A method for analyzing the hydrological similarity of river basins based on a step-by-step clustering scheme

Abstract

The present invention provides a method for analyzing the hydrological similarity of river basins based on a step-by-step clustering scheme, including: Step 1, extracting rainfall characteristics, topographic and geomorphic characteristics, and soil geological characteristics as similarity indicators; Step 2, determining the optimal number of clusters and using K-means clustering to divide similar river basin groups; Step 3, constructing the Xin'anjiang model and calibrating the model parameters; Step 4, in the final similar river basin groups, calculating the overall hydrological similarity of the third-level similar river basin groups; Step 5, extracting the flow characteristics of each experimental river basin and calculating the hierarchical control intensity of geographical characteristics at different spatial levels on the flow characteristics. The method provided by the present invention can not only quantitatively represent the overall hydrological similarity under different geographical environments, but also more clearly reveal the control effect of geographical characteristics at different spatial levels on hydrological responses, and can provide support for hydrological forecasting and water resource management in mountainous areas without data.

Description

Technical Field

[0001] The present invention relates to the field of watershed hydrological forecasting, and particularly to a method for analyzing watershed hydrological similarity based on a stepwise clustering scheme. Background Art

[0002] A watershed is a complex system with a certain organization, where the watershed morphology, drainage network, vegetation, etc., which are closely related to hydrological responses, are adapted to climate, terrain, landform, etc. This self-organization enables the identification of hydrological similarities between watersheds through climate and landscape characteristics. Generally, "3S" technology can be used to extract geographical features closely related to hydrological responses as indicators, and the hydrological similarity can be indirectly measured based on the similarity degree of these indicators, combined with watershed classification to quickly identify similar watersheds. This method of aggregating similar watersheds does not rely on measured flow data and can exhibit good stability in most regions, so it is widely used in the research of parameter regionalization in data-sparse areas.

[0003] However, surface hydrological responses are very complex due to the spatio-temporal variation of rainfall patterns and the spatial variability of landscape structures. The differences in runoff generation mechanisms and their main runoff components result in different geographical features dominating hydrological responses, and a general hydrological similarity system is still unavailable. It is necessary to select appropriate similarity indicators according to local hydrological mechanisms and separately consider their impacts and importance on hydrological responses during the hydrological simulation process. Summary of the Invention

[0004] To solve the above technical problem that is difficult to be characterized by a general hydrological similarity system, the present invention provides a method for analyzing watershed hydrological similarity based on a stepwise clustering scheme.

[0005] The present invention adopts the following technical solutions:

[0006] A method for analyzing watershed hydrological similarity based on a stepwise clustering scheme includes the following steps:

[0007] Step 1: Extract the rainfall characteristics, topographic and geomorphic characteristics, and soil and geological characteristics of the experimental watersheds, and use these three types of geographical characteristics as similarity indicators for the clustering processes at the first, second, and third levels respectively;

[0008] Step 2: In the clustering process at each level, apply principal component analysis to reduce the dimensions of different similarity indicators, namely rainfall characteristics, topographic and geomorphic characteristics, and soil and geological characteristics. For the clustering at each level, use the elbow method to determine the optimal number of clusters, and use K-means clustering to divide the similar watershed groups;

[0009] Step 3: Combine the SCE-UA algorithm to construct an automatic parameter optimization program for the Xin'anjiang model, and calibrate the model parameters of each experimental watershed;

[0010] Step 4. Use the "leave-one-out" cross-validation of model parameters in each final similar river basin group, that is, the similar river basin group at the third level, to quantify the portability of parameters with the Nash efficiency coefficient, and calculate the overall hydrological similarity of the similar river basin group at the third level;

[0011] Step 5. Extract the flow characteristics of each experimental river basin, and statistically analyze the changes in the standard deviation of the flow characteristics during the stepwise clustering process in each similar river basin group, and calculate the hierarchical control intensity of geographical characteristics at different spatial levels on the flow characteristics.

