A typhoon disaster annual risk dynamic assessment technology based on time series vector autoregressive model

Abstract

The application relates to the technical field of typhoon risk assessment, and discloses a typhoon disaster annual risk dynamic assessment technology based on a time series vector autoregressive model, which comprises the following steps: determining main disaster-causing factors affecting typhoon disaster losses by using an EOF / MVEOF method; using a VAR method to unify the positions, wind speed intensities and precipitation of historical typhoon paths to four-dimensional time series vectors, estimating parameters of a time series model, and simulating new typhoon time series; using the relationship between global sea temperature data and typhoon landing frequencies to establish a support vector machine model, and annually predicting typhoon frequencies; and using a typhoon simulation data set to evaluate the intensities of main disaster-causing factors, annually predicting typhoon losses by constructing a quantitative relationship between the main disaster-causing factors and the losses. The application can provide technical support for the management, emergency and disaster reduction of global typhoons, and has important significance for reducing global disaster losses and global sustainable development.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of typhoon risk assessment, and particularly relates to a typhoon disaster annual risk dynamic assessment technology based on a time series vector autoregressive model. BACKGROUND

[0002] Under the global warming climate conditions, the climate is gradually changing, and the natural disasters caused by extreme climate show a clear increasing trend, which has increasingly intensified negative impacts on the healthy development of various aspects of society and economy. Among various extreme weather events, the tropical cyclone causes particularly serious disaster losses. The wind disasters, rain disasters and storm surges caused by the tropical cyclone have negative impacts on the economic development and personal safety of coastal cities. The tropical cyclone is a cyclonic vortex, which generally occurs over the oceans in the tropics and subtropics, and the vortex can be divided into super typhoon, strong typhoon, typhoon, strong tropical storm, tropical storm and tropical depression according to the intensity. Frequent occurrence and landing of typhoon disasters often cause traffic interruption, residence damage, crop reduction, and more seriously, personnel casualty and property loss, etc., which bring irreversible negative impacts to the relevant regions. Mastering the characteristics and laws of the generation and development of typhoon disasters and adopting scientific methods to reduce, avoid and transfer risks not only help to improve the typhoon disaster prevention and mitigation capability, but also can reduce disaster losses, and further protect the sustainable development of economy and society.

[0003] Annual typhoon disaster risk assessment has a positive significance for guiding disaster prevention and mitigation, disaster relief and recovery and reconstruction. Annual typhoon disaster risk assessment is an analysis, judgment and simulation calculation of the influence area, intensity and loss level of typhoon disasters in a region. At present, the annual natural disaster risk assessment is still in the exploratory stage. In terms of assessment methods, the scientificity and operability of the assessment process need to be further improved, and the reliability and practicality of the assessment results need to be further improved. In terms of expression of the assessment results, the readability and popularity need to be further improved. SUMMARY

[0004] The present application aims to provide a typhoon disaster annual risk assessment method based on a time series vector autoregressive model, so as to at least partially solve the problems in the prior art.

[0005] More specifically, the typhoon disaster annual risk assessment method based on a time series vector autoregressive model of the present application comprises the following steps:

[0006] S1: obtaining relevant historical typhoon data of a study area, wherein the data includes meteorological data, typhoon path data, historical typhoon disaster loss data and sea temperature data;

[0007] S2: typhoon path simulation, comprising:

[0008] A time-series vector autoregressive model (VAR) is established based on historical typhoon data, and a VAR model library of historical typhoon paths is formed.

[0009] Based on the spatial density distribution of historical typhoon origins, the location of the typhoon origin is randomly generated. At the same time, based on the wind speed and precipitation of historical typhoon origins, the wind speed and precipitation of random origins are randomly generated using the Monte Carlo simulation method. Based on the distribution of stopping wind speeds along historical typhoon paths, the stopping wind speed is randomly generated for each typhoon path using the Monte Carlo simulation method.

[0010] Using VAR models to simulate typhoon paths includes randomly selecting VAR models from the historical typhoon path VAR model library based on the location of the randomly generated typhoon starting point, and simulating the typhoon path. When the wind speed in the random vector calculated by the VAR model is less than the stop wind speed, the VAR simulation calculation is stopped, thereby forming a random path of the future typhoon and obtaining a random typhoon path database.

[0011] S3: Establish a typhoon landfall frequency prediction model, including:

[0012] Spatial correlation analysis was performed on historical sea surface temperature (SST) data and typhoon landfall frequency. Based on the magnitude of the correlation, key seasons and key regions of SST that affect typhoon landfall frequency were identified. The SST data of these key seasons and key regions were decomposed using empirical orthogonal function (EOF) decomposition, and the first few principal components were selected as candidate independent variables for the typhoon landfall frequency prediction model. Based on the candidate independent variables, a stepwise regression method was used to screen out the factors for predicting typhoon landfall frequency by modifying the significance level or the number of regression steps. A support vector machine model was then established as the typhoon landfall frequency prediction model.

