Non-consistent regional flood frequency analysis method based on Bayesian theory

Abstract

The invention relates to the technical field of flood prediction, in particular to a non-consistent regional flood frequency analysis method based on the Bayesian theory. The method comprises the following steps: acquiring flood sequence data, meteorological data and drainage basin characteristic data of hydrological stations in a drainage basin; performing regional flood dynamic change analysis on the flood sequence data and the meteorological data to generate regional flood time sequence dynamic change data; on the basis of the regional flood time sequence dynamic change data, hydrological station flood characteristic parameters in the drainage basin are extracted from the drainage basin characteristic data, and hydrological station flood characteristic parameters in the region are obtained; and performing flood distribution parameter regression coefficient prior information derivation on the flood characteristic parameters of the hydrological stations in the region to obtain a maximum likelihood estimation value of a single hydrological station distribution parameter. According to the method, the Bayesian theory, the spatial correlation analysis and the spectral analysis are introduced, so that the accuracy and the applicability of flood frequency analysis based on the Bayesian theory are improved.

Description

Technical Field

[0001] The invention relates to the technical field of flood prediction, and in particular to a method for analyzing flood frequency in an inconsistent region based on Bayesian theory. Background Art

[0002] In the initial stage, flood frequency analysis relied on classical statistical methods, such as extreme value theory and linear regression models, which assumed that flood events were consistent between different regions, that is, the probability and intensity of floods were similar throughout the study area. However, this assumption often ignores the geographical differences between regions, climate change and other influencing factors, resulting in deviations in the results of frequency analysis in practical applications. With the development of statistics, especially the rise of non-parametric statistical methods and geostatistics, researchers gradually realized that there may be significant inconsistency in flood events between different regions. In order to better capture this regional difference, models based on Bayesian theory began to be proposed. Bayesian methods can flexibly integrate historical data, expert knowledge and other prior information to perform uncertainty analysis on flood frequencies between regions. However, the current traditional flood frequency analysis methods usually assume that flood characteristics between regions are similar, but in fact, the geographical characteristics, climate conditions and precipitation patterns of different regions vary greatly. At the same time, traditional methods usually analyze the flood distribution parameters of each hydrological station independently, ignoring the possible spatial correlation between stations, which leads to low accuracy and applicability of flood frequency analysis based on Bayesian theory. Summary of the invention

[0003] Based on this, it is necessary to provide a method for analyzing the frequency of floods in inconsistent regions based on Bayesian theory to solve at least one of the above technical problems.

[0004] To achieve the above purpose, a method for analyzing flood frequency in inconsistent regions based on Bayesian theory is provided, the method comprising the following steps: Step S1: Obtain flood sequence data, meteorological data and basin characteristic data of hydrological stations in the basin; perform regional flood dynamic change analysis on the flood sequence data and meteorological data to generate regional flood time series dynamic change data; extract flood characteristic parameters of hydrological stations in the basin from the basin characteristic data based on the regional flood time series dynamic change data to obtain flood characteristic parameters of hydrological stations in the region; Step S2: deriving the prior information of the flood distribution parameter regression coefficient for the flood characteristic parameters of the hydrological stations in the region, and obtaining the maximum likelihood estimate of the distribution parameters of a single hydrological station; importing the maximum likelihood estimate of the distribution parameters of a single hydrological station and the watershed characteristic data into a preset regional regression model to perform prior probability distribution calculation, and obtaining the prior probability distribution data of the regression coefficient of a single station; Step S3: construct a likelihood function based on the prior probability distribution data of the regression coefficient of a single site and the flood sequence data to obtain a flood distribution likelihood function; derive the posterior information of the regression coefficient of the flood distribution parameter from the flood distribution likelihood function through the Bayesian theory, thereby generating a Bayesian estimate of the regression coefficient of a single site; Step S4: construct a spatial association network based on the Bayesian estimated value of the regression coefficient of a single site to generate a site spatial association network; perform flood correlation analysis on the Bayesian estimated value of the regression coefficient of a single site using the site spatial association network to generate site flood spatial association data; Step S5: perform spatial regional flood data collection on the site flood spatial correlation data to obtain inconsistent regional flood collection data; perform fast Fourier transform on the inconsistent regional flood collection data to generate inconsistent regional flood spectrum data; perform cross-flood frequency verification on the inconsistent regional flood spectrum data based on the prior probability distribution data of the single site regression coefficient and the Bayesian estimate of the single site regression coefficient to generate a regional flood frequency distribution map.

[0005] The present invention can fully understand the temporal changes and regional characteristics of floods by acquiring flood sequence data, meteorological data and basin characteristic data of hydrological stations in the basin, and performing regional flood dynamic change analysis, so as to provide accurate basic data for subsequent flood characteristic parameter extraction and ensure the accuracy of analysis. The prior information of regression coefficients of flood characteristic parameters of hydrological stations is derived, and the prior probability distribution of regression coefficients of single stations is calculated in combination with basin characteristic data, which effectively combines the spatial characteristics of the basin with the statistical laws of flood occurrence, improves the accuracy of regression coefficient estimation in flood frequency analysis, and reduces uncertainty. By constructing a likelihood function and applying Bayesian theory to derive posterior information of regression coefficients, the estimation of regression coefficients can be updated more accurately, and the accuracy and credibility of flood frequency analysis results are enhanced by combining prior information and actual observation results of flood sequence data. By constructing a site spatial association network and performing flood correlation analysis, the spatial relationship between sites and the mutual influence of floods can be revealed, which helps to better capture the spatiotemporal distribution characteristics of floods at the regional level and provide more comprehensive data support for regional flood frequency analysis. The use of fast Fourier transform to process flood data in inconsistent areas, generate flood spectrum data, and combine Bayesian estimates to perform cross-flood frequency verification, which helps to deal with differences in flood characteristics between complex regions, improves the ability to comprehensively analyze regional flood frequencies, and generates more accurate regional flood frequency distribution maps. Therefore, the present invention improves the accuracy and applicability of flood frequency analysis based on Bayesian theory by introducing Bayesian theory, spatial correlation analysis and spectrum analysis.

[0006] Preferably, step S1 comprises the following steps: Step S11: Obtain flood sequence data, meteorological data and watershed characteristic data of hydrological stations in the watershed; Step S12: preprocessing the flood sequence data, meteorological data and watershed characteristic data to generate standard flood sequence data, standard meteorological data and standard watershed characteristic data, wherein the data preprocessing includes data cleaning, data denoising, data missing value filling and data standardization; Step S13: Analyze the regional flood dynamic change on the standard flood sequence data and the standard meteorological data to generate regional flood time series dynamic change data; Step S14: extract flood characteristic parameters of hydrological stations in the standard watershed characteristic data based on the regional flood time series dynamic change data to obtain flood characteristic parameters of hydrological stations in the region.

[0007] The present invention improves data quality and consistency by preprocessing flood sequence data, meteorological data and basin characteristic data (including cleaning, denoising, missing value filling and standardization), ensures the accuracy of subsequent analysis and calculation, and provides a highly reliable data basis for scientific research. By analyzing the dynamic changes of regional floods in flood sequence data and meteorological data, it is possible to fully grasp the trend of flood time series changes in the region, identify potential flood risk points, and provide data support for basin management and disaster warning. Extracting flood characteristic parameters of hydrological stations in the basin based on regional flood time series dynamic change data can effectively capture the dynamic change characteristics of basin characteristics and provide refined support for hydrological simulation, flood prediction and disaster prevention and mitigation decision-making in the region. Each sub-step is linked together to construct a complete closed-loop process from data collection to parameter extraction. The logic is clear and the modular design is easy to operate and expand, which is suitable for different regions and different types of flood data analysis needs. By combining dynamic change data with basin characteristic data, the extraction accuracy of flood characteristic parameters in the basin is enhanced, thereby providing a more scientific basis for flood risk prediction and management. It integrates data from three dimensions: flood sequences, meteorological conditions, and river basin characteristics, and provides a multi-angle analysis perspective, helping to reveal the causes and patterns of regional floods in more depth, and providing strong technical support for studying the mechanism of flood evolution.

[0008] Preferably, step S13 includes the following steps: Step S131: performing time series decomposition on the flood sequence data to generate flood sequence time series decomposition data; performing random fluctuation component separation on the flood sequence time series decomposition data to generate flood sequence random fluctuation component data; Step S132: performing meteorological change pattern analysis on meteorological data to generate short-term meteorological pattern data and long-term meteorological pattern data; performing extreme meteorological flood change analysis on random fluctuation component data of flood sequence according to the short-term meteorological pattern data to generate extreme meteorological flood change data; Step S133: performing common meteorological flood change analysis on the random fluctuation component data of the flood sequence through long-term meteorological model data to generate common meteorological flood change data; performing change peak extraction on the extreme meteorological flood change data to obtain extreme meteorological flood change peak data; Step S134: extract the change mean of ordinary meteorological flood change data to obtain ordinary meteorological flood change mean data; analyze the regional flood meteorological linkage effect based on the extreme meteorological flood change peak data and the ordinary meteorological flood change mean data to generate regional flood time series dynamic change data.

[0009] The present invention refines the characteristics of flood changes by performing time series decomposition and random fluctuation component separation on flood sequence data, and provides more accurate data support for analyzing the complex dynamic changes of floods, especially when dealing with uncertainty and random flood fluctuations. By analyzing the meteorological change pattern, the meteorological data is divided into short-term and long-term patterns, which not only helps to capture the impact of short-term extreme meteorological events on floods, but also reveals the role of long-term meteorological trends on flood changes, and provides refined support for flood prediction and management at different time scales. The change analysis and peak extraction of floods under extreme meteorological conditions can accurately identify the extreme change characteristics of floods, provide key data for flood risk warning and emergency response, and improve the timeliness and pertinence of disaster prevention and mitigation. The mean extraction of flood changes under normal meteorological conditions can summarize the average change law of floods, reveal the stability and predictability of floods under non-extreme conditions, and provide a reference for the long-term monitoring and management of hydrological laws in the basin. The regional flood meteorological linkage effect analysis based on the peak data of extreme meteorological flood changes and the mean data of ordinary meteorological flood changes is helpful to fully understand the multi-dimensional impact of meteorological changes on flood formation and evolution, and provide a scientific basis for the comprehensive management of the basin and the formulation of flood control measures. The final generated regional flood time series dynamic change data integrates flood change information under multiple levels, multiple time scales and multiple meteorological conditions, providing a high-quality, dynamic data foundation for subsequent hydrological model optimization, watershed management and flood prediction. The logic between each sub-step is clear, the functions are independent and highly connected, which is not only suitable for the analysis of flood dynamic changes under different meteorological conditions, but also easy to promote and apply in different watersheds or regions, with good versatility and scalability.

