A method to improve the accuracy of water storage changes based on weighted statistical correction hydrological model

Abstract

The present invention discloses a method for improving the accuracy of water storage change based on weighted statistical correction hydrological model, including: determining K hydrological models; solving and obtaining the Pearson linear correlation coefficient corresponding to each hydrological model according to the model value output by each hydrological model and the reference value corresponding to each model value; screening and obtaining M preferred hydrological models whose Pearson linear correlation coefficient is greater than a set threshold; determining the terrestrial water storage change in the Haihe River Basin output by each preferred hydrological model and the weight of each preferred hydrological model; performing weighted fusion processing according to the terrestrial water storage change in the Haihe River Basin output by each preferred hydrological model and the weight of each preferred hydrological model, and obtaining the terrestrial water storage change Merge in the Haihe River Basin after weighted fusion processing. The present invention fully considers the consistency and accuracy information between multiple hydrological models and GRACE, and the fusion of multi-source hydrological models can effectively improve the accuracy of estimating TWSA.

Description

Technical Field

[0001] The present invention belongs to the cross-technical fields of satellite gravity, hydrology, etc., and in particular relates to a method for improving the accuracy of water reserve changes based on weighted statistical correction of hydrological models. Background Art

[0002] The Haihe River Basin, which spans mountainous areas and plains, is one of the largest urban agglomerations and food production bases in China, with wheat and corn as the main crops. In the past 20 years, the Haihe River Basin has experienced dramatic changes in land type, with a large amount of land not suitable for farming being used to grow crops. However, changes in land cover have a direct impact on the water cycle, such as a significant increase in water use for human activities. Groundwater overexploitation and climate change have led to a reduction in terrestrial water storage (TWS) in the Haihe River Basin, with overexploitation being particularly severe in the plains. Therefore, exploring the temporal and spatial variations of terrestrial water storage in the Haihe River Basin is of great significance to the sustainable development of local water resources.

[0003] Accurate quantification of TWS is indispensable for food supply, energy production, daily life, social development, etc., so it is necessary to monitor the dynamic changes of TWS. The traditional method of monitoring terrestrial water storage changes (TWSA) is mainly to deploy monitoring wells. This method is time-consuming and labor-intensive, and has the limitations of scarce and uneven distribution of monitoring points. It is difficult to accurately display TWSA over a large area. In recent years, hydrological remote sensing has developed rapidly. The GRACE satellite launched in 2002 has been widely used to monitor large-scale terrestrial water resources. Domestic and foreign scholars have used GRACE data to estimate water resource changes, such as monitoring terrestrial water storage changes, groundwater storage changes, drought and flood disasters, etc.

[0004] In recent years, the use of hydrological models to assess TWSA has also seen an unprecedented growth. To date, dozens of hydrological models have been used to analyze water storage changes. TWSA is mainly simulated by land surface models (LSMs) and global hydrological water resource models (GHWRMs). The definition of LSMs originally came from the simulation of land surface processes in the atmospheric cycle and was used to calculate water flux from land to the atmosphere, such as the Global Land Data Assimilation System (GLADS) model. GHWRMs were established to cope with global water shortages and are mainly based on water balance equations, such as the WGHM and PCR-GLOBWB models. The main difference between LSMs and GHWRMs is that the former emphasizes the material basis, including water and energy balance, while the latter emphasizes the empirical method of water resource budget. In addition, LSMs only consider climate factors, while GHWRMs add the impact of climate change and human activities to water storage changes, including industrial and agricultural water use, domestic water use, and reservoir operation.

[0005] At present, with the application of hydrological models in water storage changes, scholars have compared many models with GRACE data to evaluate the feasibility and reliability of hydrological models, such as related studies at the global scale and regional scale. However, in northern China, most studies focus on the North China Plain, ignoring the mountainous areas, and rarely involve the entire Haihe River Basin. Although some studies focus on the status of groundwater, drought, and evapotranspiration in the Haihe River Basin, there are few studies involving the spatiotemporal laws and causes of TWSA and individual components. In addition, most of these studies are based on GRACE or a single model, ignoring the differences between different hydrological models or between GRACE and hydrological models. Due to the differences in GRACE satellite data versions, global models, specific water storage calculation methods, and research periods, it is necessary to conduct a comprehensive and systematic analysis of TWSA in the Haihe River Basin. Summary of the invention

[0006] The technology of the present invention solves the problem: Overcoming the shortcomings of the prior art, providing a method for improving the accuracy of water storage changes based on statistical correction hydrological model weighting, fully considering the consistency and accuracy information between multiple hydrological models and GRACE, and fusing multi-source hydrological models to effectively improve the accuracy of estimating TWSA.

[0007] In order to solve the above technical problems, the present invention discloses a method for improving the accuracy of water storage change based on statistical correction hydrological model weighting, comprising:

[0008] Determine K hydrological models;

[0009] According to the model values ​​output by each hydrological model and the reference values ​​corresponding to each model value, the Pearson linear correlation coefficient corresponding to each hydrological model is calculated;

[0010] M optimal hydrological models with Pearson linear correlation coefficient greater than the set threshold are screened out;

[0011] Determine the changes in terrestrial water storage in the Haihe River Basin output by each preferred hydrological model and the weights of each preferred hydrological model;

[0012] According to the terrestrial water storage changes in the Haihe River Basin output by each preferred hydrological model and the weights of each preferred hydrological model, a weighted fusion process is performed to obtain the terrestrial water storage changes in the Haihe River Basin after weighted fusion processing Merge.

[0013] In the above method for improving the accuracy of water storage change based on weighting of statistically corrected hydrological model, the hydrological model includes:

[0014] Noah: land surface model;

[0015] VIC: Variable Infiltration Capacity Macroscale Model, variable infiltration capacity macro model;

[0016] CLSM2.0: Catchment Land Surface Model Version 2.0, Catchment Land Surface Model Version 2.0;

[0017] CLSM2.1: Catchment Land Surface Model Version 2.1, Catchment Land Surface Model Version 2.1;

[0018] WGHM: Water Global Assessment and Prognosis hydrological Model, global assessment and prediction hydrological model;

[0019] PCR-GLOBWB: PCRaster Global Water Balance, grid global water balance model;

[0020] CLM: Community Land Model, public land surface model.

[0021] In the above method for improving the accuracy of water storage change based on weighted statistical correction hydrological model, the Pearson linear correlation coefficient corresponding to each hydrological model is calculated according to the model value output by each hydrological model and the reference value corresponding to each model value, including:

[0022] Get the model value M of the i-th month output by any k-th hydrological model k,i , and the average value of the model value of month i output by the kth hydrological model

[0023] Determine M k,i The corresponding reference value O k,i , and the average value of the reference value of i months O k,i ;

[0024] The Pearson linear correlation coefficient r of the kth hydrological model is obtained by the following formula (1): k :

[0025]

[0026] Wherein, i=1,2,…,n, n represents the total number of months; k=1,2,…,K.

