A regional frequency analysis method of marine environmental dynamic elements based on spatial interactive sampling

Abstract

The present invention discloses a regional frequency analysis method of marine environment dynamic elements based on spatial interactive sampling, which belongs to the field of coast and marine engineering technology. The method comprises the following steps: obtaining n-year time series historical data of marine environment dynamic elements of multiple sites, extracting annual extreme value samples of marine environment dynamic elements of each site by annual extreme value method; calculating sample linear moment and linear moment proportional coefficient of each site; performing consistency test on all sites according to sample linear moment proportional coefficient; clustering the remaining sites that meet the consistency test to divide homogeneous regions; selecting a suitable probability distribution model for each homogeneous region; applying exponential wave method to calculate the estimated value of specific return period of each site; calculating the confidence interval of the estimated value of specific return period of each site based on simulation results; the method can more stably characterize the frequency distribution characteristics of extreme samples in the region, and significantly reduce the instability of the estimated results that may be caused by fluctuations in single site data.

Description

Technical Field

[0001] The present invention belongs to the field of coastal and marine engineering technology, and specifically relates to a method for regional frequency analysis of marine environmental dynamic elements based on spatial interactive sampling. Background Art

[0002] The calculation of marine environmental dynamic elements is very important for marine and coastal engineering design, installation and operation. The calculation of reasonable and effective design values ​​of dynamic elements can not only ensure the safety of the project, but also reduce the cost of the project. These two points are also the key to marine and coastal engineering. However, the time span of historical data of marine environmental dynamic elements is limited, and it is impossible to cover enough extreme events to directly determine the design value of a specific return period. Therefore, people often use probability distribution models to fit these historical data for long return period calculations. In the early days, people mainly used the full sample method, that is, selecting a data set of dynamic elements (usually consisting of a series of three-hour data), and then applying the full sample of the data set to fit the probability distribution model for long return period calculations. However, extrapolation beyond the data length may provide unreliable estimates. This is because common probability distribution models (such as Weibull or lognormal distribution) are difficult to well describe the tail distribution characteristics of dynamic environmental elements.

[0003] Extreme value analysis is another method for estimating the design values ​​of marine dynamic environmental elements. This method extracts extreme samples from the time-series dynamic element data of a single site and fits the corresponding probability distribution model for inference. Due to theoretical proof and asymptotic tail properties, the tail distribution characteristics of environmental dynamic elements are better characterized, so more reasonable and reliable inferred values ​​can be obtained compared to the initial distribution method. When the estimated recurrence period is close to the data time span, the estimated value is often reliable and has a smaller uncertainty (i.e., a smaller confidence interval). However, as the estimated recurrence period increases, the uncertainty of the estimated value increases significantly, and the upper and lower confidence limits may deviate seriously from the estimated value. Therefore, this method is difficult to apply to major marine and coastal engineering design and protection that require long recurrence period estimated values. Summary of the invention

[0004] In view of this, the present invention discloses a method for regional frequency analysis of marine environmental dynamic elements based on spatial interactive sampling. The method mainly solves the problem of large confidence intervals of estimated values ​​of major marine and coastal structures under long recurrence periods due to limited historical data of marine environmental dynamic elements.

[0005] The object of the present invention is achieved by the following technical solutions:

[0006] A method for regional frequency analysis of marine environmental dynamic elements based on spatial interactive sampling includes the following steps:

[0007] S1. Obtaining dynamic elements of the marine environment at multiple sitesn Yearly time series historical data, including n More than 20 years;

[0008] S2, using the annual extreme value method to extract the annual extreme value samples of the marine environmental dynamic elements at each station;

[0009] S3, calculate the sample linear moment and linear moment proportional coefficient of each station. The sample linear moment mainly includes the first four orders (i.e. ), the linear moment proportionality coefficient mainly includes the linear deviation coefficient ( ), linear skewness coefficient ( ) and the linear kurtosis coefficient ( );

[0010] S4. Perform consistency check on all sites based on the sample linear moment proportionality coefficient and screen out sites with obviously inconsistent distribution characteristics;

[0011] S5. Based on the geographical characteristics of the sites and the distribution characteristics of the samples, the remaining sites that meet the consistency test are clustered to divide the homogeneous areas;

[0012] S6. Select an appropriate probability distribution model for each homogeneous region through goodness of fit test;

[0013] S7. Based on the selected probability distribution model, the exponential wave method is applied to calculate the estimated value of the specific return period for each station;

[0014] S8. Resimulate the homogeneous area through Monte Carlo simulation and calculate the confidence interval of the estimated value of the specific return period of each station based on the simulation results.

