A method for evaluating the impact of long-term drought on water storage capacity of a river basin based on optimal probability combination

Abstract

The present invention discloses a method for evaluating the impact of long-term drought on the water storage capacity of a watershed based on an optimal probability combination. The present invention identifies the drought start time and drought duration of a long-term meteorological drought in a watershed, uses hydrological model parameters with certain physical meanings to characterize the water storage capacity of the watershed, proposes to deduce the time point and corresponding probability value at which the water storage capacity of the watershed is most likely to change based on the principle of optimal posterior probability combination, and uses multiple evaluation criteria to evaluate the credibility of the change in water storage capacity of the watershed, the significance of the change and the robustness of the result, identifies the time point at which the water storage capacity of the watershed changes, and thus deduce the response time of the water storage capacity of the watershed to the long-term meteorological drought. The method of the present invention is scientific and reasonable, close to engineering practice, and can provide an important and highly operable reference basis for meteorological drought impact assessment, watershed hydrological forecast, government disaster relief decision-making, etc. under future climate change conditions.

Description

Technical Field

[0001] The invention relates to the technical field of watershed hydrology, and in particular to a method for evaluating the impact of long-term drought on water storage capacity of a watershed based on an optimal probability combination. Background Art

[0002] Meteorological drought is one of the most destructive environmental disasters. It not only affects the balance of aquatic and terrestrial ecosystems, but also affects many economic and social sectors, including agricultural output, industrial production, and urban water supply. Existing research on meteorological drought in my country[1] mainly focuses on the identification of meteorological drought[2], the classification of drought severity, and the transformation process between different drought types (such as the extension process from meteorological drought to hydrological drought[3]). However, there are relatively few studies on the impact of long-term meteorological drought on changes in basin conditions and its impact on hydrological forecasting and water resources planning under future climate change conditions.

[0003] One of the most important evaluation indicators of a river basin is its ability to store water, also known as its water storage capacity. The water storage capacity of a river basin is also the basic basis for hydrological simulation and hydrological forecasting. my country mainly uses (1) the water balance method and (2) the method of calibrating parameters of the hydrological model to obtain the water storage capacity of a river basin. The water balance method can only consider the basic water volume and water balance of the river basin as a constraint condition. However, the hydrological system from rainfall, river basin evaporation to the runoff at the river basin outlet section is a very complex hydrological system that requires simulation based on the internal physical mechanism. Hydrological models, whether conceptual hydrological models, semi-distributed hydrological models or distributed hydrological models, all have a physical basis to a certain extent. Among them, conceptual hydrological models are widely used in hydrological simulation and hydrological forecasting due to their simple structure, low data requirements and high simulation and forecasting accuracy. Conceptual hydrological models, such as the GR4J hydrological model and the Xin'anjiang model, all have model parameters that characterize the water storage capacity of the river basin. In practical applications, the method of calibrating model parameters using hydrological models to estimate the water storage capacity of a river basin has been widely used. However, most existing studies assume that the parameters of the hydrological model are fixed, that is, the water storage capacity of a river basin is a fixed value. This assumption fails to take into account the intrinsic impact of extreme climate change (such as long-term meteorological drought) on river basin conditions.

[0004] Due to the influence of long-term drought, the basin characteristics such as surface conditions, vegetation coverage and vegetation types, and soil moisture content under basin conditions have changed, which may lead to changes in basin water storage capacity. However, since the basin has a certain degree of resilience, that is, the ability to cope with external interference, under the influence of long-term meteorological drought conditions, the change of basin water storage capacity often lags behind the occurrence of meteorological drought.

[0005] At present, there is still a lack of research on the calculation of the response of basin water storage capacity changes to long-term meteorological drought, and there is no literature or invention to construct an optimal probability combination method based on a conceptual hydrological model and use it to deduce the response of basin water storage capacity to long-term drought.

[0006] The references involved in the above content are as follows:

[0007] [1] Wang Jinsong, Li Yaohui, Wang Runyuan, Feng Jianying, Zhao Yanxia. Review of research progress on meteorological drought in my country[J]. Arid Meteorology, 2012, 30(04): 497-508.