[0013] S4: Construct a quantitative relationship between major disaster-causing factors and typhoon losses, and establish a typhoon loss prediction model: and

[0014] S5 conducts annual risk assessments of typhoon disasters based on VAR models, typhoon landfall frequency prediction models, and typhoon loss prediction models.

[0015] According to an embodiment of the present invention, in step S1, the meteorological data includes air pressure, temperature, precipitation, evaporation, relative humidity, wind direction and wind speed, and the typhoon path data includes the intensity and latitude and longitude of the tropical cyclone in the sea area.

[0016] According to an embodiment of the present invention, in step S2, the formation of the spatial density distribution of the starting point of historical typhoons includes: classifying historical typhoon paths into five categories based on the starting, passing, and ending regions; for different categories of historical typhoon paths, generating a rectangular range of the starting point based on the boundary of the starting point; dividing the rectangular range into a latitude and longitude grid with a resolution of 1 degree; calculating the frequency of the starting point within each latitude and longitude grid; generating the starting point frequency distribution; and generating a kernel density diffusion result with a resolution of 1 degree based on the point file of the starting point frequency distribution.

[0017] According to an embodiment of the present invention, step S2 further includes verifying the accuracy of the path simulation by using the distribution of typhoon path landfall points and the intensity distribution of typhoon path landfall points.

[0018] According to an embodiment of the present invention, step S3 further includes using HoldOut test and cross-validation methods to test the prediction model of the support vector machine method.

[0019] According to the embodiment of the present invention, in step S4, based on the historical typhoon data in step S1, spatial correlation analysis is performed on the maximum precipitation, maximum wind speed, total precipitation, and total wind speed that passed through the study area in each year and within the study area, and on the historical typhoon disaster loss data. Based on the magnitude of the correlation, candidate meteorological factors are screened. If there is only one candidate meteorological factor, EOF decomposition is performed; if there are multiple candidate meteorological factors, MVEOF decomposition is performed to obtain the principal component of the candidate meteorological factor, i.e., the time coefficient PC, and the correlation coefficient between the disaster loss and the disaster loss. The principal component with the largest correlation coefficient is the main disaster-causing factor in the study area. The historical typhoon loss data in the study area is used as the dependent variable, and a univariate linear regression model is constructed with the main disaster-causing factor to obtain the typhoon loss prediction model.

[0020] According to an embodiment of the present invention, in step S4, if the first principal component of the candidate meteorological factor decomposed by EOF or MVEOF has a correlation of 0.8 or higher with the typhoon loss, and if the cumulative variance contribution rate of the first two modal variances to the total variance exceeds 80%, then the first principal component of the candidate meteorological factor is determined to be the main disaster-causing factor.

[0021] According to an embodiment of the present invention, step S5, the annual risk assessment of typhoon disasters includes:

[0022] Based on sea surface temperature (SST) data from the autumn and winter of the previous year and the spring and summer of the current year, the frequency of typhoon landfalls in the predicted year is obtained using a typhoon landfall frequency prediction model. A corresponding number of passing typhoon paths are randomly selected from a random typhoon path database, and the wind speed and precipitation distribution along each typhoon path are obtained. Based on the random typhoon paths, the intensity of candidate meteorological factors in the study area at the time of each typhoon path's formation is calculated. The overall intensity of candidate meteorological factors in the predicted year is calculated, and then combined with historical candidate meteorological factor data for each year. EOF or MVEOF decomposition is then performed again to obtain the intensity of the main disaster-causing factors in the predicted year.

[0023] By incorporating the main disaster-causing factors of the predicted year into the typhoon loss prediction model, the predicted typhoon losses are obtained, and the annual typhoon disaster risk is derived.

[0024] According to an embodiment of the present invention, in step S5, the overall intensity of the candidate meteorological factors for the predicted year is obtained by summing the intensity of the candidate meteorological factors in the study area under each typhoon path.

[0025] According to an embodiment of the present invention, in step S5, the annual risk of typhoon disaster is represented by an exceedance probability curve and an annual expected loss.

[0026] The beneficial effects of this invention's typhoon disaster annual risk assessment method based on a time-series vector autoregression model are as follows: Using Empirical Orthogonal Function (EOF) and Multivariate Empirical Orthogonal Function (MVEOF) methods, the spatial modal distribution and principal component variation characteristics of precipitation and wind speed are obtained. Based on the correlation between the principal components and direct economic losses, the main disaster-causing factors affecting typhoon disaster losses are identified. The VAR method is used to unify the location (latitude and longitude), wind speed intensity, and precipitation of historical typhoon paths onto a four-dimensional time-series vector. By estimating the parameters of the time-series model, new typhoon time series are simulated. Using the analysis of global sea surface temperature data, characteristic sea areas and variables with significant correlations to typhoon landfall frequency are selected. Stepwise regression and support vector machine models are established to achieve annual predictions of typhoon frequency in the study area before the typhoon season. Finally, the intensity of the main disaster-causing factors is assessed using a typhoon simulation dataset. The annual prediction of typhoon damage is made by establishing a quantitative relationship between the main disaster-causing factors and losses. It can provide technical support for global typhoon management, emergency response and disaster reduction. Reliable typhoon path prediction models are of great significance for mitigating global disaster losses and global sustainable development.