[0027] In the above method for improving the accuracy of water storage changes based on weighted statistical correction hydrological model, the threshold is set to 0.9.

[0028] In the above method for improving the accuracy of water storage change based on weighted statistical correction hydrological model, the terrestrial water storage change of the Haihe River Basin output by each preferred hydrological model is determined, including:

[0029] Get the soil water storage SMS output by any j-th optimal hydrological model j , groundwater reserves GWS j 、 Snow Water Equivalent SnWS j , Vegetation Water Storage CWS j and surface water storage SWS j ;

[0030] The global terrestrial water storage TWS corresponding to the jth optimal hydrological model is obtained by formula (1): j :

[0031] TWS j =SMS j +GWS j +SnWS j +CWS j +SWS j ···(2)

[0032] According to formula (2), combined with the following formula (3), the global terrestrial water storage change TWSA corresponding to the jth optimal hydrological model is calculated: j :

[0033]

[0034] in, and They represent the mean terrestrial water storage, soil water storage, groundwater storage, snow water equivalent, vegetation water storage and surface water storage from 2004 to 2009 corresponding to the jth optimal hydrological model; SMSA j GWSA j 、 SnWSA j 、CWSA j and SWSA j They represent the changes in soil water storage, groundwater storage, snow water equivalent, vegetation water storage, and surface water storage corresponding to the jth optimal hydrological model;

[0035] Ignore SnWSA j 、CWSA j and SWSA j, and obtain the terrestrial water storage change TWSA0 of the Haihe River Basin corresponding to the jth optimal hydrological model j :

[0036] TWSA0 j = SMSA j +GWSA j ···(4)

[0037] Among them, j=1,2,…,M.

[0038] In the above method for improving the accuracy of water storage change based on weighted statistical correction hydrological models, the weight of each preferred hydrological model is determined, including:

[0039] Determine the Pearson linear correlation coefficient r corresponding to the jth optimal hydrological model j ;

[0040] Determine the root mean square error RMSE corresponding to the jth optimal hydrological model j ;

[0041] According to r j and RMSE j , and the weight w of the jth optimal hydrological model is obtained by solving j .

[0042] In the above method of improving the accuracy of water storage change based on statistical correction hydrological model weighting, r j The solution formula is as follows:

[0043]

[0044] Among them, M j,i represents the model value of the i-th month output by the j-th optimal hydrological model, represents the average value of the model value of month i output by the jth preferred hydrological model, O j,i Indicates that j,i The corresponding reference value, Represents the average value of the reference value of month i; i = 1, 2, …, n, where n represents the total number of months.

[0045] In the above method of improving the accuracy of water storage change based on statistical correction hydrological model weighting, RMSE j The solution formula is as follows:

[0046]

[0047] In the above method of improving the accuracy of water storage change based on weighted statistical correction hydrological model, w j The solution formula is as follows:

[0048] wj =r j +1 / RMSE j ···(7).

[0049] In the above method of improving the accuracy of water storage change based on weighted statistical correction hydrological model, the solution formula of Merge is as follows:

[0050]

[0051] The present invention has the following advantages:

[0052] Due to the uncertainty of the hydrological model, the simulation results in a single grid have certain differences. In order to further improve the accuracy of the hydrological model, the present invention analyzes the spatiotemporal laws of TWSA in the Haihe River Basin and its sub-regions (including provinces, cities, mountainous areas and plains), and for the first time uses the GRACE mascon and the root mean square error of the hydrological model and the terrestrial water storage change trend of the correlation weighted model to construct a new statistical correction hydrological model weighting method. This method fully considers the consistency and accuracy information between multiple hydrological models and GRACE, and the fusion of multi-source hydrological models can effectively improve the accuracy of estimating TWSA. BRIEF DESCRIPTION OF THE DRAWINGS

[0053] Figure 1 is a flowchart of the steps of a method for improving the accuracy of water storage change based on weighted statistical correction hydrological model in an embodiment of the present invention;

[0054] Figure 2 2(a) is a schematic diagram of the geographical location of the Haihe River Basin in an embodiment of the present invention; wherein 2(a) is a provincial-level sub-region: Beijing, Tianjin, Hebei, Shanxi, Shandong, and Henan; 2(b) is a mountainous area and a plain;

[0055] Figure 3 is a schematic diagram of TWSA trend in the Haihe River Basin inverted by JPL and CSR in an embodiment of the present invention;

[0056] Figure 4 4(a) is a schematic diagram of GRACE data and precipitation distribution after decomposition by the STL method in an embodiment of the present invention; wherein 4(a) is the original GRACE data, 4(b) is the trend term, 4(c) is the seasonal term, and 4(d) is the residual term;

[0057] Figure 5Schematic diagram of TWSA spatial distribution in a study area of ​​a hydrological model and reference value inversion in an embodiment of the present invention; wherein, 5(a) is Noah, 5(b) is VIC, 5(c) is CLSM2.1, 5(d) is CLSM2.0, 5(e) is WGHM, 5(f) is PCR-GLOBWB, 5(g) is CLM, and 5(h) is Reference; the average value of JPL and CSR is the TWSA reference value, the color band represents the TWSA trend amplitude, and the dots represent areas without significant trends;

[0058] Figure 6 6 is a schematic diagram of TWSA time series inverted by a hydrological model and reference values ​​in the Haihe River Basin and its sub-regions in an embodiment of the present invention; wherein 6(a) is the Haihe River Basin, 6(b) is Beijing, 6(c) is Tianjin, 6(d) is Hebei, 6(e) is Shanxi, 6(f) is Shandong, and 6(g) is Henan; the shaded area represents the standard deviation of JPL and CSR;

[0059] Figure 7 It is a schematic diagram of the spatial distribution of TWSA calculated by WGHM, PCR-GLOBWB and Merge in an embodiment of the present invention;

[0060] Figure 8 8(a) is a schematic diagram of the SMSA spatial distribution of the study area inverted by a hydrological model and reference value in an embodiment of the present invention; wherein, 8(a) is Noah, 8(b) is VIC, 8(c) is CLSM2.1, 8(d) is CLSM2.0, 8(e) is WGHM, 8(f) is PCR-GLOBWB, 8(g) is CLM, and 8(h) is Reference; the average value of Noah, VIC and CLM is the SMSA reference value, the color band represents the SMSA trend amplitude, and the dots represent the area without significant trend;

