CN114580171B - A method for identification of watershed flood types and analysis of their influencing factors - Google Patents

Abstract

The invention discloses a method for identifying the type of flood in a drainage basin and analyzing the influence factors of the flood, which comprises the following steps: collecting geographic factors of weather, human activity factors, social and economic indexes and runoff data of a hydrological site in a research area; secondly, preprocessing the data; constructing a flood event dividing method for coupling the ultra-quantitative threshold sampling and the flow sequence moving variance, and dividing flood events from hydrologic site daily runoff data; step four, constructing a flood behavior characteristic index system to depict a complete flood process; step five, clustering the flood events obtained by division by using a clustering method, and identifying the main types and the flood characteristics of the flood in the drainage basin; step six, statistically analyzing the main flood types of each hydrological station, and identifying regional characteristics of the watershed flood process; and seventhly, analyzing the influence of the spatial heterogeneity of the influence factors on the flood process of the drainage basin based on the geographic detector. The method can accurately divide flood events from long-sequence runoff data, identify the main flood types and main influence factors of the watershed, and provide support for characteristic information mining, rain flood resource utilization and water resource scientific management of the watershed flood process.

Description

A method for identification of watershed flood types and analysis of their influencing factors

technical field

The invention relates to the technical field of flood type identification, in particular to a method for flood type identification in a watershed and analysis of its impact factors.

Background technique

Flood is one of the most frequent and most harmful natural disasters in the world, which seriously threatens the safety of human life and property. Affected by differences in regional climate, topography and socio-economic development, the spatial and temporal distribution of water resources is uneven, resulting in significant regional differences in flood processes. Under the background of global changes and rapid socio-economic development, flood damage is becoming more and more serious. Rapid and effective identification of the main flood types and their influencing factors in the basin can provide scientific and technological support for the management of water resources in the basin and flood control and disaster reduction.

Flood type identification requires a large number of flood event samples, and it is crucial to accurately classify flood processes from continuous runoff sequences. At present, most of the methods to divide the flood process from the continuous runoff sequence are combined with rainfall data and expert experience to artificially determine the flood process. This method mostly relies on expert experience to determine the point of rising and falling water, which accounts for a large proportion of subjective factors, and is used in large sample batch screening. Less efficient during flooding. Therefore, it is necessary to construct a reasonable method to quickly and accurately divide flood events from continuous runoff sequences, which is of great significance for the identification of flood types. In addition, identifying the main flood types and their regional characteristics in the basin is of great significance for guiding flood management. The core of identifying flood types and their regional characteristics is to construct a comprehensive feature index system to describe the flood in the basin, and based on this, use the clustering method to cluster and analyze the flood process in the basin, and identify flood types with similar processes. The current flood behavior characteristic index system can be divided into three categories: hydrometeorological indicators, hydrological indicators and indicators based on flood process. Hydrometeorological indicators and hydrological indicators mainly consider the factors that cause floods, including hydrometeorological indicators such as cyclone paths, atmospheric circulation patterns, and dynamic characteristics of weather systems, as well as hydrological and meteorological indicators such as precipitation, temperature, evaporation, snow depth, and soil moisture. index. Flood process-based indicators are calculated based on the behavioral characteristics of actual flood events. Compared with hydrometeorological indicators and hydrological indicators, floods with the same flood behavior are guaranteed to have hydrological similarity, and are increasingly obtained worldwide. more applications. However, the indicators based on the flood process used in the current research mainly focus on the magnitude characteristics of the flood, including flood peak, flood volume and runoff depth, etc., ignoring the characteristic elements of the changing process of the flood hydrograph, resulting in the flood characteristic information of the watershed in different degrees. Compression and loss. Therefore, the indicator system of future flood behavior characteristics needs to be further developed and improved, focusing on the relevant indicators of flood magnitude, time domain characteristics, rate of change, and morphological characteristics of flood hydrographs. At the same time, there are many factors that affect the flood process, and each factor has a different degree of interpretation on the flood process. Therefore, it is also crucial to accurately identify the main controlling factors among the many factors.

