Differential calibration method of hydrological model coupling rainfall-melting-snow runoff process based on runoff coefficient grading

Abstract

The application discloses a hydrological model differentiation calibration method based on runoff coefficient grading and coupling rainfall-melting-snow-runoff process, and comprises the following steps: collecting continuous hydrological data of a target basin and preprocessing; dividing out flood events of each flood according to the shape of the flow process line to form a flood sample set; calculating the runoff coefficient of each flood; determining a grading threshold according to the statistical distribution of all sample runoff coefficients, dividing the flood sample set into low and high runoff coefficient subsets; introducing a snow melting module into a traditional hydrological model; independently calibrating the model coupled with the snow melting module by using the two subsets respectively; obtaining the antecedent soil moisture of each flood, and determining the antecedent soil moisture threshold corresponding to the runoff coefficient grading threshold; in the flood forecasting stage, the antecedent soil moisture of the basin is obtained in real time, compared with the threshold, and the corresponding parameter set is dynamically selected and called for forecasting; the application improves the precision and stability of flood simulation and forecasting.

Description

Technical Field

[0001] This invention relates to the field of hydrological simulation and flood forecasting technology, specifically to a differential calibration method for hydrological models based on runoff coefficient classification of coupled rainfall-snowmelt-runoff processes. Background Technology

[0002] Flood forecasting serves as a crucial technical support for flood control and disaster prevention, and the Xin'anjiang model is widely used due to its accurate representation of runoff generation mechanisms in humid regions. However, existing technologies still have certain limitations, primarily reflected in the simplistic parameter calibration methods.

[0003] In reality, the efficiency of precipitation-to-runoff conversion is constrained by multiple factors, including soil moisture, infiltration capacity, and vegetation cover. This leads to significant differences in runoff processes during different flood events, even under similar rainfall conditions. Using a single parameter set for calibration fails to comprehensively reflect the runoff generation and concentration patterns under varying hydrological conditions, increasing forecast bias. Furthermore, the runoff coefficient, as a core indicator of precipitation-to-runoff conversion efficiency, directly reflects the runoff generation characteristics of a watershed under different hydrological conditions. A low runoff coefficient often corresponds to a drier underlying surface, with soil infiltration and evapotranspiration dominating, resulting in a significant decrease in runoff generation efficiency. Conversely, a high runoff coefficient indicates a wet or saturated watershed, reduced infiltration capacity, and a significantly enhanced efficiency in converting precipitation and snowmelt into runoff. Therefore, the runoff coefficient effectively distinguishes between different hydrological scenarios, providing a scientific basis for differentiated model calibration.

[0004] Therefore, it is necessary to classify flood events based on runoff coefficients and conduct independent parameter calibration for sample sets of different levels, so that the model parameters are more consistent with the physical mechanisms under different hydrological conditions. Summary of the Invention

[0005] Purpose of the invention: The purpose of this invention is to provide a differentiated calibration method for hydrological models that couple rainfall-snowmelt-runoff processes based on runoff coefficient classification. This classification calibration method solves the problem of insufficient forecast accuracy of the Xin'anjiang model in differentiated hydrological scenarios such as snowmelt.

[0006] Technical solution: The differential calibration method for hydrological models based on runoff coefficient classification and coupled rainfall-snowmelt-runoff processes described in this invention includes the following steps:

[0007] (1) Collect continuous hydrological data of the target watershed, including rainfall, flow, evapotranspiration and temperature data; preprocess the data and remove outliers; based on the morphological characteristics of the flow process line, segment independent flood events to form a flood sample set;

[0008] (2) For each flood, calculate its runoff coefficient; based on the statistical distribution of all sample runoff coefficients, determine the runoff coefficient classification threshold and divide the flood sample set into a low runoff coefficient sample subset and a high runoff coefficient sample subset.

[0009] (3) Introduce a snow melting module into the traditional hydrological model; the snow melting module can dynamically divide rainfall and snowfall according to the temperature threshold, and calculate the snow melting amount based on the degree-day factor and the rainfall-snow melting factor. The snow melting water volume is superimposed with the rainfall volume and then input into the hydrological model for runoff calculation.

