic parameters. Based on the set of feature parameters, the random forest algorithm is used to comprehensively analyze the cyclone intensity data and wind field structure data to determine the core influence area and intensity level distribution of the cyclone. By combining the core impact area and intensity level distribution of the cyclone with rainfall distribution data for spatial overlay analysis, a risk distribution layer is obtained and disaster element data is fused to construct the multidimensional disaster element dataset.

3. The method for assessing the risk of regional complex disaster chains under the influence of tropical cyclones according to claim 1, characterized in that, The historical trajectory of the comprehensive disaster intensity index is modeled, and anomalous patterns and spatiotemporal characteristic parameters are extracted to construct a spatiotemporal evolution prediction model for the composite disaster chain, including: Long-term records of comprehensive disaster intensity indicators are obtained, and time series analysis is used to extract features from the trajectory changes of comprehensive disaster intensity to determine key time nodes and spatial distribution patterns. Based on key time nodes and spatial distribution patterns, the deviation value between the current comprehensive disaster intensity index and historical data for the same period is calculated, and the deviation distribution results are obtained. If the deviation distribution result exceeds the preset statistical threshold, the comprehensive disaster intensity index at the current moment is determined to be an abnormal mode, and the spatiotemporal characteristic parameters of the abnormal mode are obtained. By combining the aforementioned spatiotemporal characteristic parameters and long-term recorded data, a spatiotemporal evolution prediction model for the complex disaster chain is constructed.

4. The method for assessing the risk of regional complex disaster chains under the influence of tropical cyclones according to claim 1, characterized in that, Acquire real-time multi-source observation data of tropical cyclones and input them into the spatiotemporal evolution prediction model of the complex disaster chain to predict the spatiotemporal evolution trend of the complex disaster chain, including: Acquire and integrate real-time multi-source observation data of tropical cyclones to obtain a fused dataset; The fused dataset is structured to determine the spatiotemporal distribution pattern of disaster chain-related features, and abnormal features are judged based on the spatiotemporal distribution pattern. The weight distribution of key influencing factors in the abnormal features is obtained using the random forest algorithm. The parameters of the spatiotemporal evolution prediction model of the composite disaster chain are updated based on the weight distribution of the key influencing factors. The fused dataset is then input into the updated spatiotemporal evolution prediction model of the composite disaster chain to obtain the spatiotemporal evolution trend of the composite disaster chain.

5. The method for assessing the risk of regional complex disaster chains under the influence of tropical cyclones according to claim 1, characterized in that, A quantitative analysis of the spatiotemporal evolution trend of the aforementioned complex disaster chain is conducted to determine the risk levels of different regions, including: The scope of impact is determined based on the spatiotemporal evolution trend of the aforementioned complex disaster chain; Obtain the population density of each region and analyze the degree of overlap between the affected area and each region; Based on the population density and degree of overlap, the risk level of each region is determined.

6. The method for assessing the risk of regional complex disaster chains under the influence of tropical cyclones according to any one of claims 1-5, characterized in that, The method further includes: predicting disaster losses in each region based on a preset risk loss assessment system, wherein the risk loss assessment system is constructed by the correlation between historical disaster loss data and risk levels.

Images