Method and system for evaluating corresponding relation between rainfall extreme value event and runoff peak value event

Abstract

The invention relates to the technical field of hydrological statistics, and discloses a corresponding relation evaluation method and system for a rainfall extreme value event and a runoff peak event, and the method comprises the steps: obtaining rainfall time sequence data and runoff time sequence data; according to the rainfall time sequence data and the runoff time sequence data, respectively identifying a rainfall extreme value event and a runoff peak value event; constructing a dichotomy model for quantifying the corresponding relationship between the rainfall extreme value event and the runoff peak value event under different thresholds; and a target rainfall extreme value is set, runoff peak value candidate threshold values are traversed, and a runoff peak value threshold value enabling the performance index of the dichotomy model to be optimal is selected as an optimal runoff response threshold value corresponding to the target rainfall extreme value. The method can effectively reduce the false alarm rate and the missing report rate in the evaluation process of the corresponding relation between the rainfall extreme value event and the runoff peak value event, and remarkably improves the evaluation precision of the corresponding relation.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of hydrology statistics, and more particularly, to a method and system for evaluating the corresponding relationship between precipitation extreme events and runoff peak events. BACKGROUND

[0002] In the field of watershed hydrology research, the corresponding relationship between precipitation extreme events and runoff peak events is a core issue for revealing the precipitation-runoff relationship. Traditional theory believes that runoff is a lag response to precipitation, and subsequent runoff processes can only be triggered when precipitation exceeds a critical value. Therefore, the corresponding relationship between the two can be obtained by analyzing a large amount of precipitation and runoff observation data. Accurate evaluation of this relationship is crucial for understanding the highly nonlinear characteristics of the precipitation-runoff system, which is the core scientific problem of runoff prediction and flood warning.

[0003] In the prior art, when evaluating the corresponding relationship between the two, the non-one-to-one correspondence phenomenon caused by the heterogeneity of the underlying surface of the watershed is usually converted into a binary time series two-classification prediction problem of precipitation extreme value and runoff peak occurrence / non-occurrence. By analyzing historical data, it is found that the proportion of precipitation extreme values exceeding a given level corresponding to runoff peak values exceeding a corresponding level is very low, reflecting a significant false positive rate in two-classification prediction. Although the false positive rate can be reduced by increasing the precipitation extreme value determination threshold in the business system, this will simultaneously increase the false negative rate of runoff peak events, resulting in a decrease in the overall reliability of the warning system. SUMMARY

[0004] To reduce the false positive rate and false negative rate of the evaluation and prediction of the corresponding relationship between precipitation extreme events and runoff peak events and improve the evaluation accuracy, the present application proposes the following technical solutions: In a first aspect, the present application provides a method for evaluating the corresponding relationship between precipitation extreme events and runoff peak events, comprising: obtaining precipitation time series data and runoff time series data; identifying precipitation extreme events and runoff peak events from the precipitation time series data and the runoff time series data, respectively; constructing a two-classification model for quantifying the corresponding relationship between precipitation extreme events and runoff peak events at different thresholds; setting a target precipitation extreme value, traversing runoff peak candidate thresholds, and selecting a runoff peak threshold that makes the performance index of the two-classification model optimal as the optimal runoff response threshold corresponding to the target precipitation extreme value.

[0005] As a preferred technical solution, identifying precipitation extreme events from the precipitation time series data comprises: extracting precipitation extreme events from the precipitation time series data according to the following formula:

[0006] Where, P is the precipitation identifier, Represents the study area c Time series of extreme precipitation events, Represents the study area c At the moment of precipitation, For the study area c At the moment The percentile of effective precipitation, Indicates the percentile threshold parameter of precipitation, ranging from 0 to 100.

[0007] As a preferred technical solution, after identifying the extreme precipitation event, the method further includes: For the time series of extreme precipitation events Among the points that are continuous in time, the point with the maximum precipitation value is taken as a single extreme precipitation event; For the time series of extreme precipitation events The points with temporal discontinuity in the middle are directly determined as single extreme precipitation events; Single precipitation extreme events with a time interval not less than the threshold are extracted as independent precipitation extreme events.

[0008] As a preferred technical solution, identifying a runoff peak event based on the runoff time series data includes: Runoff peak events are extracted from the runoff time series data according to the following formula:

[0009] Where, Q is the runoff identifier, Represents the study area c Time series of peak runoff events, Represents the study area c At the moment The runoff flow, For the study area c At the moment The percentile of effective runoff, Indicates the percentile threshold parameter of runoff, ranging from 0 to 100.

[0010] As a preferred technical solution, after identifying the runoff peak event, the method further includes: Time series of the peak runoff events In , the runoff peak events that meet the following constraints are extracted as independent runoff peak events:

[0011] wherein, denotes the time interval between two peak flow events of the study area c ; denotes the catchment area of the study area c ; and denote the flow of the peak flow c and the peak flow i of the study area j , respectively; denotes the minimum flow of the study area c from the peak flow i to the peak flow j .