[0010] (4) Using the low runoff coefficient sample subset and the high runoff coefficient sample subset respectively, the hydrological model of the coupled snow melting module was independently calibrated to obtain two sets of model parameter sets that are adapted to low runoff conditions and high runoff conditions respectively.

[0011] (5) Obtain the soil moisture content of each flood event in the preceding period; determine the associated soil moisture content threshold based on the flood event corresponding to the runoff coefficient classification threshold;

[0012] (6) During the flood forecasting stage, the soil moisture content of the watershed in the early stage is obtained in real time; the real-time soil moisture content in the early stage is compared with the soil moisture content threshold, and the low runoff condition parameter set or the high runoff condition parameter set is dynamically selected for flood forecasting based on the comparison results.

[0013] Furthermore, in step (2), the runoff coefficient classification threshold is determined by statistically analyzing the empirical cumulative distribution function of the runoff coefficients of all flood samples, and the runoff coefficient value corresponding to the cumulative frequency of 50% is taken as the classification threshold.

[0014] Furthermore, in step (3), the snow melting module linearly divides the total precipitation into rainfall and snowfall based on the relationship between the average temperature of the basin and the preset critical temperatures for snowfall and rainfall.

[0015] Furthermore, in step (4), the parameter calibration adopts a global optimization algorithm with the objective function of maximizing the Nash efficiency coefficient NSE. The optimization is performed on the low runoff coefficient sample subset and the high runoff coefficient sample subset respectively to obtain two sets of differentiated optimal parameter combinations.

[0016] Furthermore, in step (5), the soil moisture content in the early stage is defined as the average soil moisture content during a specific period before the flood rises, which is obtained by inversion of remote sensing data.

[0017] Furthermore, in step (6), the dynamic calling rule is as follows: when the real-time soil moisture content is greater than the soil moisture content threshold, the high runoff condition parameter set is called; when the real-time soil moisture content is less than or equal to the soil moisture content threshold, the low runoff condition parameter set is called.

[0018] The present invention provides a computer-readable storage medium having a computer program stored thereon, wherein the computer program, when executed by a processor, implements any of the methods described herein.

[0019] An electronic device according to the present invention includes a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor executes the program to implement any of the methods described herein.

[0020] Beneficial Effects: Compared with existing technologies, this invention has the following significant advantages: This invention overcomes the limitations of single-parameter calibration and adapts to diverse hydrological conditions. Addressing the problem that existing Xin'anjiang models use a single parameter set, making it difficult to match runoff differences caused by varying hydrological conditions, this invention constructs a differentiated calibration system through runoff coefficient grading. It generates suitable parameter sets for low / high runoff coefficient scenarios, reducing forecast bias caused by a one-size-fits-all approach with fixed parameters. It improves adaptability and forecast accuracy for complex underlying surfaces. Through dual-core optimization of "snowmelt module + grading calibration," this invention can effectively address the differences in hydrological processes on complex underlying surfaces (such as areas with both arid / humid conditions and areas with snow cover). Combined with the global optimization capability of the SCE-UA algorithm, it further improves the accuracy and stability of flood process simulation, providing more reliable technical support for basin flood control scheduling and disaster risk prevention. Attached Figure Description

[0021] Figure 1 This is a flowchart of the present invention;

[0022] Figure 2 This is a distribution diagram of the runoff coefficient of the flood events according to the present invention. Detailed Implementation

[0023] The technical solution of the present invention will be further described below with reference to the accompanying drawings.

[0024] like Figure 1 As shown, this embodiment of the invention provides a differential calibration method for a hydrological model of coupled rainfall-snowmelt-runoff processes based on runoff coefficient classification, including the following steps:

[0025] Step S1: Flood event data collection and preprocessing, including the following steps:

[0026] Step S1.1: Data Collection and Compilation. Collect and compile at least 30 years of continuous daily hydrological data for the target watershed, specifically including: measured rainfall data from meteorological stations within the watershed, outflow process data at the outlet section, potential evapotranspiration data, and measured daily temperature data. Ensure that the data covers different hydrological scenarios such as snowmelt floods and torrential floods to provide basic support for subsequent analysis.