[0061] Fig. 9 9 is a schematic diagram of SMSA time series inversion of a hydrological model and reference values ​​in the Haihe River Basin and its sub-regions in an embodiment of the present invention; wherein 9(a) is HRB, 9(b) is BJ, 9(c) is TJ, 9(d) is HB, 9(e) is SX, 9(f) is SD, and 9(g) is HN; the shaded area represents the standard deviation of Noah, VIC, and CLM;

[0062] Fig.1010(a) is a schematic diagram of the spatial distribution of GWSA in the study area of ​​a hydrological model and reference value inversion in an embodiment of the present invention; wherein, 10(a) is CLSM2.1, 10(b) is CLSM2.0, 10(c) is WGHM, 10(d) is PCR-GLOBWB, 10(e) is CLM, and 10(f) is Reference; the GWSA reference value is the residual between the TWSA reference value and the SMSA reference value, the color band represents the GWSA trend amplitude, and the dots represent areas without significant trends;

[0063] Fig.11 1 is a schematic diagram of the GWSA time series inverted by a hydrological model and reference values ​​in the Haihe River Basin and its sub-regions in an embodiment of the present invention; wherein 11(a) is HRB, 11(b) is BJ, 11(c) is TJ, 11(d) is HB, 11(e) is SX, 11(f) is SD, and 11(g) is HN; the shaded area represents the root mean square error of the TWSA reference value and the SMSA reference value;

[0064] Fig.12 12(a) is a schematic diagram of a time series of water storage changes of a hydrological model and a reference value in different plain and mountainous areas in an embodiment of the present invention; wherein 12(a) is a plain TWSA, 12(b) is a mountainous area TWSA, 12(c) is a plain SMSA, 12(d) is a mountainous area SMSA, 12(e) is a plain GWSA, and 12(f) is a mountainous area GWSA;

[0065] Fig.13 It is a schematic diagram of trend differences between a hydrological model and a corresponding reference value for different water reserve changes in an embodiment of the present invention;

[0066] Fig.14 14(a) is a schematic diagram of TWSA estimated by WGHM, PCR-GLOBWB, Merge and reference values ​​in four grids in an embodiment of the present invention; wherein 14(a) is a grid, 14(b) is G176, 14(c) is G209, 14(d) is G259, and 14(e) is G277;

[0067] Fig.15 15(a) is a schematic diagram of TWSA correlation between hydrological models and GRACE in different periods in an embodiment of the present invention; 15(a) is 2003-2009, and 15(b) is 2010-2014. DETAILED DESCRIPTION

[0068] In order to make the objectives, technical solutions and advantages of the present invention more clear, the embodiments disclosed in the present invention will be further described in detail below with reference to the accompanying drawings.

[0069] Traditional methods for estimating TWSA generally use the simulation results of a single hydrological model, but due to model uncertainty, there are certain differences in the model output results. Therefore, the present invention assigns weights to the selected hydrological model according to the consistency and accuracy between the model and GRACE, and combines multi-source model data to improve the accuracy of estimating TWSA.

[0070] like Figure 1 In this embodiment, the method for improving the accuracy of water storage change based on statistical correction hydrological model weighting includes:

[0071] Step 101, determining K hydrological models.

[0072] In this embodiment, the hydrological model includes but is not limited to: Noah: land surface model; VIC: Variable Infiltration Capacity Macroscale Model, variable infiltration capacity macroscale model; CLSM2.0: Catchment Land Surface Model Version 2.0, basin surface model 2.0 version; CLSM2.1: Catchment Land Surface Model Version 2.1, basin surface model 2.1 version; WGHM: Water Global Assessment and Prognosis hydrological Model, global assessment and prediction hydrological model; PCR-GLOBWB: PCRaster Global Water Balance, grid global water balance model; CLM: Community Land Model, public land surface model.

[0073] Step 102: Calculate the Pearson linear correlation coefficient corresponding to each hydrological model based on the model value output by each hydrological model and the reference value corresponding to each model value.

[0074] In this embodiment, first, the model value M of the i-th month output by any k-th hydrological model is obtained. k,i , and the average value of the model value of month i output by the kth hydrological model

[0075] Then, determine M k,i The corresponding reference value O k,i , and the average value of the reference value of i months O k,i .

[0076] Finally, the Pearson linear correlation coefficient r of the kth hydrological model is obtained by the following formula (1): k :

[0077]

[0078] Wherein, i=1,2,…,n, n represents the total number of months; k=1,2,…,K.

[0079] It should be noted that the reference values ​​corresponding to the model values ​​output by each hydrological model can be obtained from the GRACE mascon data, that is, M k,i The corresponding reference value O k,i .

[0080] Step 103, screening and obtaining M preferred hydrological models whose Pearson linear correlation coefficient is greater than a set threshold.

[0081] In this embodiment, the threshold value may be set to 0.9, that is, r obtained by solving formula (1) k Compared with 0.9, the hydrological models with a Pearson linear correlation coefficient greater than 0.9 are retained as the preferred hydrological models, and a total of M preferred hydrological models are obtained. For example, M can be equal to 2, that is, the two hydrological models of WGHM and PCR-GLOBWB are screened from the seven hydrological models of Noah, VIC, CLSM2.0, CLSM2.1, WGHM, PCR-GLOBWB and CLM as the preferred hydrological models.

[0082] Step 104, determining the change in terrestrial water storage in the Haihe River Basin output by each preferred hydrological model and the weight of each preferred hydrological model.

[0083] Preferably, the method for obtaining the change of terrestrial water storage in the Haihe River Basin output by each preferred hydrological model is as follows:

[0084] Get the soil water storage SMS output by any j-th optimal hydrological model j , groundwater reserves GWS j 、 Snow Water Equivalent SnWS j , Vegetation Water Storage CWS j and surface water storage SWS j . Where j = 1, 2,…, M.

[0085] The global terrestrial water storage TWS corresponding to the jth optimal hydrological model is obtained by formula (1): j :

[0086] TWS j =SMS j +GWS j +SnWS j +CWS j +SWS j···(2)

[0087] According to formula (2), combined with the following formula (3), the global terrestrial water storage change TWSA corresponding to the jth optimal hydrological model is calculated: j :

[0088]

[0089] in, and They represent the mean terrestrial water storage, soil water storage, groundwater storage, snow water equivalent, vegetation water storage and surface water storage from 2004 to 2009 corresponding to the jth optimal hydrological model; SMSA j GWSA j 、 SnWSA j 、CWSA j and SWSA j They respectively represent the changes in soil water storage, groundwater storage, snow water equivalent, vegetation water storage and surface water storage corresponding to the jth optimal hydrological model.