[0012] As a preferred technical solution, a binary classification model for quantifying the correspondence between precipitation extreme events and peak flow events under different thresholds is constructed, comprising: For the time series of precipitation extreme events of the identified study area c , mark the time when the precipitation extreme event occurs as 1 and the time when it does not occur as 0 to generate a precipitation binary classification sequence; For the time series of peak flow events of the identified study area c , mark the time when the peak flow event occurs as 1 and the time when it does not occur as 0 to generate a peak flow binary classification sequence; Take the precipitation binary classification sequence as the prediction item and the peak flow binary classification sequence as the observation item to construct a 2x2 contingency table, count the state counts of simultaneous occurrence of precipitation extreme events and peak flow events, occurrence of only precipitation extreme events, occurrence of only peak flow events, and non-occurrence of both precipitation extreme events and peak flow events, and obtain a binary classification model of the correspondence between precipitation extreme events and peak flow events.

[0013] As a preferred technical solution, the performance indicators of the binary classification model include the key success index and the harmonic series F1 score.

[0014] As a preferred technical solution, the calculation expression of the performance indicators of the binary classification model including the key success index and the harmonic series F1 score is as follows:

[0015]

[0016]

[0017] CSI is the key success index, which is used to quantify the overall level of the corresponding relationship between the subsequent runoff peak event triggered by the precipitation extreme event; POD is the hit rate, which is used to quantify the proportion of the precipitation extreme event correctly corresponding to all the runoff peak events; FAR is the false alarm rate, which is used to quantify the proportion of the runoff peak event that fails to correctly correspond to all the precipitation peak events; TP represents the number of states in which the precipitation extreme event and the runoff peak event occur simultaneously; FP represents the number of states in which only the precipitation extreme event occurs; FN represents the number of states in which only the runoff peak event occurs; TN represents the number of states in which neither the precipitation extreme event nor the runoff peak event occurs;

[0018]

[0019]

[0020] respectively represent the precision and recall.

[0021] As a preferred technical solution, the runoff peak candidate threshold is traversed, and the runoff peak threshold that makes the performance index of the binary classification model optimal is selected as the optimal runoff response threshold corresponding to the target precipitation extreme, comprising: The runoff peak candidate threshold that makes the key success index CSI and the harmonic series F1 score reach the maximum is selected as the optimal runoff response threshold corresponding to the target precipitation extreme.

[0022] In a second aspect, the present application also provides a precipitation extreme event and runoff peak event corresponding relationship evaluation system, which is applied to the precipitation extreme event and runoff peak event corresponding relationship evaluation method in any of the first aspect, comprising: The acquisition module is used to acquire precipitation time series data and runoff time series data; The identification module is used to identify precipitation extreme events and runoff peak events according to the precipitation time series data and the runoff time series data, respectively; The construction module is used to construct a binary classification model for quantifying the corresponding relationship between the precipitation extreme event and the runoff peak event under different thresholds; The evaluation module is used to set a target precipitation extreme, traverse the runoff peak candidate threshold, and select the runoff peak threshold that makes the performance index of the binary classification model optimal as the optimal runoff response threshold corresponding to the target precipitation extreme.

[0023] The present application has at least the following beneficial effects:​​ The application realizes accurate evaluation of the corresponding relationship between precipitation extreme value and runoff peak value event by obtaining precipitation and runoff time series data, respectively identifying extreme value events and constructing a binary classification model to quantify the corresponding relationship under different thresholds, and then selecting the optimal value by traversing the runoff peak value candidate threshold. BRIEF DESCRIPTION OF DRAWINGS

[0024] Figure 1 Figure 1 is a flowchart of the corresponding relationship evaluation method of precipitation extreme value event and runoff peak value event provided by the embodiment of the application.

[0025] Figure 2 (a) is a corresponding relationship distribution feature map of precipitation extreme value-runoff peak value under the corresponding relationship of the theoretical plane provided by the embodiment of the application.

[0026] Figure 2 (b) is a corresponding relationship change feature map of different basins under the corresponding relationship of precipitation extreme value-runoff peak value provided by the embodiment of the application.

[0027] Figure 2 (c) is a projection map of the change feature in the POD-CSI plane under the corresponding relationship of precipitation extreme value-runoff peak value provided by the embodiment of the application.

[0028] Figure 2 (d) is a projection map of the change feature in the 1-FAR plane-CSI plane under the corresponding relationship of precipitation extreme value-runoff peak value provided by the embodiment of the application.

[0029] Figure 2 (e) is a projection map of the change feature in the 1-FAR plane-POD plane under the corresponding relationship of precipitation extreme value-runoff peak value provided by the embodiment of the application.

[0030] Figure 3 (a) is a performance result map of the corresponding relationship between the upper 90% precipitation extreme value and the upper 75% runoff peak value in the precipitation-runoff process provided by the embodiment of the application.

[0031] Figure 3 (b) is a performance result map of the corresponding relationship between the upper 90% precipitation extreme value and the upper 90% runoff peak value in the precipitation-runoff process provided by the embodiment of the application.

[0032] Figure 3(c) The performance result graph of the upper 90% precipitation extreme value and the upper 99% runoff peak value corresponding relationship in the precipitation-runoff process provided by the embodiment of the present application.

[0033] Figure 4 (a) The performance result graph of the precipitation extreme value-runoff peak value optimal corresponding relationship provided by the embodiment of the present application.

[0034] Figure 4 (b) The precision-recall curve graph of the precipitation extreme value-runoff peak value optimal corresponding relationship provided by the embodiment of the present application.

[0035] Figure 5 The spatial pattern graph of the precipitation extreme value-runoff peak value optimal corresponding relationship of each basin in the United States.