[0027] Step S1.2: Data Preprocessing. Consistency checks are performed on the measured rainfall, flow, evapotranspiration, and temperature data, followed by outlier removal: For rainfall data, a threshold of three times the standard deviation of the rainfall sequence at each station within the basin is used to remove abnormal rainfall data exceeding this threshold from stations at the center of the rainstorm and remote stations in the basin; for flow, evapotranspiration, and temperature data, the same three-standard-deviation principle is followed to remove obvious outliers to ensure data reliability.

[0028] Step S1.3: Flood Event Segmentation. Based on the morphological characteristics of the flow process line, the continuous rainfall-runoff data series of the target watershed is segmented into flood events, and the hydrological data corresponding to each flood event is extracted to form an independent flood event data series.

[0029] Step S2: Runoff coefficient calculation; including the following steps:

[0030] Step S2.1: Total runoff depth ( ) represents the total runoff of a flood event ( ) and drainage area ( The ratio of () to () in mm is given by the formula:

[0031] ;

[0032] In the formula: for Real-time measured flow rate (m) 3 / s); , These represent the start and end times (s) of the floodwaters, respectively; A is the catchment area (km²). 2 ).

[0033] Step S2.2: Total Precipitation The average precipitation in the basin during the corresponding period of the flood event is expressed in mm. The formula is:

[0034] ;

[0035] In the formula: For the first Total rainfall (mm) at each rain gauge station; For the first The controlled area of ​​each rain gauge station (km²) 2 ); n represents the number of rain gauge stations within the watershed.

[0036] Step S2.3: Calculate the runoff coefficient for each flood event. Runoff coefficient Defined as the total runoff depth of a flood event With total precipitation The overall ratio directly quantifies the conversion efficiency of precipitation to runoff, and the calculation formula is as follows:

[0037] ;

[0038] Step S3: Classification and partitioning of the flood sample set; including the following steps:

[0039] Step S3.1: Sort the runoff coefficient α of all flood samples using the statistical distribution method:

[0040] ;

[0041] in, The total number of samples, Indicates the sorted order of the first... A sample of runoff coefficients. The corresponding empirical cumulative distribution function is defined as:

[0042] ;

[0043] Step S3.2: Determine the runoff coefficient threshold To balance sample size and the diversity of hydrological scenarios, the α value corresponding to a cumulative frequency of 50% was taken as... :

[0044] ;

[0045] Step S3.3: Based on the runoff coefficients of all samples The statistical distribution, based on the runoff coefficient threshold We classify runoff coefficients into low and high levels, and differentiate between hydrological scenarios corresponding to low and high precipitation-runoff conversion efficiencies. Let the sample set be... Each flood event Corresponding runoff coefficient According to the grading threshold The sample set is divided into:

[0046] ;

[0047] ;

[0048] when When this occurs, the event is classified as having a low runoff coefficient, corresponding to a dry period in the watershed, with little pre-concentration rainfall, no snowmelt, or very little snowmelt. In such cases, the soil has strong infiltration capacity, and most precipitation is consumed in infiltration and evapotranspiration, resulting in a small contribution to runoff. If the event occurs at a certain time, it is classified as a high runoff coefficient event, corresponding to a wet season in the watershed, with abundant pre-rainfall and large snowmelt. The soil is close to saturation, with low infiltration, and most of the precipitation (including snowmelt) is converted into runoff.

[0049] Step S4: Construct and couple the Xin'anjiang model of the snow melting module; details are as follows:

[0050] Step S4.1: Dynamic classification of precipitation types. Based on the average temperature of the watershed ( The relationship between total precipitation ( ) and critical temperature threshold, and the total precipitation ( ) is broken down into snowfall ( ) and rainfall ( The formula is as follows:

[0051] ;

[0052] ;

[0053] In the formula: The critical temperature for snowfall (based on experience, it can be taken as -5 to 0℃, below which all precipitation is snow); The critical temperature for rainfall (based on experience, it can be taken as 5-15℃; above this temperature, all precipitation is rain). ≤ ≤ It was raining and snowing at the time, and it was divided according to a linear ratio.

