A base flow simulation method based on a multiple linear regression model
By combining a multiple linear regression model with various baseflow segmentation methods, the problem of overestimation or underestimation in baseflow simulation was solved, achieving high-precision baseflow simulation under different geomorphic and climatic conditions and reducing uncertainty.
Patent Information
- Application Number
- CN202310477320.5
- Authority / Receiving Office
- CN · China
- Patent Type
- Patents(China)
- Current Assignee / Owner
- Filing Date
- 2023-04-28
- Publication Date
- 2025-11-21
- Estimated Expiration
- 2043-04-28
AI Technical Summary
Existing baseflow simulation methods generally suffer from overestimation or underestimation, resulting in high simulation uncertainty and an inability to accurately simulate the baseflow process.
A multiple linear regression model combined with various baseflow segmentation methods was adopted. By acquiring historical runoff data and catchment area data, a baseflow process curve was constructed, and the Nash efficiency coefficient and percentage deviation were used to evaluate the fitting accuracy, thereby reducing simulation uncertainty.
It improves the accuracy and universality of baseflow simulation, enabling accurate simulation of baseflow processes under different terrain and climate conditions, and reducing uncertainty.
Smart Images
Figure CN116502531B_ABST
Abstract
Description
TECHNICAL FIELD
[0001] The present application relates to the technical field of base flow process simulation, and particularly relates to a base flow simulation method based on a multiple linear regression model. BACKGROUND
[0002] Base flow is one of the important components of runoff, and the runoff during no rainfall is mainly supplied by base flow, which is crucial for river ecosystems and human water use. As the main source of runoff during dry season, base flow is closely related to the basic ecological maintenance of the basin, water quality, and domestic water supply. At the same time, the base flow process is also an important part of the water cycle in the basin, which participates in the water exchange process between surface runoff, unsaturated aquifer and groundwater. Base flow is mainly controlled by soil hydraulic parameters (i.e. river and aquifer conditions), precipitation and soil, land use characteristics, land use intensity and surface slope and other related factors. Understanding the contribution of base flow to runoff is crucial for studying the characteristics of runoff and water resources allocation in the basin. Under the condition that the basic hydrological data is met, using various base flow calculation methods, mastering and analyzing the base flow process of the basin can benefit the stable water source and ecological environment protection of the basin. Therefore, how to ensure the accuracy of the base flow process segmentation result through theoretical analysis and numerical simulation is a hot and difficult problem in the field of hydrology.
[0003] At present, for basins with different geomorphic characteristics and climate conditions, scholars have proposed different base flow segmentation methods. Traditionally, base flow segmentation methods mainly include numerical simulation method, graphical method, analytical method, isotope method and hydrological model method. Each base flow segmentation method has its own characteristics, but there is a common phenomenon of overestimation or underestimation, and it cannot accurately simulate the base flow process path. SUMMARY
[0004] To solve the technical deficiencies of the existing base flow simulation, the purpose of the present application is to provide a base flow simulation method based on a multiple linear regression model, so as to solve the overestimation or underestimation phenomenon of each base flow simulation result, reduce the uncertainty of base flow simulation, and thereby improve the simulation accuracy of base flow.
[0005] To achieve the above goal, a base flow simulation method based on a multiple linear regression model comprises the following steps:
[0006] A base flow simulation method based on a multiple linear regression model, characterized in that it comprises the following steps:
[0007] S1) obtaining historical runoff data, rainfall daily scale data and catchment area of each hydrological station of the basin;
[0008] S2) based on the runoff data and catchment area data obtained in S1), using multiple base flow segmentation methods to obtain different base flow total flow data;
[0009] S3) Using the data obtained in S2), multiple baseflow indices (BFIs) are simulated using various baseflow segmentation methods, and the average value of the baseflow indices is calculated. S4) Based on the data obtained in S3, simulated data is obtained using a multiple linear regression model. The base flow process is constructed using this simulated data, and a simulated base flow process curve is formed.
[0010] In addition, the following steps are also included:
[0011] S5) Based on the measured rainfall sequence data and historical runoff data obtained in S1), the actual baseflow process curve is obtained; in S5), the measured rainfall sequence data and historical runoff data are selected based on the principle that if there are no rainfall records for 15 consecutive days before a certain date, the runoff corresponding to that date is approximately equal to the baseflow. The runoff data is selected based on this principle, and the baseflow verification data is obtained.