[0036] Figure 6 (a) The POD spatial distribution graph of each basin in the United States under the precipitation extreme value-runoff peak value optimal corresponding relationship provided by the embodiment of the present application.

[0037] Figure 6 (b) The graph of the POD of each basin in the United States under the precipitation extreme value-runoff peak value optimal corresponding relationship changing with the threshold provided by the embodiment of the present application.

[0038] Figure 6 (c) The FAR spatial distribution graph of each basin in the United States under the precipitation extreme value-runoff peak value optimal corresponding relationship provided by the embodiment of the present application.

[0039] Figure 6 (d) The graph of the FAR of each basin in the United States under the precipitation extreme value-runoff peak value optimal corresponding relationship changing with the threshold provided by the embodiment of the present application.

[0040] Figure 6 (e) The CSI spatial distribution graph of each basin in the United States under the precipitation extreme value-runoff peak value optimal corresponding relationship provided by the embodiment of the present application.

[0041] Figure 6 (f) The graph of the CSI of each basin in the United States under the precipitation extreme value-runoff peak value optimal corresponding relationship changing with the threshold provided by the embodiment of the present application.

[0042] Figure 7 The architecture graph of the corresponding relationship evaluation system of the precipitation extreme value event and the runoff peak value event provided by the embodiment of the present application. DETAILED DESCRIPTION

[0043] The following will describe embodiments of the present invention with reference to the accompanying drawings and preferred technical solutions. Those skilled in the art will readily understand other advantages and benefits of the present invention from the contents disclosed in this specification. The present invention may also be implemented or applied through different specific embodiments, and the details in this specification may be modified or altered based on different viewpoints and applications without departing from the spirit of the present invention. It should be understood that the preferred technical solutions are intended only to illustrate the present invention and are not intended to limit the scope of protection of the present invention.

[0044] It should be noted that the illustrations provided in the following embodiments are merely schematic illustrations of the basic concept of the present invention. Therefore, the illustrations only show components related to the present invention and are not drawn according to the number, shape, and size of components in actual implementation. In actual implementation, the type, quantity, and proportion of each component may be changed arbitrarily, and the component layout may also be more complex.

[0045] In the following description, numerous details are discussed to provide a more thorough explanation of the embodiments of the present invention. However, it will be apparent to those skilled in the art that the embodiments of the present invention may be practiced without these specific details. In other embodiments, well-known structures and devices are shown in block diagram form rather than in detail to avoid obscuring the embodiments of the present invention.

[0046] Example 1 This embodiment proposes a method for evaluating the correspondence between extreme precipitation events and peak runoff events, such as Figure 1 As shown, Figure 1 This is a flow chart of a method for evaluating the correspondence between extreme precipitation events and peak runoff events provided in this embodiment. The method includes the following steps: S1: Obtain precipitation time series data and runoff time series data; This example uses the daily precipitation time series and runoff time series from the site observation data at the outlet section of the basin as input data. Common formats include CSV, TXT, etc. First, use Python language to read the daily precipitation time series, which is recorded as , and the daily runoff time series is recorded as , where the subscript Represents time, Represents the corresponding basin outlet section site.

[0047] S2: identifying precipitation extreme value events and runoff peak events respectively according to the precipitation time series data and the runoff time series data; S3: Construct a binary classification model to quantify the correspondence between extreme precipitation events and peak runoff events under different thresholds; S4: Set the target precipitation extreme value, traverse the runoff peak candidate thresholds, and select the runoff peak threshold that makes the performance index of the binary classification model reach the optimal value as the optimal runoff response threshold corresponding to the target precipitation extreme value.

[0048] In the specific implementation process, precipitation time series data and runoff time series data of the study area are first collected; then, according to the set effective precipitation judgment rules and percentile parameters, precipitation extreme events are identified from the precipitation time series, and at the same time, runoff peak events are extracted from the runoff time series according to the runoff percentile threshold and peak morphology conditions; subsequently, a binary classification model is constructed with the precipitation extreme event sequence as the prediction item and the runoff peak event sequence as the observation item, and the correlation relationship is quantified by statistically analyzing the corresponding states of the two in the time dimension; finally, the percentile threshold of the target precipitation extreme value is fixed, the candidate percentile thresholds of the runoff peak value are traversed, the performance indicators of the binary classification model under each threshold are calculated, and the runoff peak threshold that makes the performance indicator optimal is selected as the optimal runoff response threshold corresponding to the target precipitation extreme value, completing the evaluation of the correspondence between precipitation extreme value events and runoff peak events.

[0049] Example 2 This embodiment makes improvements based on the method for evaluating the correspondence between extreme precipitation events and peak runoff events proposed in Example 1.

[0050] Optionally, identifying precipitation extreme events according to the precipitation time series data includes: According to the following formula, precipitation extreme events are extracted from the precipitation time series data:

[0051] Where, P is the precipitation identifier, Represents the study area c Time series of extreme precipitation events, Represents the study area c At the moment of precipitation, For the study area c At the moment The percentile of effective precipitation, Indicates the percentile threshold parameter of precipitation, ranging from 0 to 100.