[0027] The features and advantages of the present invention will be described in detail through embodiments and in conjunction with the accompanying drawings. Attached Figure Description

[0028] Figure 1 This is a flowchart illustrating the dynamic annual risk assessment technology for typhoon disasters based on a time-series vector autoregression model according to an embodiment of the present invention.

[0029] Figure 2 This is a flowchart of the method for determining the main disaster-causing factors in the annual risk dynamic assessment technology of typhoon disasters based on a time series vector autoregression model according to an embodiment of the present invention.

[0030] Figure 3 This is a graph showing the typhoon landfall frequency and support vector machine results in the study area according to the embodiment of the present invention;

[0031] Figure 4 These are the typhoon landfall frequency and support vector machine HoldOut test results for the study area according to the embodiment of the present invention.

[0032] Figure 5 This is the cross-validation result of typhoon landfall frequency and support vector machine in the study area according to the embodiment of the present invention;

[0033] Figure 6 This is a correlation distribution diagram of the maximum wind speed, maximum precipitation, and direct economic loss in the study area according to the embodiment of the present invention;

[0034] Figure 7 This is a correlation distribution map of total precipitation, total wind speed, and direct economic losses in the study area according to the implementation scheme of the present invention;

[0035] Figure 8 It is the variance explained rate of the EOF decomposition of the maximum wind speed in the study area according to the embodiment of the present invention;

[0036] Figure 9 This is a correlation diagram between the top five principal components of the maximum wind speed in the study area according to the embodiment of the present invention and the direct economic loss;

[0037] Figure 10 This is a correlation diagram between the first principal component of the maximum wind speed in the study area and the direct economic loss according to the embodiment of the present invention.

[0038] Figure 11 This describes the relationship between the main disaster-causing factors and losses in the study area according to the embodiment of the present invention.

[0039] Figure 12 These are the disaster-causing factor prediction results for the study area according to the embodiment of the present invention;

[0040] Figure 13 These are the exceedance probability curves for different numbers of passing typhoons in the study area according to the embodiments of the present invention;

[0041] Figure 14 This is the expected annual loss for the study area under different numbers of passing typhoons according to the implementation scheme of the present invention. Detailed Implementation

[0042] To make the objectives, technical solutions, and advantages of this invention clearer, the invention will be further described in detail below with reference to the accompanying drawings and embodiments. However, it should be understood that the specific embodiments described herein are merely illustrative and are not intended to limit the scope of the invention.

[0043] It should be understood that the tools / models involved in this invention, such as Vector Autoregression (VAR), Maximum Likelihood Estimation, Kernel Density Diffusion, Monte Carlo Simulation Method, Empirical Orthogonal Function (EOF), Multivariate Empirical Orthogonal Function (MVEOF), Correlation Analysis, Support Vector Machine Model, Wind Speed ​​Decay Model, Exceedance Probability Curve, and Annual Expected Loss, are known in themselves. Therefore, this invention focuses on how to use and combine these technical means / features to realize the annual risk dynamic assessment technology for typhoon disasters of this invention.

[0044] Figure 1 This is a flowchart illustrating the annual risk dynamic assessment technology for typhoon disasters based on a time-series vector autoregressive model, as described in this invention. Referring to the accompanying drawings, the annual risk dynamic assessment technology for typhoon disasters based on a time-series vector autoregressive model may include the following steps:

[0045] First, historical typhoon data relevant to the study area can be obtained through various means. Historical typhoon data can include meteorological data, typhoon track data, typhoon damage data, and sea surface temperature data, etc.

[0046] More specifically, daily surface climate data datasets can be downloaded from the China Meteorological Data Network (http: / / data.cma.cn / ). The "China Daily Surface Climate Data Dataset (V3.0)" contains meteorological indicators such as air pressure, temperature, precipitation, evaporation, relative humidity, wind direction, and wind speed from 824 baseline and basic meteorological stations in China since January 1951. The CMA-STI Northwest Pacific Tropical Cyclone Optimal Track Dataset can be downloaded from the Tropical Cyclone Data Center of the China Meteorological Administration (http: / / tcdata.typhoon.org.cn / zjljsjj_zlhq.html). This dataset includes data on tropical cyclones passing through the Northwest Pacific Ocean (including the South China Sea, north of the equator, and west of 180°E) that have passed through my country since 1949. It records the intensity and latitude / longitude of tropical cyclones every 6 hours, and the intensity data includes the minimum central pressure and the 2-minute average maximum wind speed near the center. Typhoon loss data can be obtained from local (study area) statistical data, including direct economic losses and deaths. Sea surface temperature data can be obtained, for example, from the China Global Surface Temperature Data Set (China-MST).

[0047] Second, we will simulate typhoon paths, which mainly includes classifying historical typhoon paths, establishing a time series vector autoregressive model (VAR model), and simulating typhoon paths.