[0090] For the Haihe River Basin, the changes in snow water equivalent and surface water storage are relatively small and can be ignored. Furthermore, Rodell et al. pointed out that the seasonal amplitude of gravity changes caused by changes in biomass is about 5 mm, which is still an order of magnitude smaller than the current GRACE error (2 cm), so the changes in vegetation water storage can also be ignored. In other words, SnWSA can be ignored. j 、CWSA j and SWSA j , and obtain the terrestrial water storage change TWSA0 of the Haihe River Basin corresponding to the jth optimal hydrological model j :

[0091] TWSA0 j = SMSA j +GWSA j ···(4)

[0092] Preferably, the weights of the preferred hydrological models are obtained as follows:

[0093] Determine the Pearson linear correlation coefficient r corresponding to the jth optimal hydrological model j :

[0094]

[0095] Among them, M j,i represents the model value of the i-th month output by the j-th optimal hydrological model, represents the average value of the model value of month i output by the jth preferred hydrological model, O j,i Indicates that j,i The corresponding reference value, Represents the average value of the reference value of month i; i = 1, 2, …, n, where n represents the total number of months.

[0096] Determine the root mean square error RMSE corresponding to the jth optimal hydrological model j :

[0097]

[0098] According to r j and RMSE j , and the weight w of the jth optimal hydrological model is obtained by solving j :

[0099] w j =r j +1 / RMSE j ···(7)

[0100] Among them, the root mean square error and Pearson linear correlation coefficient, as effective evaluation indicators of changes in terrestrial water storage, have been widely used to study the linear relationship between different variables, and can reflect the degree of deviation between simulated values ​​and true values: the closer the Pearson linear correlation coefficient is to 1, the smaller the root mean square error is, indicating that the model is more reliable; when the Pearson linear correlation coefficient is in the range of 0 to 1, the variables are positively correlated; when the Pearson linear correlation coefficient is in the range of -1 to 0, the two variables are negatively correlated.

[0101] Step 105, performing weighted fusion processing according to the terrestrial water storage change in the Haihe River Basin output by each preferred hydrological model and the weight of each preferred hydrological model, to obtain the terrestrial water storage change in the Haihe River Basin after weighted fusion processing Merge.

[0102] In this embodiment, the solution formula of Merge is as follows:

[0103]

[0104] Based on the above embodiments, the verification process of the method for improving the accuracy of water storage change based on weighted statistical correction hydrological model described in the present invention is described below.

[0105] (1) Study area

[0106] The geographical location of the research area of ​​this invention is as follows Figure 2 The Haihe River Basin covers an area of ​​329,000km 2The Haihe River Basin is located at 111°E~121°E east longitude and 34°N~43°N north latitude, spanning eight provinces (autonomous regions and municipalities), including Beijing, Tianjin, Hebei, Shanxi, Shandong, Henan, Liaoning and Inner Mongolia Autonomous Region. Figure 2 (a) According to the terrain characteristics, the Haihe River Basin can also be divided into two main sub-regions: plains and mountains, as shown in Figure 2 (b) is shown. The basin has a temperate East Asian monsoon climate and is a semi-humid area with an average annual temperature of 1.5-14℃, an average annual precipitation of 500-600mm, and an annual evaporation of 900-1400mm. The Haihe River Basin is the most important grain production base in northern China. However, in recent years, overexploitation of water resources has led to negative changes in the ecological environment, such as groundwater depletion, drought, land subsidence, and deterioration of fresh water quality. According to the "Haihe River Basin Water Resources Bulletin", groundwater accounts for about 75% of the total water consumption. As of 2010, the cumulative unsustainable groundwater overexploitation has exceeded 90 billion cubic meters. Five of China's eight super-large groundwater overexploitation areas are in the Haihe River Basin.

[0107] (2) Data

[0108] 2.1) GRACE gravity satellite data

[0109] The present invention uses two latest GRACE mascon data to analyze the TWSA in the Haihe River Basin: JPL (https: / / grace.jpl.nasa.gov / data / get-data / jpl_global_mascons / #opennewwindow) and CSR (http: / / www.csr.utexas.edu / grace / ). The data processing of mascon products includes: using satellite laser ranging to invert the second order (C 20 ) coefficients, the first-order coefficients are estimated using the Swenson method, and the glacier isostatic adjustment correction is performed based on the ICE6G model. The JPL and CSR are averaged to reduce the uncertainty.

[0110] Since some hydrological models are only updated to 2014, such as PCR-GLOBWB, the research period of this invention is from January 2003 to December 2014. The missing months of GRACE mascon are filled by cubic spline interpolation. All monthly water storage anomalies are calculated relative to the average time baseline, that is, the average value from 2004 to 2009. The different spatial resolutions of the data will cause inconvenience in data processing and result presentation. For example, the resolution of JPL is 0.5°×0.5°, while the resolution of CSR is 0.25°×0.25°. In order to ensure the consistency of the spatial resolution of GRACE and the hydrological model, the spatial resolution of all data sets in this invention is uniformly interpolated to 0.5°×0.5°. Although interpolation may affect the accuracy of a single grid, it can be ignored at the regional scale.

[0111] 2.2) Hydrological model

[0112] Hydrological models are physical and data structures established by simulating hydrological phenomena. They are a summary of actual hydrological conditions and can help relevant departments understand, manage and predict water resources. Hydrological models usually include LSMs and GHWRMs. First, GLDAS is the most commonly used LSMs, and two versions have been updated: GLDAS-1 and GLDAS-2. Among them, GLDAS-1 includes 4 models: Noah, VIC, CLM and MOSAIC, and GLDAS-2 includes 3 models: Noah, VIC and CLSM. GLDAS-2 provides land surface data from 2000 to the present, with spatial resolutions of 0.25°×0.25° and 1°×1°, and temporal resolutions of daily and monthly scales. It improves the discontinuity of GLDAS-1. In particular, the CLM version has been updated, for example, CLM-4.0 and CLM-4.5. Secondly, WGHM and PCR-GLOBWB are two commonly used GHWRMs models in the Haihe River Basin.

[0113] The performance of seven hydrological models most widely used in the Haihe River Basin was evaluated, including GLDAS2.1_Noah, GLDAS2.1_VIC, GLDAS2.1_CLSM, GLDAS2.0_CLSM, WGHM, PCR-GLOBWB and CLM-4.5. Noah, VIC, CLSM2.1, CLSM2.0, WGHM, PCR-GLOBWB and CLM were used to simplify the description of the above seven models. The differences between hydrological models are generally caused by model structure, climate input, human activities and climate change. For example, (1) Noah and VIC lack SWS or GWS components compared with other models; (2) the number and thickness of soil layers are different; (3) except Noah, VIC, CLSM2.1 and CLSM2.0, the parameters used for climate input of other models are completely different; (4) among the seven models, only CLSM2.1 and CLSM2.0 have the same model structure. In addition, only WGHM and PCR-GLOBWB consider human water use when simulating TWSA, and only WGHM has a calibration. In summary, these differences are related to model uncertainties, which affect model performance.