[0052] Optionally, after identifying the extreme precipitation event, the method further includes: For the time series of extreme precipitation events Among the points that are continuous in time, the point with the maximum precipitation value is taken as a single extreme precipitation event; For the time series of extreme precipitation events The point of time discontinuity is directly determined as a single precipitation extreme event; The precipitation extreme event with an extraction time interval not less than a threshold single precipitation extreme event is determined as an independent precipitation extreme event.

[0053] In an example, the effective rainfall threshold is set as 1 mm / day, and the research area is screened c At the time point The precipitation satisfies The time point is determined as an effective precipitation time point; the precipitation percentile parameter of the effective precipitation time point is calculated The corresponding percentile is recorded as The precipitation extreme time sequence is constructed, wherein The value range of is 0-100; for The time-continuous is merged as a single precipitation extreme event and the maximum precipitation value thereof is selected as a representative of the event; the time-discontinuous is directly determined as a single precipitation extreme event; the event with a time interval of more than 3 days is retained as an independent precipitation extreme event; for each independent precipitation extreme event, the time point corresponding to the maximum runoff value in the runoff time sequence within 5 days after the start of the event is recorded, and the time marker of the independent precipitation extreme event is updated.

[0054] In this embodiment, on the one hand, the physical boundary of the precipitation event is determined by 1 mm / day effective rainfall, and the extreme value degree is quantified by percentile to eliminate weak precipitation interference; on the other hand, the event independence is ensured by 3-day time interval, and the time response relationship between precipitation and runoff is aligned in advance by 5-day runoff correlation window, so as to strengthen the consistency and correlation of event characteristics from the data preprocessing stage, provide accurate and aligned event sequence for subsequent binary classification model construction, and support the improvement of the evaluation accuracy of the corresponding relationship between precipitation extreme value and runoff peak value.

[0055] Optionally, according to the runoff time sequence data, a runoff peak event is identified, comprising: The runoff peak event is extracted from the runoff time sequence data according to the following formula:

[0056] In the formula, Q is a runoff identifier, indicates a runoff peak event time sequence of the research area, c indicates the runoff of the research area at the time point c is a runoff time sequence of the research area is a runoff flow of the research area at the time pointc at time the percentage of the effective runoff, unit: m 3 / s, denotes the percentile threshold parameter of runoff, is the logical operation and, the value range is 0~100.

[0057] Optionally, after identifying the runoff peak event, further comprising: in the time sequence of the runoff peak event , extract the runoff peak event satisfying the following constraints as an independent runoff peak event:

[0058] wherein, denotes the time interval between two runoff peak events in the study area c , unit: day; denotes the watershed area of the study area c , unit: km 2 ; and respectively denote the flow of runoff peak c and runoff peak i in the study area j , unit: m 3 / s; denotes the minimum flow between runoff peak c and runoff peak i in the study area j , unit: m 3 / s.

[0059] That is, between two runoff peaks, there should be at least a 5-day time interval; at the same time, the minimum flow between two runoff peaks should be less than 75% of the smaller one of the two flow peaks.

[0060] Optionally, a binary classification model for quantifying the correspondence between precipitation extreme events and runoff peak events under different thresholds is constructed, comprising: marking the time when the precipitation extreme event occurs as 1 and the time when it does not occur as 0 to generate a precipitation binary sequence for the precipitation extreme event time sequence of the study area c identified; marking the time when the runoff peak event occurs as 1 and the time when it does not occur as 0 to generate a runoff binary sequence for the runoff peak event time sequence of the study area c identified; A 2x2 contingency table is constructed with the precipitation two-class sequence as a prediction item and the runoff two-class sequence as an observation item, and state counts of a precipitation extreme event and a runoff peak event occurring simultaneously, only a precipitation extreme event occurring, only a runoff peak event occurring, and neither a precipitation extreme event nor a runoff peak event occurring are counted to obtain a two-class model of the corresponding relationship between the precipitation extreme event and the runoff peak event.

[0061] Table 1 2x2 contingency table of corresponding relationship between precipitation extreme event and subsequent runoff peak event

[0062] As shown in Table 1, TP 0 indicates correct correspondence between the precipitation extreme event and the runoff peak event; FP 0 indicates that the precipitation extreme event occurred, but no corresponding runoff peak event occurred; FN 0 indicates that the precipitation extreme event did not occur, but the runoff peak event occurred; TN 0 indicates that neither the precipitation extreme event nor the runoff peak event occurred.

[0063] Optionally, the performance indicators of the two-class model include a critical success index and a harmonic series F1 score.

[0064] Optionally, the calculation expression of the performance indicators of the two-class model including the critical success index and the harmonic series F1 score is as follows:

[0065]

[0066]

[0067]

[0068] In the formula, CSI (Critical Success Index, CSI) is a critical success index, used to quantify the overall level of the corresponding relationship between the precipitation extreme event and the subsequent runoff peak event; POD (Probability of Detection, POD) is a hit rate, used to quantify the proportion of all runoff peak events that are correctly corresponded to precipitation extreme events; FAR (False Alarm Ratio, FAR) is a false alarm rate, used to quantify the proportion of all precipitation peak events that fail to be correctly corresponded to runoff peak events; FB (Frequency Bias, FB) is a frequency bias, used to quantify the ratio of the number of runoff peak events to the number of precipitation extreme events TP represents the number of states in which the precipitation extreme event and the runoff peak event occurred simultaneously; Expressed as a nonlinear function of POD and FAR; FP Indicates the number of states where only extreme precipitation events occur; FN Indicates the number of states where only peak runoff events occur; TN The count indicates the state where neither the extreme precipitation event nor the peak runoff event occurred;

[0069]

[0070]

[0071] Where, and Represents precision and recall, respectively. Precision quantifies the proportion of all precipitation peak events that correctly correspond to runoff peak events, also known as 1-FAR. Recall is defined identically to POD. The F1 score is the harmonic sum of Precision and Recall and can be used to indicate the quality of the precipitation extreme value-runoff peak binary classification prediction at a given threshold level.