[0054] Step S4.2: Snowmelt calculation and snow water equivalent update. Calculate the snowmelt amount based on the degree-day factor when the temperature is higher than the hour. The amount of snow melt reflects the effect of temperature on snow melting:

[0055] ;

[0056] In the formula: It is a day-to-day factor, related to snow type and vegetation cover.

[0057] Step S4.3: Calculate snowmelt caused by rainfall. Considering the additional melting effect of rainfall on snow cover, this is positively correlated with rainfall amount:

[0058] ;

[0059] In the formula: Snowmelt factor;

[0060] Step S4.4: Calculate the total snowmelt. The total snowmelt is limited by the current snow storage and is updated hourly with snow water equivalent:

[0061] ;

[0062] ;

[0063] In the formula, Subscript: snow water equivalent Indicates the current hour. Indicates the next hour

[0064] Step S4.5: Calculate the snowmelt runoff. Assuming that the runoff generation mechanism of snowmelt water is similar to that of rainfall, the total liquid water volume is obtained by superimposing the snowmelt water volume and the rainfall volume, and then input into the Xin'anjiang full-storage runoff generation model that considers the snowmelt module to calculate the total runoff of snowmelt and rainfall.

[0065] Step S5: Classification and calibration process based on runoff coefficient; including:

[0066] Step S5.1: Parameter calibration uses the SCE-UA algorithm as the calibration method to simulate flow. Compared with the measured flow rate The objective function for maximizing the Nash efficiency coefficient NSE is:

[0067] ;

[0068] In the formula, m is the number of calculation steps for each flood event; The average measured flow rate (m³) 3 / s);

[0069] For sample sets of different runoff coefficient categories, the objective function is defined as the corresponding NSE set:

[0070] ;

[0071] ;

[0072] in, In the parameter vector Next, the The Nash efficiency coefficient for a flood event.

[0073] Step S5.2: Classification Calibration. Calibrate the low runoff coefficient sample set to obtain the low-level parameter set. Calibrate the high runoff coefficient sample set to obtain a high-level parameter set. In the sample sets with low and high runoff coefficients, we search for the method that maximizes the average value of events in each category. The parameters are combined to ensure the model exhibits robust performance under different hydrological conditions. The optimal calibration result can be expressed as:

[0074] ;

[0075] ;

[0076] Step S5.3: Model validation indicators include relative error of flood volume and discharge. relative error of peak flow Peak time error Nash efficiency coefficient (NSE) and determinism coefficient The formula is as follows:

[0077] ;

[0078] ;

[0079] ;

[0080] ;

[0081] ;

[0082] In the formula, To simulate flood volume, This refers to the actual measured flood volume; To simulate flood volume, This refers to the actual measured flood volume; To simulate peak occurrence time, This represents the actual peak time.

[0083] Step S6: Determine the soil moisture content threshold; details are as follows:

[0084] Step S6.1: For each flood event, define "early soil moisture content" as the average soil moisture content T hours before the flood rises (T=72 hours, which can be determined according to the watershed characteristics), used to characterize the initial wetting state of runoff generation:

[0085] ;

[0086] in, The value represents the soil moisture content in the preceding period, and n represents the number of valid observation samples within the T-hour period. Indicates the first [day] before the flood rises. The soil volumetric water content at any given time is obtained from remote sensing inversion values.

[0087] Step S6.2: Apply the determined runoff coefficient threshold. The soil moisture content threshold in the early stage is the soil moisture content threshold corresponding to the runoff coefficient threshold of the corresponding field. This is used to characterize the initial wetting state under different runoff generation conditions:

[0088] ;

[0089] Step S7: Calling up differentiated calibration parameters in flood forecasting. During the flood forecasting phase, based on processed remote sensing products (such as SMAP), the anterior soil moisture content of the watershed is acquired in real time. When the observed soil moisture content in the early stage is greater than the threshold... At that time, a high-level parameter set is invoked for simulation; when the current soil moisture content is less than or equal to a threshold... Simulation is performed by calling a lower-level parameter set. The parameter set calling logic is explicitly defined using piecewise functions:

[0090] ;

[0091] By using the above rules, the model parameters are accurately matched with the real-time initial wet state of the watershed, solving the problem of insufficient adaptability of traditional fixed parameter sets to complex runoff conditions. Ultimately, the differentiated calibration and dynamic switching of model parameters are achieved, significantly improving the accuracy of flood simulation and the reliability of forecasts.

