Flash flood susceptibility mapping using hydrological modeling and logistic regression

Abstract

The application discloses a mountain torrent susceptibility evaluation method coupled with a hydrological model and a logistic regression, constructs a hydrological model; calculates mountain torrent disaster flow of each first-order sub-basin, sets different rainfall scenarios, simulates flood processes of each first-order sub-basin under different rainfall scenarios by using the hydrological model; selects a rainfall field that can best reflect general laws of mountain torrent disasters of the basin, and collects mountain torrent disaster flow data of each first-order sub-basin under the rainfall field for mountain torrent susceptibility evaluation model construction; extracts features of each first-order sub-basin as evaluation factors, performs initial screening on mountain torrent susceptibility evaluation factors based on a Pearson correlation coefficient method, and completes screening on mountain torrent susceptibility evaluation parameters based on LASSO regression on the basis of the initial screening result; and performs mountain torrent susceptibility evaluation of the sub-basin based on the logistic regression. Compared with the prior art, the application can conveniently and efficiently perform mountain torrent susceptibility judgment of the basin.

Description

Technical Field

[0001] The present invention relates to the field of flash flood susceptibility evaluation and risk management, and in particular to a flash flood susceptibility evaluation method coupling a HEC-HMS model with logistic regression. Background Art

[0002] Flash flood susceptibility describes the likelihood of flash floods occurring in a given region or river basin. It refers to the likelihood of floods threatening human life, property, and the production and living environment. Rainfall is the primary external factor influencing flash floods, and its impact on flash floods reflects the conditional probability of flash floods occurring under the influence of inducing factors. The primary internal factors influencing the occurrence and development of flash floods are geological conditions, including topography, water network structure, and hydrogeology, reflecting the potential for flash floods in a given region.

[0003] According to the principles of flood formation, when rainfall falls on the underlying surface, runoff from the same watershed converges through slopes and river channels, converging at the basin's outlet. A watershed is the site where flash floods breed and erupt, serving as the fundamental unit for flash flood risk management and reflecting the evolution of natural characteristics. Therefore, clarifying the relationship between a watershed's topography and flash flood susceptibility is essential for improving flash flood prevention and control capabilities. Summary of the Invention

[0004] In order to solve the problem of flash flood susceptibility evaluation, the present invention aims to propose a flash flood susceptibility evaluation method that couples hydrological models with logistic regression. The sub-basin is used as the evaluation unit and the basin topography and geomorphology parameters are used as evaluation factors. A flash flood susceptibility evaluation model is established by coupling the hydrological model with the logistic regression method.

[0005] The present invention provides a flash flood susceptibility assessment method that couples a hydrological model with logistic regression, comprising:

[0006] S1, conduct hydrological analysis on the watershed and divide it into sub-basins, and build a hydrological model based on the sub-basin division results;

[0007] S2: Select the flash flood warning section of the first-level sub-basin, use the flat water level of the flash flood warning section as the flash flood warning water level, collect the flash flood warning water levels, and calculate the flash flood disaster flow of each first-level sub-basin;

[0008] S3, setting different rainfall scenarios, and using the hydrological model constructed in step S1 to simulate the flood process of each first-level sub-basin under different rainfall scenarios;

[0009] S4. Analyze the flash flood disasters in each sub-basin under different rainfall scenarios based on the flood peak discharge obtained from the flood process simulation results. Select the rainfall scenario where the peak discharge exceeds the flash flood warning discharge of each first-level sub-basin as the rainfall event that best reflects the general pattern of flash flood disasters in the basin, and construct a flash flood susceptibility assessment model for this rainfall event.

[0010] S5: Extract the characteristics of each first-level sub-watershed as evaluation factors, conduct preliminary screening of flash flood susceptibility evaluation factors based on the Pearson correlation coefficient method, and complete the screening of flash flood susceptibility evaluation parameters based on the initial screening results using LASSO regression;

[0011] S6. Construct a flash flood susceptibility evaluation model based on logistic regression, and perform flash flood susceptibility evaluation of the sub-basin according to calculation results of the flash flood susceptibility evaluation model.