[0072] Optionally, traversing the candidate runoff peak thresholds and selecting the runoff peak threshold that makes the performance index of the binary classification model reach the optimal value as the optimal runoff response threshold corresponding to the target precipitation extreme value includes: The runoff peak candidate threshold value corresponding to the maximum critical success index CSI and the harmonic series F1 score is used as the optimal runoff response threshold value corresponding to the target precipitation extreme value.

[0073] It should be noted that the value range of POD, FAR, and CSI is [0, 1], and they can take any value within the range. If 1-FAR is represented on the x-axis, POD on the y-axis, and CSI on the z-axis, all the possibilities will form a 3D surface with the following four endpoints:

[0074] function The projection on the xy plane is the Performance Diagram. In the Performance Diagram, the contour lines of CSI increase from the lower left corner (1-FAR=0, POD=0) to the upper right corner (1-FAR=1, POD=1). In the binary prediction, the perfect prediction skill is represented by POD=1, FAR=0, which falls in the upper right corner of the Performance Diagram. Among all possible theoretical values, there is a correspondence based on the quantification of the actual precipitation time series and the runoff time series. In other words, for a given basin , consider a set of percentiles ( th, th), a set of certain , and , uniquely determines a unique point on the theoretical 3D plane. When fixing the precipitation extreme, considering a series of runoff peak values (taken from different percentiles ), a series of points on the 3D surface can be obtained. When the CSI reaches the maximum, it indicates that the correspondence between the precipitation extreme event and the subsequent runoff peak event reaches the optimal threshold level:

[0075] where represents the percentile that makes the CSI reach the maximum in . Through this diagnosis, it can be obtained that, for a basin , the runoff peak value that can form the optimal correspondence with the precipitation extreme exceeding the percentile has a threshold level of the percentile .

[0076] The precision-recall curve (PRC) is used to further verify the binary classification model and the optimal correspondence. For a given basin , considering a set of percentiles (th th, th), a set of certain , and will be obtained in the binary classification prediction of the precipitation extreme-runoff peak; fixing the precipitation extreme, considering a series of runoff peak values (taken from different percentiles ), it is confirmed whether the CSI obtained by the formula satisfies the following conditions:

[0077] In addition, the above series of results will be able to form a specific PRC, corresponding to a unique area under the PRC (AUCPR). When Recall is represented on the x-axis and Precision is represented on the y-axis, AUCPR can be calculated by solving the trapezoidal area composed of discrete points:

[0078] ​​where n is the number of all points including the top-left corner (Recall=0, Precision=1) and the bottom-right corner (Recall=1, Precision=0); k is the ordinal number of all PRC points from left to right. When the catchment is considered to be under the given threshold level of the precipitation extreme value, an effective binary classification result can be formed with the runoff peak value.

[0079] It can be understood that by acquiring the precipitation and runoff time series data, respectively identifying the extreme value event and constructing the binary classification model to quantify the corresponding relationship under different threshold values, and then selecting the optimal value by traversing the runoff peak value candidate threshold, the corresponding relationship between the precipitation extreme value and the runoff peak value event is accurately evaluated.

[0080] The present application identifies the precipitation extreme value event and the runoff peak value event caused thereby by specifying the threshold level, has few input parameters, and is also high in flexibility in that the time interval of the precipitation extreme value and the runoff peak value can be specified according to different catchment characteristics; the corresponding relationship between the precipitation extreme value and the runoff peak value is quantified by means of the threshold-mediated binary classification model, a data-driven mathematical model method is used to replace the traditional hydrological model, the limitation of mechanism modeling is broken through, and the complexity of the rapid judgment of the catchment-scale precipitation-runoff relationship is facilitated; the diagnosis of a single catchment is supported, and the large-sample analysis of multiple catchments is also applicable, the calculation logic based on statistical quantization is low in demand for computing power, stable in results, and convenient to use. By means of the binary classification model and the threshold optimization mechanism, the false positive rate and the false negative rate in the evaluation process are effectively reduced, the evaluation accuracy of the corresponding relationship is significantly improved, and an efficient technical path for the rapid diagnosis of the catchment-scale precipitation-runoff response relationship is provided.

[0081] Embodiment 3 In this embodiment, the daily-scale precipitation runoff observation data of Catchment Attributes and Meteorology for Large-sample Studies (CAMELS) is used as input, and the corresponding relationship between the runoff peak value event caused by the precipitation extreme value event with a threshold level of 90th percentile is evaluated.

[0082] In this embodiment, the precipitation runoff observation data used to identify the extreme value event and the peak value event in the continental range of the United States is extracted, the data set used in this example is stored in the txt format, and the precipitation runoff observation data from 1980 to 2014 is read by using the read_table function in the Python third-party library Pandas.