[0092] Example 1:

[0093] Step S1: Sample Data Screening

[0094] Step S1.1: Data collection and compilation. Collect daily measured rainfall, flow, evapotranspiration, and temperature data for the target watershed from 1960 to 2019, a total of 60 years.

[0095] Step S1.2: Data Preprocessing. Consistency checks are performed on the measured rainfall, flow, evapotranspiration, and temperature data. For rainfall data, outliers exceeding the threshold are removed from the central and remote stations of the rainstorm center, using three times the standard deviation of the rainfall sequence for each station within the basin as the threshold. Flow, evapotranspiration, and temperature data also undergo outlier removal following the same three-standard-deviation principle. Finally, all processed data are integrated into a dataset table with a unified time-series format.

[0096] Step S1.3: Flood Event Segmentation. Based on the morphological characteristics of the flow process line, the continuous rainfall-runoff data sequence of the target watershed is segmented into flood events. The rainfall-runoff data corresponding to each flood event is extracted to form an independent flood event data sequence. After verification and screening, 30 representative flood events are finally extracted as samples for model calibration and validation. All samples are named according to the start date of the flood event.

[0097] Step S2: Runoff Coefficient Calculation. For each flood event, the average precipitation and corresponding runoff of the basin were first calculated, and then the runoff coefficient for each event was calculated. The calculation results are shown in Figure 2. The runoff coefficients for all events ranged from 0.26 to 0.77, with a large distribution range. This characteristic indicates that the basin has historically experienced both low-runoff states with relatively dry soil and strong infiltration, and high-runoff states with relatively wet soil and sufficient runoff.

[0098] Step S3: Hierarchical partitioning of the sample set:

[0099] Step S3.1: Perform statistical distribution analysis on the runoff coefficient α of the 30 extracted flood events, and determine the runoff coefficient classification threshold based on hydrological significance. Sort the α values ​​of all samples:

[0100] ;

[0101] Its corresponding empirical cumulative distribution function is defined as:

[0102] ;

[0103] in, It is 30.

[0104] Step S3.2: To balance the sample size and the differences in hydrological scenarios, the α value corresponding to a cumulative frequency of 50% is taken as... :

[0105]

[0106] Calculations showed that the 50th percentile of the runoff coefficient in the 30 samples was 0.52. Following the principle of clearly distinguishing hydrological states while ensuring a small difference in the sample size between the two groups, 0.52 was ultimately determined as the runoff coefficient classification threshold. .

[0107] Step S3.3: Based on the runoff coefficients of all samples The statistical distribution, based on the runoff coefficient threshold Classification of low and high runoff coefficients:

[0108] ;

[0109] ;

[0110] like Figure 2 As shown, when If the flow rate is low, the event is classified as having a low runoff coefficient. If the event is classified as having a high runoff coefficient, then that event is categorized as high runoff coefficient. Through this classification, the 30 flood samples were ultimately divided into two groups: a low runoff coefficient sample set containing 14 flood events, and a high runoff coefficient sample set containing 16 flood events. The two groups have relatively balanced sample sizes, suitable for both low-runoff and high-runoff hydrological scenarios in the watershed.

[0111] Step S4: Construct and couple the snow melting module of the Xin'an River model

[0112] Step S3.1: Construct and apply the Xin'anjiang model with a snowmelt module in the target watershed for calibration of all 30 floods. The model contains 20 parameters, with 4 new snowmelt module parameters added compared to the original Xin'anjiang model (as shown in Table 1) to accurately simulate the impact of snowmelt replenishment on watershed runoff.