[0008] [2] Li Shaoxuan. Research on process description and identification and response of drought events[D]. Xi'an University of Technology, 2019.

[0009] [3] Ma Lan. Dynamic changes and driving forces of the propagation of meteorological drought to hydrological drought[D]. Xi'an University of Technology, 2019.

[0010] [4] Deng Pengxin, Wang Yintang, Hu Qingfang, Liu Yong. Application of GR4J model in daily runoff simulation in Ganjiang River Basin[J]. Hydrology, 2014, 34(02): 60-65. Summary of the invention

[0011] In order to overcome the deficiencies in the prior art, the object of the present invention is to provide a method for evaluating the impact of long-term drought on the water storage capacity of a watershed based on an optimal probability combination.

[0012] The present invention firstly introduces a long-term meteorological drought identification method, defines the start time and end time of the long-term drought, and determines the drought interval. Then, based on a conceptual hydrological model, the variation characteristics of the hydrological model parameters in the watershed water storage capacity before and after the drought are constructed. Based on the Bayesian change point analysis theory, all possibilities are traversed to solve the combined probability of the two stages before and after the time point when the watershed water storage capacity may change, and the best possible change time point is selected. On the premise that the simulation results meet certain evaluation criteria, the degree of change of the watershed water storage capacity before and after the drought is evaluated, and the time point of the change of the watershed water storage capacity is determined, thereby deducing the response time and degree of the watershed water storage capacity to the long-term meteorological drought.

[0013] To achieve the above object, the present invention is implemented by the following technical solutions:

[0014] A method for evaluating the impact of long-term drought on water storage capacity of a watershed based on an optimal probability combination includes the following steps:

[0015] Step 1: Collect daily rainfall, evaporation and runoff data of the basin in the historical period and review the reliability, consistency and representativeness of the data;

[0016] Step 2: Based on the historical rainfall data reviewed in step 1, identify the drought period when long-term continuous meteorological drought occurs in the basin, including the drought onset time, drought duration and drought variability;

[0017] Step 3: Based on the conceptual hydrological model of the watershed, the rainfall, evaporation and runoff data reviewed in step 1 are used as the input of the hydrological model to construct the hydrological simulation objective function and evolutionary algorithm, and estimate and solve the parameter domain of the hydrological model;

[0018] Step 4: Based on the Bayesian change point analysis theory, the probability, significance and accuracy of the change of specific model parameters at each potential time point are solved to obtain the potential optimal change point and the potential optimal two-stage probability combination method;

[0019] Step 5, the potential optimal change point obtained in step 4 is judged to determine the actual optimal change point, and then the response time of the basin water storage capacity to long-term meteorological drought is obtained.

[0020] Preferably, in step S2, the selected watershed sequence should have only a single continuous long-term drought period.

[0021] Preferably, the start time of drought is defined as the first year in which three consecutive years of negative values ​​appear in the smoothed variation sequence; the end time of drought is determined based on the unsmoothed variation sequence, and the last year in which the variation amplitudes are negative for three consecutive years is selected; when screening for medium- and long-term continuous droughts, it should be ensured that the length of the drought period is greater than or equal to 7 years, and the average variation amplitude of the drought period should be less than -5%, and the non-drought period is the complement of the drought period.

[0022] Preferably, in step S3,

[0023] The GR4J model and model parameter θ1 are selected to characterize the changes in the actual water storage capacity of the basin;

[0024] The parameter calibration objective function is selected based on the Bayesian recursive estimation algorithm, the Gaussian distribution is used to describe the residual model, and the uniform distribution is used as the prior distribution of all unknown parameters;

[0025] The SCEM-UA algorithm was used to solve the objective function, and the posterior probability distribution of each parameter was selected using the Gelman-Rubin convergence value less than the threshold of 1.2.

[0026] Further preferably, the GR4J model includes θ1 representing the water storage capacity of the basin, the groundwater exchange coefficient θ2, the confluence reservoir capacity θ3 one day ago and the unit line confluence time θ4, wherein the change in the parameter value θ1 represents the change in the actual water storage capacity of the basin before and after the drought.