[0012] Furthermore, in Step 1, rainfall characteristics are extracted according to rainfall station data; topographic and geomorphic characteristics are extracted according to GDEMV2 30M resolution digital elevation data and Landsat4-5 TM satellite digital products; soil and geological characteristics are extracted according to the "China Soil Dataset Based on the Harmonized World Soil Database (HWSD)" and the "1:1,500,000 Geological Map of the People's Republic of China".

[0013] Furthermore, the clustering process at each level in Step 2 is specifically as follows:

[0014] Step 2.1. In the clustering process at the first level, perform principal component analysis on the rainfall characteristics of the experimental river basins to generate "principal components" of fewer rainfall characteristics; obtain the optimal number of clusters according to the elbow method for the "principal components" of the rainfall characteristics; perform K-means clustering analysis on the "principal components" of the rainfall characteristics and the optimal number of clusters to obtain the similar river basin group CⅠ-f at the first level;

[0015] Step 2.2. In the clustering process at the second level, input the topographic and geomorphic characteristics of the experimental river basins for principal component analysis to obtain "principal components" of fewer topographic and geomorphic characteristics; in each similar river basin group CⅠ-f at the first level, obtain the optimal number of clusters according to the elbow method for the "principal components" of the topographic and geomorphic characteristics; perform K-means clustering analysis on the "principal components" of the topographic and geomorphic characteristics and the optimal number of clusters to obtain the similar river basin group CⅡ-f-s at the second level;

[0016] Step 2.3. In the clustering process at the third level, perform principal component analysis on the soil and geological characteristics of the experimental river basins to obtain "principal components" of fewer soil and geological characteristics; in each similar river basin group CⅡ-f-s at the second level, obtain the optimal number of clusters according to the elbow method for the "principal components" of the soil and geological characteristics; perform K-means clustering analysis on the "principal components" of the soil and geological characteristics and the optimal number of clusters to obtain the similar river basin group CⅢ-f-s-t at the third level.

[0017] Furthermore, in Step 3, calibrate the optimal model parameters of each experimental river basin by combining the SCE-UA algorithm, specifically as follows:

[0018] Step 3.1: Set the objective function F in the SCE-UA algorithm as a combination of the Nash efficiency coefficient NSE and the relative error of flood volume FE, and calibrate it in the direction of the objective function value continuously approaching 0;

[0019]

[0020]

[0021]

[0022] where R0 and R c are the measured runoff depth and the simulated runoff depth respectively, with the unit of mm, i is the number of days in a year; Q o,i and Q c,i are the measured flow rate and the simulated flow rate respectively, with the unit of m 3 / s, is the annual average value of the measured flow rate, with the unit of m 3 / s, ω1 and ω2 are weight coefficients, here ω1 = ω2 = 0.5, and y is the number of years in the calibration period;

[0023] Step 3.2: Set the parameters of the Xin'anjiang model to be calibrated and their value ranges, and use the SCE-UA algorithm to calibrate the model parameters of all experimental basins.

[0024] Furthermore, Step 4 for calculating the overall hydrological similarity of the similar basin group includes the following steps:

[0025] Step 4.1: Conduct "leave-one-out" cross-validation of the model parameters in the third-level similar basin group. In the CIII-f-s-t basin group, there are a total of N Ⅲ (f, s, t) = n basins. Take each basin in the similar basin group as the receptor basin in turn, that is, the basin without data, and the other n - 1 similar basins in the similar basin group as its donor basins. Transplant the hydrological model parameters of the donor basins to the receptor basin for flow simulation, and evaluate the parameter transplantability using the Nash efficiency coefficient to obtain a set E i (f, s, t) of n - 1 Nash efficiency coefficients. Calculate the hydrological similarity S i (f, s, t) between the receptor basin i and other similar basins; S i (f, s, t) value greater than 0.5 indicates good hydrological similarity, and the closer it is to 1, the higher the similarity;