[0015] Preferably, the ocean dynamic environment factor of step S1 is the significant wave height, and step S2 is to extract the maximum value of the significant wave height of each year for each station as the sample to be analyzed.

[0016] Preferably, the measurement formula of step S4 is defined as follows:

[0017] ,

[0018] in, It is a site The sample linear moment ratio vector of , , , The larger the size, the more likely the site The greater the inconsistency with other sites.

[0019] Preferably, step S5 comprises the following steps:

[0020] S51, using the longitude, latitude and all effective wave height samples of the station as input, cluster the remaining stations intoK sub-regions;

[0021] S52. Perform a heterogeneity test on each sub-region to test whether each sub-region is a homogeneous region in a statistical sense.

[0022] Preferably, in the clustering process of step S51, the Euclidean distance is used to measure the difference in longitude and latitude of the two sites, and the Wasserstein distance is used to measure the difference in distribution characteristics of all samples of the two sites.

[0023] Preferably, the clustering process steps are as follows:

[0024] (1) Random assignment K sites as initial cluster centers;

[0025] (2) Calculate the sum of the Euclidean distance and Wasserstein distance between these centers and other sites, and assign these sites to different cluster centers according to the minimum distance criterion to form K clusters;

[0026] (3) For each cluster, calculate the sum of the Euclidean distance and Wasserstein distance between each site and other sites, then sum the sum at each site relative to other sites, and then select the site with the minimum sum distance as the new cluster center;

[0027] (4) Repeat the above steps until the cluster center no longer changes and stop clustering. K sub-areas.

[0028] Preferably, the heterogeneity test statistic in step S52 is The calculation formula is as follows:

[0029] ,

[0030] in, V represents the regional average linear deviation coefficient, and Represented by four parameters The distribution model is used to analyze the region The average value and standard deviation of the average linear deviation coefficient of the simulation area in the simulation, when When the value is less than 1, the sub-region can be considered to be an acceptably homogeneous region.

[0031] Preferably, step S6 is performed by calculating The statistic selects a probability distribution model for each homogeneous region. The calculation formula of the Z statistic is as follows:

[0032]

[0033] in, represents the overall linear moment kurtosis coefficient used by the selected distribution model, represents the regional average linear moment kurtosis coefficient, Indicates that the region is distributed using the selected distribution model. The standard deviation of the average linear moment kurtosis coefficient in the simulation area in the simulation, This indicates that the fitting results are acceptable and the distribution model can be used as a candidate distribution for homogeneous regions.

[0034] Preferably, the step S7 is based on The probability distribution model of the statistical test is further applied to calculate the estimated value of the specific return period of each station in the homogeneous area, and the quantile function of the frequency distribution of each station is defined as follows:

[0035] ,

[0036] in, is the dimensionless quantile function of the homogeneous region, Represents the scale factor of the station, that is, the exponential wave, which is usually the average value of the effective wave height series at each station.

[0037] Preferably, step S8 comprises the following steps:

[0038] S81. Determine the number of homogeneous sub-region sites , and the probability distribution model parameters for each site in the region;

[0039] S82. Calculate the correlation matrix for homogeneous regions , the correlations between sites in the region are considered equal, that is, the sites and The correlation between ,matrix is defined as follows:

[0040] ,

[0041] Estimated from the average correlation coefficient of extreme samples among all sites , the estimation formula is as follows:

[0042] ,

[0043] S83, Resimulate Area In each simulation, regenerate extreme samples for each station in the region and perform the following steps:

[0044] (1) Assume ,in is the maximum number of extreme samples among all stations in the region. For each time point , generates a graph with the elements A random vector , the random vector is consistent with the mean vector being 0, and the covariance matrix is Multidimensional normal distribution of;

[0045] (2) Each random vector generated in the above process , is converted into a corresponding value according to the following formula, which is the regenerated extreme sample value:

[0046] ,

[0047] in For sites in the region The quantile function of the chosen probability distribution at , is the cumulative distribution function of the standard normal distribution;

[0048] S84, for each re-simulated regional data, recalculate the estimated value of the site-specific return period through the above steps S3 to S7, and repeat the simulation of the regional M After that, arrange the estimated values ​​of the specific return period obtained by each calculation in ascending order. and It can be used as a 95% confidence interval for the estimated value of a specific return period.