[0048] To improve the accuracy of stochastic typhoon track simulations, historical typhoon tracks can be classified. Studies have revealed clear patterns in the distribution of historical typhoon tracks. Therefore, the observation area—China and the Northwest Pacific—was first divided into four regions: Region 0 mainly includes the area east of approximately 130°E and south of approximately 24°N; Region 1 mainly includes the area west of approximately 130°E and south of approximately 24°N, and the area west of approximately 120°E and north of approximately 24°N; Region 2 mainly includes the area between approximately 120°E and approximately 147°E and north of approximately 24°N; and Region 3 mainly includes the area east of approximately 147°E and north of approximately 24°N. The classification of historical typhoon tracks is determined by the regions where they originate, pass through, and end. The specific classification algorithm is shown in Table 1. Typhoon tracks that do not belong to these four categories are classified into a fifth category.

[0049] Table 1: Path Classification Algorithm Table

[0050]

[0051] Historical typhoons were classified using a typhoon classification algorithm, resulting in five categories as shown in the table above. Category 1 paths are mainly distributed in the southwestern Pacific Ocean. Category 2 paths are primarily westward-extending, originating in the northwestern Pacific and gradually extending westward until they disappear. Category 3 paths are bendable, forming relatively closer to the northwest compared to Category 5. Category 4 paths are mainly distributed in the northern Pacific Ocean; the rest are classified as Category 5.

[0052] Based on historical typhoon data, an inter-sequence vector autoregressive model (VAR model) was established, and a VAR model library of historical typhoon paths was formed.

[0053] Each point on a typhoon's path can be represented by latitude and longitude. Therefore, the position of a typhoon at any given moment is related to the latitude and longitude of the previous moment. Furthermore, the wind speed and precipitation at the previous moment can also affect the position, wind speed, and precipitation at the next moment. The specific VAR model for typhoon path simulation using a vector autoregression (VAR) model is as follows:

[0054]

[0055] in, Representing the Step time series; The value of the time series vector at time t is composed of the typhoon's path latitude and longitude and its intensity. Typhoon intensity includes both wind speed and rainfall. It is a four-dimensional vector. For longitude, For latitude, The 2-minute average maximum wind speed near the center at each point along the path, for The precipitation in the study area at any given time can be represented by the latitude, longitude, wind speed, and precipitation of the previous time t-1; a four-dimensional matrix. It is the parameter matrix to be estimated. express and The relationship between them; It is a random disturbance term.

[0056] More specifically, historical data can be processed before building the VAR model, including detrending and differencing historical typhoon tracks, and obtaining data through maximum likelihood estimation. Two parameter matrices form a VAR model library of historical typhoon paths.

[0057] In an embodiment of the present invention, a VAR time series model is used to simulate the typhoon path. The simulation of the typhoon path may include the simulation of the starting and ending points, the simulation of the path itself, and verification.

[0058] After dividing the region, the location of typhoon origins can be randomly simulated based on the spatial density distribution of historical typhoon origins. More specifically, the origins of typhoons can be determined and their spatial density distribution can be formed based on historical statistical data. The results show that origins are mainly distributed in the Northwest Pacific Ocean between 110°E and 180°E longitude, and between the equator and 30°N latitude. Then, the location of typhoon origins can be further randomly simulated, as follows:

[0059] For different types of historical typhoon paths, a rectangular range of the starting point is generated based on the boundary of the starting point. The rectangular range is then divided into latitude and longitude grids with a resolution of 1 degree. The frequency of the starting point in each latitude and longitude grid is calculated to generate the starting point frequency distribution. Based on the point file of the starting point frequency distribution, a kernel density diffusion result with a resolution of 1 degree is generated.

[0060] Kernel density diffusion calculates the density value of the data around each point. Conceptually, a smooth surface is mounted at each point. The surface value is highest at the point's location and decreases with increasing distance from the point, reaching zero at a search radius distance from the point. The kernel density value at each output raster cell is calculated by superimposing all density values ​​covering the center of the raster cell. Based on the kernel density distribution, y random starting points are generated. Since the kernel density distribution consists of point features spaced at 1 degree latitude and longitude intervals, the random starting points are also spaced at 1 degree latitude and longitude intervals.

[0061] In addition, while randomly generating the starting point of a typhoon, the wind speed and precipitation at the random starting point can be calculated using the Monte Carlo simulation method based on historical typhoon starting point wind speed and precipitation data.

[0062] The simulation of the endpoint mainly involves randomly generating the stopping wind speed for each path using the Monte Carlo simulation method, based on the historical distribution of stopping wind speeds along typhoon paths. .

[0063] In an embodiment of the present invention, a VAR time series model is used to simulate the path. When the wind speed in the random vector calculated by the VAR model is less than the stopping wind speed... The VAR simulation is stopped, thus forming a random path for future typhoons and obtaining a random typhoon path database. This invention uses a VAR time series model to establish a random simulation of typhoon paths, which not only considers geographical similarity but also the autocorrelation of time characteristics before and after the typhoon path. Its simulation results are closer to historical typhoon paths and better reflect the macroscopic characteristics of landfalling typhoons.