[0012] In some embodiments, S1 further includes calibrating and validating parameters of a semi-distributed hydrological model based on relative error of peak flow, relative error of runoff depth, peak-to-present time difference, and certainty coefficient.

[0013] In some embodiments, the expression of the flash flood discharge Q in the first-level sub-basin is as follows:

[0014]

[0015] Among them, Q is the flash flood disaster discharge of the first-level sub-basin, n is the roughness, A is the cross-sectional area corresponding to the flash flood disaster water level, R is the hydraulic radius, and J is the river gradient.

[0016] In some embodiments, the first-level sub-watershed characteristics in S5 include at least watershed scale, topography, shape, and water network structure.

[0017] In some embodiments, S5 further includes: performing preliminary screening of flash flood susceptibility evaluation factors based on the Pearson correlation coefficient method, selecting two indicator parameters with a correlation coefficient greater than or equal to 0.8 between the basin characteristic parameters, calculating the sum of the absolute values ​​of the correlation coefficients of the two indicators with all other indicators except each other, and then deleting the parameters with large corresponding result values.

[0018] In some embodiments, the correlation between two variables X and Y is defined as follows:

[0019]

[0020] Where ρ is the correlation coefficient, cov(X,Y) is the covariance between the pairwise calculation results X and Y of all characteristic parameters in the flash flood susceptibility assessment model, E(X) and E(Y) are the overall means of the variables, and σ X is the standard deviation of X, σ Yis the standard deviation of Y, X i and Y i is the value of the i-th individual in the sample, and I represents the number of individuals in the sample.

[0021] In some embodiments, the screening of flash flood susceptibility evaluation parameters based on LASSO regression in S5 further includes: selecting relatively independent evaluation factors that have an impact on flash flood evolution for flash flood susceptibility evaluation model construction, and estimating the regression coefficient of the explanatory variable. The expression is as follows:

[0022]

[0023] Among them, X is the explanatory variable matrix composed of the factors affecting the groundwater level, Y is the groundwater level prediction target variable, β j is the regression coefficient of the jth groundwater level influencing factor characteristic, λ is the penalty factor, p is the total number of explanatory variables, The minimum function value of the calculated parameter.

[0024] The flash flood susceptibility evaluation model in S6 is expressed as follows:

[0025] g(x)=b0+b1x1+b2x2+…+b n x n

[0026]

[0027] In landslide hazard assessment, g(x) is the occurrence of flash floods under the influence of flash flood disaster factors, which is represented by 0 or 1, 0 means flash floods do not occur, 1 means flash floods occur, x1, x2,…, x n is the influencing factor of flash flood susceptibility, b0 is the constant term of the flash flood susceptibility evaluation model, b1, b2, ... b n is the logistic regression coefficient corresponding to the influencing factors of flash flood susceptibility.

[0028] In some embodiments, when multiple parameters affecting flash floods are combined and calculated to obtain g(x), an evaluation of changes in the likelihood of flash flood disasters is performed.

[0029] In some embodiments, the probability value of flash flood disaster occurrence is between 0 and 1, which is used to divide the flash flood susceptibility level into five levels: very low, low, medium, high, and very high.

[0030] Compared with the prior art, the beneficial technical effects and technical progress achieved by the present invention are as follows:

[0031] 1) Using a hydrological model to simulate flood processes in each first-level sub-basin under different rainfall scenarios, obtain peak flood discharges, analyze the flash flood disasters in each sub-basin under different rainfall scenarios, select the rainfall event that best reflects the general pattern of flash flood disasters in the basin, and construct a flash flood susceptibility evaluation model under this rainfall event; on this basis, the flash flood susceptibility evaluation factors are initially screened using the Pearson correlation coefficient method, and the flash flood susceptibility evaluation parameters based on LASSO regression are used to complete the screening. The hydrological model and the flash flood susceptibility evaluation parameters based on LASSO regression interact and influence each other, and even work together to judge the flash flood susceptibility of the basin based on the landform characteristics and river system development characteristics, which is convenient and efficient;