[0113] Step S5: Classification and calibration process based on runoff coefficient

[0114] Step S5.1: Parameter calibration uses the SCE-UA algorithm as the calibration method to simulate flow. Compared with the measured flow rate The objective function is to maximize the Nash efficiency coefficient NSE.

[0115] Step S5.2: Grade-specific calibration. Calibration is performed separately for the low runoff coefficient sample set (14 floods) to obtain a parameter set suitable for low runoff scenarios. Individual calibration was performed on a high runoff coefficient sample set (16 floods) to obtain a parameter set suitable for high runoff scenarios. This enables parameter adaptation to different hydrological conditions.

[0116] Step S5.3: Verification and Comparative Analysis of Calibration Results. Select the relative error of flood volume / discharge. relative error of peak flow Peak time error Nash efficiency coefficient (NSE) and determinism coefficient As a verification indicator, calibrated 30 floods using unified parameters were first performed, and the results were compared with those of the graded calibration. As shown in Tables 2 and 3, overall, the simulation accuracy of the graded calibration is significantly better than that of the unified calibration, with outstanding optimization of core indicators: Statistical mean analysis shows that the average absolute value of the relative error of flood volume in the graded calibration is 10.93%, a decrease of 9.56% compared to the 20.58% of the unified calibration; the average absolute value of the relative error of flood peak is 14.59%, a decrease of 12.45% compared to the 27.04% of the unified calibration, with both error indicators achieving a reduction of nearly half. Furthermore, the average NSE of the graded calibration reaches 0.68, an increase of 0.15 compared to the 0.53 of the unified calibration, significantly improving the agreement between the simulation results and the measured values; although... The mean value decreased slightly from 0.84 to 0.83, but the improvement of the error index and NSE better reflects the technical advantages of the classification calibration.

[0117] Comparisons of typical flood events yielded similar results. Taking the low-yield flood No. 19990807 as an example, under unified calibration, the relative error of flood volume was -27.09%, the relative error of flood peak was -26.60%, and the NSE was only 0.22, indicating low simulation accuracy. After graded calibration, the absolute values ​​of the relative errors of flood volume and flood peak were reduced to 4.83% and -0.81%, respectively, and the NSE increased to 0.82, significantly enhancing simulation accuracy. In another high-yield flood, No. 19810523, under unified calibration, the relative errors of flood volume and flood peak were -21.53% and -35.56%, respectively, and the NSE was 0.63. After graded calibration, the absolute values ​​of the two errors were -3.33% and -19.89%, respectively, showing a significant reduction in simulation error and an increase in NSE to 0.92. Both simulation accuracy and engineering practicality were significantly improved.

[0118] Step S6: Determine the soil moisture content threshold.

[0119] Step S6.1: Obtain watershed soil moisture content using SMAP remote sensing inversion data. Define the anterior soil moisture content for each flood event. The average soil moisture content 72 hours before the flood rises is used to characterize the initial wet state of runoff generation.

[0120] ;

[0121] in, Indicates the first [day] before the flood rises. Soil volumetric water content at a given time.

[0122] Step S6.2: For a runoff coefficient threshold of 0.52, the previous soil moisture content threshold is the soil moisture content threshold for the corresponding event. This is used to characterize the initial wetting state under different runoff generation conditions:

[0123] ;

[0124] Step S7: Calling up the differential rate setting parameters in flood forecasting.

[0125] During the flood forecasting phase, the watershed soil moisture content is obtained in real time based on the processed SMAP remote sensing product. The parameter set calling logic is clearly defined using piecewise functions:

[0126] ;

[0127] When the observed soil moisture content Greater than the threshold At that time, a high-level parameter set is called for simulation; current soil moisture content Less than or equal to the threshold At that time, a lower-level parameter set is used for simulation. Through the above rules, the model parameters are matched with the real-time pre-flood wet conditions of the watershed, thus achieving accurate flood forecasting.

[0128] Table 1. Main parameters of the Xin'anjiang model with coupled snow melting module

[0129] ;

[0130] Table 2. Evaluation of Model Indicators for Sample Grading and Calibration

[0131] ;

[0132] Table 2 shows only 12 typical flood events.