[0064] The accuracy of typhoon track simulations can be verified using the distribution of typhoon landfall points and their intensity. That is, the future typhoon track results obtained from the VAR time series model are compared with historical typhoon track distributions. The accuracy of typhoon track simulations can also be measured by the maximum wind speed near the typhoon's center at landfall. This maximum wind speed near the center is calculated by taking the linear difference between the maximum wind speeds near the typhoon's center before and after landfall.

[0065] Third, a typhoon landfall frequency prediction model is established. Predicting the annual typhoon landfall frequency is fundamental to assessing the typhoon disaster risk for the following year. To predict the annual typhoon landfall frequency and ultimately assess future typhoon risks, this invention utilizes a stepwise regression model of annual typhoon landfall frequency. Variables are selected from numerous factors, with the annual typhoon landfall frequency as the primary factor. The relationship between the annual typhoon landfall frequency and the preceding sea surface temperature field is established. Significant factors are screened using stepwise regression, and a support vector machine model is established to predict the annual typhoon landfall frequency in the study area. The model is validated using the HoldOut test and cross-validation. This invention can utilize the `reg` procedure in SAS and the `fitcecoc` function in Matlab to build the model.

[0066] More specifically, spatial correlation analysis was performed on historical sea surface temperature (SST) data and typhoon landfall frequency. Correlation models were established using global SST data from the autumn and winter of the previous year and the spring and summer of the current year, respectively, to obtain the spatial correlation distribution of "SST-typhoon landfall frequency." Regions and time periods (seasons) with high correlation were selected as key time periods (seasons) and key regions affecting typhoon landfall frequency. The SST data from key time periods and key regions were decomposed using empirical orthogonal function (EOF), and the first few (e.g., ten) principal components were selected as candidate independent variables for the typhoon landfall frequency prediction model.

[0067] Based on the candidate factors, stepwise regression can be performed using the reg procedure step in SAS. By modifying the significance level (e.g., setting the significance level to 0.15) or the number of regression steps, multiple factors for forecasting the frequency of typhoon landfall can be selected. Table 2 shows the stepwise regression results of an embodiment of the present invention.

[0068] Table 2: Factors selected from the stepwise regression equation for predicted wind landfall frequency

[0069]

[0070] Finally, using the selected factors for forecasting, a support vector machine model is established as a prediction model for typhoon landfall frequency. Furthermore, the establishment of the typhoon landfall frequency prediction model also includes validation, for example, using HoldOut tests and cross-validation methods to test the prediction model of the support vector machine method. HoldOut tests and cross-validation are known to those skilled in the art and will not be elaborated upon here.

[0071] Fourth, establish a typhoon loss prediction model. Establishing a typhoon loss prediction model includes identifying the main disaster-causing factors and building the model.

[0072] Figure 2 This is a flowchart illustrating the method for determining the main disaster-causing factors in the annual risk dynamic assessment technology for typhoon disasters based on a time series vector autoregression model according to an embodiment of the present invention.

[0073] Typhoons can cause wind damage and precipitation. Therefore, total precipitation, total wind speed, and maximum precipitation and wind speed during a typhoon's passage are all potential meteorological factors contributing to typhoon losses. The calculation and processing of maximum precipitation and maximum wind speed, as well as total precipitation and total wind speed during the passage of a typhoon and within the study area, are as follows:

[0074] (1) Maximum precipitation and maximum wind speed passing through the study area and within the study area

[0075] Step 1: Using the CMA-STI Northwest Pacific Tropical Cyclone Optimal Track Dataset, statistically analyze all typhoons passing through the study area. Step 2: Analyze the time taken by all typhoons to pass through the study area. Step 3: Analyze the daily average wind speed and 20:00-20:00 precipitation at each station during the time each typhoon passed through the study area. Step 4: Calculate the maximum values ​​for the daily average wind speed and 20:00-20:00 precipitation for each typhoon during its passage through the study area each year. If there are multiple typhoons in a given year, sum the multiple maximum values ​​to obtain the maximum precipitation and maximum wind speed for each typhoon at each station in the study area each year.

[0076] (2) Total precipitation and total wind speed passing through and within the study area

[0077] Step 1: Using the CMA-STI Northwest Pacific Tropical Cyclone Optimal Track Dataset, count all typhoons passing through the study area; Step 2: Count the time all typhoons passed through the study area; Step 3: Count the daily average wind speed and 20-20 hour precipitation at each station during the time all typhoons passed through the study area each year; Step 4: Calculate the sum of the daily average wind speed and 20-20 hour precipitation at each station for each typhoon passing through the study area each year, and obtain the total precipitation and total wind speed of each station passing through and within the study area.