[0114] (3) TWSA inversion based on GRACE mascon

[0115] Figure 3 The spatial distribution of monthly TWSA inverted by JPL and CSR from 2003 to 2014 is shown. During the study period, the interannual variation of TWSA in the Haihe River Basin was obvious, with a rate of -15.66±0.67mm / year ( Figure 3 (c)). JPL( Figure 3 (a)) and CSR( Figure 3 The results of (b)) are slightly different. JPL has two descending funnels, one in northern Henan (such as Jiaozuo, Xinxiang, Puyang and Handan), and the other in eastern Hebei (such as Hengshui and Cangzhou), while CSR has only one descending funnel, at the junction of Hebei and Henan. In addition, CSR has a higher resolution and can reflect the gradual change characteristics between grids. This difference is caused by the leakage error that may occur when GRACE measures TWSA in a small area. So far, the TWSA inverted by GRACE has been widely accepted. In order to improve the signal-to-noise ratio of TWSA, the present invention uses the average of two GRACE mascon data as the reference value of TWSA. The research of Sakumura et al. shows that the use of a comprehensive average value of different institutions can effectively reduce errors.

[0116] Figure 4 The raw data of TWSA after STL decomposition is shown ( Figure 4 (a)), trend item ( Figure 4 (b)), Seasonal term ( Figure 4 (c)) and the residual term ( Figure 4 (d)). It can be seen that after removing the seasonal term and the residual term, the trend term is smoother than the original data. From the time series point of view, the TWSA of the two institutions has a similar trend of change, both showing a long-term downward trend, especially since 2005, the decline has been more drastic. From the spatial distribution point of view, TWSA shows a downward trend from north to south and gradually intensifies. The change in the southern part of the Haihe River Basin reaches -30 to -35 mm / year. However, the TWSA inverted by JPL (-17.4±0.7 mm / year) has a larger variation range than CSR (-13.7±0.5 mm / year), and related studies have also found similar results. The TWSA inverted by GRACE also reflects two extreme drought events that occurred in 2009 and 2011. These two droughts were mainly caused by the Arctic Oscillation and El Niño events. From Figure 4 The monthly precipitation distribution in (a) shows that the precipitation was basically stable from 2003 to 2009, and then began to increase. Since El Nino and La Nina usually occur alternately and have opposite effects on the climate, there is more precipitation in the second year after the El Nino event, that is, there was a brief increase in 2010 and 2012.

[0117] (4) Evaluation of TWSA trends simulated by hydrological models

[0118] 4.1) TWSA Trends

[0119] The TWSA inverted by GRACE is an integrated variable, which is often used to evaluate the reliability of hydrological model output results. Figure 5 Seven hydrological models were drawn ( Figure 5 (a)~(g)) and reference value( Figure 5 (h)) Spatial distribution of simulated TWSA trend. Dots indicate areas where the trend is not significant. It can be seen that except for WGHM (-17.11±0.44mm / year) and PCR-GLOBWB (-23±0.69mm / year) (Figures 5(e) and 5(f)), other models underestimate the downward trend of TWSA. This is because the two GHWRMs (WGHM and PCR-GLOBWB) have all water storage components and take into account water use for human activities. As for the reason why PCR-GLOBWB has a larger slope than WGHM, the WGHM2.2a version was calibrated. The present invention uses the recently updated WGHM2.2d version, which significantly improves the simulation results, but there is still overestimation. In addition, the four GLDAS models (Noah, VIC, CLSM2.1 and CLSM2.0) ( Figure 5The TWSA variation trends of (a), (b), (c) and (d)) are not obvious, which are -2.60±0.69mm / year, -1.17±0.45mm / year, -1.32±0.72mm / year and -1.08±0.88mm / year respectively, and the GLDAS model has more points without significant trend than the GHWRMs. In contrast, GRACE does not have any points without trend. Finally, although CLM( Figure 5 (g)) is consistent with the spatial distribution of GRACE in the southern Haihe River Basin, but they show the opposite trend in the northern Haihe River Basin. Therefore, the temporal and spatial patterns of TWSA simulated by different hydrological models are quite different, and this difference mainly comes from the uncertainty of a single component.

[0120] Figure 6 The TWSA time series for the Haihe River Basin and six sub-regions are shown. The error is calculated based on the standard deviation between the two GRACEmascon data, which is indicated by the shadow. It can be seen that there are differences in the TWSA trends simulated by different hydrological models. In general, GHWRMs overestimate TWSA, while LSMs underestimate TWSA. For the entire Haihe River Basin area ( Figure 6 (a)), the decline rate reached -15.66mm / year. For the six sub-regions, the decline rates were: Beijing (-7.28±0.48mm / year, Figure 6 (b)), Tianjin (-10.73±0.88mm / year, Figure 6 (c)), Hebei (-13.79±0.74mm / year, Figure 6 (d)), Shanxi (-19.91±0.91mm / year, Figure 6 (e)), Shandong (-21.28±1.51mm / year, Figure 6 (f)), Henan (-34.57±2.06mm / year, Figure 6 (g)). However, CLSM2.1 and CLSM2.0 show that Henan has a slight TWSA change trend, which is -7.28±1.24mm / year and -11.48±1.5mm / year, respectively, which is much lower than the reference value (-34.57±2.06mm / year). In addition, the TWSA change trend of the two CLSM models in Beijing and Tianjin is not obvious. The TWSA inversion of WGHM is most consistent with GRACE (r=0.91, RMSE=26.02mm, NSE=0.86), and the consistency of CLSM2.0 with GRACE is the lowest (r=0.49, RMSE=67.18mm, NSE=-0.11).

[0121] For the sub-regions of the Haihe River Basin, GHWRMs generally overestimate TWSA, except for Shanxi. Specifically, the following three points can be summarized: (1) Compared with GRACE, LSMs (Noah, VIC, CLSM2.1, CLSM2.0 and CLM) underestimate TWSA in Beijing and Hebei; (2) LSMs and WGHM underestimate TWSA in Tianjin, Shandong and Henan. (3) Both LSMs and GHWRMs underestimate TWSA in Shanxi. Model structure may be the main factor leading to the difference between LSMs and GHWRMs. Both GHWRMs have good consistency with GRACE in Beijing and Hebei (r>0.84). Among them, PCR-GLOBWB has good correlation in Tianjin, Shanxi and Shandong, with correlation coefficients of 0.79, 0.75 and 0.85, respectively. If RMSE, NSE and r are analyzed comprehensively, WGHM is most consistent with GRACE in Henan, with a correlation coefficient as high as 0.91.