[0083] In this embodiment, the 90th percentile precipitation extreme threshold is set to identify the precipitation extreme event, and the 50th to 99th series runoff peak threshold is set to identify the runoff peak event. According to the paired combination of the 90th percentile precipitation threshold and the 50th to 99th percentile runoff threshold, the corresponding binary classification prediction model is constructed, and the prediction performance of the binary classification model under different threshold pairings is quantified.

[0084] In this embodiment, the Numpy and Pandas libraries are mainly called and encapsulated as functions to process and evaluate the input data, and the results are saved in csv format. The specific steps are as follows: Step ①: Traverse all the basins in the CAMELS dataset, record the occurrence time of precipitation extreme events and runoff peak events according to different threshold pairings, and store them in csv files according to the basin number.

[0085] Step ②: Generate CSI theoretical values in the range of POD ∈ [0, 1] and FAR ∈ [0, 1], and generate 1000 groups of (POD, FAR, CSI) combinations.

[0086] Step ③: Quantify the corresponding relationship based on the binary classification model, and save the TP, TN, FP, and FN results of different basins under each threshold pairing as a csv file.

[0087] Step ④: Calculate the POD, FAR, and CSI of each basin, select the runoff threshold that maximizes the CSI, determine the optimal runoff threshold corresponding to the 90th precipitation threshold of the basin, and save the results.

[0088] In this embodiment, as shown in Figure 2 (a)- Figure 2 (e), the actual observed corresponding relationship is compared with the theoretical value, and the joint variation characteristics of POD, FAR, and CSI are plotted by Matplotlib. Among them, Figure 2 (a) The pyramid-shaped shadow 3D surface is the theoretical CSI value, the color scatter points are the measured corresponding relationship, and the black thick solid line is the theoretical CSI contour line; Figure 2 (b) to Figure 2 (e) show the change of the corresponding relationship of each basin with the runoff threshold, the measured CSI distribution shows a bounded convergence characteristic, the scatter points are concentrated in the interval of POD and 1-FAR being 0.4-1, and the maximum CSI value is about 0.6, indicating that it is difficult to achieve perfect prediction performance.

[0089] In this embodiment, Figure 3(a), (b), (c) show the correspondence between the upper 90% precipitation extreme and the upper 75%, 90%, 99% runoff peak value, respectively: at the 75% runoff threshold, the runoff response is more "sensitive", but the balance between false positives and false negatives of precipitation extreme is insufficient; at the 90% runoff threshold, the precipitation-runoff trigger-response accuracy is optimal, and the CSI performs best; at the 99% runoff threshold, the runoff response is extremely "strict", and the false negative rate of runoff peak value increases significantly, although the false positive rate is still high, but the overall accuracy falls.

[0090] In this embodiment, the No. 12013500 Willapa basin with the maximum AUCPR (0.799) is selected for analysis: as the runoff threshold increases, the number of runoff peak values decreases, the false negative rate decreases but the false positive rate increases, resulting in the simultaneous increase of POD, FAR and FB. The CSI increases between the 75th and 90th percentiles, and decreases between the 90th and 99th percentiles, indicating that there is an optimal corresponding relationship threshold.

[0091] In this embodiment, the performance chart and precision-recall curve are plotted for the Willapa basin, as shown in Figure 4 (a) and Figure 4 (b), the CSI value starts to rise from the 50th percentile, reaches a maximum of 0.601 at the 85th percentile, and then decreases to 0.2 at the 99th percentile. At this time, the POD is 0.732 (about 70% of the runoff peak values are triggered by precipitation extremes), the FAR is 0.229 (about 77% of the precipitation extremes correspond to runoff peak values), and the FB is 0.949. The number of precipitation and runoff peak values is close, and the precision-recall curve further verifies this conclusion.

[0092] In this embodiment, as shown in Figure 5 , the large sample test of the continental basin of the United States shows that at the 90th precipitation threshold, the optimal runoff threshold of 61% of the basins is higher than the 80th percentile (distributed in the west coast, the middle and southern parts, and the northeast), and the optimal runoff threshold of 22% of the basins is lower than the 60th percentile (mainly distributed in the Rocky Mountains, the north and the southeast).

[0093] In this example, as shown in Figure 6 , of the runoff peak values at the 90th percentile, 24% to 79% (median 54%) correspond to the 90th precipitation extreme, and of the 90th precipitation extreme, 34% to 80% (median 62%) correspond to the 90th runoff peak value; Figure 6 (a) shows the spatial distribution of POD of each basin in the United States under the optimal correspondence between precipitation extreme and runoff peak value, and the POD value of the west coast and the middle and southern basins is significantly higher due to the more direct precipitation-triggered runoff; Figure 6(b) The variation characteristics of POD with the runoff peak threshold (50th to 99th) under this optimal relationship are revealed. POD shows a stable upward trend with the increase of the threshold, which confirms that the underreporting of extreme rainfall values ​​is less when the threshold is high; Figure 6 (c) The spatial distribution of FAR under the optimal correspondence relationship is presented. Due to the complex hydrological processes in the Rocky Mountains and northern basins, the phenomenon of "precipitation extremes falsely triggering runoff peaks" is more prominent, and the FAR values ​​are relatively higher; Figure 6 (d) reflects the variation of FAR with the runoff peak threshold. FAR shows a decreasing trend as the threshold increases, because the high threshold can filter out false alarms with no runoff response; Figure 6 (e) The spatial distribution of CSI under the optimal correspondence is depicted. The west coast and northeastern basins have better CSI performance due to the clearer precipitation-runoff correspondence; Figure 6 (f) Verify the changing characteristics of CSI with the runoff peak threshold. CSI first increases and then decreases with the increase of the threshold, reaching the optimal value in the 80th~90th percentile range.