[0026] E i (f, s, t) = {NSE(1), NSE(2), …, NSE(i - 1), NSE(i + 1), …, NSE(n)}

[0027] (where \(i = 1, 2, \ldots, n\)),

[0028] S i (f, s, t)=MEDIAN(E i (f, s, t))

[0029] In the formula, \(i\) is the \(i\)-th receptor basin in CⅢ-f-s-t;

[0030] Step 4.2: According to the set \(E(f, s, t)\) of Nash efficiency coefficients of all receptor basins in the CIII-f-s-t similar basin group, calculate the overall hydrological similarity \(S(f, s, t)\) of this similar basin group; the larger the value of \(S(f, s, t)\), the more similar hydrological processes can be generated under the corresponding geographical environment of the CIII-f-s-t basin group;

[0031] E(f, s, t)=\{E1(f, s, t), E2(f, s, t), \ldots, E n (f, s, t)\},

[0032] S(f, s, t)=MEDIAN(E(f, s, t)).

[0033] Furthermore, the specific steps of step 5 are as follows:

[0034] Step 5.1: According to the description of the flow characteristics, extract the flow characteristics from the hydrological station data in the "Hydrological Yearbook" to describe the annual, monthly, and daily scale flow states within each experimental basin for many years;

[0035] Step 5.2: Statistically analyze the changes in the standard deviation of the flow characteristics within the third-level similar basin group (CIII-f-s-t) during the stepwise clustering process, and calculate the hierarchical control intensity of geographical characteristics at different spatial levels on the flow characteristics;

[0036]

[0037] D Ⅰ (f)=MEDIAN(A Ⅰj (f)),

[0038]

[0039] D Ⅱ (f, s)=MEDIAN(A Ⅱj (f, s)),

[0040]

[0041] D Ⅲ (f, s, t)=MEDIAN(A Ⅲj (f, s, t)),

[0042] where j is the jth flow feature, and b j 0 is the standard deviation of the jth flow feature for all experimental basins, i.e., the original state, A Ⅰj (f) represents the change in the standard deviation of the jth flow feature relative to the original state in the similar basin group CⅠ-f at the first hierarchical clustering level; similarly, A Ⅱj (f, s) and A Ⅲj (f, s, t) represent the changes in the standard deviation of the jth flow feature relative to the original state in the similar basin groups CⅡ-f-s and CⅢ-f-s-t at the second and third hierarchical clustering levels, respectively; A Ⅰj (f), A Ⅱj (f, s) and A Ⅲj (f, s, t) The larger the value, the stronger the control effect of the geographical features at the corresponding spatial level on the specific hydrological response. A value greater than 0 indicates that this type of geographical feature plays a positive control role in the hydrological response, can effectively identify similar runoff processes, and cause the flow features to aggregate; D Ⅰ (f), D Ⅱ (f, s) and D Ⅲ (f, s, t) The larger the value, the stronger the control effect of the geographical features at the corresponding spatial level on the hydrological response represented by all flow features.

[0043] Advantages of the present invention: The present invention comprehensively considers the influence and relative importance of various geographical features on the hydrological response, and proposes a simple step-by-step clustering method to reveal the geographical distribution law of hydrological similarity. Moreover, since this method simulates the natural spatial levels of the basins and follows the movement path of rainfall, that is, rainfall, topography, and soil geological characteristics are sequentially incorporated as similarity indicators into the basin classification process, this clustering process enhances the interpretability of basin classification, quantitatively describes the control effect of geographical features at different spatial levels on the hydrological response, further reveals the role of different geographical features in indicating hydrological similarity, and provides support for more efficient hydrological prediction and water resource management in data-scarce basins. BRIEF DESCRIPTION OF THE DRAWINGS

[0044] Figure 1 are the 64 experimental basins of the present invention.

[0045] Figure 2 is the main flow chart of the present invention.

[0046] Figure 3 is the process of subdividing the basin group based on the step-by-step clustering scheme of the present invention.

[0047] Figure 4The overall hydrological similarity of the similar basin groups (CIII-f-s-t) at the third level of the present invention.