[0078] By analyzing meteorological station data and Northwest Pacific tropical cyclone data, physical quantities (meteorological factors) that may affect typhoon losses can be obtained, such as maximum precipitation and maximum wind speed passing through the study area, and total precipitation and total wind speed passing through the study area. Spatial correlation analysis is performed on these physical quantities with historical typhoon loss data, and physical quantities with high correlation to typhoon losses are selected as candidate meteorological factors. EOF or MVEOF decomposition is performed on the candidate meteorological factors to obtain the spatiotemporal distribution characteristics of the meteorological data. Then, correlation analysis is performed between the time coefficient (PC) and typhoon loss data, and the principal components (time coefficients) with high correlation are selected as the main disaster-causing factors in the study area. For example, if the first principal component of the candidate meteorological factor's EOF or MVEOF decomposition has a correlation of 0.8 or higher with typhoon losses, and the cumulative variance contribution rate of the first two modal variances to the total variance exceeds 80%, then the first principal component of that candidate meteorological factor can be identified as the main disaster-causing factor. Therefore, candidate meteorological factors can include maximum wind speed, total precipitation, and total wind speed.

[0079] For example, when the candidate meteorological factor is the maximum wind speed, the maximum wind speed at 10 meters above sea level at each point along the typhoon path can be used according to the wind speed attenuation model, and the distance between the latitude and longitude of each point along the typhoon path and the center of the latitude and longitude of the study area can be used. Calculate the maximum wind speed in the study area when the typhoon path is generated. :

[0080]

[0081] Where R is the radius of maximum wind speed; r is the distance between the latitude and longitude of each point on the typhoon path and the center of the study area; V10 is the wind speed at 10 meters above sea level at each point on the typhoon path; X is the attenuation index, which can be a uniformly distributed random number in the range of [0.5, 0.7].

[0082] Then, using historical typhoon loss data of the study area as the dependent variable, a univariate linear regression model was constructed with the main disaster-causing factors to obtain the typhoon loss prediction model.

[0083] Fifth, conduct annual typhoon risk assessments, which mainly include forecasts of typhoon landfall frequency, major disaster-causing factors, and typhoon damage.

[0084] Based on sea surface temperature (SST) data from the autumn and winter of the previous year and the spring and summer of the current year, the frequency of typhoon landfalls in the predicted year is obtained using a typhoon landfall frequency prediction model. A corresponding number of passing typhoon paths are randomly selected from a random typhoon path database, and the wind speed and precipitation distribution along the typhoon paths are obtained. Based on the random typhoon paths, the intensity of candidate meteorological factors in the study area at the time of typhoon path generation is calculated. The overall intensity of candidate meteorological factors in the predicted year is calculated. For example, if the candidate meteorological factor is the total annual precipitation, the total precipitation in the study area under each typhoon path is summed according to the typhoon landfall frequency in the predicted year to obtain the overall intensity of candidate meteorological factors in the predicted year. This is then combined with the candidate meteorological factor data of historical years, and EOF or MVEOF decomposition is performed again to obtain the intensity of the main disaster-causing factors in the predicted year.

[0085] By incorporating the predicted major disaster-causing factors into the S5 typhoon loss prediction model, the predicted typhoon losses are obtained, and the annual typhoon disaster risk is derived. In other words, the loss corresponding to the intensity of the major disaster-causing factors is calculated using the typhoon loss prediction model.

[0086] By repeatedly performing the above process, we can obtain the exceedance probability curve and annual expected loss for different numbers of passing typhoons, which is also the annual risk of typhoon disasters.

[0087] Example: A province prone to typhoon disasters was selected as the study area, and the method of this invention was used for evaluation.

[0088] Data from 1956 to 2016 in the study area were collected, including historical observation datasets and meteorological datasets such as the CMA-STI Northwest Pacific Tropical Cyclone Optimal Track Dataset, daily precipitation and wind speed data from relevant meteorological stations, and global sea surface temperature datasets, as well as historical typhoon disaster datasets compiled and statistically analyzed by relevant departments.

[0089] A VAR model was built based on historical typhoon data, and a support vector machine model was built based on historical sea surface temperature data and historical typhoon landfall frequency data, i.e., a typhoon landfall frequency prediction model. The VAR model and the typhoon landfall frequency prediction model were used to predict the typhoon landfall frequency and typhoon paths in the study area, and the results were validated. See [link to results]. Figure 3 , Figure 4 and Figure 5The results show that the support vector machine (SVM) can obtain the predicted values ​​for each sample point. Correlation analysis between the original data and the predicted values ​​yields a correlation coefficient exceeding 0.8 and a sum of squared errors of 0.5424. Using the HoldOut test to validate the SVM predictions, correlation analysis between the original data and the test values ​​yields a correlation coefficient exceeding 0.97 and a sum of squared errors of 0.0847. Using the cross-validation test to validate the SVM predictions, correlation analysis between the original data and the test values ​​yields a correlation coefficient of 0.6382 and a sum of squared errors of 1.1017.

[0090] Next, a typhoon loss prediction model was established, as follows:

[0091] Correlation analysis was conducted between the annual direct economic losses caused by typhoon disasters in the province and the maximum precipitation, maximum wind speed, total precipitation, and total wind speed during the typhoon season at various stations in the province. Correlation coefficients were calculated, and then the differences between the correlation coefficients of each station were processed to obtain the spatial correlation distribution between typhoon disaster losses in the province and these meteorological factors. See [link to details]. Figure 6 and Figure 7 .