SUMMARY OF THE INVENTION

The purpose of the present invention is to provide a method for identifying flood types in a watershed and analyzing its impact factors, and to construct a flood event division method that couples ultra-quantitative threshold sampling and flow sequence moving variance, so as to quickly and effectively obtain accurate data from the daily runoff data of hydrological stations. Divide flood events, identify the main flood types and their representative characteristics in the basin, analyze the impact of spatial heterogeneity of influencing factors on the flood process, and provide support for the information mining of flood process characteristics in the basin, the utilization of rainwater resources and the scientific management of water resources.

To achieve the above object, the technical scheme adopted by the present invention is as follows:

Step 1) Collect and download continuous daily-scale runoff data of hydrological stations in the study area; collect and download meteorological and geographical factors, human activity factors and socio-economic factors that affect the flood process in the basin, including: precipitation, temperature, wind speed, DEM, NDVI, slope, urban change rate, population density, GDP and carbon emissions, etc.;

Step 2) Data preprocessing, including one or more of data spatial interpolation, sampling, reclassification, data format conversion, and regional statistics; the spatial interpolation of the data adopts the inverse distance weighted average method; the sampling is performed through the Fishnet function in the GIS software. , generate uniform sampling points in space at a resolution of 1km, and obtain various data at the location of the sampling points; reclassification adopts the natural breakpoint method; data format conversion mainly refers to the mutual conversion of vector and raster data according to research needs; Regional statistics use the zonal statistics function in GIS software to count the mean, maximum and minimum values of various influencing factors in each region;

Step 3) Construct a flood event classification method that couples over-quantitative threshold sampling and flow sequence moving variance, and divide flood events from daily runoff data: 3-1. Use over-quantitative threshold sampling method to determine flood peaks; 3-2 Move based on flow sequence The variance determines the rising point and the retreat point of the flood process, and the interval between the rising point and the retreat point is a flood process;

Step 4) Build a flood behavior characteristic index system to describe the complete flood process: 4-1. Construct a flood behavior characteristic index system that can completely describe the flood process. The flood behavior characteristic index includes flood magnitude, time domain, rate of change and flood process. 4-2. Flood behavior characteristic indexes of flood events obtained from the division of calculation step 3);

Step 5) Use the clustering method to cluster the flood events obtained from the division, and identify the main types of floods in the watershed and their flood characteristics: use the k-means clustering algorithm to cluster the flood events in the divided study area to obtain the number of watershed categories, Identify the main types of floods in the basin and the characteristics of the flood process;

Step 6) Statistically analyze the main flood types of each hydrological station in the study area, and identify the regional characteristics of the flood process in the watershed: calculate the proportion of different flood types of each hydrological station, and use the type with the largest proportion to characterize the flood type of the site. Divide into the same area to identify the regional characteristics of the flood process in the watershed;

Step 7) Analyze the influence of the spatial heterogeneity of influencing factors on the flood process of the river basin based on the geographic detector: use the factor detection module in the geographic detector to detect the regional distribution characteristics of the flood process in the river basin, and analyze the influence of the spatial heterogeneity of the influencing factors on the flood process of the river basin. Influence of flood zone characteristics.

Further, in step 3), the judgment formula of the independent flood peak of the ultra-quantitative threshold sampling method described in 3-1 is:

In the formula, θ is the time interval between two adjacent flood peaks, the unit is day; A is the basin area, the unit is km ² ; Q _min is the minimum flow between the two adjacent flood peaks Q ₁ and Q ₂ , the unit is m ³ / s.

Further, the adaptive threshold of the ultra-quantitative threshold sampling method described in step 3) is determined by the following six steps:

Ⅰ. The median value of the annual maximum sequence is used as the initial threshold u ₀ to form a threshold space: u _i =u ₀ +(i-1)/5×σ _x , where i is the number of threshold samples, and σ _x is the daily average Standard deviation of flow; taking 5+ln(A) days as the sampling block time, and A as the area of the watershed (km ² ), i constituted a sequence of over-limit flood peaks. The number of over-threshold flood peaks should obey a Poisson distribution:

P(x=k)=e ^-λλk / ^k ! (2)

In the formula, k = 0, 1, 2, ..., λ is the annual average number of over-threshold values.

Ⅱ. The chi-square test is carried out on the number of samples of each over-limit flood peak sequence, and the significant level is 0.05.

Ⅲ. Draw the average exceeding function graph of each over-limit flood peak sequence, and judge the reasonable range of the threshold u. where the average exceeding function is a linear function of the threshold u, which can be expressed as:

where X is a random variable, u is a threshold, ξ is a morphological parameter, and σ is a scale parameter.