[0094] It can be understood that the present invention realizes the quantitative evaluation of the correspondence between precipitation extremes and runoff peaks through a threshold-mediated binary classification model: data-driven statistical modeling replaces the traditional hydrological model, and the input parameters only require precipitation and runoff time series, which is adaptable to the characteristics of different watersheds and has high flexibility; through threshold traversal and CSI optimization mechanism, the false alarm rate and missed alarm rate are effectively reduced, and the measured CSI in the Willapa watershed reaches 0.601, and the evaluation accuracy is significantly improved; it supports single-basin diagnosis and large-sample spatial analysis, and the calculation logic based on the Python library has low computing power requirements and strong result stability, which can directly serve hydrological forecasting and flood warning.

[0095] Example 4 like Figure 7 As shown, this embodiment proposes a correspondence evaluation system for extreme precipitation events and runoff peak events, which is applied to the correspondence evaluation method for extreme precipitation events and runoff peak events as described in the above embodiment, including: an acquisition module 100, an identification module 200, a construction module 300 and an evaluation module 400.

[0096] Among them, the acquisition module 100 is used to obtain precipitation time series data and runoff time series data; the identification module 200 is used to identify precipitation extreme events and runoff peak events respectively based on the precipitation time series data and the runoff time series data; the construction module 300 is used to construct a binary classification model for quantifying the correspondence between precipitation extreme events and runoff peak events under different thresholds; the evaluation module 400 is used to set the target precipitation extreme value, traverse the candidate runoff peak thresholds, and select the runoff peak threshold that makes the performance index of the binary classification model reach the optimal value as the optimal runoff response threshold corresponding to the target precipitation extreme value.

[0097] It should be noted that the aforementioned explanation of the embodiment of the method for evaluating the corresponding relationship between the precipitation extreme event and the runoff peak event is also applicable to the system for evaluating the corresponding relationship between the precipitation extreme event and the runoff peak event of this embodiment, and thus will not be repeated here.

[0098] In the description of the present specification, the description of the terms "one embodiment", "some embodiments", "an example", "a specific example", or "some examples" and the like means that the specific features, structures, materials or characteristics described in connection with the embodiment or example are included in at least one embodiment or example of the present application. In the present specification, the illustrative description of the above terms is not necessarily directed to the same embodiment or example. Moreover, the specific features, structures, materials or characteristics described can be combined in any one or N embodiments or examples in a suitable manner. In addition, the person skilled in the art can combine and combine the different embodiments or examples described in the present specification and the features of the different embodiments or examples without contradiction.

[0099] In addition, the terms "first", "second" are only for descriptive purposes and cannot be understood as indicating or implying relative importance or implicitly indicating the number of the indicated technical features. Therefore, the features defined with "first", "second" can explicitly or implicitly include at least one of the features. In the description of the present application, the meaning of "N" is at least two, for example, two, three, etc., unless otherwise explicitly specified.

[0100] Any process or method descriptions in flow charts or otherwise described herein can be understood as representing code modules, segments, or portions of code that include one or more executable instructions for implementing specific logic functions (or steps) in the process, and the various embodiments of the present application include additional implementations in which the order of steps can differ from those shown or discussed, including a step can occur at substantially the same time with other steps or can be performed in reverse order, and additional steps can be included, all of which have been contemplated to be within the scope of the present application.

[0101] It should be understood that the parts of the present application can be implemented in hardware, software, firmware or their combination. In the above embodiments, the N steps or methods can be implemented by software or firmware stored in the memory and executed by a suitable instruction execution system. As in another embodiment, if implemented in hardware, any of the following technologies known in the art or their combination can be used: discrete logic circuit with logic gate circuit for implementing logic functions on data signals, application specific integrated circuit with suitable combination logic gate circuit, programmable gate array, field programmable gate array, etc.

[0102] Those skilled in the art can understand that all or part of the steps of the method carried out by the above-mentioned embodiments can be completed by a program instructing related hardware, and the program can be stored in a computer readable storage medium. When the program is executed, it includes one of the steps of the method embodiments or a combination thereof.

[0103] Obviously, the above embodiments of the present application are only examples for clearly illustrating the present application, and are not intended to limit the implementation manner of the present application. Based on the above description, other different forms of changes or variations can be made by those skilled in the art. Here, all the implementation manners do not need to be exhausted. Any modification, equivalent replacement and improvement made within the spirit and principle of the present application shall be included in the protection scope of the claims of the present application.

Claims

1. A method for evaluating the correspondence between extreme precipitation events and peak runoff events, characterized in that: include: Obtain precipitation time series data and runoff time series data; identifying precipitation extreme value events and runoff peak events respectively based on the precipitation time series data and the runoff time series data; A binary classification model was constructed to quantify the correspondence between extreme precipitation events and peak runoff events at different thresholds. The target precipitation extreme value is set, the runoff peak candidate thresholds are traversed, and the runoff peak threshold that makes the performance index of the binary classification model reach the optimal value is selected as the optimal runoff response threshold corresponding to the target precipitation extreme value.

