Lumped hydrological model parameter grading calibration method considering spatial heterogeneity

Abstract

The invention discloses a lumped hydrological model parameter grading calibration method considering spatial heterogeneity. According to the method, multi-source data such as remote sensing and satellite monitoring soil humidity and evapotranspiration are introduced, a target function set considering spatial heterogeneity of hydrological state variables such as evapotranspiration and soil humidity is established, and a correlation matrix of local hydrological state variables and simulation values is deduced; on the basis, the lumped hydrological model is divided into a runoff production module, a confluence module and the like, and a lumped hydrological model multi-stage grading calibration framework considering spatial heterogeneity is provided. According to the method, the spatial heterogeneity of the hydrological state variables is considered, the accuracy of parameter calibration of the lumped hydrological model is improved, the problem that the classical lumped hydrological model does not consider the spatial-temporal variation of the hydrological state variables is solved, and the accuracy of parameter calibration is improved.

Description

Technical Field

[0001] The present invention belongs to the field of runoff forecasting, and more specifically, relates to a lumped hydrological model parameter classification calibration method considering spatial heterogeneity. Background Art

[0002] Calibration of hydrological model parameters is a key link in rainfall-runoff modeling. Accurate parameter estimation is an important basis for high-precision hydrological forecasting. With the continuous development of "3S" and computer technology, satellite, remote sensing, machine learning and other technologies have been widely used in hydrological forecasting modeling. Introducing large-scale raster data sets to carry out data assimilation and joint parameter calibration is an important technical means to improve the accuracy of hydrological forecasting.

[0003] Calibration of parameters based only on the flow process at the outlet of the basin may lead to large errors in the simulation of hydrological state variables. Large-scale, long-period, gridded soil moisture datasets provide detailed information on the spatiotemporal distribution of state variables such as evaporation and soil water, and are therefore widely used in the calibration of distributed or semi-distributed hydrological model parameters. However, existing studies generally apply gridded data products to distributed or semi-distributed hydrological models, and few studies focus on the application value of the spatiotemporal information covered by raster data products to conceptual lumped hydrological models. In addition, although there are technical attempts to introduce satellite and remote sensing data to calibrate the parameters of conceptual hydrological models, the arithmetic mean of the basin is generally used, ignoring the spatial dynamic information of the hydrological state variables, which may lead to potential errors.

[0004] Therefore, the existing conceptual lumped hydrological model parameter calibration method has the technical problem of not considering the spatial variability characteristics of the hydrological state variables, resulting in distortion in the estimation of hydrological model parameters. Summary of the invention

[0005] In view of the above defects or improvement needs of the prior art, the purpose of the present invention is to provide a lumped hydrological model parameter graded calibration method considering spatial heterogeneity, which aims to solve the technical problems that the spatial heterogeneity of the hydrological process is not considered in the existing lumped hydrological model parameter calibration and the hydrological state variables cannot be accurately simulated.

[0006] To achieve the above object, the present invention adopts the following technical solutions:

[0007] A lumped hydrological model parameter classification calibration method considering spatial heterogeneity includes:

[0008] S1. Introduce remote sensing, satellite monitoring evapotranspiration and soil moisture data, divide the basin into Thiessen polygons according to the rainfall stations in the basin, extract the average evapotranspiration of the target basin, soil water series and the hydrological model state variable series corresponding to the observation station;

[0009] S2. Select the runoff generation mode and calibrate the hydrological model parameters, extract the simulated values ​​of the watershed hydrological model state variables (soil moisture, evapotranspiration), deduce the correlation matrix between the local hydrological state variables and the simulated values, and select the hydrological station state variable sequence with the best correlation;

[0010] S3. Based on the optimal hydrological station state variables, a set of accuracy evaluation indicators covering objective functions such as evapotranspiration, soil water, and runoff is constructed;

[0011] S4. The lumped hydrological model is divided into a runoff generation module, a runoff confluence module, etc. Based on the above-mentioned set of accuracy evaluation indicators, a multi-stage hierarchical calibration framework of the lumped hydrological model considering spatial heterogeneity is obtained;

[0012] The calibration process of the proposed lumped hydrological model hierarchical calibration framework includes two stages: calibration of runoff module parameters and calibration of runoff confluence module parameters. Specifically, the runoff module parameters are first calibrated according to evapotranspiration and soil moisture; secondly, the runoff module parameters are kept unchanged and the confluence module parameters are calibrated according to the hydrological station flow; finally, the runoff module parameters are kept unchanged and the confluence module parameters are calibrated. The above calibration steps are repeated and iterated until the model converges.