[0122] In the Haihe River Basin, PCR-GLOBWB and WGHM have the best consistency with GRACE (r>0.9), so they were used as input models for statistical calibration. Figure 7 The spatial distribution of WGHM, PCR-GLOBWB and Merge is shown. The present invention uses the GRACEmascon inversion results as the true value. It can be seen that the spatial distribution of the Merge results estimated by the new statistical correction hydrological model weighting method is more consistent with GRACE. However, PCR-GLOBWB obviously has an over-simulation phenomenon in the middle area of ​​the Haihe River Basin; WGHM has an under-simulation phenomenon in the eastern area of ​​the Haihe River Basin. The average trend of Merge is better than that of a single model.

[0123] 4.2) SMSA Trends

[0124] SMSA usually has three acquisition sources, namely, field oil well observations, satellite remote sensing estimates and model simulations. However, field measurements are point-scale and sparsely distributed in the Haihe River Basin. Although remote sensing soil moisture estimates can meet spatial requirements, they usually represent the soil moisture of the top ~5cm, which is much lower than the soil depth of the hydrological model. Therefore, the present invention believes that SMSA simulated by the hydrological model is relatively reliable. In addition, there is no method to determine which model has the highest accuracy. In order to reduce the impact of model uncertainty on the reference value, the reference value of SMSA adopts the average value of multi-source data (Noah, VIC and CLM) because these three models have more soil layers than other models. SMSA uncertainty is equal to the standard deviation between Noah, VIC and CLM.

[0125] Figure 8 There are seven hydrological models ( Figure 8 (a)~(g)) and the corresponding reference values ​​( Figure 8 (h) Spatial distribution of simulated SMSA trends. For the entire Haihe River Basin, the SMSA trends estimated by Noah, VIC and CLM are highly consistent with the results of the reference values, with correlation coefficients greater than 0.9, NSE greater than 0.5 and RMSE less than 18.33 mm. However, in terms of spatial distribution, WGHM shows almost no obvious trend of change, and CLM significantly overestimates the SMSA downward trend in the northern part of the Haihe River Basin. In addition, the four models, CLSM2.0, CLSM2.1, WGHM and PCR-GLOBWB, match the SMSA reference values ​​poorly. This result is more easily obtained by analyzing statistical indicators. The statistical results of these four models are generally lower than those of the other three models, which also proves that it is appropriate to use the average value of Noah, VIC and CLM as the reference value.

[0126] Fig. 9 The SMSA time series for the Haihe River Basin and six sub-regions are shown. For the entire Haihe River Basin, SMSA was basically stable from 2003 to 2014, with a variation range of -8 to 6 mm / year. For the sub-regions, Noah was the most reliable in most sub-regions (r>0.9), indicating that Noah can simulate soil moisture well. From the temporal and spatial variation of Noah, it can be seen that SMSA showed a significant downward trend in Shandong (-5.23±1.44 mm / year), Henan (-7.67±1.21 mm / year), southern Hebei, and southern Shanxi, while there was no significant downward trend in Beijing, Tianjin, northern Hebei, and northern Shanxi. Similar patterns were observed in the SMSA reference values. In addition, the following points can be obtained: (1) Compared with the reference values, in Beijing, CLM and VIC overestimated their SMSA, while Noah had the highest correlation with a correlation coefficient of 0.9. (2) In Hebei, Shanxi, and Shandong, Noah, VIC, and PCR-GLOBWB overestimated their SMSA. (3) Only CLM overestimates the SMSA of Tianjin, and only Noah overestimates the SMSA of Henan. On the contrary, for the other underestimations, it is because the soil thickness of these models is not enough to fully accommodate soil moisture.

[0127] 4.3) GWSA Trends

[0128] Similar to TWSA and SMSA, GWSA can also be measured by on-site drilling, but there are limitations such as insufficient and unevenly distributed monitoring holes. According to the terrestrial water balance principle (Equation (3)), the GWSA reference value can be calculated by the residual between the TWSA inverted by GRACE and the SMSA simulated by the model. The uncertainty of GWSA is calculated by the root mean square of the uncertainty of the TWSA and SMSA reference values. Figure 10 shows the spatial distribution of the GWSA trend in the Haihe River Basin. It can be seen that the reference value shows a decreasing trend that gradually strengthens from north to south. However, CLSM2.0, CLSM2.1 and CLM all underestimated GWSA, and even had no interannual variation trend. In contrast, the two GHWRMs models overestimated GWSA. In addition, WHGM showed the largest variation range, about -152 to 54 mm, and a highly correlated statistical result (RMSE = 20.71 mm, NSE = 0.87, r = 0.94), because it simulated the part of groundwater extracted and recharged from soil and surface water, accurately considering the actual domestic water consumption.

[0129] Fig.11 The GWSA time series of the Haihe River Basin and six sub-regions are shown. It can be seen that for the entire Haihe River Basin, although there were brief increases in 2003-2004 and 2010-2012, the GWSA still decreased significantly at a rate of -15.23±0.87mm / year. For the sub-regions, in fact, the spatial distribution of GWSA is basically consistent with that of TWSA, but the time series value is weaker than that of TWSA. In other words, the GHWRMs have the best fitting effect with the reference value, because according to the water balance formula, the difference between GWSA and TWSA is small when SMSA is stable. This also reflects that the decline in TWSA is mainly caused by the decline in GWSA. Among the sub-regions, the GWSA in Shandong (-20.05±1.03mm / year) and Henan (-30.73±1.35mm / year) decreased the fastest. Both GHWRMs models showed weak trends in Shanxi. The large area of ​​​​the reference value shows that the groundwater changes in the sub-regions are extremely uneven. By comparing the shaded areas of the three water storage changes, it can be found that the error of SMSA is relatively stable and seasonal, while the error of GWSA is about twice that of TWSA, among which GWSA has the largest error in Tianjin.