[0010] Preferably, the regional flood-meteorological linkage effect analysis based on the extreme meteorological flood change peak data and the ordinary meteorological flood change mean data includes: The flood-meteorological linkage effect index is calculated by using the linkage effect quantification formula for the peak data of extreme meteorological flood changes and the mean data of ordinary meteorological flood changes to obtain the flood-meteorological linkage effect index; the linkage effect regional sensitivity analysis is conducted on the hydrological stations in the basin based on the flood-meteorological linkage effect index to generate regional flood linkage sensitivity data; Based on the regional flood linkage sensitivity data, the flood-meteorological linkage effect index is ranked by regional effects to generate regional sensitivity ranking data; the flood intensity change trend of extreme meteorological flood change peak data and ordinary meteorological flood change mean data is analyzed through regional sensitivity ranking data to generate regional flood time series dynamic change data.

[0011] The present invention calculates the flood-meteorological linkage effect index by using the linkage effect quantification formula, which can reveal the influence of extreme meteorological conditions and ordinary meteorological conditions on flood changes in a quantitative form, and provide scientific and systematic indicator support for flood management and early warning systems. By performing linkage effect regional sensitivity analysis on the flood-meteorological linkage effect index, the response sensitivity of different hydrological stations to meteorological changes can be accurately identified, which provides a basis for formulating targeted prevention and control measures and regional flood risk assessment, and helps to improve regional flood control capabilities. By sorting the effects of regional flood linkage sensitivity data, the differences in flood-meteorological linkage effects in different regions can be revealed, which provides support for refined regional flood risk management and resource scheduling. High-sensitivity areas are given priority to effectively optimize the configuration of flood control resources. Based on the regional sensitivity sorting data, the flood intensity change trend analysis of the extreme meteorological flood change peak data and the ordinary meteorological flood change mean data can accurately identify the intensity change trend of future floods, help decision makers make emergency plans in advance, and improve disaster response efficiency. The generated regional flood time series dynamic change data can fully reflect the time series characteristics of flood changes under extreme and ordinary meteorological conditions, provide strong data support for the optimization of flood prediction models and future basin management, and further improve prediction accuracy and decision-making effects. This process can not only be flexibly applied to different meteorological conditions and regions, but also adapt to the needs of various river basins through modular design. It has high application value in actual flood warning, disaster management, and coordinated analysis of meteorology and hydrology.

[0012] Preferably, the linkage effect quantification formula is used to calculate the flood meteorological linkage effect index for the extreme meteorological flood change peak data and the ordinary meteorological flood change mean data. The linkage effect quantification formula is as follows: In the formula, It is expressed as the flood-meteorological linkage effect index, Expressed as the peak value of extreme meteorological floods, It is expressed as the average value of ordinary meteorological flood. It is expressed as the mean temperature under extreme meteorological conditions. It is expressed as the mean temperature under normal meteorological conditions. Expressed as extreme rainfall, Expressed as normal rainfall, Expressed as the duration of extreme rainfall, It is expressed as the duration of normal rainfall, Expressed as the impact weight of flood peak, Expressed as the impact weight of temperature change, Expressed as the impact weight of rainfall change, Expressed as the impact weight of rainfall duration.

[0013] The present invention analyzes and integrates a linkage effect quantification formula, which is based on the changes in flood characteristics under different meteorological conditions (extreme weather and ordinary weather), considers the impact of meteorological factors (temperature, rainfall, rainfall duration) on flood changes, and quantifies these impacts into a linkage effect index. The numerator in the formula (reflecting the changes in floods) is: Indicates flood peaks under extreme weather conditions Compared with the flood average under normal weather conditions The difference between them represents the magnitude of flood changes and reflects the direct impact of extreme meteorological conditions on floods. Indicates the temperature under extreme weather conditions Temperature under normal weather conditions The difference between the two reflects the indirect impact of temperature change on floods. Temperature changes can indirectly change flood conditions by affecting precipitation, evaporation, snowmelt and other processes. The denominator (reflecting changes in meteorological factors): Indicates the amount of rainfall under extreme meteorological conditions The rainfall under normal weather conditions The difference between the two reflects the driving effect of rainfall changes on floods. Rainfall is the direct cause of floods, and extreme rainfall may significantly increase the intensity and frequency of floods. Indicates the duration of rainfall under extreme meteorological conditions Duration of normal rainfall The difference between reflects the impact of rainfall duration changes on floods. The duration of rainfall directly affects the accumulation of ground water, thereby affecting the occurrence and change of floods. The impact weight of flood peak value indicates the contribution of flood peak value changes under extreme meteorological conditions to the linkage effect. If the flood peak value changes greatly under extreme meteorological conditions, it means that the area is more sensitive to extreme meteorological changes, and the parameter The impact weight of temperature change indicates the impact of temperature change on flood changes. If the temperature changes significantly, it may cause stronger precipitation or snowmelt, thus affecting the formation of floods. The size of reflects the indirect impact of temperature change on floods. The impact weight of rainfall change represents the direct impact of rainfall change on floods. The greater the rainfall, the higher the intensity and frequency of floods may be. The larger the value, the more important the rainfall is in flood generation. The impact weight of the change in rainfall duration indicates the impact of rainfall duration on floods. If the rainfall duration is longer, it usually means a higher flood risk. The weight indicates the degree of influence of rainfall duration on floods. When using the conventional linkage effect quantification formula in this field, the flood-meteorological linkage effect index can be obtained. By applying the linkage effect quantification formula provided by the present invention, the flood-meteorological linkage effect index can be calculated more accurately. The linkage effect quantification formula comprehensively considers the linkage relationship between flood changes and meteorological factors, and uses weight parameters to quantify the impact of each factor on flood risks, so as to provide a multi-dimensional and comprehensive flood-meteorological linkage effect index. By accurately calculating flood changes under different meteorological conditions, it can provide an important basis for flood warning, risk assessment, and disaster prevention and mitigation.

[0014] Preferably, step S2 comprises the following steps: Step S21: Calculate the maximum likelihood estimate of the flood sequence distribution parameters for the flood characteristic parameters of the hydrological stations in the region to obtain the maximum likelihood estimate of the distribution parameters of a single hydrological station, wherein the formula for calculating the maximum likelihood estimate of the flood sequence distribution parameters is as follows: ; In the formula, is the regression coefficient of the flood sequence distribution parameter, represents the distribution parameter estimated based on the flood series of a single site only, Indicates i Site distribution parameters H non-consistent covariates, and denote the link functions of the first and second distribution parameters, respectively, is a linear function; Step S22: construct a regional regression model; the regional regression model is represented by a model matrix to obtain a matrixed regional regression model; wherein the construction formula of the regional regression model is as follows: ; ; In the formula, Indicates i Sites The variance of the error caused by the sample estimate of Indicates the site i With site j The correlation coefficient between the sample errors is the same as that under the consistency condition; The formula for the model matrix representation is as follows: ; In the formula, X is given by K Site C The regional covariates K ×( C +1) rank matrix; The flood sequence λ The regional regression coefficients of the distribution parameters; Indicated in t Time by C The vector of random effect terms of the basin characteristic indicators is assumed to have a mean of 0, and the covariance matrix is The multivariate normal distribution of Indicated in t Time by K The flood sequence of the station λ A vector of regional regression model errors consisting of distribution parameters; Indicated in t Time by K The flood sequence of the station λ A vector of sample errors for the estimates of the distribution parameters; Indicated in t The total error at time is and The sum of .

[0015] Step S23: importing the maximum likelihood estimation value of the distribution parameter of a single hydrological station and the watershed characteristic data into the matrixed regional regression model to perform prior probability distribution calculation, and obtain the prior probability distribution data of the regression coefficient of the single station; The formula for calculating the prior probability distribution is as follows: ; In the formula, Data expressed as a priori probability distributions of regression coefficients for individual sites.

[0016] The present invention calculates the flood sequence distribution parameters of each hydrological station by the maximum likelihood estimation (MLE) method. These parameters reflect the flood characteristics of a specific station under different climatic conditions. The regression coefficients shown in the formula, etc., represent the impact of different inconsistency covariates (such as precipitation, temperature, etc.) in the flood sequence of the station on the flood change. These regression coefficients help to quantify how these factors affect flood events and estimate the possibility and scale of floods based on historical data. The regional regression model integrates the flood data of multiple hydrological stations to establish a regional model to capture the correlation between stations. The model construction formula represents the error covariance structure between different stations, where the errors between stations may be correlated (represented by the correlation coefficient ρ_ij). After the regional regression model is matrixed, it helps to combine the distribution parameter estimates of each station with the regional characteristic data, so as to make a wider range of flood predictions. The distribution parameters and basin characteristic data obtained from a single hydrological station are input into the matrixed regional regression model to calculate the prior probability distribution. The calculation of the prior probability distribution can help derive the prior information of the regression coefficient of a single station. This information will serve as the basis for further flood forecasting, and the model prediction effect will be optimized by fusing data and prior information through Bayesian methods.

[0017] Preferably, step S3 comprises the following steps: Step S31: construct a likelihood function based on the prior probability distribution data of the regression coefficient of a single station and the flood sequence data to obtain a flood distribution likelihood function; wherein the construction formula of the likelihood function is as follows: ; In the formula, It is expressed as flood distribution likelihood function; Step S32: Using Bayesian theory to derive the flood posterior probability distribution function from the flood distribution likelihood function, thereby obtaining the posterior probability distribution data of the regression coefficient of a single station; wherein the formula for deriving the flood posterior probability distribution function is as follows: ; In the formula, The posterior probability distribution data of the regression coefficient of a single site is expressed as The value range of Step S33: Performing Bayesian estimation calculation on the prior probability distribution data of the regression coefficient of a single site through the posterior probability distribution data of the regression coefficient of a single site, to obtain the Bayesian estimation value of the regression coefficient of a single site; wherein the Bayesian estimation value calculation is as follows: .