[0049] Beneficial effects: Taking the significant wave height as an example, the present invention provides a regional frequency analysis method based on spatial interactive sampling. This method mainly solves the problem of large confidence intervals of estimated values ​​of major marine and coastal structures under long recurrence periods caused by limited historical data of marine environmental dynamic elements. In this method, a new clustering algorithm is proposed to autonomously divide the stations in a certain area into different clusters to obtain homogeneous areas. Benefiting from the longitude, latitude and all sample information of the station as the input of the clustering algorithm and the measurement methods of Euclidean distance and Wasserstein distance, the geographical characteristics and sample distribution characteristics between stations in the homogeneous area can be guaranteed to be as similar as possible. Through the exponential wave method, the sample information of all stations in the homogeneous area can be used to construct a dimensionless frequency distribution curve of a region. Compared with the single-station extreme value analysis, due to the use of sample distribution information of all stations in the homogeneous area, this method can more stably characterize the frequency distribution characteristics of extreme samples in the region, significantly reducing the instability of the estimated results that may be caused by fluctuations in single-station data. Benefiting from the increase in the information of effective samples, the uncertainty of the estimation of marine environmental dynamic elements is significantly reduced, and the estimation period is significantly extended. The proposed method can provide important reference for the design and construction of marine and coastal engineering. BRIEF DESCRIPTION OF THE DRAWINGS

[0050] In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the drawings required for use in the embodiments or the description of the prior art will be briefly introduced below. Obviously, the drawings described below are only embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on the provided drawings without paying creative work.

[0051] Figure 1 It is a schematic flow chart of the method for regional frequency analysis of marine environmental dynamic elements based on spatial interactive sampling disclosed in the present invention;

[0052] Figure 2 A distribution map of the research area and research points of the embodiment;

[0053] Figure 3 The final division result of the homogeneous region after clustering in the embodiment;

[0054] Figure 4 is the dimensionless quantile function of three homogeneous regions. DETAILED DESCRIPTION

[0055] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0056] Embodiment: The embodiment of the present invention discloses a method for regional frequency analysis of marine environment dynamic elements based on spatial interactive sampling. Figure 1 : is a flow chart of this embodiment. The research area is the Bohai Sea, the Yellow Sea and the northern East China Sea in China. The research points are A1 (120.2°E, 39.0°N), A2 (122.5°E, 35.0°N) and A3 (123.8°E, 29.2°N). Point A1 is located in the Bohai Sea, point A2 is located in the middle of the Yellow Sea, and point A3 is located in the northern East China Sea. Figure 2 shown.

[0057] This embodiment includes the following steps:

[0058] S1. Obtaining dynamic elements of the marine environment at multiple sites Yearly time series historical data, including In this example, the ocean dynamic environment factor is the significant wave height, based on the significant wave height hindcast data simulated by the SWAN model. The time range is 42 years, from 1979 to 2020, with a time step of 1 hour and a spatial accuracy of 0.1°. There are a total of 6306 stations in the study area.

[0059] S2. The annual extreme value method is used to extract the annual extreme value samples of the marine environment dynamic elements of each station. In this example, the maximum value of the annual significant wave height is extracted for each station as the sample to be analyzed; the annual maximum value sample is extracted from the 42-year time series wave height data of each station to form the sample to be analyzed.

[0060] S3, calculate the sample linear moment and linear moment proportional coefficient of each station. The sample linear moment mainly includes the first four orders (i.e. ), the linear moment proportionality coefficient mainly includes the linear deviation coefficient ( ), linear skewness coefficient ( ) and the linear kurtosis coefficient ( ). In this example, the sample linear moment and linear moment proportionality coefficient of the annual maximum wave height data of each station in the study area are calculated.