[0027] It is further preferred that the hydrological simulation calculation method is based on a Bayesian recursive estimation algorithm, and specifically adopts the following function as the parameter domain calibration objective function:

[0028]

[0029] in,

[0030]

[0031] in, Γ(.) is the gamma function, is the prior distribution; Q represents the measured runoff; c represents the basin number; ξ represents the hydrological model input, including rainfall and potential evapotranspiration; T represents the sequence length; t represents the step size; the parameter τ represents the residual model type; τ = 0, indicating that the residual model follows a Gaussian distribution;

[0032] Preferably, in step S4, the model parameters of the non-drought period are calibrated based on the algorithm in step 3. Model parameters after calibration of drought period; Model parameters during calibration of drought period In this stage, the range of the parameters θ′2, θ′3, and θ′4 follows the range of the corresponding parameters in the non-drought period, that is, θ′2 = θ2, θ′3 = θ3, and θ′4 = θ4. The unknown quantity θ′1 still needs to be calibrated to characterize the change in the water storage capacity of the basin during the drought period. The Bayesian change point analysis method is used to traverse all possible points to obtain the optimal probability combination method and the potential optimal change point.

[0033] It is further preferred that the specific process of obtaining the optimal probability combination mode and the potential optimal change point is:

[0034] Assume that the unknown time point k is the potential time point at which the water storage capacity of the basin changes;

[0035] Assume that the rainfall-runoff relationship before the change in watershed storage capacity is Estimate the four unknown model parameters corresponding to each time point k and the likelihood function value of this stage;

[0036] Assuming that the basin water storage capacity changes, the rainfall-runoff relationship is in, Estimate the model parameters θ′1 and the likelihood function value at this stage;

[0037] For time point k, the likelihood function L(Y; k) is the product of the two-stage likelihood functions and is expressed as The posterior probability density function k|Y is expressed as Repeat the above calculation process to obtain the likelihood function at each time point. The potential best change point is the time k with the maximum p(k|Y).

[0038] Preferably, the criterion for determining the actual optimal change point in step S5 is: the following requirements need to be met simultaneously:

[0039] The two-stage NSE result conformity requirement: The NSE values ​​of the two stages before and after the change point must be greater than or equal to a certain value, that is, NSE 变点前 ≥NSE 阈 , NSE 变点后 ≥NSE 阈 ;

[0040] The significance requirement of the two-stage parameter change is that the relative difference between the basin water storage capacity θ1 and θ′1 in the two stages must be greater than or equal to a certain threshold α, that is, the parameter significant change level requirement

[0041] The difference between the two-stage NSE results should be less than or equal to a certain threshold β to ensure that the two stages have the same degree of simulation, that is, |NSE 变点前 -NSE 变点后 |≤β;

[0042] The robustness requirement of the two-stage simulation results is: by changing the initial conditions of the parameters and repeatedly calculating the simulation results for multiple times, only when the basin is determined to have a significant change in its water storage capacity in each calculation, can the basin be identified as a significantly changed basin, and the corresponding change point is the time point when the basin water storage capacity is most likely to change.

[0043] Compared with the prior art, the present invention has the following advantages and beneficial effects:

[0044] (1) Scientific and reasonable, close to actual application:

[0045] The present invention uses hydrological model parameters to measure the size and changes of the basin's water storage capacity, and based on the principle of optimal posterior probability combination, derives the time point when the basin's water storage capacity is most likely to change. It has a strong statistical basis and physical properties, can objectively reflect the characteristics of the hydrological system and basin's response to climate drought, and fully considers the stability, repeatability and practicality of the data results.

[0046] (2) Provide important and highly operational reference for basin hydrological research under future climate change conditions:

[0047] Make full use of the existing measured data on climate change, adopt hydrological model methods to estimate the basin water storage capacity under different climatic conditions, deduce the time point when the basin water storage capacity changes through the most likely combination method, obtain the response time of the basin water storage capacity to long-term meteorological drought, and use a variety of evaluation indicators to evaluate the rationality of the results, providing an important and highly operational reference for basin hydrological research under future climate change conditions. BRIEF DESCRIPTION OF THE DRAWINGS

[0048] Figure 1 It is a specific flow chart of the method of the present invention;