[0012] S6) Compare the baseflow process curves obtained in S4) and S5) to determine the fitting accuracy and goodness of fit. The method for determining the fitting accuracy and goodness of fit in S6) is as follows: Use the Nash-Sutcliffe Efficiency coefficient (NSE) and the Percent bias (Pbias) to evaluate each baseflow simulation value. The range of NSE is from negative infinity to 1. When it is close to 1, the higher the quality of the simulation is considered. The closer Pbias is to 0, the better the performance of the simulation is considered.
[0013] Prioritize the baseflow index BFI and the average baseflow index in S3). The calculation formula is as follows:
[0014]
[0015] In the formula, V B V represents the total base flow rate within the specified time period. S The total runoff volume within the time period, BFI1, BFI2, ..., BFI n These are the baseflow index values simulated by n methods, where n is the number of baseflow segmentation methods, and the BFI value is between 0 and 1.
[0016] Prior to this, the construction method of the Multiple Linear Regression Model (MLRM) in S4 is as follows:
[0017] BF MLRM = a1BF1 + a2BF2 + ... + a n BF n
[0018]
[0019] BF = a BF + b BF + c BF + d BF + e BF + f BF + g BF + h BF + i BF + j BF + k BF + l BF + m BF + n BF + o BF + p BF + q BF + r BF + s BF + t BF + u BF + v BF + w BF + x BF + y BF + z BF MLRM BF = a BF + b BF + c BF + d BF + e BF + f BF + g BF + h BF + i BF + j BF + k BF + l BF + m BF + n BF + o BF + p BF + q BF + r BF + s BF + t BF + u BF + v BF + w BF + x BF + y BF + z BF n BF = a BF + b BF + c BF + d BF + e BF + f BF + g BF + h BF + i BF + j BF + k BF + l BF + m BF + n BF + o BF + p BF + q BF + r BF + s BF + t BF + u BF + v BF + w BF + x BF + y BF + z BF n BF = a BF + b BF + c BF + d BF + e BF + f BF + g BF + h BF + i BF + j BF + k BF + l BF + m BF + n BF + o BF + p BF + q BF + r BF + s BF + t BF + u BF + v BF + w BF + x BF + y BF + z BF n BF = a BF + b BF + c BF + d BF + e BF + f BF + g BF + h BF + i BF + j BF + k BF + l BF + m BF + n BF + o BF + p BF + q BF + r BF + s BF + t BF + u BF + v BF + w BF + x BF + y BF + z BF
[0020] Preferably, the S2 base flow separation method is the HYSEP method, the one-parameter digital filter method or the recursive digital filter method.
[0021] By using the above technical means, the application has the following advantages: 1) the application is based on three widely used base flow separation methods, uses a multiple linear regression model to establish a new base flow simulation method and process, reduces the uncertainty of base flow simulation, and thus improves the simulation accuracy of base flow; 2) the base flow simulation method established by the application can improve the universality of the base flow simulation method without being affected by the differences in hydro-meteorological conditions and underlying surface conditions of a basin; and 3) the runoff corresponding to the consecutive 15-day date without rainfall record defined by the application is equal to the base flow, which makes up for the lack of measured base flow values in the base flow simulation evaluation. BRIEF DESCRIPTION OF DRAWINGS
[0022] Figure 1 is a specific flowchart of the application;
[0023] Figure 2 is a base flow simulation accuracy comparison chart of a hydrological site 1;
[0024] Figure 3 is a base flow simulation accuracy comparison chart of a hydrological site 2;
[0025] Figure 4 is a runoff and base flow hydrograph in a certain period of a hydrological site 1;
[0026] Figure 5 is a runoff and base flow hydrograph in a certain period of a hydrological site 2. DETAILED DESCRIPTION
[0027] The technical solutions of the application will be further described in detail below with reference to the drawings. In order to highlight the advantages of the application, the three widely used base flow separation methods, namely the HYSEP method, the one-parameter digital filter method (OPDF) and the recursive digital filter method (RDF), are used as cases for specific implementation.
[0028] As Figure 1As shown, a base flow simulation method based on a multiple linear regression model of the application comprises the following steps:
[0029] S1) Data collection: Collecting measured daily runoff data, daily rainfall data and rainfall area data of hydrological stations in a certain basin from 1981 to 1995.
[0030] S2) Base flow segmentation: Using three widely used base flow segmentation methods, HYSEP, OneParameter Digital Filter (OPDF) and Recursive Digital Filter (RDF), to obtain different base flow total data.
[0031] S3) Model construction: First, calculate the base flow indices BFI of each base flow simulation value obtained by the above three methods, which are 0.393, 0.41 and 0.418 for hydrological station 1, and 0.464, 0.483 and 0.484 for hydrological station 2; second, calculate the average value of the base flow indices of the two hydrological stations The results are 0.407 and 0.477, respectively.