[0018] The present invention constructs a likelihood function of flood distribution by combining the prior probability distribution data of the regression coefficient of a single station with the flood sequence data. The likelihood function represents the probability of the observed flood sequence data under a given regression coefficient. This function obtains the likelihood of the entire flood sequence by multiplying the probability values ​​of all time points. By constructing the likelihood function, the degree of fit of the model to the observed data can be quantified, providing a basis for subsequent Bayesian updates. The likelihood function helps to understand how a specific regression coefficient affects the performance of the flood prediction model. Applying Bayesian theory, the posterior probability distribution of the flood regression coefficient is derived through the likelihood function and the prior probability distribution. The Bayesian formula combines prior knowledge (i.e., prior distribution) and observed data (i.e., likelihood function) to obtain the posterior probability distribution of the regression coefficient under the condition of a given flood sequence. The posterior probability distribution can fully consider the prior information and new observed data, and provide a more accurate estimate of the regression coefficient. The Bayesian method can handle uncertainty and provide more comprehensive results than point estimation, and is suitable for problems with high uncertainty such as flood prediction. In this step, the posterior probability distribution is used to calculate the Bayesian estimate of the regression coefficient. The Bayesian estimate obtains the most likely estimate of the regression coefficient by taking a weighted average of all possible values ​​of the regression coefficient, where the weight is given by the posterior distribution. The Bayesian estimate provides a comprehensive estimate of the regression coefficient, which is more robust than the simple maximum likelihood estimate. The Bayesian estimate can more accurately predict the probability and characteristics of floods, especially when there are few samples or incomplete data, and can provide more reliable results.

[0019] Preferably, step S4 comprises the following steps: Step S41: constructing a site feature vector according to the Bayesian estimation value of the regression coefficient of a single site to generate site feature matrix data; Step S42: performing parameter space search on the site characteristic matrix data to generate prior parameter set data; Step S43: construct a spatial association network for the prior parameter set data to generate a site spatial association network; use the site spatial association network to perform flood correlation analysis on the Bayesian estimate of the regression coefficient of a single site to generate site flood spatial association data.

[0020] The present invention constructs a site feature vector through the Bayesian estimate of the regression coefficient of a single site, and finally generates a site feature matrix data. The site feature matrix combines the regression coefficient with other relevant features of the site to provide a complete feature description for subsequent analysis. Through the generation of the site feature matrix, multiple relevant characteristics of the site can be fully reflected, providing basic data for subsequent analysis. It is helpful to integrate information from different sources, associate the regression coefficient of the site with other features (such as geographical location, historical flood records, etc.), and enhance the multi-dimensional analysis ability of the model. Perform parameter space search on the site feature matrix data to generate prior parameter set data. This process involves systematically searching the possible parameter space to find the parameter configuration that best suits the current flood prediction model. Parameter space search can optimize the parameter selection of the model to ensure that the selected parameters are most suitable for the characteristics of the current data, thereby improving the accuracy of the model. Through this search method, a suitable parameter set can be found in a wide parameter space, avoiding local optimal solutions and providing more comprehensive optimization results. Using the prior parameter set data, a site spatial association network is constructed, and flood correlation analysis is performed on the Bayesian estimate of the regression coefficient of a single site, and finally site flood spatial association data is generated. The construction and analysis of spatial correlation networks help identify the relationships between different sites and their impact on floods. By constructing a site spatial correlation network, we can better understand the spatial relationship between different sites and provide intuitive data support for analyzing the mutual impact between sites. Flood correlation analysis can reveal the similarities and differences in flood occurrence patterns at different sites, helping to optimize the model's prediction results, especially when there is a certain degree of spatial dependence. This step helps to improve the accuracy of the model in spatial distribution, making flood risk predictions in different geographical locations more accurate.

[0021] Preferably, constructing a spatial association network for the prior parameter set data includes: Perform multi-scale decomposition on the prior parameter set data to generate feature decomposition data; perform time series window processing on the feature decomposition data to generate time series feature data; perform spatial mapping processing on the time series feature data to generate feature decomposition matrix data; perform coordinate transformation on the feature decomposition matrix data to generate standardized coordinate data; Perform distance calculation on the standardized coordinate data to generate distance matrix data; perform weight assignment on the distance matrix data to generate spatial distance feature data; perform triangulation on the spatial distance feature data to generate initial network data; perform spatial partitioning on the initial network data to generate partition structure data; The degree distribution of the partition structure data is processed to generate network topology structure data; based on the graph neural network algorithm, the spatial association network of the network topology structure data is constructed to generate a site spatial association network.

[0022] The present invention helps to extract multi-level information in data through multi-scale decomposition, and enhances the model's processing ability for features of different scales. The model is enabled to identify the relationships at different scales, which helps to capture potential spatial patterns and trends. Time series window processing can extract dynamic patterns and trends in time series, helping to identify short-term or long-term change patterns. It helps to improve the prediction ability of the model, especially when the data has obvious time series characteristics. Through spatial mapping, time series characteristics and spatial characteristics can be combined, improving the analysis ability of the model in the dual dimensions of space and time. It helps to further identify the time change characteristics in spatial distribution, especially in complex geographical environments. Coordinate transformation helps to ensure spatial consistency between different data sets and avoid analysis errors caused by differences in coordinate systems. It improves the standardization of data processing and facilitates subsequent spatial association analysis. The distance matrix provides spatial relationship data between sites, which can reveal the relative positions between different sites and help determine the scope of influence and the strength of association. It makes spatial analysis have a quantitative basis, which is helpful for subsequent calculations and network construction. Weight allocation can adjust the strength of association between sites according to spatial distance, and improve the accuracy of spatial association networks. The relationship between different sites is made more dynamic and can be adjusted according to the actual distance and degree of influence. Triangulation lays the foundation for subsequent network analysis and helps form a preliminary spatial association network. Through triangulation, the network structure is closer to the actual spatial distribution, which improves the authenticity of the network. Spatial partitioning can help analyze the relationship differences within and outside the region and improve the performance of the spatial association network in different regions. It provides a clear data framework for subsequent network optimization and regional analysis. Through degree distribution analysis, key nodes in the network can be identified and the contribution of different sites to the entire network can be understood. This step helps to optimize the network structure and improve the stability and efficiency of the network. The graph neural network algorithm can effectively capture the complex spatial relationships between sites and optimize the spatial reasoning ability of the network by learning the associations between different nodes. Through graph neural networks, the adaptability and flexibility of the network can be enhanced, and the ability of the spatial association network to process new data can be improved.

[0023] Preferably, step S5 includes the following steps: Step S51: performing spatial regional flood data collection on the site flood spatial correlation data to obtain non-consistent regional flood collection data; performing fast Fourier transform on the non-consistent regional flood collection data to generate non-consistent regional flood spectrum data; Step S52: performing cross-flood frequency verification on the inconsistent regional flood spectrum data according to the prior probability distribution data of the single-station regression coefficient and the Bayesian estimation value of the single-station regression coefficient, and generating verified regional flood spatial distribution data; Step S53: Visualize the flood frequency of the verified regional flood spatial distribution data to generate a regional flood frequency distribution map.

[0024] The present invention can effectively identify areas with complex patterns and abnormal behaviors by specially collecting data from inconsistent areas, which helps to improve the attention and analysis accuracy of special areas. FFT can convert flood data from the time domain to the frequency domain, extract the frequency characteristics in flood events, and thus reveal periodic or repetitive patterns. The generation of spectrum data simplifies the representation of flood data and provides efficient numerical representation for subsequent analysis, reducing the burden of big data processing. Combining prior probability distribution and Bayesian estimation can accurately correct flood spectrum data to make it more reliable and statistically significant. This not only improves the accuracy of the data, but also strengthens the robustness of the model. Through cross-validation, the deviation caused by inconsistent data or noise can be reduced, and the reliability and accuracy of spectrum verification can be improved. The generated verified regional flood spatial distribution data is helpful to analyze the regional changes of floods and provide support for accurate prediction and regional flood management. By generating a flood frequency distribution map, the flood frequency distribution of each region can be clearly displayed, helping decision makers to identify high-risk areas and optimize resource allocation and emergency management. The visualized data makes complex statistical and spectrum analysis results easier to understand and apply, which helps relevant personnel to more effectively predict and plan floods. The frequency distribution map provides a visual estimate of the probability of future floods, making the prediction model more accurate and able to provide early warning of potential risks. BRIEF DESCRIPTION OF THE DRAWINGS

[0025] Figure 1 A schematic diagram of the steps of a method for analyzing flood frequency in inconsistent regions based on Bayesian theory; Figure 2 for Figure 1 Detailed implementation steps of step S2 in the flowchart; Figure 3 for Figure 1 Detailed implementation steps of step S3 in FIG. The realization of the purpose, functional features and advantages of the present invention will be further explained in conjunction with embodiments and with reference to the accompanying drawings. DETAILED DESCRIPTION

[0026] The technical method of the present invention is described clearly and completely below in conjunction with the accompanying drawings. Obviously, the described embodiments are 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 technicians in this field without creative work are within the scope of protection of the present invention.

[0027] In addition, the accompanying drawings are only schematic illustrations of the present invention and are not necessarily drawn to scale. The same reference numerals in the figures represent the same or similar parts, and their repeated description will be omitted. Some of the block diagrams shown in the accompanying drawings are functional entities and do not necessarily correspond to physically or logically independent entities. The functional entities can be implemented in software form, or implemented in one or more hardware modules or integrated circuits, or implemented in different networks and / or processor methods and / or microcontroller methods.

[0028] It should be understood that, although the terms "first", "second", etc. may be used herein to describe various units, these units should not be limited by these terms. These terms are used only to distinguish one unit from another unit. For example, without departing from the scope of the exemplary embodiments, the first unit may be referred to as the second unit, and similarly the second unit may be referred to as the first unit. The term "and / or" used herein includes any and all combinations of one or more of the listed associated items.