[0061] S4. Perform consistency check on all sites based on the sample linear moment proportionality coefficient and screen out sites with obviously inconsistent distribution characteristics.

[0062] The measurement formula of step S4 is defined as follows:

[0063] ,

[0064] in, It is a site The sample linear moment ratio vector of , , , The larger the size, the more likely the site The greater the inconsistency with other sites. Usually if the number of sites is greater than 15, then The critical value is set to 3, that is, when When it is greater than 3, it can be determined that the site is inconsistent with other sites.

[0065] In this example, 1036 nearshore stations failed the consistency test, probably due to water depth and topographic factors. The sample linear moment characteristics of these stations were obviously inconsistent with the entire area. Therefore, these stations were screened out and the remaining 5270 stations were further analyzed.

[0066] S5. Based on the geographical characteristics of the sites and the distribution characteristics of the samples, the remaining sites that meet the consistency test are clustered to divide the homogeneous areas.

[0067] S51, using the longitude, latitude and all effective wave height samples of the station as input, cluster the remaining stations into KIn the clustering process, the Euclidean distance is used to measure the difference in longitude and latitude between two stations, and the Wasserstein distance is used to measure the difference in the distribution characteristics of all samples of the two stations. The specific steps of the clustering process are as follows:

[0068] (1) Random assignment K sites as initial cluster centers;

[0069] (2) Calculate the sum of the Euclidean distance and Wasserstein distance between these centers and other sites. According to the minimum distance criterion, these sites are assigned to different cluster centers to form K clusters.

[0070] (3) For each cluster, calculate the sum of the Euclidean distance and Wasserstein distance between each site and other sites, and then sum the sum at each site relative to other sites. Then, select the site with the minimum sum distance as the new cluster center.

[0071] (4) Repeat the above steps until the cluster center no longer changes and stop clustering. K In this example, the input data for clustering is the standardized longitude, latitude and 42 maximum wave height samples for each station.

[0072] S52. Perform a heterogeneity test on each sub-region to test whether each sub-region is a homogeneous region in a statistical sense. Heterogeneity test statistic The calculation formula is as follows:

[0073] ,

[0074] in, represents the regional average linear deviation coefficient, and Represented by four parameters The distribution model is used to analyze the region The average value and standard deviation of the average linear deviation coefficient of the simulation area in the simulation, when When the value is less than 1, the sub-region can be considered as an acceptably homogeneous region. It is recommended that the number of simulations is greater than 1000.

[0075] In this example, clustering from K The test starts when the value is equal to 3 and stops when each region passes the heterogeneity test. K When is 3, the three regions obtained all pass the heterogeneity test. The statistics are -1.5, -0.53 and 0.94 respectively, which means that these three areas are all acceptable homogeneous areas. The final division results of the homogeneous areas are as follows Figure 3 shown.

[0076] S6. Select an appropriate probability distribution model for each homogeneous region through goodness-of-fit test.

[0077] By calculation The statistic selects a suitable probability distribution model for each homogeneous region. This method measures the goodness of fit of a given distribution by comparing the similarity between the overall linear moment kurtosis coefficient of the selected distribution line type and the regional average linear moment kurtosis coefficient. The calculation formula of the Z statistic is as follows:

[0078]

[0079] in, represents the overall linear moment kurtosis coefficient used by the selected distribution model, represents the regional average linear moment kurtosis coefficient, Indicates that the region is distributed using the selected distribution model. The standard deviation of the average linear moment kurtosis coefficient in the simulation area in the simulation, This indicates that the fitting results are acceptable and the distribution model can be used as a candidate distribution for homogeneous regions. It is recommended that the number of simulations be greater than 1000.

[0080] In this example, the candidate distribution models are five commonly used three-parameter probability distribution models, namely, generalized logistic distribution (GLO), generalized extreme value distribution (GEV), three-parameter lognormal distribution (LN3), Pearson type III distribution (PE3) and generalized Pareto distribution (GPA). The Z statistic results of the goodness of fit test of the above five probability distribution models for the three regions are shown in Table 1. It can be seen that for regions 1 and 2, GEV, LN3 and PE3 distributions perform better. The statistics are all less than 1.64, and only the GEV distribution performs well for region 3. In this example, the GEV distribution is selected to construct the frequency distribution curve for all three regions.