[0013] Furthermore, the calculation formula for extracting the state variables of the basin hydrological model is:

[0014]

[0015] Where V = {W,…,E} is the set of hydrological state variables, W is soil moisture, and E is evapotranspiration; is the average value sequence of state variables (such as represent the basin-average soil moisture and evapotranspiration series, respectively); A is a representative observation station i The corresponding state variable sequence (i=1,2,…,N); N is the number of observation stations in the basin.

[0016] Furthermore, step S2 specifically includes:

[0017] S21. Establish a watershed hydrological model and select the optimal flow generation mode for the target watershed based on the watershed outlet flow target;

[0018] S22. According to the above hydrological model simulation results, extract the simulated value V of the state variable of the basin hydrological model sim On this basis, the correlation matrix R of local hydrological state variables, average values ​​and simulated values ​​is derived

[0019]

[0020] Where, V sim Simulated values ​​for hydrological state variables;

[0021] S23. Based on the correlation evaluation results, select the hydrological station state variable sequence V with the best correlation

[0022] V = max{r(X i ,X sim )}

[0023] Where V is the observed value of the station state variable that has the strongest correlation with the simulated value.

[0024] Furthermore, the state variable objective function is constructed based on the optimized state variable observation values:

[0025] F V =f(V,V sim )

[0026] In the formula, F V is the objective function that takes into account the spatial heterogeneity of state variables.

[0027] Furthermore, step S4 specifically includes:

[0028] S41. Based on the state variable objective function F considering spatial heterogeneity V Determine the parameters of the flow generation module;

[0029] S42. Keep the flow generation module parameters unchanged, according to the flow objective function F Q Determine the parameters of the confluence module;

[0030] S43. Keep the parameters of the confluence module unchanged, recalibrate the parameters of the flow generation module, repeat the above calibration process, and iterate repeatedly until the model converges.

[0031] In general, compared with the prior art, the above technical solution conceived by the present invention can achieve the following beneficial effects.

[0032] In order to solve the problem of multi-source uncertainty errors in the calibration of hydrological model parameters, the present invention introduces a large-scale raster data set constructed by satellite and remote sensing technology, derives the correlation matrix of local hydrological state variables and simulation values, constructs a set of objective functions covering evapotranspiration, soil moisture, and flow processes, and proposes a runoff classification calibration framework that considers the spatial heterogeneity of hydrological state variables. This solves the problem of distortion in the simulation of hydrological state variables in lumped hydrological models and improves the accuracy of parameter calibration. BRIEF DESCRIPTION OF THE DRAWINGS

[0033] Figure 1 A flow chart of a lumped hydrological model parameter classification calibration method considering spatial heterogeneity according to the present invention;

[0034] Figure 2A study area map provided for an embodiment of the present invention;

[0035] Figure 3 Thiessen polygon division for the study area;

[0036] Figure 4 It is the time series diagram of average soil moisture and local soil moisture in the study area;

[0037] Figure 5 This is the correlation matrix and significance level of local soil moisture and simulated values ​​from 2017 to 2020. DETAILED DESCRIPTION

[0038] In order to make the purpose, technical solutions and advantages of the present invention more clearly understood, the present invention is further described in detail below in conjunction with the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are only used to explain the present invention and are not intended to limit the present invention. In addition, the technical features involved in the various embodiments of the present invention described below can be combined with each other as long as they do not conflict with each other.

[0039] refer to Figure 1 The present invention provides a lumped hydrological model parameter classification calibration method considering spatial heterogeneity, comprising the following steps:

[0040] (1) Remote sensing and satellite monitoring data are introduced. First, the basin is divided into Thiessen polygons according to the rainfall stations in the basin, and the weight of each rainfall station is calculated to extract the average evapotranspiration, soil water series and hydrological model state variable series corresponding to the observation stations in the target basin;

[0041] (2) Select the runoff generation mode and calibrate the hydrological model parameters, extract the simulated values ​​of the state variables of the basin hydrological model, and deduce the correlation matrix between the local hydrological state variables and the simulated values, and select the hydrological station state variable sequence with the best correlation;

[0042] According to the preferred runoff mode described in step (2) and the hydrological model parameter calibration, taking soil moisture as an example, the soil moisture calculation expressions for the super-infiltration, full storage, and mixed runoff modes are as follows:

[0043] W t =W0+f t (1)

[0044]

[0045]

[0046] Where W t is the soil moisture at time t; f tis the soil infiltration rate during period t; WMM is the maximum water storage capacity of a single point in the basin; B is the coefficient of uneven distribution of tension water storage capacity; PE t is the net rainfall in period t; FA t is the infiltration amount from the surface to the soil during period t.