[0029] To achieve this, please refer to Figures 1 to 3 , a method for analyzing flood frequency in inconsistent regions based on Bayesian theory, the method comprising the following steps: Step S1: Obtain flood sequence data, meteorological data and basin characteristic data of hydrological stations in the basin; perform regional flood dynamic change analysis on the flood sequence data and meteorological data to generate regional flood time series dynamic change data; extract flood characteristic parameters of hydrological stations in the basin from the basin characteristic data based on the regional flood time series dynamic change data to obtain flood characteristic parameters of hydrological stations in the region; Step S2: deriving the prior information of the flood distribution parameter regression coefficient for the flood characteristic parameters of the hydrological stations in the region, and obtaining the maximum likelihood estimate of the distribution parameters of a single hydrological station; importing the maximum likelihood estimate of the distribution parameters of a single hydrological station and the watershed characteristic data into a preset regional regression model to perform prior probability distribution calculation, and obtaining the prior probability distribution data of the regression coefficient of a single station; Step S3: construct a likelihood function based on the prior probability distribution data of the regression coefficient of a single site and the flood sequence data to obtain a flood distribution likelihood function; derive the posterior information of the regression coefficient of the flood distribution parameter from the flood distribution likelihood function through the Bayesian theory, thereby generating a Bayesian estimate of the regression coefficient of a single site; Step S4: construct a spatial association network based on the Bayesian estimated value of the regression coefficient of a single site to generate a site spatial association network; perform flood correlation analysis on the Bayesian estimated value of the regression coefficient of a single site using the site spatial association network to generate site flood spatial association data; Step S5: perform spatial regional flood data collection on the site flood spatial correlation data to obtain inconsistent regional flood collection data; perform fast Fourier transform on the inconsistent regional flood collection data to generate inconsistent regional flood spectrum data; perform cross-flood frequency verification on the inconsistent regional flood spectrum data based on the prior probability distribution data of the single site regression coefficient and the Bayesian estimate of the single site regression coefficient to generate a regional flood frequency distribution map.

[0030] The present invention can fully understand the temporal changes and regional characteristics of floods by acquiring flood sequence data, meteorological data and basin characteristic data of hydrological stations in the basin, and performing regional flood dynamic change analysis, so as to provide accurate basic data for subsequent flood characteristic parameter extraction and ensure the accuracy of analysis. The prior information of regression coefficients of flood characteristic parameters of hydrological stations is derived, and the prior probability distribution of regression coefficients of single stations is calculated in combination with basin characteristic data, which effectively combines the spatial characteristics of the basin with the statistical laws of flood occurrence, improves the accuracy of regression coefficient estimation in flood frequency analysis, and reduces uncertainty. By constructing a likelihood function and applying Bayesian theory to derive posterior information of regression coefficients, the estimation of regression coefficients can be updated more accurately, and the accuracy and credibility of flood frequency analysis results are enhanced by combining prior information and actual observation results of flood sequence data. By constructing a site spatial association network and performing flood correlation analysis, the spatial relationship between sites and the mutual influence of floods can be revealed, which helps to better capture the spatiotemporal distribution characteristics of floods at the regional level and provide more comprehensive data support for regional flood frequency analysis. The use of fast Fourier transform to process flood data in inconsistent areas, generate flood spectrum data, and combine Bayesian estimates to perform cross-flood frequency verification, which helps to deal with differences in flood characteristics between complex regions, improves the ability to comprehensively analyze regional flood frequencies, and generates more accurate regional flood frequency distribution maps. Therefore, the present invention improves the accuracy and applicability of flood frequency analysis based on Bayesian theory by introducing Bayesian theory, spatial correlation analysis and spectrum analysis.

[0031] In the embodiment of the present invention, reference Figure 1 FIG. 1 is a schematic diagram of the steps of a method for analyzing flood frequency in inconsistent regions based on Bayesian theory of the present invention. In this example, the method for analyzing flood frequency in inconsistent regions based on Bayesian theory includes the following steps: Step S1: Obtain flood sequence data, meteorological data and basin characteristic data of hydrological stations in the basin; perform regional flood dynamic change analysis on the flood sequence data and meteorological data to generate regional flood time series dynamic change data; extract flood characteristic parameters of hydrological stations in the basin from the basin characteristic data based on the regional flood time series dynamic change data to obtain flood characteristic parameters of hydrological stations in the region; In an embodiment of the present invention, long-term flood sequence data is collected from various hydrological stations in the basin. The data should include parameters such as the time of flood occurrence, flood flow, and water level. The data can be obtained from national or regional meteorological and water conservancy departments, or relevant hydrological data can be obtained through remote sensing technology. Meteorological data in the basin are collected, especially parameters such as precipitation, temperature, and humidity. These data are usually provided by meteorological stations, and can also be obtained through remote sensing satellite data or regional meteorological bureaus. Meteorological data should cover a long span in the basin, especially data on meteorological extreme events related to floods, such as heavy rainfall events. The geographical feature data of the basin are obtained, including basin area, land use type, slope, soil type, hydrological and climatic conditions, etc. The elevation data and land use data of the basin can be obtained using a geographic information system (GIS). Time series analysis is performed on the flood sequence data to identify the changing trends and periodic characteristics of regional floods. Trend analysis of flood data is performed using methods such as autoregressive moving average model (ARMA), sliding average method (SMA), and Hopkins method to identify seasonal and interannual changes. Analyze extreme flood events in flood sequences and identify potential flood risk points. Based on flood sequence data, establish a time series dynamic change model of floods, taking into account factors that affect flood changes (such as precipitation, climate change, etc.). Use machine learning algorithms such as regression analysis, neural networks, and support vector machines to perform time series modeling to predict future dynamic changes in regional floods. The generated regional flood time series dynamic change data should be able to reveal the trends and fluctuations of flood changes on different time scales, and help understand long-term and short-term flood changes. Extract relevant flood characteristic parameters based on flood sequence data. These parameters include but are not limited to: flood peak flow (i.e., maximum flood flow), flood frequency (probability of exceeding a certain flow threshold), flood duration (referring to the length of time the flood lasts), flood probability (frequency of flood occurrence), and flood risk index (a flood risk assessment parameter that comprehensively considers multiple factors). Based on the data obtained from the analysis of the dynamic changes in regional flood time series, combined with the basin characteristic data, quantitatively extract the flood characteristics of each hydrological station. For example, the peak method can be used to extract the maximum flow of the flood, or statistical methods (such as least squares fitting) can be used to analyze the long-term change trend of the flood and extract the flood characteristics of different hydrological stations. Through regression analysis or other parameterization methods, based on the time series dynamic change data and meteorological data, the flood characteristic parameters of each hydrological station in the basin are derived. The basin characteristic data (such as basin area, slope, land use, etc.) are combined with the regional flood time series dynamic change data, and the characteristic parameters closely related to the occurrence of floods are extracted through multiple regression analysis, weighted average and other methods. GIS tools can be used for spatial analysis, combined with hydrological models and the spatial distribution characteristics of the basin, to extract the geographical and hydrological characteristic parameters of the station.

[0032] Step S2: deriving the prior information of the flood distribution parameter regression coefficient for the flood characteristic parameters of the hydrological stations in the region, and obtaining the maximum likelihood estimate of the distribution parameters of a single hydrological station; importing the maximum likelihood estimate of the distribution parameters of a single hydrological station and the watershed characteristic data into a preset regional regression model to perform prior probability distribution calculation, and obtaining the prior probability distribution data of the regression coefficient of a single station; In the embodiment of the present invention, the flood characteristic parameters of each hydrological station in the region, such as flood peak flow, flood frequency, flood duration, etc., are obtained from step S1. These flood characteristic parameters will be used for subsequent regression analysis. Select a suitable probability distribution function to describe the distribution of flood flow or water level data. Commonly used flood distribution functions include Weibull distribution, Gumbel distribution, Log-normal distribution, etc. The maximum likelihood estimation method is used to solve the maximum likelihood estimation value based on the flood characteristic parameter data of each hydrological station. The characteristic data of the basin (such as basin area, land use, climate conditions, etc.) are combined with the maximum likelihood estimation value of a single hydrological station as input to construct a regression model. The basin characteristic data is used as an explanatory variable in the regression model. Design a regional regression model (such as linear regression, polynomial regression or other appropriate regression model) to calculate the regression coefficient of each station. The maximum likelihood estimation value of the obtained single station distribution parameter and the basin characteristic data are used in the regression model to calculate the prior probability distribution. The prior distribution can be estimated by Bayesian inference or Markov Chain Monte Carlo (MCMC) methods. Through the regional regression model and Bayesian inference, the prior probability distribution data of the regression coefficient of each hydrological station is obtained. These data provide prior information for the subsequent flood model parameter estimation.

[0033] Step S3: construct a likelihood function based on the prior probability distribution data of the regression coefficient of a single site and the flood sequence data to obtain a flood distribution likelihood function; derive the posterior information of the regression coefficient of the flood distribution parameter from the flood distribution likelihood function through the Bayesian theory, thereby generating a Bayesian estimate of the regression coefficient of a single site; In an embodiment of the present invention, by combining the prior probability distribution data of the regression coefficient of a single station generated in step S2 with the historical flood sequence data, a suitable probability distribution function (such as Weibull distribution, Gumbel distribution, Log-normal distribution, etc.) is selected to model the flood data. The likelihood function is used to describe the possibility of the regression coefficient parameters under the given observation data (such as flood sequence data). Combined with the prior distribution of the regression coefficient of a single hydrological station derived in step S2, the relationship between the flood distribution parameters such as the parameters of the likelihood function and the observed flood sequence data can be further refined to obtain a specific likelihood function. Under the Bayesian statistical framework, the posterior distribution of the regression coefficient can be derived from the prior distribution and the likelihood function by Bayes' theorem. By combining the prior probability distribution and the likelihood function, the posterior distribution is calculated using the Bayesian formula: For Bayesian estimation, the maximum a posteriori estimation (MAP) is generally used to calculate the most likely regression coefficient, that is, to maximize the posterior distribution, and the posterior distribution can be maximized by an optimization algorithm (such as gradient descent, MCMC sampling, etc.). If you need to obtain detailed distribution information of the regression coefficient, you can sample the posterior distribution through the Markov Chain Monte Carlo (MCMC) method to obtain the regression coefficients of multiple samples. These samples reflect the potential uncertainty of the regression coefficient. The Bayesian estimate (i.e., the posterior mean) can be used as the optimal estimate of the regression coefficient. If the posterior distribution is a Gaussian distribution, its mean can be directly taken as the Bayesian estimate. The Bayesian estimate can not only provide the optimal value of the regression coefficient, but also give the distribution of the regression coefficient and reveal the uncertainty of the parameter. The uncertainty can be evaluated by the variance or standard deviation of the posterior distribution, which will help analyze the credibility of the model prediction.