[0048] Figure 5 The change (relative to the initial state) of the standard deviation of each flow characteristic of the similar basin groups (CIII-f-s-t) at the third level of the present invention.

[0049] Figure 6 The median of the changes in the standard deviations of all flow characteristics (relative to the initial state) of the similar basin groups (CIII-f-s-t) at the third level of the present invention. Detailed implementation mode

[0050] The following further describes the present invention in conjunction with embodiments. The embodiments are only used to illustrate the present invention and do not constitute a limitation on the scope of the claims. Other alternative means that can be conceived by those skilled in the art are within the scope of the claims of the present invention.

[0051] Embodiment 1

[0052] As Figure 1 shown, 64 basins in the mountainous areas of southern China are selected as the implementation objects in this embodiment.

[0053] A method for analyzing the hydrological similarity of basins based on a stepwise clustering scheme, the flowchart is as Figure 2 shown, and includes the following steps:

[0054] Step 1: Extract 17 rainfall characteristics according to the rainfall station data; extract 13 topographic and geomorphic characteristics according to the GDEMV2 30M resolution digital elevation data and Landsat4-5 TM satellite digital products; extract 5 soil and geological characteristics according to the "China Soil Dataset Based on the Harmonized World Soil Database (HWSD)" and the "1:1,500,000 Geological Map of the People's Republic of China", and use these three types of geographical characteristics as the similarity indicators for the clustering processes at the first, second, and third levels respectively;

[0055] Step 2: In the clustering process of each level, apply principal component analysis to reduce the dimension of different similarity indicators, namely rainfall characteristics, topographic and geomorphic characteristics, and soil and geological characteristics. For the clustering of each level, use the elbow method to determine the optimal number of clusters, and use K-means clustering to divide similar basin groups;

[0056] Step 3: Combine the SCE-UA algorithm to construct an automatic parameter optimization program for the Xin'anjiang model, and calibrate the model parameters of each experimental basin;

[0057] Step 4. In each of the final similar basin groups, i.e., the third-level similar basin groups, use the "leave-one-out" cross-validation of the model parameters to quantify the portability of the parameters with the Nash efficiency coefficient, and calculate the overall hydrological similarity of the third-level similar basin groups;

[0058] Step 5. Extract 35 flow characteristics of each experimental basin, and statistically analyze the changes in the standard deviation of the flow characteristics during the stepwise clustering process in each similar basin group, and calculate the hierarchical control intensity of geographical characteristics at different spatial levels on the flow characteristics.

[0059] Specifically, three types of geographical characteristics are extracted in Step 1 above, as follows:

[0060] Step 1.1. As shown in Table 1 below, according to the descriptions of 17 rainfall characteristics (Table 1), extract the rainfall characteristics from the rainfall station data in the "Hydrological Yearbook" to describe the monthly average rainfall, heavy rainfall, and rainfall intensity in each basin over many years;

[0061] Table 1 Descriptions of rainfall characteristics

[0062]

[0063]

[0064] Step 1.2. As shown in Table 2, according to the descriptions of 13 topographic and geomorphic characteristics, extract the topographic and geomorphic characteristics from the GDEM V2 30M resolution digital elevation data and Landsat 4-5 TM satellite digital products to describe the terrain undulation, basin shape, and ground cover in each basin;

[0065] Table 2 Descriptions of topographic and geomorphic characteristics

[0066]

[0067] Step 1.3. As shown in Table 3, according to the descriptions of 5 soil and geological characteristics, extract the soil and geological characteristics from the "China Soil Dataset Based on the Harmonized World Soil Database (HWSD)" and the "1:1,500,000 Geological Map of the People's Republic of China" to describe the soil properties and geological conditions in each basin.