[0032] 2) In the process of completing the screening of flash flood susceptibility evaluation parameters based on LASSO regression, the logistic regression method is further combined to construct a flash flood susceptibility evaluation model. By comprehensively evaluating the flash flood susceptibility level of the basin, the ability to prevent, control and control flash floods is improved, providing a quantitative basis for the construction of a flash flood defense system. BRIEF DESCRIPTION OF THE DRAWINGS

[0033] Figure 1 This is an overall flow chart of a flash flood susceptibility assessment method that couples a hydrological model with logistic regression according to the present invention;

[0034] Figure 2 This is an algorithm block diagram of an embodiment of a flash flood susceptibility assessment method that couples a hydrological model with logistic regression according to the present invention;

[0035] Figure 3 This is an example diagram of the sub-basin division of the study basin;

[0036] Figure 4 This is an example diagram of the rate-based flood process of the HEC-HMS hydrological model constructed for the study basin;

[0037] Figure 5 This is an example diagram of the flood process during the validation period of the HEC-HMS hydrological model constructed in the study basin;

[0038] Figure 6 It is to study the distribution map of the warning sections in each sub-basin of the basin;

[0039] Figure 7 This is an example diagram of the flood process of each first-level sub-basin simulated by the HEC-HMS hydrological model under different design rainfall scenarios;

[0040] Figure 8 This is an example of the calculation results of the correlation coefficient of the watershed characteristic parameters;

[0041] Figure 9 It is a graded distribution map of the susceptibility of mountain torrents in the study basin. DETAILED DESCRIPTION

[0042] The technical solution of the present invention is described below in conjunction with specific embodiments and drawings.

[0043] like Figure 1 As shown, the specific steps of the flash flood susceptibility assessment method of the present invention by coupling the hydrological model with logistic regression are as follows:

[0044] Step 1: Conduct hydrological analysis of the target basin based on the digital elevation model (DEM) and divide it into sub-basins. Combined with the sub-basin division results, a hydrological model, specifically the HEC-HMS model, is constructed to calibrate and validate the model parameters. This includes:

[0045] The HEC-HMS model is as follows:

[0046] According to the rainfall-runoff characteristics of the study basin, the SCS curve method, SCS unit line method, exponential decay method and Muskingum method were selected to calculate runoff generation, confluence, base flow and river channel evolution.

[0047] (1) SCS curve method

[0048]

[0049] In the above formula, R t is the accumulated net rainfall at time t, mm; P is the rainfall depth at time t, mm; I a is the initial rainfall loss, mm; S is the maximum soil water storage capacity, which is a measure of the ability of the sub-basin to absorb and intercept rainfall, mm.

[0050] (2) SCS unit line method

[0051]

[0052] In the above formula, Q p is the peak flow rate of the unit line, T p is the time of peak flow rate per unit line, C is the conversion constant, which is 2.08 in the international system of units; A is the basin area, km 2 .

[0053] (3) Exponential decay method

[0054] The exponential decay model assumes that the relationship between the base flow at a certain moment and the initial base flow is as follows:

[0055] Q bt =Q b0 D t

[0056] In the above formula, Q bt is the base flow at time t, m 3 / s;Qb0 is the base flow at the initial moment, m 3 / s; D is the attenuation index, which is the ratio of the base flow at time t to the base flow of the previous day, and its value range is 0 to 1.

[0057] (4) Muskingum Method

[0058] Q2=C1I2+C2I1+C3Q1

[0059]

[0060] C1+C2+C3=1

[0061] In the above formula, Δt is the time step, h; I1 and I2 are the starting and ending flow rates of the upstream section, m 3 / s; Q1 and Q2 are the starting and ending flow rates of the downstream section, m 3 / s; K is the slope of the channel storage curve, which is the time it takes for the flood wave to pass through the river section, h; x is the flow density factor, a dimensionless number.