[0081] S7. Based on the selected probability distribution model, the exponential wave method is applied to calculate the estimated value of the specific return period for each station.

[0082] Based on The probability distribution model of the statistical test is further applied to calculate the estimated value of the specific return period of each station in the homogeneous area, and the quantile function of the frequency distribution of each station is defined as follows:

[0083] ,

[0084] in, is the dimensionless quantile function of the homogeneous region, Represents the scale factor of the station, that is, the exponential wave, which is usually the average value of the effective wave height series at each station.

[0085] Table 1

[0086] Distribution Model GLO GEV LN3 PE3 GPA Region 1 2.94 0.77 1.17 1.10 15.64 Area 2 3.03 1.47 1.21 1.63 155.75 Area 3 3.28 1.54 2.00 2.13 58.99

[0087] In this example, by matching the theoretical linear moment proportionality coefficient of the GEV distribution with the regional average linear moment proportionality coefficient, the dimensionless quantile function of each region can be obtained, as Figure 4 Based on the quantile function and the scale factor of each station, the estimated values ​​of different return periods for each station can be obtained.

[0088] S8. Resimulate the homogeneous area through Monte Carlo simulation and calculate the confidence interval of the estimated value of the specific return period of each station based on the simulation results. It includes the following steps:

[0089] S81. Determine the number of homogeneous sub-region sites , and the probability distribution model parameters of each station in the region. In this example, the probability distribution model parameters of each station in the region are calculated by the linear moment method, and it can be determined that the number of stations in region 1 is 534, region 2 is 2613, and region 3 is 2123. Further, for each region, the three parameters of the GEV distribution model of each station are calculated by the linear moment method. Time application can also be achieved by other methods.

[0090] S82. Calculate the correlation matrix for homogeneous regions The correlations between sites in the region can be considered equal, that is, and The correlation between ,matrix is defined as follows:

[0091] ,

[0092] It can be estimated from the average correlation coefficient of extreme samples among all sites. , the estimation formula is as follows:

[0093]

[0094] S83, Resimulate Area In each simulation, regenerate extreme samples for each station in the region and perform the following steps:

[0095] (1) Assume ,in is the maximum number of extreme samples among all stations in the region. For each time point , generates a graph with the elements A random vector , the random vector is consistent with the mean vector being 0, and the covariance matrix is Multidimensional normal distribution of;

[0096] (2) Each random vector generated in the above process , is converted into a corresponding value according to the following formula, which is the regenerated extreme sample value:

[0097]

[0098] in For sites in the region The quantile function of the chosen probability distribution at , is the cumulative distribution function of the standard normal distribution. In this embodiment, is the quantile function of the GEV distribution.

[0099] S84. For each re-simulated regional data, recalculate the estimated value of the site-specific recurrence period through the above steps S3 to S7. Repeat the simulation regional M After that, arrange the estimated values ​​of the specific return period obtained by each calculation in ascending order. and It can be used as a 95% confidence interval for the estimated value of a specific return period. M Greater than 1000.

[0100] In this embodiment, the number of simulations M is set to 5000 times, that is, after the estimated values ​​calculated 5000 times are arranged in ascending order, the 125th value and the 4875th value are the 95% confidence interval boundary values ​​of the estimated values.