[0034] Step S4: construct a spatial association network based on the Bayesian estimated value of the regression coefficient of a single site to generate a site spatial association network; perform flood correlation analysis on the Bayesian estimated value of the regression coefficient of a single site using the site spatial association network to generate site flood spatial association data; In an embodiment of the present invention, each hydrological station in the region is regarded as a node in the network, and the attributes of each node include the Bayesian regression coefficient estimate of the station, the geographical location coordinates (such as longitude and latitude), and the basin characteristic data of the station (such as basin area, slope, etc.). The preliminary association weight is calculated according to the geographical distance between the nodes. Generally, the spatial distance between two stations is calculated using a distance formula (such as Euclidean distance). The closer the distance, the greater the association weight. The association weight is further corrected in combination with the similarity between the Bayesian regression coefficient estimates of the stations. The closer the regression coefficients are, the stronger the similarity between the two stations in flood characteristics. By combining the geographical distance weight and the regression coefficient similarity weight, a comprehensive association weight between stations is generated to quantify the spatial association strength between stations. According to the comprehensive association weight between stations, a station spatial association matrix is ​​constructed. Each element in the matrix represents the spatial association strength between two stations, and the main diagonal element of the matrix is ​​zero (that is, the association between the station itself and itself is zero). Based on the station spatial association matrix, it is converted into a network structure to generate a station spatial association network, in which the nodes in the network represent stations and the edge weights represent the association strength between stations. In the site spatial association network, the flood characteristic correlation index between sites is calculated by analyzing the time series of flood sequence data. Specifically, the correlation degree of flood characteristics between two sites is quantified by statistically analyzing the temporal correlation of site flood sequences (such as the Pearson correlation coefficient). Based on the above analysis results, the flood sequence correlation is combined with the weight of the spatial association network to further generate flood spatial association data between each pair of sites. The analysis results are collated to generate a complete data set, recording the spatial correlation strength, flood characteristic correlation strength, and possible correlation path or impact range between each pair of sites.

[0035] Step S5: perform spatial regional flood data collection on the site flood spatial correlation data to obtain inconsistent regional flood collection data; perform fast Fourier transform on the inconsistent regional flood collection data to generate inconsistent regional flood spectrum data; perform cross-flood frequency verification on the inconsistent regional flood spectrum data based on the prior probability distribution data of the single site regression coefficient and the Bayesian estimate of the single site regression coefficient to generate a regional flood frequency distribution map.

[0036] In an embodiment of the present invention, target sites in a region are screened based on site flood spatial correlation data, and flood sequence data are collected through historical records and real-time observations. The collected flood sequence data of each site are spatially integrated to form a basic data set of regional flood dynamic changes. Non-consistent areas with obvious differences in flood characteristics are identified, and flood data in these areas are extracted to form non-consistent regional flood collection data. De-trending processing (such as removing long-term trends or outliers) is performed on the non-consistent regional flood collection data to ensure the accuracy of frequency domain analysis. Fast Fourier transform is performed on the pre-processed flood data sequence to obtain the representation of flood data in the frequency domain and generate non-consistent regional flood spectrum data. The main frequency component and energy distribution in the spectrum are extracted to analyze the periodicity and burst characteristics of flood events in time. The prior probability distribution data of the regression coefficient of a single site and the Bayesian estimate of the regression coefficient of the site are used as input parameters for frequency verification. By jointly analyzing the Bayesian estimate, the non-consistent regional flood spectrum data and the prior distribution of the regression coefficient, a cross-validation model is constructed to evaluate the reliability of the flood frequency characteristics. Compare the predicted frequency distribution with the actual spectrum data, adjust the frequency verification model, and generate verified regional flood spatial distribution data. Based on the verified regional flood spatial distribution data, perform spatial visualization mapping of flood frequency characteristics. The horizontal axis represents the spatial position (such as longitude and latitude). The vertical axis or color depth represents the magnitude of the flood frequency or characteristic value. Combined with GIS (geographic information system) tools or frequency distribution mapping algorithms, the flood frequency distribution is presented as a regional frequency distribution map to intuitively display the differences in flood risks in different areas of the basin.

[0037] Preferably, step S1 comprises the following steps: Step S11: Obtain flood sequence data, meteorological data and watershed characteristic data of hydrological stations in the watershed; Step S12: preprocessing the flood sequence data, meteorological data and watershed characteristic data to generate standard flood sequence data, standard meteorological data and standard watershed characteristic data, wherein the data preprocessing includes data cleaning, data denoising, data missing value filling and data standardization; Step S13: Analyze the regional flood dynamic change on the standard flood sequence data and the standard meteorological data to generate regional flood time series dynamic change data; Step S14: extract flood characteristic parameters of hydrological stations in the standard watershed characteristic data based on the regional flood time series dynamic change data to obtain flood characteristic parameters of hydrological stations in the region.

[0038] In an embodiment of the present invention, historical flood data is obtained from hydrological monitoring stations in the basin, including information such as the time of occurrence, flow, and duration of the flood. Meteorological data such as precipitation, temperature, wind speed, humidity, etc. are collected in the basin area, and these data can be obtained through meteorological stations or remote sensing data. Including information such as the topography, soil type, land use, and vegetation coverage of the basin. These data can be obtained from a geographic information system (GIS) or a remote sensing data source to help understand the hydrological behavior of the basin. Identify and delete erroneous or incomplete data points. For example, missing timestamps or outliers are processed. For continuous meteorological data or flood series data, filtering techniques (such as low-pass filtering) are used to remove random noise to ensure that the data is smooth and reliable. Interpolation methods (such as linear interpolation, KNN interpolation, etc.) are used to fill in missing values ​​to ensure the continuity of time series data. All types of data (such as flood flow, precipitation, etc.) are normalized or standardized to eliminate unit differences and facilitate subsequent analysis and modeling. Time series analysis is performed based on flood series data and meteorological data. Time series modeling methods (such as ARIMA, LSTM, etc.) can be used to analyze the relationship between the law of flood occurrence and meteorological factors. Through the analysis results, the time series change data describing the regional floods, such as the frequency, intensity, and duration of floods, are generated. Based on the regional flood time series dynamic change data of step S13, the flood characteristic parameters of each hydrological station in the basin are extracted. For example, by analyzing the spatiotemporal distribution, intensity, and frequency of floods at different stations, flood characteristics such as peak flow and flow duration are extracted. The flood characteristic parameters of each hydrological station in the basin are integrated to generate regional hydrological station flood characteristic parameters, providing basic data for subsequent flood warning, prediction, and other work.

[0039] Preferably, step S13 includes the following steps: Step S131: performing time series decomposition on the flood sequence data to generate flood sequence time series decomposition data; performing random fluctuation component separation on the flood sequence time series decomposition data to generate flood sequence random fluctuation component data; Step S132: performing meteorological change pattern analysis on meteorological data to generate short-term meteorological pattern data and long-term meteorological pattern data; performing extreme meteorological flood change analysis on random fluctuation component data of flood sequence according to the short-term meteorological pattern data to generate extreme meteorological flood change data; Step S133: performing common meteorological flood change analysis on the random fluctuation component data of the flood sequence through long-term meteorological model data to generate common meteorological flood change data; performing change peak extraction on the extreme meteorological flood change data to obtain extreme meteorological flood change peak data; Step S134: extract the change mean of ordinary meteorological flood change data to obtain ordinary meteorological flood change mean data; analyze the regional flood meteorological linkage effect based on the extreme meteorological flood change peak data and the ordinary meteorological flood change mean data to generate regional flood time series dynamic change data.

[0040] In an embodiment of the present invention, by using a time series analysis method (such as STL decomposition method, Hodrick-Prescott filtering method, etc.), the flood sequence data is decomposed into long-term trends, seasonal changes and random fluctuation components. This step can help identify long-term change trends and periodic fluctuations in the data. By using random fluctuation component separation technology (such as wavelet transform or empirical mode decomposition EMD), the short-term random fluctuation components in the flood sequence are extracted. This data is used to capture abnormal fluctuations during the occurrence of floods, especially fluctuations under extreme weather or emergencies. Statistical analysis or machine learning methods (such as principal component analysis PCA, factor analysis, etc.) are used to extract change patterns from meteorological data. Short-term meteorological pattern data (such as daily or monthly change patterns) and long-term meteorological pattern data (such as interannual changes or climate change trends) are generated. Based on the short-term meteorological pattern data, the random fluctuation component data of the flood sequence is analyzed, especially focusing on extreme meteorological conditions (such as extreme precipitation, strong winds, etc.), evaluating the impact of these meteorological changes on floods, generating extreme meteorological flood change data, and revealing the relationship between floods and extreme weather. Based on long-term meteorological pattern data (such as seasonal changes, annual climate change, etc.), the random fluctuation component data of the flood sequence is analyzed, the impact of meteorological factors (such as seasonal precipitation changes, temperature changes, etc.) on flood fluctuations is identified, and ordinary meteorological flood change data is generated. The change peaks are extracted from the extreme meteorological flood change data, and the focus is on analyzing those events with particularly abnormal flood intensity or duration. These peaks can help identify the most extreme fluctuations and influencing factors in floods. The average of ordinary meteorological flood change data is calculated to obtain the mean of flood fluctuations under long-term meteorological conditions, reflecting the characteristics of flood changes under conventional meteorological conditions. The extreme meteorological flood change peak data and the ordinary meteorological flood change mean data are combined, and the interaction and influence between meteorological factors and flood occurrence are analyzed using linkage analysis methods (such as multivariate regression, correlation analysis, etc.) to generate regional flood time series dynamic change data. This data reveals the dynamic change law of regional floods under different meteorological modes.

[0041] Preferably, the regional flood-meteorological linkage effect analysis based on the extreme meteorological flood change peak data and the ordinary meteorological flood change mean data includes: The flood-meteorological linkage effect index is calculated by using the linkage effect quantification formula for the peak data of extreme meteorological flood changes and the mean data of ordinary meteorological flood changes to obtain the flood-meteorological linkage effect index; the linkage effect regional sensitivity analysis is conducted on the hydrological stations in the basin based on the flood-meteorological linkage effect index to generate regional flood linkage sensitivity data; Based on the regional flood linkage sensitivity data, the flood-meteorological linkage effect index is ranked by regional effects to generate regional sensitivity ranking data; the flood intensity change trend of extreme meteorological flood change peak data and ordinary meteorological flood change mean data is analyzed through regional sensitivity ranking data to generate regional flood time series dynamic change data.