The empirical mean excess function of the sample _en (u) can be estimated by the following formula:

where n is the number of samples.

Ⅳ. Carry out the Anderson-Darling test for each over-limit flood peak sequence, and select the appropriate threshold u in combination with the above average over-function graph results. The statistic of the Anderson-Darling test for k samples can be expressed as:

的P_AD值，当P_AD>α，接受原假设，否则拒绝原假设。计算中以逐个超限洪峰序列的经验分布和理论累积分布作为两个独立样本计算统计量，进而计算P_AD值。In the formula, n _i is the number of samples; F(x) is the cumulative distribution function of the sample x, N= _∑ni is the number of all samples; H _N (x) is the distribution function of all N samples; B _N = {x∈R:H _N (x)<1}. Anderson-Darling statistics can be obtained by interpolation and extrapolation in actual calculations

The value of _PAD , when _PAD >α, accept the null hypothesis, otherwise reject the null hypothesis. In the calculation, the empirical distribution and the theoretical cumulative distribution of each over-limit flood peak sequence are used as two independent samples to calculate the statistics, and then the _PAD value is calculated.

V. For the selected threshold u and the over-limit flood peak sequence corresponding to the threshold, use the maximum likelihood method to estimate the parameters of the Generalized Pareto Distribution. The generalized Pareto distribution can be expressed as:

In the formula, ξ is the morphological parameter, σ is the scale parameter, u is the position parameter, and u is the threshold in this method. When the morphological parameter ξ is zero, the generalized Pareto distribution corresponds to the exponential distribution; when the morphological parameter ξ is less than zero, it is a constant Pareto distribution; when the morphological parameter ξ is greater than zero, it is a Pareto-II distribution.

Ⅵ. According to the research needs, select the flow value corresponding to the suitable return period as the threshold value of the sample selection method for over-quantitative threshold value.

Further, in step 3) described in 3-2 based on the movement variance of the flow sequence to determine the rising point and the receding point of the flood process, comprising the following three steps: A, select the appropriate moving window days, calculate the movement of the flow sequence Variance Var; B. Determine the threshold value TH _var of the moving variance of the flow sequence; C. Compare the moving variance and the threshold value of the flow sequence to determine the rising point and the ebb point of the flood.

Further, according to 3-2 in step 3), the starting point and the receding point of the flood process are determined based on the moving variance of the flow sequence, and the calculation formulas of the moving variance of the flow sequence and its threshold are respectively:

Var(i)=Var(Q _i ,Q _i+1 ,...,Q _i+n ) (7)

是该流量序列移动方差的均值，σ(Var)是该流量序列移动方差的方差，θ是一系数，用于控制该算法识别洪水过程的敏感度，θ越小，能够识别出的洪水过程越多。where Q is the flow sequence value, n is the number of days in the moving window,

is the mean value of the moving variance of the flow sequence, σ(Var) is the variance of the moving variance of the flow sequence, and θ is a coefficient used to control the sensitivity of the algorithm to identify the flood process. The smaller the θ, the more flood process can be identified. many.

Further, the characteristic indicators of flood behavior described in 4-1 in step 4) include coefficient of variation, skewness coefficient, kurtosis, proportion of flood time, high pulse time proportion, standardized flood peak, flood rate, and flood rate. , flood peak, flood total, peak time, flood occurrence time, duration, flood peak number and flood peak modulus.

Further, in step 4) in 4-2, the flood behavior characteristic index of the flood event divided by the calculation step 3 can be expressed as:

X=[x ₁ ,x ₂ ,...,x _n ] (9)

In the formula, X is the behavior characteristic index matrix of a flood event; x is a flood behavior characteristic index; n is the number of indexes.

Further, the distance adopted by the k-means clustering algorithm described in step 5) is the Euclidean distance, and its calculation formula is:

In the formula, d is the Euclidean distance between two points X and Y in the n-dimensional Euclidean space, the X coordinate is (x ₁ , x ₂ ,...,x _n ), and the Y coordinate is (y ₁ ,y ₂ ,...,y _n ).