[0101] In the embodiment of the present invention, the results of the single-point extreme value analysis and the results of the regional frequency analysis method are further compared at the research points A1, A2 and A3. These three research points represent three homogeneous regions. According to Table 2, the estimated values ​​of the significant wave height of 50 to 200 years are basically the same, while the 95% confidence interval width of the regional frequency analysis method is significantly smaller than that of the single-point extreme value analysis. In particular, for the A1 and A3 research points, the 95% confidence interval width of the estimated value of the significant wave height of 200 years is reduced by about 2 times. The 95% confidence interval of the estimated value of the significant wave height under a certain return period means that the probability of the expected wave height occurring within the interval is 95%. Therefore, for the design and risk analysis of offshore structures, a narrower confidence interval is desirable, which indicates that the extrapolation considering the confidence interval boundary is stable and can avoid over-protection or under-protection. Since the width of the confidence interval increases with the increase of the return period, the estimated value of the significant wave height may become unreliable under a longer return period. For example, the limit of the reliable return period is defined as the longest return period for which the width of the 95% confidence interval is less than one-quarter of the estimated value. The results of the single-point extreme value analysis show that this limit based on 42 years of data is about 50 years at the three study points. However, the significant reduction in the width of the confidence interval obtained by the regional frequency analysis method leads to a significant extension of the limit of the reliable return period. For example, the limit of the reliable return period at study point A3 can reach about 300 years. In summary, the proposed method can obtain more reliable estimated values ​​of significant wave height, which provides an important reference for the design, construction and protection of marine and coastal engineering, especially for the design and protection of major marine engineering projects that require longer design values ​​of return periods.

[0102] Table 2

[0103] Return period A1 single point extreme value analysis Frequency analysis of A1 region A2 single point extreme value analysis A2 region frequency analysis A3 Single Point Extreme Value Analysis A3 Region Frequency Analysis 50 4.47 / 0.63 4.41 / 0.42 6.75 / 2.19 6.68 / 1.51 10.38 / 2.47 10.22 / 1.88 100 4.62 / 0.85 4.54 / 0.49 7.45 / 2.78 7.29 / 1.95 11.03 / 3.38 10.81 / 2.13 150 4.71 / 0.99 4.61 / 0.53 7.89 / 3.85 7.67 / 2.25 11.38 / 4.08 11.13 / 2.24 200 4.77 / 1.11 4.66 / 0.55 8.23 / 4.48 7.95 / 2.43 11.63 / 4.64 11.36 / 2.31

[0104] In this specification, each embodiment is described in a progressive manner, and each embodiment focuses on the differences from other embodiments. The same or similar parts between the embodiments can be referred to each other. For the device disclosed in the embodiment, since it corresponds to the method disclosed in the embodiment, the description is relatively simple, and the relevant parts can be referred to the method part.

[0105] The above description of the disclosed embodiments enables one skilled in the art to implement or use the present invention. Various modifications to these embodiments will be apparent to one skilled in the art, and the general principles defined herein may be implemented in other embodiments without departing from the spirit or scope of the present invention. Therefore, the present invention will not be limited to the embodiments shown herein, but rather to the widest scope consistent with the principles and novel features disclosed herein.

Claims

1. A method for regional frequency analysis of marine environmental dynamic elements based on spatial interactive sampling, characterized in that: The following steps are involved: S1. Obtain n-year time series historical data of marine environmental dynamic elements at multiple sites, where n is greater than 20 years; S2, using the annual extreme value method to extract the annual extreme value samples of the marine environmental dynamic elements at each station; S3, calculate the sample linear moment and linear moment proportional coefficient of each station. The sample linear moment mainly includes the first four orders (i.e. ), the linear moment proportionality coefficient mainly includes the linear deviation coefficient ( )、Linear skewness coefficient( ) and the linear kurtosis coefficient ( ); S4. Perform consistency check on all sites based on the sample linear moment proportionality coefficient and screen out sites with obviously inconsistent distribution characteristics; S5. Based on the geographical characteristics of the sites and the distribution characteristics of the samples, the remaining sites that meet the consistency test are clustered to divide the homogeneous areas; The steps include: S51, using the longitude, latitude and all effective wave height samples of the station as input, cluster the remaining stations into K sub-regions; S52, conduct a heterogeneity test on each sub-region to test whether each sub-region is a homogeneous region in a statistical sense; In the clustering process of step S51, the Euclidean distance is used to measure the difference in longitude and latitude of the two sites, and the Wasserstein distance is used to measure the difference in distribution characteristics of all samples of the two sites; The clustering process steps are as follows: (1) Random assignment K sites as initial cluster centers; (2) Calculate the sum of the Euclidean distance and Wasserstein distance between these centers and other sites, and assign these sites to different cluster centers according to the minimum distance criterion to form K clusters; (3) For each cluster, calculate the sum of the Euclidean distance and Wasserstein distance between each site and other sites, then sum the sum at each site relative to other sites, and then select the site with the minimum sum distance as the new cluster center; (4) Repeat the above steps until the cluster center no longer changes and stop clustering. K sub-regions; S6. Select an appropriate probability distribution model for each homogeneous region through goodness of fit test; S7. Based on the selected probability distribution model, the exponential wave method is applied to calculate the estimated value of the specific return period for each station; S8. Resimulate the homogeneous area through Monte Carlo simulation and calculate the confidence interval of the estimated value of the specific return period of each station based on the simulation results.