[0042] In an embodiment of the present invention, a formula for the linkage effect between floods and meteorological factors is established by using a mathematical model or a regression analysis method. The peak value data of extreme meteorological flood changes and the mean value data of ordinary meteorological flood changes are substituted into the above formula, and the weight coefficient is determined by statistical analysis and modeling to obtain the flood-meteorological linkage effect index of each hydrological station or region. Based on the flood-meteorological linkage effect index, the sensitivity of different hydrological stations or regions to meteorological changes is evaluated by sensitivity analysis methods (such as local sensitivity analysis, Monte Carlo simulation, etc.). The flood-meteorological linkage effect index is compared with the meteorological and hydrological data of each hydrological station in the region to evaluate the response of each hydrological station to flood changes under extreme meteorological and ordinary meteorological conditions. Each station is scored for sensitivity to generate a sensitivity value for each hydrological station. By sorting the flood-meteorological linkage effect index of each hydrological station, sensitivity ranking data of different hydrological stations in the region are generated. The linkage effect index can be ranked using a ranking method or a ranking algorithm (such as Kendall rank correlation coefficient, Spearman rank correlation analysis, etc.). The generated regional sensitivity ranking data reflects which areas are most sensitive to flood meteorological changes, helping to identify high-risk areas of focus. Combined with regional sensitivity ranking data and flood change data, the trend analysis of flood intensity in the region is carried out through regression analysis or time series prediction (such as ARIMA model, LSTM, etc.). The priority analysis areas are determined according to the ranking, and trend prediction is carried out using the peak data of extreme meteorological flood changes and the mean data of ordinary meteorological flood changes. The change trend of flood intensity in the region is obtained, and the upward or downward trend of flood intensity and its correlation with meteorological changes are analyzed. The trend analysis results are combined with the regional sensitivity ranking data to generate regional flood time series dynamic change data, which describes the flood change laws in different regions under different meteorological conditions and provides them to decision makers for flood warning and disaster prevention planning.

[0043] Preferably, the linkage effect quantification formula is used to calculate the flood meteorological linkage effect index for the extreme meteorological flood change peak data and the ordinary meteorological flood change mean data. The linkage effect quantification formula is as follows: In the formula, It is expressed as the flood-meteorological linkage effect index, Expressed as the peak value of extreme meteorological floods, It is expressed as the average value of ordinary meteorological flood. It is expressed as the mean temperature under extreme meteorological conditions. It is expressed as the mean temperature under normal meteorological conditions. Expressed as extreme rainfall, Expressed as normal rainfall, Expressed as the duration of extreme rainfall, It is expressed as the duration of normal rainfall, Expressed as the impact weight of flood peak, Expressed as the impact weight of temperature change, Expressed as the impact weight of rainfall change, Expressed as the impact weight of rainfall duration.

[0044] The present invention analyzes and integrates a linkage effect quantification formula, which is based on the changes in flood characteristics under different meteorological conditions (extreme weather and ordinary weather), considers the impact of meteorological factors (temperature, rainfall, rainfall duration) on flood changes, and quantifies these impacts into a linkage effect index. The numerator in the formula (reflecting the changes in floods) is: Indicates flood peaks under extreme weather conditions Compared with the flood average under normal weather conditions The difference between them represents the magnitude of flood changes and reflects the direct impact of extreme meteorological conditions on floods. Indicates the temperature under extreme weather conditions Temperature under normal weather conditions The difference between the two reflects the indirect impact of temperature change on floods. Temperature changes can indirectly change flood conditions by affecting precipitation, evaporation, snowmelt and other processes. The denominator (reflecting changes in meteorological factors): Indicates the amount of rainfall under extreme meteorological conditions The rainfall under normal weather conditions The difference between the two reflects the driving effect of rainfall changes on floods. Rainfall is the direct cause of floods, and extreme rainfall may significantly increase the intensity and frequency of floods. Indicates the duration of rainfall under extreme meteorological conditions Duration of normal rainfall The difference between reflects the impact of rainfall duration changes on floods. The duration of rainfall directly affects the accumulation of ground water, thereby affecting the occurrence and change of floods. The impact weight of flood peak value indicates the contribution of flood peak value changes under extreme meteorological conditions to the linkage effect. If the flood peak value changes greatly under extreme meteorological conditions, it means that the area is more sensitive to extreme meteorological changes, and the parameter The impact weight of temperature change indicates the impact of temperature change on flood changes. If the temperature changes significantly, it may cause stronger precipitation or snowmelt, thus affecting the formation of floods. The size of reflects the indirect impact of temperature change on floods. The impact weight of rainfall change represents the direct impact of rainfall change on floods. The greater the rainfall, the higher the intensity and frequency of floods may be. The larger the value, the more important the rainfall is in flood generation. The impact weight of the change in rainfall duration indicates the impact of rainfall duration on floods. If the rainfall duration is longer, it usually means a higher flood risk. The weight indicates the degree of influence of rainfall duration on floods. When using the conventional linkage effect quantification formula in this field, the flood-meteorological linkage effect index can be obtained. By applying the linkage effect quantification formula provided by the present invention, the flood-meteorological linkage effect index can be calculated more accurately. The linkage effect quantification formula comprehensively considers the linkage relationship between flood changes and meteorological factors, and uses weight parameters to quantify the impact of each factor on flood risks, so as to provide a multi-dimensional and comprehensive flood-meteorological linkage effect index. By accurately calculating flood changes under different meteorological conditions, it can provide an important basis for flood warning, risk assessment, and disaster prevention and mitigation.

[0045] As an example of the present invention, refer to Figure 2 As shown, in this example, step S2 includes: Step S21: Calculate the maximum likelihood estimate of the flood sequence distribution parameters for the flood characteristic parameters of the hydrological stations in the region to obtain the maximum likelihood estimate of the distribution parameters of a single hydrological station, wherein the formula for calculating the maximum likelihood estimate of the flood sequence distribution parameters is as follows: ; In the formula, is the regression coefficient of the flood sequence distribution parameter, represents the distribution parameter estimated based on the flood series of a single site only, Indicates i Site distribution parameters H non-consistent covariates, and denote the link functions of the first and second distribution parameters, respectively, is a linear function; Step S22: construct a regional regression model; the regional regression model is represented by a model matrix to obtain a matrixed regional regression model; wherein the construction formula of the regional regression model is as follows: ; ; In the formula, Indicates i Sites The variance of the error caused by the sample estimate of Indicates the site i With site j The correlation coefficient between the sample errors is the same as that under the consistency condition; The formula for the model matrix representation is as follows: ; In the formula, X is given by K Site C The regional covariates K ×( C +1) rank matrix; The flood sequence λ The regional regression coefficients of the distribution parameters; Indicated in t Time by C The vector of random effect terms of the basin characteristic indicators is assumed to have a mean of 0, and the covariance matrix is The multivariate normal distribution of Indicated in t Time by K The flood sequence of the station λ A vector of regional regression model errors consisting of distribution parameters; Indicated in t Time by K The flood sequence of the station λ A vector of sample errors for the estimates of the distribution parameters; Indicated in t The total error at time is and The sum of .

[0046] Step S23: importing the maximum likelihood estimation value of the distribution parameter of a single hydrological station and the watershed characteristic data into the matrixed regional regression model to perform prior probability distribution calculation, and obtain the prior probability distribution data of the regression coefficient of the single station; The formula for calculating the prior probability distribution is as follows: ; In the formula, Data expressed as a priori probability distributions of regression coefficients for individual sites.

[0047] In the embodiment of the present invention, based on the given flood sequence data and covariates, the maximum likelihood estimation (MLE) method is used to calculate the flood sequence distribution parameters of a single hydrological station using a regression model. The model is set as: , obtain time series flood data and related meteorological covariate data from each hydrological station in the region. For each hydrological station, the maximum likelihood estimation method is used for regression analysis to calculate the optimal regression coefficient of each distribution parameter. The maximum likelihood estimate of the flood series distribution parameter of each hydrological station is obtained through the above formula. A multi-station and multi-covariate regression model is used to combine the distribution parameters of the hydrological station with the basin characteristics to construct a regional regression model. The construction formula is: ; ; Here, Indicates i The sample error of a hydrological station at time t is: A matrix representing the regional regression model is constructed by using the estimated distribution parameters of all hydrological stations for the correlation coefficients between different stations. and random effects As variables, a matrix representation containing all basin characteristics and regional covariates is formed. The error terms and covariance matrices of all hydrological stations are calculated and matrixed to obtain the matrix representation of the regional regression model. Based on the regression coefficients of the hydrological stations and the basin characteristic data, the prior probability distribution of the regression coefficients is calculated. The prior probability distribution calculation formula is: = Prior distribution model, where is the regression coefficient, describing the relationship between the hydrological station and the regional characteristics. The distribution parameters and watershed characteristics data obtained by the maximum likelihood estimation method are imported into the regional regression model. The prior distribution data of the regression coefficient are calculated according to the model, and these data will be used for further Bayesian inference or other statistical analysis methods to evaluate future flood changes.

[0048] As an example of the present invention, refer to Figure 3 As shown, in this example, step S3 includes: Step S31: construct a likelihood function based on the prior probability distribution data of the regression coefficient of a single station and the flood sequence data to obtain a flood distribution likelihood function; wherein the construction formula of the likelihood function is as follows: ; In the formula, It is expressed as flood distribution likelihood function; Step S32: Using Bayesian theory to derive the flood posterior probability distribution function from the flood distribution likelihood function, thereby obtaining the posterior probability distribution data of the regression coefficient of a single station; wherein the formula for deriving the flood posterior probability distribution function is as follows: ; In the formula, The posterior probability distribution data of the regression coefficient of a single site is expressed as The value range of Step S33: Performing Bayesian estimation calculation on the prior probability distribution data of the regression coefficient of a single site through the posterior probability distribution data of the regression coefficient of a single site, to obtain the Bayesian estimation value of the regression coefficient of a single site; wherein the Bayesian estimation value calculation is as follows: .