[0068] Table 3 Descriptions of soil and geological characteristics

[0069]

[0070]

[0071] Specifically, as Figure 3 shown, the clustering process at each level in Step 2 is as follows:

[0072] Step 2.1. During the first-level clustering process, input the 17 rainfall characteristics of all experimental basins into the "Statistical Product and Service Solutions" software SPSS for principal component analysis to generate the "principal components" of rainfall characteristics with fewer quantities. Input the rainfall "principal components" into the Python program of the elbow method to obtain the optimal number of clusters, as shown in Table 4 below.

[0073] Table 4 Optimal number of clusters

[0074]

[0075] After that, input the rainfall "principal components" and the optimal number of clusters into the "Statistical Product and Service Solutions" software SPSS for K-means clustering analysis to obtain the similar basin groups CⅠ-f at the first level;

[0076] Step 2.2. During the second-level clustering process, input the 13 topographic and geomorphic characteristics of all experimental basins into the "Statistical Product and Service Solutions" software SPSS for principal component analysis to obtain the "principal components" of topographic and geomorphic characteristics with fewer quantities. In each first-level similar basin group (CⅠ-f), input the topographic and geomorphic "principal components" into the Python program of the elbow method to obtain the optimal number of clusters, as shown in Table 4. Input the topographic and geomorphic "principal components" and the optimal number of clusters into the "Statistical Product and Service Solutions" software SPSS for K-means clustering analysis to obtain the similar basin groups CⅡ-f-s at the second level, as Figure 3 shown;

[0077] Step 2.3. During the third-level clustering process, input the 5 soil and geological characteristics of all experimental basins into the "Statistical Product and Service Solutions" software SPSS for principal component analysis to obtain the "principal components" of soil and geological characteristics with fewer quantities. In each second-level similar basin group (CⅡ-f-s), input the soil and geological "principal components" into the Python program of the elbow method to obtain the optimal number of clusters, as shown in Table 4. Input the soil and geological "principal components" and the optimal number of clusters into the "Statistical Product and Service Solutions" software SPSS for K-means clustering analysis to obtain the similar basin groups CⅢ-f-s-t at the third level, that is, the final similar basin groups, as Figure 3 shown. The number and serial numbers of basins within each group (consistent with the basin serial numbers in Figure 1 ) are shown in Table 5 below.

[0078] Table 5 Number and serial numbers of basins in the third-level similar basin groups (CIII-f-s-t)

[0079]

[0080] Specifically, in step 3, the optimization model parameters of each experimental basin are calibrated by combining with the SCE-UA algorithm, which is specifically as follows:

[0081] Step 3.1: Set the objective function (F) in the SCE-UA algorithm as a combination of the Nash efficiency coefficient (NSE) and the relative error of flood volume (FE), and the calibration proceeds in the direction of the objective function value continuously approaching 0;

[0082]

[0083]

[0084]

[0085] In the formula, R0 and R c are the measured runoff depth and the simulated runoff depth respectively, with the unit of mm, i is the number of days in a year; Q o,i and Q c,i are the measured flow rate and the simulated flow rate respectively, with the unit of m 3 / s, is the annual average value of the measured flow rate, with the unit of m 3 / s, ω1 and ω2 are weight coefficients, here ω1 = ω2 = 0.5, and y is the number of years in the calibration period;

[0086] Step 3.2: Set the parameters of the Xin'anjiang model to be calibrated and their value ranges as shown in Table 6, and use the SCE-UA algorithm to calibrate the model parameters of all experimental basins for subsequent parameter transplantation.

[0087] Table 6 Parameters of the Xin'anjiang model to be calibrated and their value ranges

[0088]

[0089]

[0090] Specifically, step 4 is specifically as follows:

[0091] Step 4.1: Conduct "leave-one-out" cross-validation of the model parameters in the third-level similar basin group. For example, in the CIII-f-s-t basin group, there are a total of N Ⅲ (f, s, t) = n basins. Each basin in the group is successively used as the receptor basin (data-deficient basin), and the other n - 1 similar basins in the group are used as its donor basins. Transplant the hydrological model parameters of the donor basins to the receptor basin for flow simulation, and evaluate the parameter transplantability using the Nash efficiency coefficient. Obtain a set of n - 1 Nash efficiency coefficients (E i (f, s, t)), and calculate the hydrological similarity S between receptor basin i and other similar basins i(f, s, t). S i If the value of (f, s, t) is greater than 0.5, it indicates that the hydrological similarity degree is good, and the closer it is to 1, the higher the similarity degree. Calculate the overall hydrological similarity (S(f, s, t)) of the CIII-f-s-t river basin group as follows Figure 4 shown