[0092] from Figures 6-7 It can be seen that the spatial correlation between different precipitation and wind speed data and direct economic losses from typhoon disasters varies. Specifically, the maximum wind speed, total wind speed, and direct economic losses are highly correlated when the typhoon passes through the province and is within its territory, with correlation coefficients across the province remaining above 0.5, especially for maximum wind speed. Precipitation shows a lower correlation with direct economic losses. Therefore, maximum wind speed is initially selected as a candidate meteorological factor.

[0093] Subsequently, the candidate meteorological factors were decomposed using EOF to obtain the principal components (time coefficient PC) of each candidate meteorological factor and their correlation coefficients with disaster losses. Based on the magnitude of the correlation coefficients, the maximum wind speed was determined to be a suitable candidate meteorological factor for the study area. More specifically, the maximum wind speed passing through the province and within its territory was decomposed using EOF, and the variance explained by each mode is shown in the figure. Figure 8 As shown, the correlation coefficients between the obtained principal components and the direct economic losses from typhoon disasters in the province are calculated. According to the variance explanation rate, the first mode variance contributes as much as 80.4% to the total variance, and the second mode variance contributes as much as 7.1% to the total variance. The cumulative variance contribution of the first two modes to the total variance exceeds 80%. Therefore, the first and second modes can be selected to represent the spatiotemporal distribution of the maximum wind speed passing through the study area. Figure 9 The figure shows the correlation coefficients between the first five principal components and the direct economic loss after performing EOF decomposition on the maximum wind speed passing through the province and within the province. The correlation between the first principal component and the direct economic loss of the typhoon disaster in the province reached 0.8. Figure 10The graphs show line plots of the first principal component and direct economic losses, revealing that the fluctuations of the first principal component and direct economic losses are largely consistent. Therefore, the candidate meteorological factor affecting typhoon losses in this province is the maximum wind speed passing through and within the province, with the first principal component of the maximum wind speed EOF decomposition serving as the primary disaster-causing factor.

[0094] A predictive model for the main disaster-causing factors and direct economic losses in the study area was obtained through univariate linear regression. (See appendix.) Figure 11 .

[0095] Finally, a risk simulation assessment is conducted, including: based on the obtained typhoon landfall frequency, randomly selecting a corresponding number of passing typhoon paths from the random typhoon path database; calculating the maximum wind speed in the study area at the time of typhoon path formation (candidate meteorological factor) according to the wind speed attenuation model; and then combining this with historical candidate meteorological factor data to re-perform EOF decomposition to obtain the intensity of the predicted main disaster-causing factor time coefficient PC, i.e., the main disaster-causing factor. (See Appendix) Figure 12 Substituting the predicted results of the main disaster-causing factors (intensity of the time coefficient PC) into the typhoon loss prediction model, the predicted typhoon loss is obtained. Repeating the above process, the exceedance probability curves and annual expected losses for different numbers of passing typhoons are obtained, such as... Figure 13 and Figure 14 .

[0096] This invention, when using a VAR model to predict typhoon paths, quantitatively assesses the wind speed and precipitation distribution in the study area at various time points along the typhoon path. Utilizing the statistical relationship between hazard-causing factors and losses, it calculates the economic losses of typhoons in the study area from the perspective of major hazard-causing factors. Since losses from multiple typhoons may collectively impact the losses in the study area, this invention predicts typhoon losses by considering both frequency and the intensity of major hazard-causing factors simultaneously, thereby achieving a quantitative risk assessment of the annual economic loss risk from typhoon disasters in the study area.

[0097] The above embodiments are illustrative of the present invention and are not intended to limit the present invention. Any simple modifications to the present invention are within the scope of protection of the present invention.