[0062] Step 1-1: Based on the basin DEM, perform hydrological analysis on the target basin and divide it into sub-basins, build a semi-distributed hydrological model HEC-HMS, and use the model to simulate the formation process of flash floods. Figure 3 As shown in Figure 1, the first-level sub-basin was selected as the key research object to obtain the sub-basin division results of the study basin. The rainfall runoff data of the study area from 1981 to 2019 were collected, and 10 representative flood processes with complete data and induced flash flood disasters were selected. Six of them were used to calibrate the model parameters, and four were used to verify the forecast effect of the model. The specific flood processes used for calibration and verification are as follows: Figure 4 and Figure 5 As shown;

[0063] Step 1-2: Evaluate the accuracy of the HEC-HMS model constructed in step 1-1 based on the relative error of peak flow, relative error of runoff depth, peak-to-present time difference, and certainty coefficient as model accuracy evaluation indicators to achieve hydrological model parameter calibration and verification;

[0064] The calculation formulas for each indicator are as follows:

[0065] Peak flow relative error RE Q :

[0066]

[0067] Among them, Q O is the measured peak flow, Q S is the peak flow calculated by HEC-HMS model, and the relative error of peak flow RE Q The allowed margin of error is 20%;

[0068] Runoff depth relative error RE R :

[0069]

[0070] Among them, R O is the measured runoff depth, R S The runoff depth is calculated by the HEC-HMS model. When the runoff depth is greater than 20 mm, the allowable error is 20 mm; when it is less than 3 mm, the allowable error is 3 mm.

[0071] Peak time difference ΔT:

[0072] ΔT=T S -T O (3)

[0073] Among them, T O is the measured flood peak time, T S is the peak time obtained by model simulation; ΔT is 30% of the time interval between the predicted time and the measured peak time as the allowable error. When the allowable error is less than 3 hours or the length of one period calculated by the model, 3 hours or one calculation period is used as the allowable error.

[0074] Determinism coefficient DC:

[0075]

[0076] Among them, n is the length of the data time series, i is the time series data index, Q c (i) is the flow rate of the i-th sample calculated by the model, Q o (i) is the measured flow rate of the i-th sample, is the mean of the measured flow rate.

[0077] Table 1 shows the calibration and validation results of the HEC-HMS model (basin-wide flood simulation system). The model performs well in flash flood simulation. The coefficients of certainty for flood events during the calibration period were consistently above 0.7, with an average of 0.803; the coefficients of certainty for the validation period were consistently above 0.8, with an average of 0.889. The peak and runoff depth errors for both the calibration and validation periods were within 20%, with peak-to-present time differences of less than 2 hours, demonstrating high simulation accuracy for flood events.

[0078] Table 1

[0079]

[0080] Step 2: Select the first-level sub-basin flash flood warning section and collect the flash flood warning water level, and use the Manning formula to calculate the flash flood disaster flow in the sub-basin; specifically include

[0081] The first-level sub-basin is used as the flash flood susceptibility evaluation unit, and the flash flood disaster flow of each first-level sub-basin is calculated; the warning section is generally selected as the section with the lowest elevation within 1km of the basin outlet. If a section is not the lowest elevation, but the residential areas along the section are clustered, it can also be considered to be set as a warning section. The location of the warning section in each sub-basin is as follows: Figure 6 shown.

[0082] The flat water level of the warning section is used as the flash flood warning water level, and the flash flood disaster flow of each first-level sub-basin is calculated using the Manning formula. The Manning formula is as follows:

[0083]

[0084] in, Q is the flash flood disaster discharge of the first-level sub-basin, n is the roughness, A is the cross-sectional area corresponding to the flash flood disaster water level, R is the hydraulic radius, and J is the river gradient.

[0085] According to the calculation results of Manning's formula, the disaster-causing flow of each sub-basin is shown in Table 2.

[0086] Table 2

[0087]

[0088] Step 3: Set different rainfall scenarios and use the hydrological model to simulate the flood process of each first-level sub-basin under different rainfall scenarios. The specific description is as follows:

[0089] According to the historical rainfall pattern of short duration and concentrated rainfall in the basin, the rainfall duration is set to 1h, 2h, and 3h, and the cumulative rainfall is set to 30mm, 40mm, and 50mm. The HEC-HMS model is used to simulate the flood process of each first-level sub-basin, and the flood process simulation results are obtained, as shown in the following figure: Figure 7 As shown;