[0092] E i (f, s, t) = {NSE(1), NSE(2), …, NSE(i - 1), NSE(i + 1), …, NSE(n)}

[0093] (i = 1, 2, …, n),

[0094] S i (f, s, t) = MEDIAN(E i (f, s, t))

[0095] In the formula, i is the i-th receptor river basin in CⅢ-f-s-t

[0096] Step 4.2: According to the set (E(f, s, t)) of Nash efficiency coefficients of all receptor river basins in the CIII-f-s-t river basin group, calculate the overall hydrological similarity (S(f, s, t)) of this river basin group. The larger the value of S(f, s, t), the more similar hydrological processes can be generated under the corresponding geographical environment of the CIII-f-s-t river basin group

[0097] E(f, s, t) = {E1(f, s, t), E2(f, s, t), …, E n (f, s, t)},

[0098] S(f, s, t) = MEDIAN(E(f, s, t))

[0099] Specifically, the specific steps of step 5 are as follows

[0100] Step 5.1: As shown in Table 7, according to the description of 35 flow characteristics, extract the flow characteristics from the hydrological station data in the "Hydrological Yearbook" to describe the annual, monthly, and daily scale flow states within each river basin

[0101] [[ID=4~2]]Table 7 Description of 35 flow characteristics

[0102]

[0103]

[0104] Step 5.2. Statistically analyze the change in the standard deviation of the flow characteristics within the third-level similar river basin group (CIII-f-s-t) during the progressive clustering process, and calculate the hierarchical control intensity of geographical characteristics at different spatial levels on the flow characteristics.

[0105]

[0106] D Ⅰ (f) = MEDIAN(A Ⅰj (f)),

[0107]

[0108] D Ⅱ (f, s) = MEDIAN(A Ⅱj (f, s)),

[0109]

[0110] D Ⅲ (f, s, t) = MEDIAN(A Ⅲj (f, s, t)).

[0111] In the formula, j represents the jth flow characteristic, b j 0 is the standard deviation (original state) of the jth flow characteristic of all experimental river basins, A Ⅰj (f) represents the change in the standard deviation of the jth flow characteristic in the first-level similar river basin group (CⅠ-f) after the first-level clustering (relative to the original state). Similarly, A Ⅱj (f, s) and A Ⅲj (f, s, t) respectively represent the changes in the standard deviation of the jth flow characteristic in the second- and third-level similar river basin groups (CⅡ-f-s and CⅢ-f-s-t) after the second and third-level clusterings (relative to the original state). A Ⅰj (f), A Ⅱj (f, s) and A Ⅲj (f, s, t) are shown as Figure 5 indicated. The larger the value, the stronger the control effect of the geographical characteristics at the corresponding spatial level on the specific hydrological response. A value greater than 0 indicates that this type of geographical characteristic plays a positive control role in the hydrological response, can effectively identify similar runoff processes, and cause the flow characteristics to aggregate. D Ⅰ (f), D Ⅱ (f, s) and D Ⅲ (f, s, t) are shown as Figure 6 indicated. The larger the value, the stronger the control effect of the geographical characteristics at the corresponding spatial level on the hydrological response represented by all flow characteristics.

[0112] It should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention rather than to limit them. Although the present invention has been described in detail with reference to the preferred embodiments, those of ordinary skill in the art should understand that the technical solutions of the present invention can be modified or equivalently replaced without departing from the spirit and scope of the technical solutions of the present invention, and they should all be covered within the scope of the claims of the present invention.