[0133] Table 3. Evaluation of Model Indicators for Sample Grading and Calibration

[0134] ;

[0135] Table 3 shows only 12 typical flood events.

Claims

1. A differential calibration method for a hydrological model of coupled rainfall-snowmelt-runoff processes based on runoff coefficient classification, characterized in that, Includes the following steps: (1) Collect continuous hydrological data of the target watershed, including rainfall, flow, evapotranspiration and temperature data; preprocess the data and remove outliers; Based on the morphological characteristics of the flow process line, independent flood events are segmented to form a flood sample set; (2) Calculate the runoff coefficient for each flood event; Based on the statistical distribution of runoff coefficients in all samples, the runoff coefficient classification threshold is determined, and the flood sample set is divided into a low runoff coefficient sample subset and a high runoff coefficient sample subset. (3) Introduce a snow melting module into the traditional hydrological model; the snow melting module can dynamically divide rainfall and snowfall according to the temperature threshold, and calculate the snow melting amount based on the degree-day factor and the rainfall-snow melting factor. The snow melting water volume is superimposed with the rainfall volume and then input into the hydrological model for runoff calculation. (4) Using the low runoff coefficient sample subset and the high runoff coefficient sample subset respectively, the hydrological model of the coupled snow melting module was independently calibrated to obtain two sets of model parameter sets that are adapted to low runoff conditions and high runoff conditions respectively. (5) Obtain the soil moisture content of each flood event in the preceding period; determine the associated soil moisture content threshold based on the flood event corresponding to the runoff coefficient classification threshold; (6) During the flood forecasting stage, the soil moisture content of the watershed in the early stage is obtained in real time; the real-time soil moisture content in the early stage is compared with the soil moisture content threshold, and the low runoff condition parameter set or the high runoff condition parameter set is dynamically selected for flood forecasting based on the comparison results.

2. The method for differential calibration of a hydrological model based on runoff coefficient classification for coupled rainfall-snowmelt-runoff processes according to claim 1, characterized in that, In step (2), the runoff coefficient classification threshold is determined by the empirical cumulative distribution function of the runoff coefficients of all flood samples, and the runoff coefficient value corresponding to the cumulative frequency of 50% is taken as the classification threshold.

3. The method for differential calibration of a hydrological model based on runoff coefficient classification for coupled rainfall-snowmelt-runoff processes according to claim 1, characterized in that, In step (3), the snow melting module linearly divides the total precipitation into rainfall and snowfall based on the relationship between the average temperature of the basin and the preset critical temperatures for snowfall and rainfall.

4. The method for differential calibration of a hydrological model based on runoff coefficient classification for coupled rainfall-snowmelt-runoff processes according to claim 1, characterized in that, In step (4), the parameter calibration adopts a global optimization algorithm with the objective function of maximizing the Nash efficiency coefficient NSE. The optimization is performed on the low runoff coefficient sample subset and the high runoff coefficient sample subset respectively to obtain two sets of differentiated optimal parameter combinations.

5. The method for differential calibration of a hydrological model based on runoff coefficient classification for coupled rainfall-snowmelt-runoff processes according to claim 1, characterized in that, In step (5), the soil moisture content in the early stage is defined as the average soil moisture content during a specific period before the flood rises, which is obtained by inversion of remote sensing data.

6. The method for differential calibration of a hydrological model based on runoff coefficient classification of coupled rainfall-snowmelt-runoff processes according to claim 1, characterized in that, In step (6), the dynamic calling rule is as follows: when the real-time soil moisture content is greater than the soil moisture content threshold, the high runoff condition parameter set is called; when the real-time soil moisture content is less than or equal to the soil moisture content threshold, the low runoff condition parameter set is called.

7. A computer-readable storage medium having a computer program stored thereon, characterized in that, When the computer program is executed by a processor, it implements the method as described in any one of claims 1-6.

8. An electronic device comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, characterized in that, When the processor executes the program, it implements the method as described in any one of claims 1-6.

Images