[0090] Step 4: Analyze the peak flow rate based on the flood process simulation results. Whether the peak flow rate exceeds the flash flood warning flow rate of each first-level sub-basin is used as the judgment standard. Analyze the flash flood disaster situation of each sub-basin under different rainfall scenarios. Select the rainfall event that best reflects the general pattern of flash flood disasters in the basin. Collect flash flood disaster flow data for each first-level sub-basin in this rainfall event for the construction of the flash flood susceptibility assessment model. The specific description is as follows:

[0091] Using whether the peak flow in the flood simulation results exceeds the basin's flash flood warning flow as the criterion, we analyzed the flash flood disasters in each sub-basin under different rainfall scenarios. We selected the rainfall events that best reflect the general pattern of flash flood disasters in the basin, namely, the flash flood disaster flow data from rainfall events with a rainfall duration of 1 hour and a cumulative rainfall of 40 mm. Among the 22 target first-level sub-basins, 12 were disaster-prone sub-basins and 10 were non-disaster-prone sub-basins. These intermediate disaster samples were used to construct the flash flood susceptibility assessment model in step 6. Step 5 involves model parameter screening, and step 6 is model construction.

[0092] Step 5: Taking the first-level sub-basin as the evaluation unit, extract the characteristics of each first-level sub-basin as the evaluation factor, and conduct a preliminary screening of the flash flood susceptibility evaluation factors based on the Pearson correlation coefficient method. Based on the preliminary screening results, the flash flood susceptibility evaluation parameters are screened based on the LASSO regression. The process of this step is the parameter screening operation of the flash flood susceptibility evaluation model. The specific parameters are shown in Table 3.

[0093] Table 3

[0094]

[0095]

[0096]

[0097] Step 5-1: Preliminary screening of flash flood susceptibility evaluation factors is performed based on the Pearson correlation coefficient method between watershed characteristic parameters. By filtering out parameters with high correlation, the multicollinearity problem is preliminarily solved. The Pearson correlation coefficient is used to measure the correlation between two variables X and Y, and is defined as shown in formula (6):

[0098]

[0099] In the above formula, ρ is the correlation coefficient, cov(X,Y) is the covariance between the pairwise calculation results X and Y of all characteristic parameters in the flash flood susceptibility assessment model, E(X) and E(Y) are the overall means of any two basin characteristic parameter variables X and Y, respectively, and σ X is the standard deviation of X, σ Y is the standard deviation of Y, X i and Y i is the value of the i-th individual in the sample, I represents the number of individuals in the sample; the pairwise calculation results X and Y between all characteristic parameters in the flash flood susceptibility assessment model are as follows: Figure 8 shown.

[0100] After completing the calculation of the correlation coefficients between parameters, two indicator parameters with correlation coefficients greater than or equal to 0.8 were selected, and the sum of the absolute values ​​of the correlation coefficients of the two indicators with all other indicators except each other was calculated. The parameters with larger corresponding result values ​​were then deleted to complete the preliminary screening of parameters. Based on the preliminary screening principle of parameters and the results of correlation coefficient calculation, the parameters involved in the construction of flash flood susceptibility were preliminarily screened, and 15 parameters were obtained as independent variables to input into the Lasso regression model, namely: average elevation of the basin, undulation, Melton ratio, average slope of the basin, slope ratio, area elevation integral, basin morphological factor, roundness rate, river length, river frequency, geomorphic structure ratio, branching ratio, river channel maintenance constant, river fineness ratio, and main channel length.

[0101] Step 5-2: Select the flash flood susceptibility evaluation factors based on LASSO regression, select relatively independent evaluation factors that have an impact on flash flood evolution for the construction of flash flood susceptibility evaluation model, and estimate the regression coefficient of the explanatory variable. The expression is as follows:

[0102]

[0103] Among them, X is the explanatory variable matrix composed of the factors affecting the groundwater level, Y is the groundwater level prediction target variable, β j is the regression coefficient of the jth groundwater level influencing factor characteristic, λ is the penalty factor, p is the total number of explanatory variables, is the minimum function value of the calculated parameter;

[0104] The 15 characteristic parameters obtained through the initial correlation screening are used as independent variables, 40mm -1 The disaster data of each sub-basin under the h rainfall scenario is used as the dependent variable, and the model finally retains two evaluation factors, namely the landform structure ratio and the basin morphology factor;

