Hydrological model performance evaluation method based on combination of panel regression and hydrological process analysis

Abstract

The invention relates to the technical field of runoff prediction and analysis, provides a hydrological model performance evaluation method and system based on combination of panel regression and hydrological process analysis, and utilizes meteorological data and basin attributes of each basin to drive a preset hydrological model to obtain simulated runoff. And based on a seasonal trend decomposition method, respectively decomposing the surface average rainfall observed quantity and the surface average rainfall analog quantity, respectively calculating the relative deviation of each sequence, and analyzing the processing performance of the preset hydrological model on the drainage basin rainfall deviation according to the relative deviation of each sequence in combination with the hydrological process. And finally, through a panel regression model, respectively and quantitatively evaluating the contribution of rainfall deviation to simulated runoff deviation. According to the method, the influence of the rainfall deviation on the runoff simulation performance of the hydrological model can be quantified and analyzed, and an important scientific basis and targeted guidance are provided for subsequent model parameter optimization and performance improvement.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of runoff prediction analysis, and more particularly to a hydrological model performance evaluation method based on panel regression combined with hydrological process analysis. BACKGROUND

[0002] A hydrological model is a model that simulates runoff at an outlet of a watershed by quantitatively describing processes such as precipitation, snowmelt, interception, depression storage, evaporation, infiltration, runoff generation, slope flow concentration, and river flow concentration, based on the principle of hydrological cycle and through mathematical simulation. Currently, the performance evaluation of a hydrological model usually relies on comprehensive statistical indicators such as Nash-Sutcliffe efficiency coefficient and Kling-Gupta efficiency coefficient. Such indicators can reflect the overall fitting degree between the simulation results and the observed values, but it is difficult to reveal the formation mechanism of the hydrological model error and its transmission mechanism in different hydrological processes. In addition, as the main input of the hydrological model, the spatial and temporal bias of precipitation often leads to important sources of runoff simulation error. Traditional precision evaluation focuses on runoff and fails to quantify and analyze the influence of precipitation bias on the runoff simulation performance of the hydrological model. SUMMARY

[0003] To solve the problem that the influence of precipitation bias on the runoff simulation performance of the hydrological model cannot be quantified and analyzed in the prior art, the present application provides a hydrological model performance evaluation method based on panel regression combined with hydrological process analysis, which is used to quantify and analyze the influence of precipitation bias on the runoff simulation performance of the hydrological model.

[0004] In order to achieve the above technical effects, the technical solutions of the present application are as follows: A hydrological model performance evaluation method based on panel regression combined with hydrological process analysis, comprising the following steps: S10: According to the acquired meteorological data of each watershed and the watershed attributes of each watershed, the runoff is simulated by using a preset hydrological model to obtain the simulated runoff of each watershed; the meteorological data includes the surface average precipitation observation and the surface average precipitation simulation of each watershed; S20: The surface average precipitation observation is decomposed to obtain the observed precipitation trend sequence, the observed precipitation seasonal sequence, and the observed precipitation residual sequence of each watershed, and the surface average precipitation simulation of each watershed is decomposed to obtain the simulated precipitation trend sequence, the simulated precipitation seasonal sequence, and the simulated precipitation residual sequence of each watershed; S30: Based on the sequence obtained in S20, calculate the relative deviation of the simulated precipitation trend sequence, the relative deviation of the simulated precipitation seasonal sequence, the relative deviation of the simulated precipitation residual sequence, and the relative deviation of the simulated runoff for each watershed. Using the relative deviations, combine the hydrological model to simulate runoff and analyze the performance of the preset hydrological model in handling watershed precipitation deviations. S40: Based on the relative deviation, construct a panel regression model to quantitatively evaluate the contribution of the relative deviation to the runoff deviation.

[0005] Preferably, in S10, the meteorological data further includes: air temperature and potential evapotranspiration of each watershed, and the process for obtaining the observed and simulated areal average precipitation of each watershed is as follows: S101: Divide the watershed into grids based on the boundaries of each watershed, and calculate the area of ​​each grid. S102: Obtain precipitation observation data and precipitation simulation data, and use the area of ​​each grid as the weight to calculate the surface average precipitation observation of each watershed based on the precipitation observation data and the surface average precipitation simulation of each watershed based on the precipitation simulation data.

[0006] Preferably, in S10, based on the acquired meteorological data and watershed attributes of each watershed, a preset hydrological model is used to simulate and obtain the simulated runoff of each watershed. The process is as follows: C1: Based on the meteorological data, the watershed attributes, and the long short-term memory neural network, obtain the static and dynamic parameters of the hydrological model:

[0007] in, The static parameters of watershed b; The dynamic parameters of watershed b; For watershed b, this is a watershed attribute; This refers to the Long Short-Term Memory neural network; This represents the period from day 1 to day t. For network parameters; Meteorological data for basin b, including simulated areal average precipitation, temperature and potential evapotranspiration for basin b; C2: Based on the static and dynamic parameters, the hydrological model is driven to obtain the simulated runoff for each watershed:

[0008] in, For the simulated runoff of watershed b, Preset HBV Hydrological model.

[0009] Preferably, the network parameters of the long short-term memory neural network are updated by gradient descent method, and the expression is: Preferably, the network parameters of the long short-term memory neural network are updated by gradient descent method, and the expression is:

[0010] wherein, is the network parameters of the long short-term memory neural network after the i-th optimization; is the network parameters of the long short-term memory neural network after the (i+1)-th optimization; is the learning rate; is the loss function; is the gradient of the loss function with respect to the network parameters after the i-th optimization; is the meteorological data, wherein the meteorological data includes a simulated surface mean precipitation, air temperature and potential evapotranspiration; is the observed runoff.