[0032] Finally, establish a Multiple Linear Regression Model (MLRM) with the following construction method:
[0033] BF MLRM = a1BF HYSEP +a2BF OPDF +a3BF3
[0034]
[0035]
[0036]
[0037] According to the calculation, the coefficients of the three linear regression models of the two hydrological stations are 0.5, 0.11, 0.39 and 0.5, 0.05, 0.45, respectively. Therefore, the base flow simulation model established is divided into: 0.5BF HYSEP +0.11BF OPDF +0.39BF3 and 0.5BF HYSEP +0.05BF OPDF +0.45BF3.
[0038] S4) Model verification: The baseflow of two hydrological sites was simulated by the linear regression model established in the above steps, and the runoff corresponding to the date of no rainfall record in the previous 15 days was defined as the baseflow according to the measured rainfall sequence data, and the discontinuous baseflow sequence obtained was taken as the baseflow verification value and the simulation value for comparison. As shown in Figure 2 and Figure 3 The baseflow simulated by the MLRM used in the application is better than the results of the other three baseflow separation methods in terms of Nash efficiency coefficient and percentage bias (Pbias). In particular, when MLRM is applied to hydrological site 1, NSE reaches more than 0.7, and Pbias is about -13%, and when MLRM is applied to hydrological site 1, NSE reaches more than 0.8, and Pbias is about -5%. The simulation evaluation value of MLRM is much better than that of the other three simulation methods.
[0039] Figure 4 and Figure 5 The runoff and baseflow hydrograph of two hydrological sites in a certain period are illustrated, and it can be seen from the figure that the baseflow process simulated by MLRM can well capture the peak and valley changes of baseflow, and its value is between the simulation values of the three methods, which well makes up for the overestimation or underestimation phenomenon of each baseflow separation method, and greatly reduces the uncertainty of baseflow simulation.
[0040] The above only describes the example implementation of the application and is not used to limit the application. The number of baseflow separation methods in the application can also be specifically customized according to different research areas. Any modification, equivalent replacement, improvement, etc. within the scope of the claims of the application should be within the protection scope of the application.
Claims
1. A baseflow simulation method based on a multiple linear regression model, characterized by, Comprise the following steps: S1) obtaining historical runoff data, measured rainfall sequence data and catchment area of each hydrological station of the river basin; S2) based on the data obtained in S1), using multiple base flow separation methods to obtain different base flow total data based on historical runoff data and catchment area data; S3) calculating a plurality of base flow indices BFI from the data obtained by S2) and finding a base flow index average value S3) Base flow index BFI and average of base flow indices The formula for calculating the base flow index BFI is as follows: where V B is the total amount of base flow in the time period, V S is the total amount of runoff in the time period, BFI1, BFI2, …, BFI n are the base flow index values simulated by n methods respectively, n is the number of base flow separation methods, and the value of BFI is between 0 and 1; S4) based on the data obtained in S3), using a multiple linear regression model to obtain simulation data, and constructing a base flow process by using the simulation data to form a simulated base flow process curve; Wherein: the construction method of the multiple linear regression model MLRM is as follows: BF MLRM = a1BF1+ a2BF2+... + a n BF n where BF MLRM is the sequence of base flows simulated by the MLRM method, BF n is the sequence of base flows simulated by the n-th base flow segmentation method, a n is the regression coefficient of BF n .
2. The baseflow simulation method based on a multiple linear regression model according to claim 1, characterized by, Further comprising the following steps, S5) according to the measured rainfall sequence data and historical runoff data obtained in S1), obtaining the actual base flow process curve; S6) comparing the base flow process curves obtained in S4) and S5) respectively to determine the fitting accuracy and the degree of coincidence.
3. The baseflow simulation method based on a multiple linear regression model according to claim 2, characterized by, The method for determining the fitting accuracy and the degree of coincidence in S6) is as follows: using the Nash efficiency coefficient NSE and the percentage bias Pbias to evaluate each base flow simulation value; the range of NSE is from negative infinity to 1, when it approaches 1, the higher the quality of simulation is determined, and the closer Pbias is to 0, the better the performance of simulation is determined.
4. The baseflow simulation method based on a multiple linear regression model according to claim 2, characterized by, In S5), the selection of measured rainfall sequence data and historical runoff data, if there is no rainfall record for 15 consecutive days before a certain date, the runoff corresponding to the date is equal to the base flow, and the runoff data is selected according to this principle to obtain the base flow verification data.
5. The baseflow simulation method based on a multiple linear regression model according to any one of claims 1 to 4, characterized by, The base flow separation method in S2) is HYSEP method, single parameter digital filtering method and recursive digital filtering method.