[0105] Step 6: A flash flood susceptibility evaluation model is constructed based on logistic regression. The flash flood susceptibility of the first-level sub-basin is divided into five levels according to the model calculation results. The specific descriptions are as follows:

[0106] In step 6-1, the logistic regression model is a commonly used multivariate regression statistical method in natural disaster susceptibility assessment. It is a typical regression analysis method for binary dependent variables. Logistic regression is based on the theory of linear regression. By introducing the logistic function, the predicted value is constrained to the range of 0 to 1, making it suitable for binary classification problems. When the independent variable in the logistic regression is significant, the relationship between the occurrence of flash floods and the parameter is expressed as follows:

[0107] g(x)=b0+b1x1+b2x2+…+b n x n (8)

[0108]

[0109] In landslide hazard assessment, g(x) represents the occurrence of flash floods under the influence of flash flood disaster factors, and is represented by 0 or 1, 0 means flash floods do not occur, 1 means flash floods occur, x1, x2,…, x n is the influencing factor of flash flood susceptibility, b0 is the constant term of the flash flood susceptibility evaluation model, b1, b2, ... b n is the logistic regression coefficient corresponding to the influencing factor of flash flood susceptibility. The probability value of flash flood disaster occurrence is between 0 and 1. The probability of flash flood disaster not occurring is obtained as follows:

[0110]

[0111] The ratio of the probability of a flash flood disaster occurring to the probability of it not occurring, odds, is expressed as follows:

[0112] odds = probability of a flash flood disaster occurring / probability of a flash flood disaster not occurring = e g(x) (11)

[0113] When using the probability of no flash flood disaster to evaluate flash flood disaster, in order to facilitate the analysis of the degree of change in the probability of flash flood disaster, the logarithms of both sides are taken:

[0114] ln(odds)=g(x)=b0+b1x1+b2x2+…+b n x n (12)

[0115] After taking the logarithms of both sides, it is found that the degree of change in the probability of flash flood disasters is g(x). That is to say, when multiple parameters affecting flash floods are combined and calculated to obtain g(x), the change in the possibility of flash flood disasters can be reflected.

[0116] The parameters of Dali River sub-basin geomorphic structure ratio and basin morphological factor were compared with the parameters of Dali River sub-basin geomorphic structure ratio and basin morphological factor at 40mm. -1 Logistic regression analysis was performed on the corresponding sub-basin disaster data under the h rainfall scenario, and the parameter estimation results of the flash flood susceptibility assessment model were obtained as follows:

[0117] Table 4

[0118]

[0119] The flash flood susceptibility assessment model is constructed as follows:

[0120] f(x)=-11.824+1.937x1+2.115x2 (25)

[0121] In the above formula, x1 is the morphological factor of the flash flood susceptibility evaluation factor, and x2 is the landform structure ratio of the flash flood susceptibility evaluation factor.

[0122] In step 6-2, the probability P value of flash flood disasters is between 0 and 1, and the flash flood susceptibility level is divided into five levels: very low, low, medium, high and very high with a step size of 0.2; according to the flash flood susceptibility level classification standard, the flash flood susceptibility results of each sub-basin of the basin are studied, such as Figure 9 shown.

[0123] The above implementation steps are intended only to facilitate understanding of the specific methods and core concepts of the present invention and are not intended to limit the present invention. Any equivalent substitutions, combinations, and modifications made by researchers in this field based on the concepts of the present invention without departing from the principles of the present invention shall also be deemed to fall within the scope of protection of the present invention.