Further, in step 5), the criterion for judging that the clustering reaches the optimal result is that the Davies-Bouldin Index (DBI) reaches the minimum, and the calculation formula of the index is:

In the formula, n is the number of categories divided by the clustering algorithm, c _i is the cluster center of the i-th category, σ _i is the distance of all flood events in the i-th category from its cluster center, and d is the i-th category of cluster centers and the distance between the cluster center of the jth class, c _j is the cluster center of the _jth class, and σj is the distance of all flood events in the jth class from its cluster center.

Further, the input of the k-means clustering algorithm described in step 5) is the flood behavior characteristic index of all site flood events in the watershed, which can be expressed as:

For a watershed with m observation stations, the matrix W formed by the flood behavior characteristic indicators of flood events of all stations in the watershed can be expressed as:

W=[Y ₁ ,Y ₂ ,...,Y _m ] (12)

In the formula, Y _m is the matrix composed of the flood behavior characteristic indexes of all flood events at the m-th station in the watershed, and the subscripts 1, 2,...,m represent the first to m-th stations.

For a site Y where k flood events have occurred, the matrix composed of all the flood event behavioral characteristic indicators can be expressed as:

Y=[X ₁ , X ₂ ,...,X _k ] (13)

In the formula, X is the behavior characteristic index matrix of a flood event in the above formula (9). And formula (13) can be further expressed as:

According to formula (14), then W can be further expressed as:

In the formula, the first subscript 1,2,…,m represents the 1st to mth stations; the second subscript 1,2,…,k represents the 1st to kth flood events; the third The subscripts 1,2,…,n represent the 1st to nth flood behavior characteristic indicators.

Further, in the factor detection and interaction detection module of the geographic detector described in step 7), the degree of interpretation of the flood area characteristics by the influencing factors can be expressed as:

和σ²分别为洪水区域特征Y在第h区域和全区的方差。In the formula, h=1,...,L, is the partition of the influence factor X; N _h and N are the number of units in the h-th area and the whole area, respectively;

and σ ² are the variances of the flood area characteristic Y in the h-th area and the whole area, respectively.

Beneficial effects of the present invention:

The method for identifying the type of flood in the river basin and analyzing its impact factor of the present invention can quickly and effectively divide the flood event from the daily runoff data of the hydrological station by constructing a flood event division method that couples ultra-quantitative threshold sampling and flow sequence moving variance. Identify the main flood types and their representative characteristics in the basin, and analyze the influence of spatial heterogeneity of influencing factors on the flood process. The method of the invention can accurately divide flood events from long-sequence runoff data, identify main flood types and main influencing factors in the river basin, and provide support for information mining of flood process characteristics in the river basin, utilization of rainwater resources and scientific management of water resources.

The present invention will be further described in detail below with reference to the accompanying drawings and specific embodiments.

Description of drawings

Fig. 1 is a flow chart of basin flood type identification and its impact factor analysis method of the present invention;

Fig. 2 is a schematic flow chart of the flood event classification method coupled with ultra-quantitative threshold sampling and flow sequence moving variance;

Figure 3 is a schematic diagram of the average exceeding function of the out-of-limit sample;

Figure 4 is a schematic diagram of the method for determining the rising point and the retreat point of the flood process based on the moving variance of the flow sequence;

Figure 5 is a schematic diagram of the behavioral characteristic indicators of flood events;

Figure 6 Schematic diagram of the identification process of flood types and their regional distribution characteristics in the basin.

Detailed ways

Example 1

The present invention proposes a method for identifying the type of flood in a watershed and analyzing its impact factors. The following will take a certain watershed in my country as a case area to further illustrate the specific application of the technical solution of the present invention, and the specific application includes the following steps:

Step 1) Collect meteorological and geographical factors, human activity factors, socioeconomic indicators and runoff data of hydrological stations in the study area:

The daily runoff data of the hydrological stations in the study area are from the "Hydrological Yearbook of the People's Republic of China", and these stations are evenly distributed in the whole basin; the daily precipitation data of the meteorological stations are from the National Meteorological Science Data Center (http://data.cma.cn). ); DEM (90m) and NDVI (1km) were obtained from the Resource and Environmental Science and Data Center, Chinese Academy of Sciences (https://www.resdc.cn/); socioeconomic data including urbanization rate, population density, GDP and carbon emissions , extracted from China Urban Statistical Yearbook and China Carbon Accounting Database (CEADs) respectively; the impact factor data are shown in Table 1:

Table 1 Impact factor data situation

serial number Impact factor short name unit type of data 1 Annual cumulative precipitation P_total mm site 2 Annual maximum daily precipitation P_max mm site 3 Annual maximum 3-day precipitation P_max3 mm site 4 slope Slope — grid 5 Elevation Elev m grid 6 NDVI NDVI — grid 7 urbanization rate CR % grid 8 Population density HD People/km² grid 9 GDP GDP Billion yuan grid 10 intensity of human activity IHA Million tons grid

Step 2) One or more items of spatial interpolation, sampling, reclassification, data format conversion and regional statistics are performed on the above-mentioned precipitation, DEM, NDVI and socioeconomic data.