[0130] (5) Analysis of water storage changes in plain and mountainous areas

[0131] The Haihe River Basin can be divided into two parts, mountainous area and plain area, according to its terrain characteristics. The trends of the three hydrological variables (TWSA, SMSA and GWSA) are different under different terrain conditions. Fig.12The time series changes of the three hydrological variables in mountainous areas and plains are shown separately. In general, the amplitude of change in the plains is greater than that in the mountains. For example, the TWSA rate in the plains is -17.17±1.05mm / year, while that in the mountains is -14.19±0.69mm / year, indicating that the deep groundwater in the plains is seriously overexploited. Fig.12 It can be seen that the seasonal changes of TWSA and SMSA are obvious, while GWSA is chaotic. In addition, the reliability of the models was analyzed based on three hydrological variables. First, GHWRMs overestimated the trend of TWSA, while LSMs underestimated the trend of TWSA ( Fig.12 (a) and 12(b)). Secondly, in the SMSA simulation results ( Fig.12 (c) and 12(d)), Noah has the highest correlation coefficient (plain: 0.89, mountain: 0.96), and WGHM has the lowest correlation coefficient (plain: 0.6, mountain: 0.5). Finally, the inversion results of WGHM in the plain are higher than the reference value, while the other six hydrological models underestimate the GWSA trend ( Fig.12 (e) and 12(f)).

[0132] (6) Model uncertainty

[0133] Uncertainties are the main reason for the differences between GRACE and hydrological models in TWSA trends. Uncertainties include model structure, climate inputs, human activities, and climate change. Fig.13The difference between the hydrological model and the corresponding reference value for different water storage changes is shown, and this value will be used as an indicator to measure the uncertainty of water storage. For example, if the value is negative, it means that the model underestimates TWSA. It can be seen that, except for GHWRMs, LSMs underestimate the TWSA trend in the Haihe River Basin. Different hydrological models have completely different model structures. The two GHWRMs overestimate the trend of TWSA due to overestimation of GWSA, while groundwater overdraft leads to a decrease in GWSA. On the contrary, the reasons for the underestimation of LSMs can be divided into two categories: (1) the model ignores the groundwater component, such as Noah and VIC; (2) the model has insufficient ability to capture groundwater changes, such as CLSM2.1, CLSM2.0 and CLM. Therefore, the model structure plays an important role in the difference in TWSA trends between GRACE and hydrological models. Then, taking CLSM2.0 and CLSM2.1 as examples, the present invention verifies the impact of climate input on uncertainty. Although the time series of CLSM2.0 varies slightly more than that of CLSM2.1, the two models have similar spatial layouts and have high correlations in 2003-2009 (r = 0.97) and 2010-2014 (r = 0.92) (Figure 15). Therefore, climate input is not the key factor in uncertainty.

[0134] For TWSA, the two GHWRMs have better consistency with GRACE than other hydrological models, and the correlation coefficients of the two GHWRMs models are greater than 0.98 during the study period (e.g. Fig.15 ). This means that human water use is a major source of uncertainty. Finally, TWSA can also be driven by climate change (both natural and anthropogenic). Figure 6 It can be seen that the TWSA simulated by the model decreases rapidly in the dry years (2005-2009) and rebounds in the wet years (2010-2012). Figure 4 As shown in (a), the present invention speculates that the increase in precipitation after 2010 is related to the two La Niña events, and accordingly, the TWSA simulated by LSMs gradually deviates from the GRACE results after 2010. The above shows that climate change (such as precipitation) is another important cause of uncertainty.

[0135] Based on the above embodiments, the application of the method for improving the accuracy of water reserve change based on weighted statistical correction hydrological model described in the present invention is described below.

[0136] (1) Comparative analysis of TWSA trends in the Haihe River Basin

[0137] The present invention selects four grids (G176, G209, G259 and G277) in the southern and northern parts of the Haihe River Basin to demonstrate the applicability of the new statistical correction hydrological model weighting method. Fig.14 The time series of WGHM, PCR-GLOBWB, Merge and reference values ​​in four selected grids are shown. It can be seen that although the individual models all show a similar downward trend as GRACE, there are large differences in the magnitude of change between the two. WGHM overestimates the TWSA trend of G176 and G209 and underestimates the TWSA trend of G259 and G277. PCR, on the contrary, underestimates the TWSA trend of G176 and G209 and overestimates the TWSA trend of G259 and G277. After statistical correction, the time series of Merge is more consistent with GRACE.

[0138] (2) Analysis of driving factors of TWSA changes in the Haihe River Basin

[0139] 2.1) Correlation of TWSA in different periods

[0140] Fig.15 is the correlation coefficient between each model and reference value at different stages of TWSA. The study shows that the TWSA is divided into two stages ( Figure 6 ), in 2003-2009, all hydrological models fit well with GRACE, and in 2010-2014, except for WGHM and PCR-GLOBWB, other LSMs fit poorly with GRACE. Taking CLM as an example, the correlation coefficient between CLM and GRACE in 2003-2009 is 0.90, the largest among all models, while the correlation coefficient in 2010-2014 is only 0.34, the smallest among all models. This is because LSMs are more sensitive to soil moisture changes caused by precipitation.

[0141] 2.2) Reasons why GRACE and hydrological models over- / underestimate TWSA

[0142] Scanlon's research results mentioned that most hydrological models underestimated the change in water reserves. In contrast, the present invention found that some models overestimated the TWSA in the Haihe River Basin, and there are traces to follow. Therefore, targeted analysis is needed in specific areas. In the entire Haihe River Basin, the TWSA inverted by GRACE showed a sharp downward trend, and there was no trend-free point. It gradually increased from north to south, with an average slope of -15.66±0.67mm / year. Correspondingly, the GHWRMs model overestimated TWSA, while the LSMs underestimated TWSA. Because LSMs are more effective in simulating SMSA, while GHWRMs can better simulate GWSA. The Haihe River Basin has serious groundwater depletion, so GHWRMs with stronger groundwater capture capacity are more consistent with the GRACE results. In addition, the spatial distribution of the seven hydrological models has more or less trend-free points.

[0143] 2.3) Driving factors of groundwater depletion in the Haihe River Basin

[0144] Water resources and energy fluxes in different terrain areas are affected by both climate change and human activities. Excessive groundwater extraction and reduced precipitation have led to a decline in GWSA, while winter wheat irrigation is the main cause of excessive groundwater extraction in the plain area of ​​the Haihe River Basin. Irrigation led to a decrease in groundwater reserves in the first half of the year, and in the second half of the year, groundwater reserves were restored to a certain extent after replenishment during the rainy season from June to September. There is a large amount of grain planting area in the plain area of ​​the Haihe River Basin, about 18% of which is land that needs irrigation. Agricultural water use accounts for 68% of the total water use in the region, of which 37% comes from groundwater extraction. Concentrated irrigation has accelerated groundwater depletion in the region. Therefore, human activities are likely to be the main cause of groundwater changes in most of the plain areas of the Haihe River Basin. In mountainous areas, it was found that the mountain runoff decreased in terms of water budget, and the evaporation in the mountainous areas increased from the perspective of energy structure. These two aspects indicate that climate change is also the cause of water shortage in the basin. The changes in water reserves in the Haihe River Basin caused by different factors have made the water problem more complicated, and the water conservancy department needs to conduct comprehensive management of the basin.