[0049] Figure 2 This is a schematic diagram of the possible changes and significance of the basin water storage capacity before and after the "change point" inferred by this method;

[0050] Figure 3 This is a schematic diagram of the conformity judgment of the two-stage NSE simulation results before and after the "change point" obtained by this method;

[0051] Figure 4 This is a schematic diagram of the difference between the NSE simulation results before and after the "change point" obtained by this method;

[0052] Figure 5 The schematic diagram is a diagram of the posterior probability density distribution of different "change points" obtained by this method and the selection of the best change point position;

[0053] Figure 6 shows the significance of the change in water storage capacity of the three test basins A, B, and C, the posterior probability distribution of different change points, and the schematic diagram of the two-stage NSE simulation results. Fig. 6A The significance of the change in water storage capacity of the experimental basin A (left), the posterior probability distribution of different change points (middle) and the two-stage NSE simulation results (right); Figure 6B The significance of the change in water storage capacity of the experimental basin B (left), the posterior probability distribution of different change points (middle) and the two-stage NSE simulation results (right); Figure 6C The significance of the change in water storage capacity of the experimental basin C (left) and the two-stage NSE simulation results (right). DETAILED DESCRIPTION

[0054] The present invention identifies the drought period of long-term meteorological drought in the basin, uses hydrological model parameters to characterize the water storage capacity of the basin, and based on the principle of optimal posterior probability combination, derives the time point and corresponding probability value when the water storage capacity of the basin is most likely to change, evaluates the significance of the change, determines the time point when the water storage capacity of the basin changes significantly, and thus obtains the response time of the water storage capacity of the basin to long-term meteorological drought.

[0055] Combine the following Figure 1Describe the specific steps of the present invention:

[0056] Step 1: Collect daily rainfall, evaporation and runoff data of the basin in the historical period and review the reliability, consistency and representativeness of the data;

[0057] Step 2, based on the historical rainfall data, identify the drought period when long-term continuous meteorological drought occurs in the basin, including the drought start time, drought duration and drought variability;

[0058] For ease of understanding, the process of step 2 is described in detail below.

[0059] (2.1) To exclude interference, the selected watershed sequence should have only a single continuous long-term drought period;

[0060] (2.2) Based on the historical annual rainfall sequence of the basin, the mean annual rainfall is obtained, the variation sequence of the annual rainfall compared to the annual average rainfall is obtained, and the variation sequence is smoothed with a three-year sliding window; the start time of the drought is defined as the first year in which the smoothed variation sequence has three consecutive negative values; the end time of the drought is determined based on the unsmoothed variation sequence, and the last year in which the variation is negative for three consecutive years. However: (1) If the positive variation of a year in the unsmoothed variation sequence after the start of the drought is greater than 15%, the year before that year is determined as the end time of the drought; (2) If the variation of two consecutive years is positive but less than 15%, the year before those two years is determined as the end time of the drought.

[0061] (2.3) To ensure that the duration of the drought sequence and the degree of meteorological drought during the drought period have practical significance, when screening for medium- and long-term continuous droughts, the length of the drought period should be greater than or equal to 7 years, and the average variation of the drought period should be less than -5%. The normal period (non-drought period) is the complement of the drought period.

[0062] The definition of long-term meteorological drought is a basic problem in drought analysis and calculation. For drought analysis, based on the results of practical application and statistical experience, different regions may choose the most suitable meteorological drought definition method and long-duration drought selection criteria according to actual conditions. The selection of meteorological drought definition methods generally requires sufficient theoretical basis, simple and convenient application, flexible and robust form, and easy acceptance. Such as the Standardized Precipitation Index (SPI) based on rainfall, and the Palmer Drought Severity Index (PDSI) based on rainfall and potential evapotranspiration. In the definition of long-term meteorological drought, in order to exclude the influence of a single wet year that occurs occasionally during the drought period on the judgment of the length of the drought period, it is usually necessary to perform certain smoothing on the obtained data series. Moreover, due to the existence of catchment resilience, short-duration droughts usually do not cause fundamental changes to the characteristics of the basin. The calculation process of the long-term drought definition method adopted in the present invention is simpler and more convenient than SPI and PDSI, and has been widely recognized by the international community and has been applied in many continents.