Claims

1. A dynamic annual risk assessment method for typhoon disasters based on a time-series vector autoregressive model, characterized in that, Includes the following steps: S1: Obtain historical typhoon data related to the study area, including meteorological data, typhoon path data, historical typhoon disaster loss data, and sea surface temperature data; S2: Typhoon path simulation, including: establishing a time series vector autoregressive model (VAR) based on historical typhoon data, and forming a library of historical typhoon path VAR models; Based on the spatial density distribution of historical typhoon origins, the location of the typhoon origin is randomly generated. At the same time, based on the wind speed and precipitation of historical typhoon origins, the wind speed and precipitation of random origins are randomly generated using the Monte Carlo simulation method. Based on the distribution of stopping wind speeds along historical typhoon paths, the stopping wind speed is randomly generated for each typhoon path using the Monte Carlo simulation method. Using VAR models to simulate typhoon paths includes randomly selecting VAR models from the historical typhoon path VAR model library based on the location of the randomly generated typhoon starting point, and simulating the typhoon path. When the wind speed in the random vector calculated by the VAR model is less than the stop wind speed, the VAR simulation calculation is stopped, thereby forming a random path of the future typhoon and obtaining a random typhoon path database. S3: Establish a typhoon landfall frequency prediction model, including: Spatial correlation analysis was performed on historical sea surface temperature (SST) data and typhoon landfall frequency. Based on the magnitude of the correlation, key seasons and key regions of SST that affect typhoon landfall frequency were identified. The SST data of these key seasons and key regions were decomposed using empirical orthogonal function (EOF), and the top few principal components were selected as candidate independent variables for the typhoon landfall frequency prediction model. Based on the candidate independent variables, a stepwise regression method was used to screen out the factors for predicting typhoon landfall frequency by modifying the significance level or the number of regression steps. A support vector machine model was then established as the typhoon landfall frequency prediction model. S4: Construct a quantitative relationship between major disaster-causing factors and typhoon losses, and establish a typhoon loss prediction model: and S5. Based on the VAR model, the typhoon landfall frequency prediction model, and the typhoon loss prediction model, an annual typhoon disaster risk assessment is conducted. This annual typhoon disaster risk assessment includes: Based on sea surface temperature (SST) data from the autumn and winter of the previous year and the spring and summer of the current year, the frequency of typhoon landfalls in the predicted year is obtained using a typhoon landfall frequency prediction model. A corresponding number of passing typhoon paths are randomly selected from a random typhoon path database, and the wind speed and precipitation distribution along each typhoon path are obtained. Based on the random typhoon paths, the intensity of candidate meteorological factors in the study area at the time of each typhoon path's formation is calculated. The overall intensity of candidate meteorological factors in the predicted year is calculated, and then combined with historical candidate meteorological factor data for each year. EOF or MVEOF decomposition is then performed again to obtain the intensity of the main disaster-causing factors in the predicted year. By incorporating the main disaster-causing factors of the predicted year into the typhoon loss prediction model, the predicted typhoon losses are obtained, and the annual typhoon disaster risk is derived.

2. The method for dynamic annual risk assessment of typhoon disasters based on a time series vector autoregressive model as described in claim 1, characterized in that, In step S1, the meteorological data includes air pressure, temperature, precipitation, evaporation, relative humidity, wind direction and wind speed, and the typhoon path data includes the intensity and latitude and longitude of the tropical cyclone in the sea area.

3. The method for dynamic annual risk assessment of typhoon disasters based on a time series vector autoregressive model as described in claim 1, characterized in that, In step S2, the formation of the spatial density distribution of the starting point of historical typhoons includes: classifying historical typhoon paths into five categories based on the starting, passing, and ending regions; for different categories of historical typhoon paths, generating a rectangular range of the starting point based on the boundary of the starting point; dividing the rectangular range into a latitude and longitude grid with a resolution of 1 degree; calculating the frequency of the starting point within each latitude and longitude grid; generating the starting point frequency distribution; and generating a kernel density diffusion result with a resolution of 1 degree based on the point file of the starting point frequency distribution.

4. The method for dynamic annual risk assessment of typhoon disasters based on a time series vector autoregressive model as described in claim 1, characterized in that, Step S2 also includes verifying the accuracy of the path simulation by using the distribution of typhoon path landfall points and the intensity distribution of typhoon path landfall points.

5. The method for dynamic annual risk assessment of typhoon disasters based on a time series vector autoregressive model as described in claim 1, characterized in that, Step S3 also includes using HoldOut test and cross-validation methods to test the prediction model of the support vector machine method.

6. The method for dynamic annual risk assessment of typhoon disasters based on a time series vector autoregressive model as described in claim 1, characterized in that, In step S4, based on the historical typhoon data from step S1, spatial correlation analysis is performed on the maximum precipitation, maximum wind speed, total precipitation, and total wind speed that passed through the study area in each year and within the study area, and the historical typhoon disaster loss data. Based on the magnitude of the correlation, candidate meteorological factors are selected. If there is only one candidate meteorological factor, perform EOF decomposition. If there are multiple candidate meteorological factors, perform multivariate empirical orthogonal function MVEOF decomposition to obtain the principal components of the candidate meteorological factors, namely the time coefficient PC, and the correlation coefficient between them and disaster losses. The principal component with the largest correlation coefficient is the main disaster-causing factor in the study area. Using the historical typhoon loss data of the study area as the dependent variable, construct a univariate linear regression model with the main disaster-causing factor to obtain the typhoon loss prediction model.

7. The method for dynamic annual risk assessment of typhoon disasters based on a time series vector autoregressive model as described in claim 6, characterized in that: In step S4, if the first principal component of the candidate meteorological factor decomposed by EOF or MVEOF has a correlation of 0.8 or higher with the typhoon loss, and the cumulative variance contribution rate of the first two modal variances to the total variance exceeds 80%, then the first principal component of the candidate meteorological factor is determined to be the main disaster-causing factor.

8. The method for dynamic annual risk assessment of typhoon disasters based on a time series vector autoregressive model as described in claim 1, characterized in that: In step S5, the overall intensity of the candidate meteorological factors for the predicted year is obtained by summing the intensity of the candidate meteorological factors in the study area under each typhoon path.

9. The method for dynamic annual risk assessment of typhoon disasters based on a time series vector autoregressive model as described in claim 1, characterized in that: In step S5, the annual risk of typhoon disaster is represented by the exceedance probability curve and the annual expected loss.

Images