Step 3) Divide flood events and write corresponding programs using matlab software. The specific algorithm is as follows:

3-1. Determine the flood peak of the flood event according to the over-quantitative threshold sampling method (Fig. 2):

3-1-1 The following steps are used to determine the flood threshold. Taking a hydrological station in the downstream of the basin as an example, the median value of the annual maximum sequence of 32800m ³ /s is used as the initial threshold u ₀ , and the threshold space is 32800～51556m ³ / s. Taking 24 days as the sampling block time, 15 over-limit flood peak sequences are formed. See Table 2 for the thresholds and the number of overruns for each sequence:

Table 2 Threshold and parameter estimation results of generalized Pareto distribution

3-1-2 Perform chi-square test on the number of samples of each over-limit flood peak sequence, and the significance level is 0.05. The results show that 13 of the 15 over-limit flood peak sequences conform to the Poisson distribution (see the column H _{0 in Table 1, where H 0 is} 0 to accept the null hypothesis, that is, the samples obey the Poisson distribution; H ₀ A value of 1 indicates rejection of the null hypothesis).

3-1-3 Draw the average exceeding function graph of each over-limit flood peak sequence, and judge the reasonable threshold u range. According to Fig. 3, it can be found that the average over-function graph shows that the linear trend appears near 43500m ³ /s. Considering the over-limit data volume and the convergence of the over-limit distribution, 43500m ³ /s is chosen as the reference for determining the threshold.

3-1-4 The Anderson-Darling test is carried out on each over-limit flood peak sequence, and the results are shown in Table 2. Combined with the results of the above-mentioned average exceeding function graph, the threshold is determined to be 43518m ³ /s.

3-1-5 For the selected threshold of 43518m ³ /s and the over-limit flood peak sequence corresponding to the threshold, the maximum likelihood method is used to estimate the parameters of the Generalized Pareto Distribution. The results are shown in Table 1. shown;

3-1-6 According to the research needs, select the flow value corresponding to the recurrence period of 5 years as the threshold value of the over-quantitative threshold sampling method.

3-1-7 Use the findpeaks function to calculate the maximum value in the runoff sequence, and write the corresponding judgment sentence according to the above formula (1) and the threshold value, and filter the flood peak T _pk of the flood event that meets the conditions.

3-2. Determine the ups and downs of the flood process based on the moving variance of the flow sequence. Figure 4 shows the specific implementation process of determining the rising and retreating points of the flood process based on the moving variance of the flow sequence. The moving variance and threshold of the flow sequence are calculated according to the above formulas (7) and (8). The moving variance and threshold of , determine the ups and downs of the flood process. Between the rising point and the receding point is a flood process. In this division, the hydrological station divided 30 flood events.

Step 4) Calculate the flood behavior characteristic index:

4-1. Screen the indicators used to describe the characteristics of flood behavior. The indicators of flood behavior characteristics include indicators related to flood magnitude, time domain, rate of change, and morphological characteristics of flood hydrographs. The flood behavior characteristic indicators used in this case area There are 15 in total, as shown in Table 3:

Table 3 Flood behavior characteristic indicators

σ、μ₃、μ₄分别是洪水事件的平均流量(m³/s)、方差、三阶中心矩和四阶中心矩；t_0.75pk是洪水过程中流量大于0.75倍洪峰的时间长。Note: Q _t is the flow rate (m ³ /s) in the t-th period; A is the area of the watershed (km ² ); t _start and t _end are the rising and retreating points of flood events,

σ, μ ₃ , μ ₄ are the mean flow (m ³ /s), variance, third-order central moment and fourth-order central moment of the flood event, respectively; t _0.75pk is the time when the flow is greater than 0.75 times the flood peak during the flooding process.