[0063] Step 3, based on the hydrological model, using the rainfall, evaporation and runoff in step 1 as model input, constructing the objective function and estimating the solution parameter values.

[0064] Step 3 further comprises the following steps:

[0065] (3.1) This method uses the conceptual hydrological model of the daily scale basin, the GR4J model, to analyze the changes in the basin's water storage capacity. The model contains four model parameters, namely, θ1 (mm) representing the basin's water storage capacity, the groundwater exchange coefficient θ2 (mm), the reservoir capacity of the confluence reservoir 1 day ago θ3 (mm) and the unit line confluence time θ4 (days). In this method, the change in the parameter value θ1 is used to represent the change in the actual basin's water storage capacity before and after the drought.

[0066] The GR4J model is a conceptual rainfall-runoff model on a daily scale [4]. The model was originally proposed by the French. After being developed and improved by Nascimento, Edijatno, Perrin and others, it has been tested in hundreds of regions with different climatic conditions, such as France, Australia, and the United States. Due to its excellent performance, it is widely used in hydrological forecasting, water resources planning and other fields. The model has a simple structure and only has four parameters. Moreover, within its model structure, θ1, which represents the water storage capacity of the basin, is a sensitive parameter, while the other three parameters are relatively insensitive. This provides a good model basis for those who hope to use the changes in the estimated values ​​of the model parameters to represent the changes in the water storage capacity of the basin.

[0067] (3.2) The hydrological simulation calculation method is based on the Bayesian Recursive Estimation algorithm, and the following function is used as the parameter domain calibration objective function:

[0068]

[0069] in,

[0070]

[0071] in, Γ(.) is the gamma function, is the prior distribution; Q represents the measured runoff; c represents the basin number; ξ represents the hydrological model input, including rainfall and potential evapotranspiration; T represents the sequence length; t represents the step size; the parameter τ represents the residual model type. In the present invention, τ = 0, indicating that the residual model obeys the Gaussian distribution;

[0072] (3.3)Use uniform distribution as the prior distribution of all unknown quantities.

[0073] The choice of prior distribution is of great significance for the estimation of unknown quantities. Uniform distribution is widely used because of its characteristics of traversing the interval range and taking uniform values.

[0074] (3.4) The SCEM-UA (Shuffled Complex Evolution Metropolis) algorithm is used to solve the objective function, and the Gelman-Rubin convergence value less than the threshold value of 1.2 is used as the posterior distribution screening criterion for each variable.

[0075] SCEM-UA is a widely used algorithm based on Markov Chain Monte Carlo to calculate the posterior probability distribution of simulation parameters. Compared with the traditional Metropolis-Hasting algorithm, this method can effectively reduce the number of calculations required to derive the posterior distribution of parameters.

[0076] Step 4: Based on the Bayesian change point analysis theory, the probability, significance and accuracy of changes in specific model parameters (representing the water storage capacity of the basin) at each potential time point are solved.

[0077] Step 4 further includes the following steps:

[0078] (4.1) Based on the Bayesian change point analysis theory, we traverse all possible points to find the optimal probability combination and potential optimal change point. Assume that the unknown time point k is the time point when the water storage capacity of the basin changes, that is, k is taken from the set Θ = {1, 2, ..., T}. If k = T, it means "no change".

[0079] (4.2) Assume that the rainfall-runoff relationship before the change in basin water storage capacity is

[0080]

[0081] Therefore, for all possible time points k, the corresponding model parameter calibration values ​​are All of them can be obtained according to step 3. At this stage, a total of 4 unknown quantities need to be calibrated and solved, namely

[0082] (4.3) Similar to 4.2, assuming that the rainfall-runoff relationship after the basin water storage capacity changes is

[0083]

[0084] in, The values ​​of the parameters θ2, θ3, and θ4 in this stage follow the values ​​before the change in the water storage capacity of the basin, that is, the two-stage posterior probability p c and p′ c The connection between the two stages is the parameter values ​​θ′2=θ2, θ′3=θ3, θ′4=θ4. The inherent reason for this setting is that in the GR4J model, θ2, θ3, θ4 are non-sensitive parameters, that is, parameters that are relatively insensitive to changes in basin climate conditions. Therefore, at this stage, only the unknown quantity θ′1 needs to be calibrated.