Claims

1. A watershed hydrological similarity analysis method based on a stepwise clustering scheme, characterized in that: The following steps are involved: Step 1: Extract the rainfall characteristics, topographic features, and soil geological characteristics of the experimental basin, and use these three types of geographical characteristics as similarity indicators for the first-level, second-level, and third-level clustering processes respectively; Step 2: In the clustering process at each level, principal component analysis is applied to reduce the dimension of different similarity indicators, namely rainfall characteristics, topographic characteristics, and soil geological characteristics. For each level of clustering, the elbow rule is used to determine the optimal number of clusters, and K-means clustering is used to divide similar watershed groups. Step 3: Combined with the SCE-UA algorithm, an automatic parameter optimization program for the Xin'anjiang model was constructed to calibrate the model parameters for each experimental basin, as follows: Step 3.1: Set the objective function F in the SCE-UA algorithm to be the combination of the Nash efficiency coefficient NSE and the flood relative error FE, and calibrate in the direction that the objective function value continuously approaches 0; Where R0, R c are the measured runoff depth and the simulated runoff depth, respectively, in mm, i is the number of days in a year; Q o,i , Q c,i They are measured flow and simulated flow, respectively, in m 3 / s, The annual average value of the measured flow, unit is m 3 / s, ω1 and ω2 are weight coefficients, where ω1=ω2=0.5, and y is the number of years in the rate period; Step 3.2: Set the Xin'anjiang model parameters to be calibrated and their value ranges, and use the SCE-UA algorithm to calibrate the model parameters of all experimental basins; Step 4: In the final similarity basin groups, i.e., the third-level similarity basin groups, the model parameters are cross-validated using the leave-one-out cross-validation method. The portability of the parameters is quantified using the Nash efficiency coefficient, and the overall hydrological similarity of the third-level similarity basin groups is calculated. Step 5: Extract the flow characteristics of each experimental basin, statistically analyze the changes in the standard deviation of flow characteristics in each similar basin group during the stepwise clustering process, and calculate the hierarchical control intensity of geographical characteristics at different spatial levels on flow characteristics.

2. The method for analyzing watershed hydrological similarity based on a stepwise clustering scheme according to claim 1, characterized in that: In step 1, rainfall characteristics were extracted based on rain gauge data; topographic and geomorphological characteristics were extracted based on GDEMV2 30M resolution digital elevation data and Landsat4-5 TM satellite digital products; and soil geological characteristics were extracted based on the "Chinese Soil Dataset Based on the World Soil Database (HWSD)" and the "1:1.5 Million Geological Map of the People's Republic of China".

3. The method for analyzing watershed hydrological similarity based on a stepwise clustering scheme according to claim 1, characterized in that: The clustering process at each level in step 2 is as follows: Step 2.1: In the first-level clustering process, the rainfall characteristics of the experimental watershed are subjected to principal component analysis to generate a smaller number of "principal components" of rainfall characteristics. The "principal components" of rainfall characteristics are then classified according to the elbow rule to obtain the optimal number of clusters. The "principal components" of rainfall characteristics and the optimal number of clusters are then subjected to K-means cluster analysis to obtain the first-level similar watershed clusters CⅠ-f. Step 2.2: In the second-level clustering process, the topographic and geomorphological characteristics of the experimental watershed are input into principal component analysis to obtain a smaller number of "principal components" of topographic and geomorphological characteristics. In each first-level similar watershed cluster CⅠ-f, the "principal components" of topographic and geomorphological characteristics are used to obtain the optimal number of clusters according to the elbow rule. The "principal components" of topographic and geomorphological characteristics and the optimal number of clusters are then subjected to K-means cluster analysis to obtain the second-level similar watershed cluster CⅡ-fs. In step 2.3, during the third-level clustering process, the soil geological characteristics of the experimental basins were subjected to principal component analysis to obtain a smaller number of "principal components" of soil geological characteristics. In each second-level similar basin group CⅡ-fs, the "principal components" of soil geological characteristics were used to obtain the optimal number of clusters according to the elbow rule. The "principal components" of soil geological characteristics and the optimal number of clusters were subjected to K-means cluster analysis to obtain the third-level similar basin group CⅢ-fst.