4-2. The flood behavior characteristic index of the flood event divided according to the calculation step 3): the process of calculating the flood behavior characteristic index according to the flood event flow hydrograph is shown in FIG. 5 .

Step 5) Use the clustering method to cluster the flood events obtained from the division, and identify the main types of floods in the basin and their flood characteristics: the process of identifying the types of floods in the basin is shown in the upper part of Figure 6, and the k-means clustering algorithm is used to divide the The obtained flood events in the study area are clustered, the number of watershed categories is obtained, and the main types of watershed floods and flood process characteristics are identified:

The matrix composed of the behavior characteristics of flood events is input into matlab, and the kmeans function is used to cluster the flood events of the site. The DBI index is selected as the criterion for judging the clustering effect, and the calculation formula of the DBI index is shown in formula (10). The smaller the DBI index value, the greater the gap between categories, and the better the clustering effect. In this case, site flooding events are classified into 3 categories.

Step 6) Statistically analyze the main flood types of each hydrological station in the study area, and identify the regional characteristics of the flood process in the basin: The process of identifying the regional distribution characteristics of the flood process in the basin is shown in the lower part of Figure 6. By calculating the proportion of different flood types in each hydrological station , the flood type of the site is characterized by the type with the largest proportion, and the flood type of the site is divided into the same area to identify the regional distribution characteristics of the flood process in the basin.

Step 7) Use the geographic detector method to analyze the influence of spatial heterogeneity of influencing factors on the regional distribution characteristics of flood types in the basin:

The natural breakpoint method is used to classify the influencing factors. Meteorological factors such as annual cumulative precipitation, annual maximum daily precipitation, and annual maximum 3-day precipitation are divided into 10 grades, and geographical factors such as slope and elevation are divided into 12 grades. NDVI The factors are divided into 6 levels, and the urbanization rate, population density, GDP and human activity intensity factors are divided into 8 levels.

Use the factor detection module in the geographic detector to detect the influencing factors of the flood area characteristics of the river basin, and determine the influence degree of each influencing factor on the flood area characteristics. The results are shown in Table 4:

Table 4 Geodetector results

The above geographic detector results show that the influence and interaction of meteorological conditions, geographical factors, land use and human activities in the basin are important reasons for the characteristics of flood areas. For a single factor, meteorological conditions (q=0.40-0.57), geographical factors (q=0.08-0.54) and human activities (q=0.24-0.48) were the main drivers of flood area characteristics, while vegetation cover (q=0.08-0.48) 0.15) has less effect. For the multi-factor interaction, the interaction of P _max3 and Elev had the greatest impact on the characteristics of flood areas in the basin (q=0.79), followed by the interaction of P _total and P _max (q=0.72).

The above description is only for the implementation of an example of the present invention, and is not intended to limit the present invention. The data conversion and the selection of the clustering algorithm in the present invention can be set according to requirements and specific research areas. Any modification, equivalent replacement, improvement, etc. made within the scope of the claims of the present invention shall fall within the protection scope of the present invention.

Claims (5)

1. A method for identifying the flood type of a drainage basin and analyzing the influence factors of the flood type is characterized in that: the method comprises the following steps:

step 1) collecting continuous daily-scale radial flow data of hydrological stations in a downloading research area; collecting and downloading meteorological geographic factors, human activity factors and social and economic factors which influence the flood process of the drainage basin;

step 2), data preprocessing, including one or more items of data spatial interpolation, sampling, reclassification, data format conversion and region statistics;