[0085] 4.4 For time point k, the likelihood function L(Y; k) is the product of the two-stage likelihood functions expressed as

[0086]

[0087] 4.5 Therefore, the posterior probability density function k|Y is expressed as

[0088]

[0089] The potential optimal change point is the time k with the maximum p(k|Y).

[0090] Step 5: Determination of the actual optimal change point.

[0091] Based on the above calculation process, each basin can obtain a potential optimal change point and a potential optimal two-stage probability combination. However, these potential change points need to be further evaluated to determine whether there is a significant difference in the basin water storage capacity values ​​(i.e., θ1 and θ′1) before and after the two stages. The evaluation criteria are as follows:

[0092] (1) Conformity requirements of two-stage NSE results: The NSE values ​​of the two stages before and after the change point must be greater than or equal to a certain value to ensure the reliability of the simulation results. Here, NSE 阈 NSE 阈 Need to be given in advance, NSE result judgment requires two-stage NSE 变点前 ≥NSE 阈 , NSE 变点后 ≥NSE 阈 ;

[0093] (2) Significance requirements for parameter changes in the two stages: The relative difference between the watershed water storage capacity θ1 and θ′1 in the two stages must be greater than or equal to a certain threshold α. The parameter significant change level requirement

[0094] (3) Difference requirement of two-stage NSE results: The difference of two-stage NSE results should be less than or equal to a certain threshold β to ensure that the two stages have the same degree of simulation, that is, |NSE 变点前 -NSE 变点后 |≤β;

[0095] (4) Robustness requirements for two-stage simulation results: By changing the initial conditions of the parameters and repeatedly calculating the simulation results, a basin can be identified as a significantly changed basin only if it is determined in each calculation that the basin’s water storage capacity has changed significantly. The corresponding change point is the time point when the basin’s water storage capacity is most likely to change.

[0096] For details of the evaluation criteria (1), (2), and (3), please see the attached Figure 2-4 The optimal change point and corresponding probability analysis are detailed in the attached Figure 5 .

[0097] The following takes three river basins A, B, and C as examples to describe the specific implementation of the present invention in detail.

[0098] Step 1: Collect daily rainfall, evaporation and runoff data for the basin over the historical period and review the data for reliability, consistency and representativeness.

[0099] In this embodiment, the measured daily rainfall, evaporation and runoff series data of three basins A, B and C from January 1, 1976 to December 31, 2011 for a total of 36 years are selected.

[0100] Step 2, based on the historical rainfall data, identify the drought period when long-term continuous meteorological drought occurs in the basin, including the drought start time, drought duration and drought variability;

[0101] In this specific implementation, the annual average rainfall of each basin is calculated, the variation sequence of rainfall over the years relative to the annual average rainfall is obtained, and the variation sequence is smoothed with a sliding window of three years. The drought start time and duration of each basin are estimated respectively, and the results are shown in Table 1.

[0102] Table 1 Statistical table of meteorological drought results in each basin

[0103]

[0104]

[0105] Step 3, based on the conceptual hydrological model of the watershed, uses the rainfall, evaporation and runoff that have passed the "three-property review" in step 1 as the input of the hydrological model, constructs the hydrological simulation objective function and evolutionary algorithm, and estimates and solves the parameter domain of the hydrological model.

[0106] In this specific implementation, the GR4J model and the model parameter θ1 are selected to characterize the change of the actual water storage capacity of the basin. The parameter calibration objective function is selected based on the Bayesian recursive estimation algorithm, the Gaussian distribution is used to describe the residual model, and the uniform distribution is used as the prior distribution of all unknown parameters. The SCEM-UA algorithm is used to solve the objective function, and the posterior probability distribution of each parameter is selected using the Gelman-Rubin convergence value less than the threshold value of 1.2.

[0107] Step 4: Based on the Bayesian change point analysis method, the probability, significance and accuracy of changes in specific model parameters (representing the water storage capacity of the basin) at each potential time point are solved.