2. The method for regional frequency analysis of marine environmental dynamic elements based on spatial interactive sampling according to claim 1 is characterized in that: The ocean dynamic environment element in step S1 is the significant wave height, and step S2 is to extract the maximum value of the significant wave height of each year for each station as a sample to be analyzed.

3. The method for regional frequency analysis of marine environmental dynamic elements based on spatial interactive sampling according to claim 1 is characterized in that: The measurement formula of step S4 is defined as follows: ， in, It is a site The sample linear moment ratio vector of , , , The larger the size, the more likely the site The greater the inconsistency with other sites.

4. The method for regional frequency analysis of marine environmental dynamic elements based on spatial interactive sampling according to claim 1 is characterized in that: The heterogeneity test statistic in step S52 The calculation formula is as follows: ， in, represents the regional average linear deviation coefficient, and Represented by four parameters The distribution model is used to analyze the region The average value and standard deviation of the average linear deviation coefficient of the simulation area in the simulation, when When the value is less than 1, the sub-region can be considered to be an acceptably homogeneous region.

5. The method for regional frequency analysis of marine environment dynamic elements based on spatial interactive sampling according to claim 1 is characterized in that: The step S6 is to calculate The statistic selects a probability distribution model for each homogeneous region. The calculation formula of the Z statistic is as follows: ， in, represents the overall linear moment kurtosis coefficient used by the selected distribution model, represents the regional average linear moment kurtosis coefficient, Indicates that the region is distributed using the selected distribution model. m The standard deviation of the average linear moment kurtosis coefficient in the simulation area in the simulation, This indicates that the fitting results are acceptable and the distribution model can be used as a candidate distribution for homogeneous regions.

6. The method for regional frequency analysis of marine environmental dynamic elements based on spatial interactive sampling according to claim 1 is characterized in that: The step S7 is based on Z The probability distribution model of the statistical test is further applied to calculate the estimated value of the specific return period of each station in the homogeneous area, and the quantile function of the frequency distribution of each station is defined as follows: ， in, is the dimensionless quantile function of the homogeneous region, Represents the scale factor of the station, that is, the exponential wave, which is usually the average value of the effective wave height series at each station.

7. The method for regional frequency analysis of marine environment dynamic elements based on spatial interactive sampling according to claim 1 is characterized in that: The step S8 comprises the following steps: S81. Determine the number of homogeneous sub-region sites , and the probability distribution model parameters for each site in the region; S82. Calculate the correlation matrix for homogeneous regions , the correlations between sites in the region are considered equal, that is, the sites and The correlation between ,matrix is defined as follows: ， Estimated from the average correlation coefficient of extreme samples among all sites , the estimation formula is as follows: ， S83, Resimulate Area In each simulation, regenerate extreme samples for each station in the region and perform the following steps: (1) Assume ,in is the maximum number of extreme samples among all stations in the region. For each time point , generates a graph with the elements A random vector , the random vector is consistent with the mean vector being 0, and the covariance matrix is Multidimensional normal distribution of; (2) Each random vector generated in the above process , is converted into a corresponding value according to the following formula, which is the regenerated extreme sample value: ， in For sites in the region The quantile function of the chosen probability distribution at , is the cumulative distribution function of the standard normal distribution; S84, for each re-simulated regional data, recalculate the estimated value of the site-specific return period through the above steps S3 to S7, and repeat the simulation of the regional M After that, arrange the estimated values ​​of the specific return period obtained by each calculation in ascending order. and It can be used as a 95% confidence interval for the estimated value of a specific return period.

Images