[0047] On this basis, the correlation matrix between local hydrological state variables and simulated values ​​is derived to explore the spatial variability of soil moisture in the basin. Taking soil moisture as an example, the correlation matrix expression of local soil moisture, average soil moisture and simulated values ​​is as follows:

[0048]

[0049] Where, W sim is the simulated value of the hydrological state variable, which can be calculated by equations (1) to (3); are the soil moisture observation values ​​corresponding to different stations; is the average soil moisture in the basin.

[0050] On this basis, the hydrological station state variable sequence W with the strongest correlation is selected, which is expressed as follows

[0051] W = max{r(W i ,W sim )} (5)

[0052] Where W is the observed value of the station state variable that has the strongest correlation with the simulated value.

[0053] (3) Based on the optimal hydrological station state variable W, a set of accuracy evaluation indicators for objective functions such as soil water and runoff is constructed;

[0054] According to step (3), the accuracy evaluation index expression of the objective function of soil water, runoff, etc. constructed is as follows:

[0055] F W =1-r(W sim ,W) (6)

[0056] F Q =1-E NS (7)

[0057] In the formula, F W is the objective function considering the spatial heterogeneity of soil moisture; F Q is the objective function of runoff evaluation, E NS is the Nash efficiency coefficient of the measured flow rate and the simulated value.

[0058] (4) The lumped hydrological model is divided into runoff generation module and runoff confluence module. Based on the above accuracy evaluation index set, a multi-stage hierarchical calibration framework of the lumped hydrological model considering spatial heterogeneity is proposed;

[0059] The calibration process of the lumped hydrological model hierarchical calibration framework in step (4) includes two stages: calibration of runoff module parameters and calibration of runoff module parameters. Specifically, firstly, the runoff module parameters are calibrated according to evapotranspiration and soil moisture; secondly, the runoff module parameters are kept unchanged and the runoff module parameters are calibrated according to the flow rate of the hydrological station; finally, the runoff module parameters are kept unchanged and the runoff module is calibrated, and the above calibration steps are repeated, and the iterations are repeated until the model converges.

[0060] Taking the Xunhe River Basin in the Hanjiang River sub-basin as an example, the runoff classification rate was determined considering the spatial heterogeneity of soil moisture. Figure 2 ), collected daily runoff, precipitation and evapotranspiration data in the Xunhe River Basin from 2001 to 2020, and carried out runoff forecast in the Xunhe River Basin with daily scale as the time step, of which 2001 to 2015 was the training period and 2017 to 2020 was the test period. The data in 2016 had large errors and were therefore not considered. The SMCI1.0 soil moisture dataset established by in-situ observation combined with machine learning algorithm was introduced, and the Thiessen polygons were divided according to the distribution of rainfall stations in the basin ( Figure 3 ), the weights of the Thiessen polygon ratios of each interval are shown in Table 1; further, the average soil moisture of the target basin and the local soil moisture series corresponding to the rainfall observation station are extracted ( Figure 4 ).

[0061] Table 1 Rainfall stations and weights in the Xunhe River basin above Xiangjiaping

[0062]

[0063] On this basis, the Xin'anjiang model was adopted as the hydrological modeling framework, and the super-infiltration, full storage, and vertical and lateral coupled mixed runoff modes were taken as alternative runoff modes. The results showed that the Xunhe River Basin was dominated by a mixed runoff mode. Therefore, the soil moisture simulated by the mixed runoff mode was adopted. Figure 5 The correlation matrix and significance level of local soil moisture and simulated values ​​during the test period (flood season from 2017 to 2020) are given. The results show that there are significant differences in the correlation of soil moisture between different sites. The soil moisture changes of geographically adjacent sites are relatively consistent. However, there are large spatial variability of soil moisture in some local areas. Table 2 gives the correlation coefficient and significance test results of soil moisture during the flood season from 2001 to 2020. The results show that compared with the average soil moisture in the basin, the correlation between the soil moisture at the site and the simulated soil moisture is more significant. When the soil moisture at all sites has a significant linear relationship with the simulated value, the correlation between the average soil moisture in the basin and the simulated value is also strong.

[0064] Table 2 Correlation evaluation results between simulated values ​​and local soil moisture and average soil moisture

[0065]

[0066] Note: The gray shading in the table indicates that the correlation coefficient between the soil moisture at this site and the simulated value is the largest, * indicates that the p value is less than 0.05, ** indicates that the p value is less than 0.01, *** indicates that the p value is less than 0.001, and the unmarked correlation coefficient indicates that there is no significant linear relationship between the soil moisture at this site and the simulated soil moisture.