4. The method for analyzing watershed hydrological similarity based on a stepwise clustering scheme according to claim 1, characterized in that: Step 4: Calculating the overall hydrological similarity of similar watershed groups includes the following steps: Step 4.1: Perform a leave-one-out cross-validation of the model parameters in the third level of similar watershed groups. In the CIII-fst watershed group, there are N Ⅲ (f, s, t) = n basins, each basin in the similar basin group is taken as the recipient basin, that is, the basin without data, and the other n-1 similar basins in the similar basin group are taken as its donor basins. The hydrological model parameters of the donor basins are transplanted to the recipient basins for flow simulation, and the Nash efficiency coefficient is used to evaluate the parameter portability, and the set of n-1 Nash efficiency coefficients E is obtained. i (f, s, t), based on which the hydrological similarity S between the recipient basin i and other similar basins is calculated i (f,s,t); S i The value of (f, s, t) greater than 0.5 indicates good hydrological similarity, and the closer it is to 1, the higher the similarity; E i (f,s,t)={NSE(1),NSE(2),…,NSE(i-1),NSE(i+1),…,NSE(n)}, S i (f,s,t)=MEDIAN(E i (f,s,t)) Where i is the i-th receptor basin in CⅢ-fst, i = 1, 2, …, n; Step 4.2: Calculate the overall hydrological similarity S(f,s,t) of the CIII-fst similarity basin group based on the set of Nash efficiency coefficients E(f,s,t) of all receptor basins within the CIII-fst similarity basin group. A larger value of S(f,s,t) indicates that more similar hydrological processes can be generated under the geographical environment corresponding to the CIII-fst basin group. E(f,s,t)={E1(f,s,t),E2(f,s,t),…,E n (f,s,t)}, S(f,s,t)=MEDIAN(E(f,s,t)).

5. The method for analyzing watershed hydrological similarity based on a stepwise clustering scheme according to claim 1, characterized in that: The step 5 is specifically as follows: Step 5.1: Extract flow characteristics from the hydrological station data in the Hydrological Yearbook according to the description of flow characteristics, and describe the flow status of each experimental basin at the annual, monthly, and daily scales for many years; Step 5.2: Count the changes in the standard deviation of flow characteristics within the third-level similar watershed group CIII-fst during the stepwise clustering process, and calculate the hierarchical control strength of geographical characteristics at different spatial levels on flow characteristics; D Ⅰ (f)=MEDIAN(A Ⅰj (f)), D Ⅱ (f,s)=MEDIAN(A Ⅱj (f,s)), D Ⅲ (f,s,t)=MEDIAN(A Ⅲj (f,s,t)), Where j is the jth flow feature, b j 0 is the standard deviation of the jth flow characteristics of all experimental basins, that is, the original state, A Ⅰj (f) represents the change of the standard deviation of the j-th flow characteristic relative to the original state in the first-level similar basin group CⅠ-f after the first-level clustering; similarly, A Ⅱj (f,s) and A Ⅲj (f, s, t) represents the change of the standard deviation of the jth flow characteristic relative to the original state in the second and third level similar basin groups CⅡ-fs and CⅢ-fst after the second and third level clustering respectively; A Ⅰj (f), A Ⅱj (f,s) and A Ⅲj The larger the value of (f,s,t), the stronger the control effect of the geographical features at the corresponding spatial level on the specific hydrological response. A value greater than 0 indicates that this type of geographical feature plays a positive control role on the hydrological response, can effectively identify similar runoff processes, and causes the flow characteristics to cluster; D Ⅰ (f), D Ⅱ (f,s) and D Ⅲ The larger the value of (f,s,t), the stronger the control effect of the geographical characteristics of the corresponding spatial level on the hydrological response represented by all flow characteristics.

Images