Claims

1. A flash flood susceptibility assessment method that couples a hydrological model with logistic regression, characterized in that: include: S1: Conduct hydrological analysis on the watershed and divide it into sub-basins. Combine the sub-basin division results to build a hydrological model. The hydrological model is as follows: According to the rainfall-runoff characteristics of the study basin, the SCS curve method, SCS unit line method, exponential decay method and Muskingum method were selected to calculate runoff generation, confluence, base flow and river channel evolution. S2: Select the flash flood warning section of the first-level sub-basin, use the flat water level of the flash flood warning section as the flash flood warning water level, collect the flash flood warning water levels, and calculate the flash flood disaster flow of each first-level sub-basin; S3, setting different rainfall scenarios, and using the semi-distributed hydrological model constructed in step S1 to simulate the flood process of each first-level sub-basin under different rainfall scenarios; S4. Analyze the flash flood disasters in each sub-basin under different rainfall scenarios based on the flood peak discharge obtained from the flood process simulation results. Select the rainfall scenario where the peak discharge exceeds the flash flood warning discharge of each first-level sub-basin as the rainfall event that best reflects the general pattern of flash flood disasters in the basin, and construct a flash flood susceptibility assessment model for this rainfall event. S5: Extract the characteristics of each first-level sub-watershed as evaluation factors, conduct preliminary screening of flash flood susceptibility evaluation factors based on the Pearson correlation coefficient method, and complete the screening of flash flood susceptibility evaluation parameters based on the initial screening results using LASSO regression; The characteristics of each first-level sub-basin include at least the basin scale, topography, shape and water network structure; The flash flood susceptibility evaluation factors were initially screened based on the Pearson correlation coefficient method. Two index parameters with a correlation coefficient greater than or equal to 0.8 between the basin characteristic parameters were selected. The sum of the absolute values ​​of the correlation coefficients of the two indicators with all other indicators except each other was calculated, and the parameters with the largest corresponding values ​​were deleted. The screening of flash flood susceptibility evaluation parameters based on LASSO regression further includes: selecting relatively independent evaluation factors that have an impact on flash flood evolution for flash flood susceptibility evaluation model construction, and estimating the regression coefficient of the explanatory variable. The expression is as follows: ; in, is the explanatory variable matrix composed of factors affecting groundwater level, is the groundwater level prediction target variable, For the The regression coefficient of the characteristics of the factors affecting groundwater level, is the penalty factor, is the total number of explanatory variables, is the minimum function value of the calculated parameter; S6. Construct a flash flood susceptibility evaluation model based on logistic regression, and use the calculation results of the flash flood susceptibility evaluation model to evaluate the flash flood susceptibility of the sub-basin; the flash flood susceptibility evaluation model is expressed as follows: ; ; ; In landslide hazard assessment, It is the occurrence of flash floods under the influence of flash flood disaster factors, represented by 0 or 1, 0 means flash floods do not occur, 1 means flash floods occur, is the influencing factor of flash flood susceptibility, is the constant term of the flash flood susceptibility assessment model, is the logistic regression coefficient corresponding to the influencing factors of flash flood susceptibility.

2. The flash flood susceptibility assessment method of coupling hydrological model and logistic regression according to claim 1 is characterized in that: S1 further includes calibration and verification of semi-distributed hydrological model parameters based on relative error of peak discharge, relative error of runoff depth, peak-to-present time difference and certainty coefficient.

3. The flash flood susceptibility assessment method of coupling hydrological model and logistic regression according to claim 1 is characterized in that: First-level sub-basin flash flood discharge The expression is as follows: ; in, The first-level sub-basin mountain torrent disaster flow, is the roughness, is the cross-sectional area corresponding to the flash flood disaster water level, is the hydraulic radius, It is the gradient of the river channel.

4. The flash flood susceptibility assessment method of coupling a hydrological model with logistic regression according to claim 1, characterized in that: in, Two variables and The correlation between them is defined as follows: ; in, is the correlation coefficient, The pairwise calculation results between all characteristic parameters in the flash flood susceptibility assessment model and The covariance between and are the overall means of the variables, for The standard deviation of for The standard deviation of and For the sample The value of each individual in the sample, is the number of individuals in the sample.

5. The flash flood susceptibility assessment method of coupling hydrological model and logistic regression according to claim 1 is characterized in that: Multiple parameters affecting flash floods are obtained through combined calculation When the disaster risk is high, the change of possibility of flash flood disaster should be evaluated.

6. The flash flood susceptibility assessment method of coupling hydrological model and logistic regression according to claim 1 is characterized in that: The probability value of flash flood disasters is between 0 and 1, which is used to divide the flash flood susceptibility level into five levels: very low, low, medium, high and very high.

Images