2. The method for evaluating the correspondence between extreme precipitation events and peak runoff events according to claim 1, characterized in that: Identifying precipitation extreme events based on the precipitation time series data, including: According to the following formula, precipitation extreme events are extracted from the precipitation time series data: Where, P is the precipitation identifier, Represents the study area c Time series of extreme precipitation events, Represents the study area c At the moment of precipitation, For the study area c At the moment The percentile of effective precipitation, Indicates the percentile threshold parameter of precipitation, ranging from 0 to 100.

3. The method for evaluating the correspondence between extreme precipitation events and peak runoff events according to claim 2, characterized in that: After identifying the extreme precipitation event, the method further includes: For the time series of extreme precipitation events Among the points that are continuous in time, the point with the maximum precipitation value is taken as a single extreme precipitation event; For the time series of extreme precipitation events The points with discontinuous time in the middle are directly determined as single extreme precipitation events; Single precipitation extreme events with a time interval not less than the threshold are extracted as independent precipitation extreme events.

4. The method for evaluating the correspondence between extreme precipitation events and peak runoff events according to claim 1, characterized in that: Identifying a runoff peak event based on the runoff time series data includes: Runoff peak events are extracted from the runoff time series data according to the following formula: Where, Q is the runoff identifier, Represents the study area c Time series of peak runoff events, Represents the study area c At the moment The runoff flow, For the study area c At the moment The percentile of effective runoff, Indicates the percentile threshold parameter of runoff, ranging from 0 to 100.

5. The method for evaluating the correspondence between extreme precipitation events and peak runoff events according to claim 4, characterized in that: After identifying a peak runoff event, the method further includes: Time series of the peak runoff events In , the runoff peak events that meet the following constraints are extracted as independent runoff peak events: Where, Represents the study area c the time interval between two peak runoff events; Represents the study area c the drainage area; and Represents the study area c Peak runoff i and peak runoff j Traffic volume; Represents the study area c Peak runoff i To peak runoff j The minimum flow rate between .

6. The method for evaluating the correspondence between extreme precipitation events and peak runoff events according to claim 1, characterized in that: A binary classification model is constructed to quantify the relationship between extreme precipitation events and peak runoff events under different thresholds, including: For the identified study area c The precipitation extreme value event time series is generated by marking the moment when the precipitation extreme value event occurs as 1 and the moment when it does not occur as 0, thus generating a precipitation binary classification sequence; For the identified study area c The time series of runoff peak events is generated by marking the moment when the runoff peak event occurs as 1 and the moment when it does not occur as 0, generating a runoff binary classification sequence; The precipitation binary classification sequence is used as the prediction item and the runoff binary classification sequence is used as the observation item. A 2×2 contingency table is constructed to count the states of simultaneous occurrence of precipitation extreme events and runoff peak events, occurrence of only precipitation extreme events, occurrence of only runoff peak events, and neither precipitation extreme events nor runoff peak events occur. A binary classification model of the corresponding relationship between precipitation extreme events and runoff peak events is obtained.

7. The method for evaluating the correspondence between extreme precipitation events and peak runoff events according to claim 6, characterized in that: The performance indicators of the binary classification model include the key success index and the harmonic fraction F1 score.

8. The method for evaluating the correspondence between extreme precipitation events and peak runoff events according to claim 7, characterized in that: The performance indicators of the binary classification model, including the key success index and the harmonic series F1 score, are calculated as follows: Where, CSI is the critical success index, which is used to quantify the overall level of the correspondence between extreme precipitation events and subsequent runoff peak events; POD is the hit rate, which is used to quantify the proportion of all runoff peak events that correctly correspond to extreme precipitation events; FAR is the false alarm rate, which is used to quantify the proportion of all precipitation peak events that fail to correctly correspond to runoff peak events; TP Indicates the number of times when extreme precipitation events and peak runoff events occur simultaneously; FP Indicates the number of states where only extreme precipitation events occur; FN Indicates the number of states where only peak runoff events occur; TN The count indicates the state where neither the extreme precipitation event nor the peak runoff event occurred; Where, and represent precision and recall respectively.

9. The method for evaluating the correspondence between extreme precipitation events and peak runoff events according to claim 8, characterized in that: Traverse the candidate runoff peak thresholds and select the runoff peak threshold that makes the performance index of the binary classification model reach the optimal value as the optimal runoff response threshold corresponding to the target precipitation extreme value, including: The runoff peak candidate threshold value corresponding to the maximum critical success index CSI and the harmonic series F1 score is used as the optimal runoff response threshold value corresponding to the target precipitation extreme value.

10. A system for evaluating the correspondence between extreme precipitation events and peak runoff events, characterized in that: include: Acquisition module, used to obtain precipitation time series data and runoff time series data; an identification module, configured to identify precipitation extreme value events and runoff peak events, respectively, based on the precipitation time series data and the runoff time series data; A construction module is used to construct a binary classification model for quantifying the corresponding relationship between extreme precipitation events and peak runoff events under different thresholds; The evaluation module is used to set the target precipitation extreme value, traverse the candidate runoff peak thresholds, and select the runoff peak threshold that makes the performance index of the binary classification model reach the optimal value as the optimal runoff response threshold corresponding to the target precipitation extreme value.

Images