[0108] In this specific implementation method, let k be the potential time point that may cause the change of basin water storage capacity after the occurrence of meteorological drought, and let the rainfall-runoff relationship before the change of basin water storage capacity be Estimate the four unknown model parameters corresponding to each time point k and the likelihood function value of this stage. At the same time, let the rainfall-runoff relationship after the basin water storage capacity changes be Estimate the model parameters θ′1 and the likelihood function value at this stage. Therefore, according to the calculation formula, the posterior probability density function p(k|Y) at time point k can be obtained. Repeat the above calculation process to obtain the likelihood function at each time point. The potential optimal change point is the time k with the maximum p(k|Y). Therefore, each basin obtains a potential optimal change point and the corresponding potential optimal two-stage probability combination method.

[0109] Step 5: Identification of potential optimal change points.

[0110] This implementation method uses four evaluation criteria to determine whether the potential optimal change point is the time point when the basin water storage capacity actually changes. Figure 6 shows the estimated values ​​of the two-stage basin water storage capacity of each basin, the posterior likelihood probability of each change point, and the NSE simulation results. It was found that the basin water storage capacity of basins A and B changed significantly (such as Fig. 6A and 6B ), and its potential optimal change point is the actual optimal change point, where the water storage capacity of basin A increases significantly after the meteorological drought occurs, the water storage capacity of basin B decreases significantly after the meteorological drought occurs, and the water storage capacity of basin C does not change significantly after the meteorological drought occurs (e.g. Figure 6C ). In summary, the present invention identifies the drought start time and drought duration of a long-term meteorological drought in a basin, uses hydrological model parameters with certain characteristics to characterize the water storage capacity of the basin, and based on the principle of optimal posterior probability combination, derives the time point and corresponding probability value when the water storage capacity of the basin is most likely to change, and uses multiple criteria to evaluate the credibility of the change in the water storage capacity of the basin, the significance of the change and the robustness of the results, identifies the time point when the water storage capacity of the basin changes significantly, and thus obtains the response time of the water storage capacity of the basin to long-term meteorological drought. The present invention has a strong statistical basis and physical properties, can objectively reflect the characteristics of the hydrological system and the response of the basin, and fully considers the stability, repeatability and practicality of the data results, makes full use of the existing measured data of climate change, and uses multiple evaluation indicators to evaluate the rationality of the results, which can provide an important and highly operational reference for basin hydrological research under future climate change conditions.

[0111] The above description is only a preferred embodiment of the present invention and does not limit the present invention in any form. Any simple modification or equivalent change made to the above embodiment based on the technical essence of the present invention shall fall within the protection scope of the present invention.

Claims

1. A method for evaluating the impact of long-term drought on water storage capacity of a basin based on optimal probability combination, characterized in that: The steps include: Step 1: Collect daily rainfall, evaporation and runoff data of the basin in the historical period and review the reliability, consistency and representativeness of the data; Step 2: Based on the historical rainfall data reviewed in step 1, identify the drought period when long-term continuous meteorological drought occurs in the basin, including the drought onset time, drought duration and drought variability; Step 3: Based on the conceptual hydrological model of the watershed, the rainfall, evaporation and runoff data reviewed in step 1 are used as the input of the hydrological model to construct the hydrological simulation objective function and evolutionary algorithm, and estimate and solve the parameter domain of the hydrological model; In step 3, the GR4J model and the model parameter θ1 are selected to characterize the change of the actual water storage capacity of the basin; the parameter calibration objective function is selected based on the Bayesian recursive estimation algorithm, the Gaussian distribution is used to describe the residual model, and the uniform distribution is used as the prior distribution of all unknown parameters; the SCEM-UA algorithm is used to solve the objective function, and the posterior probability distribution of each parameter is selected using the Gelman-Rubin convergence value less than the threshold value 1.2; Step 4: Based on the Bayesian change point analysis theory, the probability, significance and accuracy of the change of specific model parameters at each potential time point are solved to obtain the potential optimal change point and the potential optimal two-stage probability combination method; Step 5, the potential optimal change point obtained in step 4 is judged to determine the actual optimal change point, and then the response time of the basin water storage capacity to long-term meteorological drought is obtained.