step 3) constructing a flood event partitioning method coupling the over-quantitative threshold sampling and the flow sequence moving variance, and partitioning flood events from daily runoff data: 3-1, judging the flood peak by using a super-quantitative threshold sampling method; the threshold value of the super quantitative threshold value sampling method is determined by the following six steps: i, taking the median of the annual maximum sequence as an initial threshold value u ₀ And forming a threshold space: u. of _i ＝u ₀ +(i-1)/5×σ _x Where i is the number of threshold samples, σ _x Standard deviation of daily average flow; taking 5+ ln (A) days as sampling block time, A is drainage basin area and unit is km ² I overrun peak sequences are formed; II, performing chi-square test on the sample number of each overrun flood peak sequence, wherein the significance level is 0.05; III, drawing an average over function graph of each over-limit peak sequence, and judging the range of the threshold value u; IV, performing Anderson-Darling test on the i transfinite flood peak sequences, and selecting a threshold u by combining the average transfinite function graph result; v, performing parameter estimation of generalized pareto distribution on the selected threshold u and the overrun flood peak sequence corresponding to the threshold by using a maximum likelihood method; VI, according to the research requirements,selecting a flow value corresponding to the suitable recurrence period as a threshold value of the over-quantitative threshold value sampling method; 3-2, determining a rising point and a water withdrawal point of a flood process based on the flow sequence moving variance, wherein a flood process is performed between the rising point and the water withdrawal point, and the rising point and the water withdrawal point of the flood process are determined based on the flow sequence moving variance, and the method comprises the following three steps: A. selecting the number of days of a moving window, and calculating the moving variance Var of the flow sequence; B. determining a threshold TH for a variance of a flow sequence movement _var (ii) a C. Comparing the moving variance of the flow sequence with a threshold value, and determining a rising point and a falling point of the flood; the calculation formula of the threshold value of the flow sequence moving variance is as follows:

in the formula (I), the compound is shown in the specification,

is the mean value of the flow sequence moving variance, sigma (Var) is the variance of the flow sequence moving variance, theta is a coefficient for controlling the sensitivity of identifying the flood process, and the smaller theta is, the more flood processes can be identified;

step 4), constructing a flood behavior characteristic index system to describe a complete flood process: 4-1, constructing a flood behavior characteristic index system capable of completely depicting a flood process, wherein the flood behavior characteristic indexes comprise relevant indexes of flood magnitude, time domain, change rate and flood process line state characteristics; 4-2, calculating flood behavior characteristic indexes of the flood events obtained by dividing in the step 3);

step 5) clustering the flood events obtained by division by using a clustering method, and identifying the main types and the flood characteristics of the flood in the drainage basin: clustering flood events of the divided research areas by using a k-means clustering algorithm to obtain the number of types of the drainage basin, and identifying the main types and the process characteristics of the flood of the drainage basin;

step 6), carrying out statistical analysis on main flood types of all hydrological stations in the research area, and identifying regional characteristics of the flood process of the drainage basin: calculating the occupation ratios of different flood types of each hydrological station, representing the flood type of the station by the type with the largest occupation ratio, dividing the station with the same flood type into the same area, and identifying the area characteristics of the watershed flood process;

step 7) analyzing the influence of the spatial heterogeneity of the influence factors on the flood process of the drainage area based on the geographic detector: and detecting the influence factors of the regional distribution characteristics of the watershed flood process by using a factor detection module in the geographic detector, and analyzing the influence of the spatial heterogeneity of the influence factors on the regional characteristics of the flood.

2. The method of claim 1, wherein the method comprises the steps of: the independent flood peak in the over-quantitative threshold sampling method in the step 3) should meet the following conditions:

in the formula, theta is the time interval between two adjacent flood peaks, and the unit is day; a is the area of the drainage basin and the unit is km ² ；Q _min For two adjacent peaks Q ₁ And Q ₂ Minimum flow rate in m ³ /s。

3. The method of claim 1, wherein the method comprises the steps of: the flood behavior characteristic indexes in the step 4) comprise relevant indexes of flood magnitude, time domain, change rate and flood process line state characteristics, and specifically comprise variation coefficients, state bias coefficients, kurtosis, flood rise time ratio, high pulse time ratio, standardized flood peak, flood rise rate, flood fall rate, flood peak, total flood amount, peak occurrence time, flood occurrence time, duration, flood peak number and flood peak modulus.

4. The method of claim 1, wherein the method comprises the steps of: the distance adopted by the k-means clustering algorithm in the step 5) is an Euclidean distance, the standard for judging that the clustering reaches the optimal result is that the Daisenberg Digital index DBI reaches the minimum, and the calculation formula of the index is as follows:

wherein n is the number of categories divided by the clustering algorithm, c _i Is the cluster center of the ith class, σ _i Is the distance of all flood events in the ith class from its cluster center, d is the distance between the class center of the ith class and the class center of the jth class, c _j Is the cluster center of class j, σ _j Is the distance of all flood events in class j from its cluster center.

5. The method of claim 1, wherein the method comprises the steps of: and 7) detecting the interpretation degree of the influence factors on the characteristics of the flood area by adopting a factor detection module and an interaction detection module in the geographic detector.

Images