[0049] In the embodiment of the present invention, the flood distribution is described using a likelihood function based on the prior probability distribution data of the regression coefficient of a single station and the flood sequence data. The likelihood function is used to measure the probability of the observed data (flood sequence) when the regression coefficient is given. Formula: ;in Expressed as flood distribution likelihood function, is the flood data with a given regression coefficient The probability density function under the condition of . Using Bayesian theory, combined with the prior probability distribution and the likelihood function, the posterior probability distribution function of flood distribution is derived. The posterior probability distribution function represents the updated distribution of the regression coefficient based on the observed data. ,in The posterior probability distribution data of the regression coefficient of a single site is expressed as The value range of The prior distribution of the regression coefficient is expressed as follows. The Bayesian estimate of the regression coefficient for a single station is calculated using the posterior probability distribution. This process uses the posterior distribution to perform a weighted average of the regression coefficients to obtain the most likely regression coefficient. ,in is the Bayesian estimate of the regression coefficient for a single site.

[0050] Preferably, step S4 comprises the following steps: Step S41: constructing a site feature vector according to the Bayesian estimation value of the regression coefficient of a single site to generate site feature matrix data; Step S42: performing parameter space search on the site characteristic matrix data to generate prior parameter set data; Step S43: construct a spatial association network for the prior parameter set data to generate a site spatial association network; use the site spatial association network to perform flood correlation analysis on the Bayesian estimate of the regression coefficient of a single site to generate site flood spatial association data.

[0051] In an embodiment of the present invention, relevant data including meteorological data (such as precipitation, temperature, etc.), geographic information (such as altitude, land use type, etc.) and historical flood data are collected from each site. Combined with Bayesian estimates, this information is converted into a feature vector. The feature vector of each site may include: precipitation, temperature change, land use type, river basin information and Bayesian estimates of regression coefficients. The feature vectors of all sites are integrated into a site feature matrix. This matrix will be used in subsequent analysis and modeling processes. According to the site feature matrix, a parameter space is defined, including possible regression coefficients, flood probability, precipitation pattern, soil type and other factors. A search algorithm (such as particle swarm optimization, grid search, simulated annealing, etc.) is used to search in the parameter space. This process will find a set of prior parameters that best describe the relationship between site characteristics and flood occurrence. Through the optimization process of the search algorithm, a set of prior parameters that meet the requirements is obtained. These parameters will provide a basis for the next step of spatial association network construction. The spatial association network is constructed using geographic information and flood patterns between sites. A graph theory approach can be used, where each site represents a node in the graph and the edges between nodes represent the spatial correlation between sites (such as geographic distance or similarity of hydrological characteristics). The relationship between sites is represented by constructing an adjacency matrix or a weight matrix. The site spatial association network is used to analyze the flood correlation between sites, combined with the Bayesian estimate of the regression coefficient and flood sequence data. Methods such as correlation analysis, cluster analysis or spatial regression models can be used to evaluate the similarity of flood occurrence between sites. Based on the analysis results, site flood spatial association data are generated, which can help understand the spatiotemporal patterns of flood occurrence between different sites and how to improve flood warning and risk assessment through data from adjacent sites.

[0052] Preferably, constructing a spatial association network for the prior parameter set data includes: Perform multi-scale decomposition on the prior parameter set data to generate feature decomposition data; perform time series window processing on the feature decomposition data to generate time series feature data; perform spatial mapping processing on the time series feature data to generate feature decomposition matrix data; perform coordinate transformation on the feature decomposition matrix data to generate standardized coordinate data; Perform distance calculation on the standardized coordinate data to generate distance matrix data; perform weight assignment on the distance matrix data to generate spatial distance feature data; perform triangulation on the spatial distance feature data to generate initial network data; perform spatial partitioning on the initial network data to generate partition structure data; The degree distribution of the partition structure data is processed to generate network topology structure data; based on the graph neural network algorithm, the spatial association network of the network topology structure data is constructed to generate a site spatial association network.

[0053] In an embodiment of the present invention, a priori parameter sets (such as regression coefficients, precipitation data, etc.) are decomposed into features of multiple scales by using wavelet transform, Fourier transform or similar multi-scale analysis methods. Multi-scale decomposition is performed on each parameter set to extract feature information at different scales. For example, for precipitation data, the data can be decomposed into high-frequency and low-frequency components by wavelet transform, representing detail information and overall trends, respectively. Multi-scale feature data is output for further analysis. According to the time characteristics of the data, a suitable time window is defined. For example, a fixed time span (such as one week or one month) can be selected as the window length. Time series window processing is performed on the feature decomposition data, and the data is divided into time periods to obtain data in each window. Features (such as mean, variance, trend, etc.) are extracted in each time window. A time series feature vector is generated for each time window to represent the change characteristics in the time period. The time series feature data is mapped to a spatial coordinate system according to the geographic coordinates, features and other information of the site. Interpolation methods or mapping algorithms such as Lagrange interpolation and Kriging interpolation can be used. Map the time series feature data of each site to the spatial coordinates to form a feature matrix in the spatial coordinate system. Through spatial mapping processing, the matrix data obtained contains spatiotemporal information and can be further processed. Convert the spatial coordinate data to a standardized coordinate system through linear transformation or coordinate standardization method. Common methods include normalization, Z-score standardization, etc. Process the coordinates of each site through a conversion formula or method to obtain standardized coordinate data. The standardized coordinate data will be used for subsequent distance calculation and network construction. According to the geographic coordinates of the site, select a suitable distance measurement method. Commonly used methods include Euclidean distance, Manhattan distance, etc. Calculate the distance between all sites to generate a symmetric distance matrix. Each element represents the spatial distance between two sites. The distance matrix obtained provides the basis for subsequent weight allocation and spatial association analysis. According to the distance matrix, use certain functions (such as Gaussian kernel function, inverse distance weighting, etc.) to assign weights to each pair of sites. Generally, the closer the distance between sites, the greater the weight. For each pair of sites, calculate its spatial weight. Generally, the weight can be calculated by the following formula: ;in, For Site and Site The distance between is the scale factor of the weight. Through weight allocation, a matrix is ​​generated to represent the spatial distance characteristics between sites. The Delaunay triangulation algorithm is used to generate the initial network structure based on the spatial coordinates of the site. Delaunay triangulation can ensure that the inner point of each triangle does not contain other sites, thereby avoiding unreasonable connections between data. The sites are connected by triangulation to form a preliminary spatial network. The initial network data obtained represents the preliminary spatial connection relationship between sites. The initial network can be divided into several functional areas using spectral clustering algorithms, community detection algorithms (such as Louvain algorithm), etc. The sites are divided into different areas according to the spatial location and network structure of the site. The partitioned structural data is output, and each partition contains several sites, representing the spatial relationship in the area. The degree of each site (that is, the number of connections with other sites) is calculated, and the degree distribution is analyzed. The degree distribution diagram can be drawn to analyze the importance of the site in the network. According to the degree distribution data, the topological structure between sites is established to clarify which sites have higher connectivity and which are hubs in the network. The final network topology data is obtained, which will be used for the subsequent construction of the spatial association network. Select appropriate GNN models, such as GCN (graph convolutional network) and GAT (graph attention network), to learn the relationship between nodes (sites) in the network. Use network topology data to train the GNN model to learn the spatial correlation and flood correlation between sites. After training, use the model to generate the final site spatial correlation network to reveal the spatial flood correlation between sites.

[0054] Preferably, step S5 includes the following steps: Step S51: performing spatial regional flood data collection on the site flood spatial correlation data to obtain non-consistent regional flood collection data; performing fast Fourier transform on the non-consistent regional flood collection data to generate non-consistent regional flood spectrum data; Step S52: performing cross-flood frequency verification on the inconsistent regional flood spectrum data according to the prior probability distribution data of the single-station regression coefficient and the Bayesian estimation value of the single-station regression coefficient, and generating verified regional flood spatial distribution data; Step S53: Visualize the flood frequency of the verified regional flood spatial distribution data to generate a regional flood frequency distribution map.

[0055] In an embodiment of the present invention, flood-related inconsistency areas are identified based on site flood spatial correlation data. These areas may have inconsistent flood characteristics due to factors such as topography, climate, and urbanization. Site data in these inconsistency areas are collected to obtain time series flood data in the area. The collected data may include precipitation, flow, water level, etc. The collected flood data is preprocessed, including removing outliers, filling missing data, smoothing, etc. The spatiotemporal consistency of the data is ensured so that it can be used for spectrum analysis. Fast Fourier transform (FFT) is performed on the processed flood data to convert from the time domain to the frequency domain. FFT can help reveal periodic features in the data, such as the frequency, period, and amplitude of flood events. The output result is a spectrum data set that describes the frequency distribution characteristics of flood events in the inconsistency area. By analyzing the spectrum data, the periodicity and variation of flood events at different time scales can be revealed. The obtained flood spectrum data is cross-validated with the prior distribution of Bayesian estimation. This step verifies the accuracy of the model by comparing the flood frequency data between different sites. After cross-validation, a corrected regional flood spatial distribution data is generated. This data describes the flood frequency and distribution of each station in the region, and takes into account the Bayesian estimates of each station. This data can be used to further analyze the flood risks in different regions and help decision makers understand the spatial variation of floods. Use appropriate visualization tools (such as heat maps, spatial distribution maps, spectrum maps, etc.) to display the frequency distribution of regional floods. You can use GIS software (such as ArcGIS) or programming libraries (such as Matplotlib, Seaborn, Plotly, etc.) for graphics. Convert the verified regional flood spatial distribution data into a visualization chart. The frequency of floods can be represented by the depth of color, or the flood risk level of different stations can be represented by the size of symbols. For example, a heat map is used to represent the frequency of floods, and the frequency of floods at different stations is displayed by the gradient of regional colors. The final regional flood frequency distribution map will clearly show the frequency of floods in different regions and stations, providing intuitive visual support for regional flood risk assessment. This map can help post-disaster reconstruction, early warning systems, and regional flood control decisions.