2. The method for evaluating the impact of long-term drought on water storage capacity of a river basin based on optimal probability combination according to claim 1, characterized in that: In step 2, the selected watershed sequence should have only a single continuous long-term drought period.

3. The method for evaluating the impact of long-term drought on water storage capacity of a river basin based on optimal probability combination according to claim 1 or 2, characterized in that: The start time of drought is defined as the first year in which three consecutive years of negative values ​​appear in the smoothed variation sequence; the end time of drought is determined based on the unsmoothed variation sequence, and the last year in which the variation amplitudes are negative for three consecutive years is selected; when screening for medium- and long-term continuous droughts, the length of the drought period should be greater than or equal to 7 years, and the average variation amplitude of the drought period should be less than -5%, and the non-drought period is the complement of the drought period.

4. The method for evaluating the impact of long-term drought on water storage capacity of a river basin based on optimal probability combination according to claim 1, characterized in that: The GR4J model contains θ1, which represents the water storage capacity of the basin, the groundwater exchange coefficient θ2, the confluence reservoir capacity θ3 one day ago, and the unit line confluence time θ4. The change in the parameter value θ1 represents the change in the actual water storage capacity of the basin before and after the drought.

5. The method for evaluating the impact of long-term drought on water storage capacity of a river basin based on optimal probability combination according to claim 1, characterized in that: The following function is used as the parameter domain to calibrate the objective function: in, in, Γ(.) is the gamma function, is the prior distribution; Q represents the measured runoff; c represents the basin number; ξ represents the hydrological model input, including rainfall and potential evapotranspiration; T represents the sequence length; t represents the step size; the parameter τ represents the residual model type; τ = 0, indicating that the residual model follows a Gaussian distribution; 6. The method for evaluating the impact of long-term drought on water storage capacity of a river basin based on optimal probability combination according to claim 1, characterized in that: In step 4, the model parameters of the non-drought period are calibrated based on the algorithm in step 3. Then calibrate the parameters of the drought period model; use the Bayesian change point analysis method to traverse all possible points to obtain the optimal probability combination and potential optimal change point.

7. The method for evaluating the impact of long-term drought on water storage capacity of a river basin based on optimal probability combination according to claim 6, characterized in that: The specific process of obtaining the optimal probability combination method and the potential optimal change point is: Assume that the unknown time point k is the potential time point at which the water storage capacity of the basin changes; Assume that the rainfall-runoff relationship before the change in watershed storage capacity is The model parameters to be solved in this stage are Estimate the four unknown model parameters corresponding to each time point k and the likelihood function value of this stage; Assuming that the basin water storage capacity changes, the rainfall-runoff relationship is in, Estimate the model parameters θ'1 and the likelihood function value at this stage; For time point k, the likelihood function L(Y; k) is the product of the two-stage likelihood functions and is expressed as The posterior probability density function k|Y is expressed as Repeat the above calculation process to obtain the likelihood function at each time point. The potential best change point is the time k with the maximum p(k|Y).

8. The method for evaluating the impact of long-term drought on water storage capacity of a river basin based on optimal probability combination according to claim 1 or 6, characterized in that: The criterion for determining the actual optimal change point in step 5 is: the following requirements must be met simultaneously: The two-stage NSE result conformity requirement: The NSE values ​​of the two stages before and after the change point must be greater than or equal to a certain value, that is, NSE 变点前 ≥NSE 阈 , NSE 变点后 ≥NSE 阈 ; The significance requirement of the two-stage parameter change is that the relative difference between the basin water storage capacity θ1 and θ'1 in the two stages must be greater than or equal to a certain threshold α, that is, the parameter significant change level requirement The difference between the two-stage NSE results should be less than or equal to a certain threshold β to ensure that the two stages have the same degree of simulation, that is, |NSE 变点前 -NSE 变点后 |≤β; The robustness requirement of the two-stage simulation results is: by changing the initial conditions of the parameters and repeatedly calculating the simulation results for multiple times, only when the basin is determined to have a significant change in its water storage capacity in each calculation, can the basin be identified as a significantly changed basin, and the corresponding change point is the time point when the basin water storage capacity is most likely to change.

Images