[0067] Finally, the stations with the strongest correlation were selected to carry out the lumped model hierarchical calibration, and compared with the results of the classical calibration method that did not consider the spatial heterogeneity of soil moisture, the soil moisture correlation coefficient was used as the evaluation index, and the evaluation results are shown in Table 3. The results show that after considering the spatial heterogeneity of soil moisture, the average soil moisture correlation coefficient of the classical calibration method increased to 0.49, and after adopting the hierarchical calibration scheme considering the spatial heterogeneity of soil moisture, the average value of the soil moisture correlation coefficient was 0.54 and the median was 0.60. Therefore, the soil moisture correlation of the calibration results using the hierarchical calibration framework is stronger.

[0068] Table 3 Comparison of the calibration results between the runoff classification framework and the classical calibration method

[0069]

[0070]

[0071] Note: The grey shading in the table indicates that the correlation coefficient between the soil moisture at this site and the simulated value is better.

[0072] It will be easily understood by those skilled in the art that the above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions and improvements made within the spirit and principles of the present invention should be included in the protection scope of the present invention.

Claims

1. A lumped hydrological model parameter classification calibration method considering spatial heterogeneity, characterized in that: The following steps are involved: S1. Introduce remote sensing, satellite monitoring evapotranspiration and soil moisture data, divide the basin into Thiessen polygons according to the rainfall stations in the basin, extract the average evapotranspiration of the target basin, soil water series and the hydrological model state variable series corresponding to the observation station; S2. Select the runoff mode and calibrate the hydrological model parameters, extract the simulated values ​​of the state variables of the basin hydrological model, and deduce the correlation matrix between the local hydrological state variables and the simulated values; S3. Based on the optimal hydrological station state variables, a set of accuracy evaluation indicators covering the objective functions of evapotranspiration, soil water, and runoff is constructed; S4. Divide the lumped hydrological model into an evapotranspiration module, a runoff generation module, and a runoff confluence module, and obtain a multi-stage hierarchical calibration framework of the lumped hydrological model considering spatial heterogeneity based on the accuracy evaluation index set; The calibration process of the lumped hydrological model hierarchical calibration framework includes two stages: calibration of runoff module parameters and runoff confluence module parameters. First, the runoff module parameters are calibrated according to evapotranspiration and soil moisture; second, the runoff module parameters are kept unchanged and the runoff confluence module parameters are calibrated according to the hydrological station flow; finally, the runoff module is calibrated while the confluence module parameters are kept unchanged. The above calibration steps are repeated and iterated repeatedly until the model converges.

2. According to the method of claim 1, the lumped hydrological model parameter classification method considering spatial heterogeneity is characterized in that: The calculation formula for extracting the state variables of the basin hydrological model is: Where V = {W,…,E} is the set of hydrological state variables, W is soil moisture, and E is evapotranspiration; is the average value sequence of state variables (such as represent the basin-average soil moisture and evapotranspiration series, respectively); A is a representative observation station i The corresponding state variable sequence, i = 1, 2, ..., N; N is the number of observation stations in the basin.

3. According to claim 2, a lumped hydrological model parameter classification calibration method considering spatial heterogeneity is characterized in that: Step S2 specifically includes: S21. Establish a watershed hydrological model and select the optimal flow generation mode for the target watershed based on the watershed outlet flow target; S22. Extract the simulated value V of the state variable of the basin hydrological model according to the hydrological model simulation results sim On this basis, the correlation matrix R of local hydrological state variables, average values ​​and simulated values ​​is derived Where, V sim Simulated values ​​for hydrological state variables; S23. Based on the correlation evaluation results, select the hydrological station state variable sequence V with the best correlation V=max{r(X i ,X sim )} Where V is the observation sequence of the station state variable that has the strongest correlation with the simulated value.

4. A lumped hydrological model parameter classification calibration method considering spatial heterogeneity according to claim 3, characterized in that: Construct the state variable objective function based on the state variable observations: F V =f(V,V sim ) In the formula, F V is the objective function that takes into account the spatial heterogeneity of state variables.

5. A lumped hydrological model parameter classification calibration method considering spatial heterogeneity according to claim 4, characterized in that: The parameter calibration process of the graded calibration method is as follows: S41. Based on the state variable objective function F considering spatial heterogeneity V Determine the parameters of the flow generation module; S42. Keep the flow generation module parameters unchanged, according to the flow objective function F Q Determine the parameters of the confluence module; S43 keeps the parameters of the confluence module unchanged, recalibrates the parameters of the flow generation module, repeats the above calibration process, and iterates repeatedly until the model converges.

Images