Claims

1. A method for analyzing flood frequency in inconsistent regions based on Bayesian theory, characterized in that: The following steps are involved: Step S1: Obtain flood sequence data, meteorological data and basin characteristic data of hydrological stations in the basin; perform regional flood dynamic change analysis on the flood sequence data and meteorological data to generate regional flood time series dynamic change data; Based on the regional flood time series dynamic change data, the flood characteristic parameters of the hydrological stations in the basin are extracted from the basin characteristic data to obtain the flood characteristic parameters of the hydrological stations in the region; Step S2: deriving the prior information of the flood distribution parameter regression coefficient for the flood characteristic parameters of the hydrological stations in the region, and obtaining the maximum likelihood estimate of the distribution parameters of a single hydrological station; importing the maximum likelihood estimate of the distribution parameters of a single hydrological station and the watershed characteristic data into a preset regional regression model to perform prior probability distribution calculation, and obtaining the prior probability distribution data of the regression coefficient of a single station; Step S3: construct a likelihood function based on the prior probability distribution data of the regression coefficient of a single site and the flood sequence data to obtain a flood distribution likelihood function; derive the posterior information of the regression coefficient of the flood distribution parameter from the flood distribution likelihood function through the Bayesian theory, thereby generating a Bayesian estimate of the regression coefficient of a single site; Step S4: construct a spatial association network based on the Bayesian estimated value of the regression coefficient of a single site to generate a site spatial association network; perform flood correlation analysis on the Bayesian estimated value of the regression coefficient of a single site using the site spatial association network to generate site flood spatial association data; Step S5: collecting spatial regional flood data on the site flood spatial correlation data to obtain inconsistent regional flood collection data; The non-uniform area flood collection data is subjected to fast Fourier transform to generate the non-uniform area flood spectrum data; the non-uniform area flood spectrum data is cross-flood frequency verified according to the prior probability distribution data of the single station regression coefficient and the Bayesian estimation value of the single station regression coefficient to generate the regional flood frequency distribution map.

2. The method for analyzing the frequency of flood in inconsistent regions based on Bayesian theory according to claim 1 is characterized in that: Step S1 includes the following steps: Step S11: Obtain flood sequence data, meteorological data and watershed characteristic data of hydrological stations in the watershed; Step S12: preprocessing the flood sequence data, meteorological data and watershed characteristic data to generate standard flood sequence data, standard meteorological data and standard watershed characteristic data, wherein the data preprocessing includes data cleaning, data denoising, data missing value filling and data standardization; Step S13: Analyze the regional flood dynamic change on the standard flood sequence data and the standard meteorological data to generate regional flood time series dynamic change data; Step S14: extract flood characteristic parameters of hydrological stations in the standard watershed characteristic data based on the regional flood time series dynamic change data to obtain flood characteristic parameters of hydrological stations in the region.

3. The method for analyzing the frequency of flood in inconsistent regions based on Bayesian theory according to claim 2 is characterized in that: Step S13 includes the following steps: Step S131: performing time series decomposition on the flood sequence data to generate flood sequence time series decomposition data; performing random fluctuation component separation on the flood sequence time series decomposition data to generate flood sequence random fluctuation component data; Step S132: performing meteorological change pattern analysis on meteorological data to generate short-term meteorological pattern data and long-term meteorological pattern data; performing extreme meteorological flood change analysis on random fluctuation component data of flood sequence according to the short-term meteorological pattern data to generate extreme meteorological flood change data; Step S133: performing common meteorological flood change analysis on the random fluctuation component data of the flood sequence through long-term meteorological model data to generate common meteorological flood change data; performing change peak extraction on the extreme meteorological flood change data to obtain extreme meteorological flood change peak data; Step S134: extract the change mean of ordinary meteorological flood change data to obtain ordinary meteorological flood change mean data; analyze the regional flood meteorological linkage effect based on the extreme meteorological flood change peak data and the ordinary meteorological flood change mean data to generate regional flood time series dynamic change data.

4. The method for analyzing the frequency of flood in inconsistent regions based on Bayesian theory according to claim 3 is characterized in that: The regional flood-meteorological linkage effect analysis based on the extreme meteorological flood change peak data and the ordinary meteorological flood change mean data includes: The flood-meteorological linkage effect index is calculated by using the linkage effect quantification formula for the peak data of extreme meteorological flood changes and the mean data of ordinary meteorological flood changes to obtain the flood-meteorological linkage effect index; the linkage effect regional sensitivity analysis is conducted on the hydrological stations in the basin based on the flood-meteorological linkage effect index to generate regional flood linkage sensitivity data; Based on the regional flood linkage sensitivity data, the flood-meteorological linkage effect index is ranked by regional effects to generate regional sensitivity ranking data; the flood intensity change trend of extreme meteorological flood change peak data and ordinary meteorological flood change mean data is analyzed through regional sensitivity ranking data to generate regional flood time series dynamic change data.

5. The method for analyzing the frequency of flood in inconsistent regions based on Bayesian theory according to claim 4 is characterized in that: The quantitative formula of linkage effect is used to calculate the flood meteorological linkage effect index for the peak data of extreme meteorological flood changes and the mean data of ordinary meteorological flood changes. The quantitative formula of linkage effect is as follows: In the formula, It is expressed as the flood-meteorological linkage effect index, Expressed as the peak value of extreme meteorological floods, It is expressed as the average value of ordinary meteorological flood. It is expressed as the mean temperature under extreme meteorological conditions. It is expressed as the mean temperature under normal meteorological conditions. Expressed as extreme rainfall, Expressed as normal rainfall, Expressed as the duration of extreme rainfall, It is expressed as the duration of normal rainfall, Expressed as the impact weight of flood peak, Expressed as the impact weight of temperature change, Expressed as the impact weight of rainfall change, Expressed as the impact weight of rainfall duration.

6. The method for analyzing the frequency of flood in inconsistent regions based on Bayesian theory according to claim 1 is characterized in that: Step S2 includes the following steps: Step S21: Calculate the maximum likelihood estimate of the flood sequence distribution parameters for the flood characteristic parameters of the hydrological stations in the region to obtain the maximum likelihood estimate of the distribution parameters of a single hydrological station, wherein the formula for calculating the maximum likelihood estimate of the flood sequence distribution parameters is as follows: ； In the formula, is the regression coefficient of the flood sequence distribution parameter, represents the distribution parameter estimated based on the flood series of a single site only, Indicates i Site distribution parameters H non-consistent covariates, and denote the link functions of the first and second distribution parameters, respectively, is a linear function; Step S22: construct a regional regression model; the regional regression model is represented by a model matrix to obtain a matrixed regional regression model; wherein the construction formula of the regional regression model is as follows: ； ； In the formula, Indicates i Sites The variance of the error caused by the sample estimate of Indicates the site i With site j The correlation coefficient between the sample errors is the same as that under the consistency condition; The formula for the model matrix representation is as follows: ； In the formula, X is given by K Site C The regional covariates K ×( C +1) rank matrix; The flood sequence λ The regional regression coefficients of the distribution parameters; Indicated in t Time by C The vector of random effect terms of the basin characteristic indicators is assumed to have a mean of 0, and the covariance matrix is The multivariate normal distribution of Indicated in t Time by K The flood sequence of the station λ A vector of regional regression model errors consisting of distribution parameters; Indicated in t Time by K The flood sequence of the station λ A vector of sample errors for the estimates of the distribution parameters; Indicated in t The total error at time is and The sum of ; Step S23: importing the maximum likelihood estimation value of the distribution parameter of a single hydrological station and the watershed characteristic data into the matrixed regional regression model to perform prior probability distribution calculation, and obtain the prior probability distribution data of the regression coefficient of the single station; The formula for calculating the prior probability distribution is as follows: ； In the formula, Data expressed as a priori probability distributions of regression coefficients for individual sites.

7. The method for analyzing the frequency of flood in inconsistent regions based on Bayesian theory according to claim 1 is characterized in that: Step S3 includes the following steps: Step S31: construct a likelihood function based on the prior probability distribution data of the regression coefficient of a single station and the flood sequence data to obtain a flood distribution likelihood function; wherein the construction formula of the likelihood function is as follows: ； In the formula, It is expressed as flood distribution likelihood function; Step S32: Using Bayesian theory to derive the flood posterior probability distribution function from the flood distribution likelihood function, thereby obtaining the posterior probability distribution data of the regression coefficient of a single station; wherein the formula for deriving the flood posterior probability distribution function is as follows: ； In the formula, The posterior probability distribution data of the regression coefficient of a single site is expressed as The value range of Step S33: Performing Bayesian estimation calculation on the prior probability distribution data of the single site regression coefficient through the posterior probability distribution data of the single site regression coefficient to obtain the Bayesian estimation value of the single site regression coefficient; wherein the Bayesian estimation value calculation is as follows: 。 8. The method for analyzing the frequency of flood in inconsistent regions based on Bayesian theory according to claim 1 is characterized in that: Step S4 includes the following steps: Step S41: constructing a site feature vector according to the Bayesian estimation value of the regression coefficient of a single site to generate site feature matrix data; Step S42: performing parameter space search on the site characteristic matrix data to generate prior parameter set data; Step S43: construct a spatial association network for the prior parameter set data to generate a site spatial association network; use the site spatial association network to perform flood correlation analysis on the Bayesian estimate of the regression coefficient of a single site to generate site flood spatial association data.

9. The method for analyzing the frequency of flood in inconsistent regions based on Bayesian theory according to claim 8 is characterized in that: The construction of spatial association network for prior parameter set data includes: Perform multi-scale decomposition on the prior parameter set data to generate feature decomposition data; perform time series window processing on the feature decomposition data to generate time series feature data; perform spatial mapping processing on the time series feature data to generate feature decomposition matrix data; perform coordinate transformation on the feature decomposition matrix data to generate standardized coordinate data; Perform distance calculation on the standardized coordinate data to generate distance matrix data; perform weight assignment on the distance matrix data to generate spatial distance feature data; perform triangulation on the spatial distance feature data to generate initial network data; perform spatial partitioning on the initial network data to generate partition structure data; The degree distribution of the partition structure data is processed to generate network topology structure data; based on the graph neural network algorithm, the spatial association network of the network topology structure data is constructed to generate a site spatial association network.

10. The method for analyzing the frequency of flood in inconsistent regions based on Bayesian theory according to claim 1, characterized in that: Step S5 includes the following steps: Step S51: performing spatial regional flood data collection on the site flood spatial correlation data to obtain non-consistent regional flood collection data; performing fast Fourier transform on the non-consistent regional flood collection data to generate non-consistent regional flood spectrum data; Step S52: performing cross-flood frequency verification on the inconsistent regional flood spectrum data according to the prior probability distribution data of the single-station regression coefficient and the Bayesian estimation value of the single-station regression coefficient, and generating verified regional flood spatial distribution data; Step S53: Visualize the flood frequency of the verified regional flood spatial distribution data to generate a regional flood frequency distribution map.

Images