[0145] There are two reasons for the phased recovery of groundwater in the Haihe River Basin. On the one hand, the precipitation increased from 2010 to 2012, which was also the year with the highest precipitation from 2003 to 2014. During this period, the groundwater declined relatively slowly. On the other hand, the South-to-North Water Diversion Project was planned since 2003 (middle route) and started operation in December 2014. During the implementation of the project, the Miyun Reservoir continuously diverted water to replenish the groundwater. According to the Haihe River Basin Water Resources Bulletin, after 2014, groundwater consumption continued to decline, which reflects the contribution of the South-to-North Water Diversion Project to reducing groundwater depletion. There are also literatures in the early stage that show that the project has a positive impact on groundwater recovery. However, the recovery of shallow groundwater is a long process. Although the amount of water transferred to the Haihe River Basin through reservoirs has increased in the past decade, groundwater is still the most important water source due to its universality, flexibility and low cost, especially in extreme drought periods when groundwater is severely exploited. Therefore, the water crisis in the Haihe River Basin is still severe and still deserves great attention.

[0146] Although the present invention has been disclosed as above in the form of a preferred embodiment, it is not intended to limit the present invention. Any person skilled in the art may make possible changes and modifications to the technical solution of the present invention by using the methods and technical contents disclosed above without departing from the spirit and scope of the present invention. Therefore, any simple modifications, equivalent changes and modifications made to the above embodiments based on the technical essence of the present invention without departing from the content of the technical solution of the present invention shall fall within the protection scope of the technical solution of the present invention.

[0147] The contents not described in detail in the specification of the present invention belong to the common knowledge of the professionals in this field.

Claims

1. A method for improving the accuracy of water storage changes based on weighted statistical correction hydrological models, characterized in that: include: Determine K hydrological models; According to the model values ​​output by each hydrological model and the reference values ​​corresponding to each model value, the Pearson linear correlation coefficient corresponding to each hydrological model is calculated; M optimal hydrological models with Pearson linear correlation coefficient greater than the set threshold are screened out; Determine the changes in terrestrial water storage in the Haihe River Basin output by each preferred hydrological model and the weights of each preferred hydrological model; According to the terrestrial water storage changes in the Haihe River Basin output by each preferred hydrological model and the weights of each preferred hydrological model, a weighted fusion process is performed to obtain the terrestrial water storage changes in the Haihe River Basin after weighted fusion processing Merge; According to the model values ​​output by each hydrological model and the reference values ​​corresponding to each model value, the Pearson linear correlation coefficient corresponding to each hydrological model is calculated, including: Get the model value M of the i-th month output by any k-th hydrological model k,i , and the average value of the model value of month i output by the kth hydrological model Determine M k,i The corresponding reference value O k,i , and the average value of the reference value of i months O k,i ; The Pearson linear correlation coefficient r of the kth hydrological model is obtained by the following formula (1): k : Where, i = 1, 2, ..., n, n represents the total number of months; k = 1, 2, ..., K; Determine the changes in terrestrial water storage in the Haihe River Basin output by each preferred hydrological model, including: Get the soil water storage SMS output by any j-th optimal hydrological model j , groundwater reserves GWS j , Snow Water Equivalent SnWS j , Vegetation Water Storage CWS j and surface water storage SWS j ; The global terrestrial water storage TWS corresponding to the jth optimal hydrological model is obtained by formula (1): j : TWS j =SMS j +GWS j +SnWS j +CWS j +SWS j ···(2) According to formula (2), combined with the following formula (3), the global terrestrial water storage change TWSA corresponding to the jth optimal hydrological model is calculated: j : in, and They represent the mean terrestrial water storage, soil water storage, groundwater storage, snow water equivalent, vegetation water storage and surface water storage from 2004 to 2009 corresponding to the jth optimal hydrological model; SMSA j GWSA j 、SnWSA j 、CWSA j and SWSA j They represent the changes in soil water storage, groundwater storage, snow water equivalent, vegetation water storage, and surface water storage corresponding to the jth optimal hydrological model; Ignore SnWSA j 、CWSA j and SWSA j , and obtain the terrestrial water storage change TWSA0 of the Haihe River Basin corresponding to the jth optimal hydrological model j : TWSA0 j =SMS j +GWSA j ···(4) Among them, j=1,2,…,M.

2. The method for improving the accuracy of water storage change based on weighted statistical correction hydrological model according to claim 1, characterized in that: Hydrological models, including: Noah: land surface model; VIC: Variable Infiltration Capacity Macroscale Model, variable infiltration capacity macro model; CLSM2.0: Catchment Land Surface Model Version 2.0, Catchment Land Surface Model Version 2.0; CLSM2.1: Catchment Land Surface Model Version 2.1, Catchment Land Surface Model Version 2.1; WGHM: Water Global Assessment and Prognosis hydrological Model, global assessment and prediction hydrological model; PCR-GLOBWB: PCRaster Global Water Balance, grid global water balance model; CLM: Community Land Model, public land surface model.

3. The method for improving the accuracy of water storage change based on weighted statistical correction hydrological model according to claim 1, characterized in that: Set the threshold to 0.

9.

4. The method for improving the accuracy of water storage change based on weighted statistical correction hydrological model according to claim 1, characterized in that: Determine the weight of each preferred hydrological model, including: Determine the Pearson linear correlation coefficient r corresponding to the jth optimal hydrological model j ; Determine the root mean square error RMSE corresponding to the jth optimal hydrological model j ; According to r j and RMSE j , and the weight w of the jth optimal hydrological model is obtained by solving j .

5. The method for improving the accuracy of water storage change based on weighted statistical correction hydrological model according to claim 4, characterized in that: r j The solution formula is as follows: Among them, M j,i represents the model value of the i-th month output by the j-th optimal hydrological model, represents the average value of the model value of month i output by the jth optimal hydrological model, O j,i Indicates that j,i The corresponding reference value, Represents the average value of the reference value of month i; i = 1, 2, …, n, where n represents the total number of months.

6. The method for improving the accuracy of water storage change based on weighted statistical correction hydrological model according to claim 5, characterized in that: RMSE j The solution formula is as follows:

7. The method for improving the accuracy of water storage change based on weighted statistical correction hydrological model according to claim 6, characterized in that: w j The solution formula is as follows: w j =r j +1 / RMSE j ···(7)。 8. The method for improving the accuracy of water storage change based on weighted statistical correction hydrological model according to claim 4, characterized in that: The solution formula of Merge is as follows:

Images