[0011] Preferably, the loss function is a weighted combination of root mean square errors, and the expression of the loss function is:

[0012] wherein, is the loss function, is the flow weight coefficient, and respectively represent the simulated runoff and the observed runoff of the basin b on the t-th day before transformation, and respectively represent the simulated runoff and the observed runoff of the basin b on the t-th day after transformation, , , is 10 -6 , B is the total number of basins, T is the total time length.

[0013] Preferably, in S20, the observed surface mean precipitation is decomposed to obtain an observed precipitation trend sequence, an observed precipitation seasonal sequence and an observed precipitation residual sequence of each basin, and the simulated surface mean precipitation is decomposed to obtain a simulated precipitation trend sequence, a simulated precipitation seasonal sequence and a simulated precipitation residual sequence of each basin, and the expression is:

[0014] wherein, represents from the 1st day to the t-th day, and respectively represent the observed surface mean precipitation and the simulated surface mean precipitation, and respectively represent the observed precipitation trend sequence and the simulated precipitation trend sequence, and respectively, are the observed and simulated seasonal series of precipitation, and respectively, are the observed and simulated residual series of precipitation.

[0015] Preferably, the S30 process is: S301: calculating the relative bias of the simulated trend series of precipitation of each basin The calculation expression is:

[0016] wherein, represents the observed trend series of precipitation of basin b on day t, represents the simulated trend series of precipitation of basin b on day t, b represents a basin, y represents a hydrological year, and Y represents the time length of a hydrological year; S302: calculating the relative bias of the simulated seasonal series of precipitation of each basin The calculation expression is:

[0017] wherein, represents the observed seasonal series of precipitation of basin b on day t, represents the simulated seasonal series of precipitation of basin b on day t, b represents a basin, y represents a hydrological year, and Y represents the time length of a hydrological year; S303: calculating the relative bias of the simulated residual series of precipitation of each basin The calculation expression is:

[0018] wherein, represents the observed residual series of precipitation of basin b on day t, represents the simulated residual series of precipitation of basin b on day t, b represents a basin, y represents a hydrological year, and Y represents the time length of a hydrological year; S304: calculating the relative bias of the simulated runoff of each basin The calculation expression is:

[0019] wherein, represents the observed runoff of basin b on day t, represents the simulated runoff of basin b on day t, b represents a basin, y represents a hydrological year, and Y represents the time length of a hydrological year; S305: based on the relative bias , the relative bias , the relative bias , and the relative bias , the process of runoff simulation combined with the hydrological model is analyzed to analyze the processing method and effect of the preset hydrological model on the precipitation bias of the basin.

[0020] Preferably, the S40 process is: based on the relative deviation , the relative deviation , the relative deviation and the relative deviation Panel data is constructed, panel regression analysis is performed by a fixed effect model, a panel regression model is constructed, and the relative deviation , and are fitted to obtain the contribution of the relative deviation to the simulated runoff deviation , and satisfy the expression:

[0021] wherein, is the fixed effect of the basin, is the time fixed effect, is the error term; is the contribution of the precipitation trend sequence, i.e., the relative deviation changes 1% causes the simulated runoff relative deviation to change ; is the contribution of the precipitation seasonal sequence, i.e., the relative deviation changes 1% causes the simulated runoff relative deviation to change ; is the contribution of the precipitation residual sequence, i.e., the relative deviation changes 1% causes the simulated runoff relative deviation to change .

[0022] The present application also proposes a hydrological model performance evaluation system based on the combination of panel regression and hydrological process analysis, for realizing the hydrological model performance evaluation method based on the combination of panel regression and hydrological process analysis as described above, and the system comprises: a precipitation and runoff acquisition module, configured to simulate runoff by using a preset hydrological model according to the acquired meteorological data of each basin and the basin attributes of each basin, and acquire simulated runoff of each basin; the meteorological data comprises surface average precipitation observation and surface average precipitation simulation of each basin; a precipitation decomposition module, configured to decompose the surface average precipitation observation to obtain observed precipitation trend sequences, observed precipitation seasonal sequences and observed precipitation residual sequences of each basin, and decompose the surface average precipitation simulation of each basin to obtain simulated precipitation trend sequences, simulated precipitation seasonal sequences and simulated precipitation residual sequences of each basin; ​The relative deviation analysis module is configured to calculate the relative deviation of the simulated precipitation trend sequence of each basin, the relative deviation of the simulated precipitation seasonal sequence of each basin, the relative deviation of the simulated precipitation residual sequence of each basin, and the relative deviation of the simulated runoff of each basin based on the sequences output by the precipitation decomposition module, and analyze the processing performance of the preset hydrological model on the precipitation deviation of the basin in combination with the process of runoff simulation by the hydrological model using the relative deviations. The panel regression module is configured to construct a panel regression model based on the relative deviations and quantitatively evaluate the contribution of the relative deviations to the runoff deviation.

[0023] The present application also provides an electronic device, which comprises a memory, a processor, and a program stored in the memory and executable on the processor; and the processor is configured to read the program in the memory to implement the steps in the method for evaluating the performance of a hydrological model based on the combination of panel regression and hydrological process analysis.

[0024] Compared with the prior art, the technical scheme of the present application has the following beneficial effects: The present application provides a method and system for evaluating the performance of a hydrological model based on the combination of panel regression and hydrological process analysis, which uses the meteorological data and basin attributes of each basin to drive a preset hydrological model to obtain simulated runoff. Based on a seasonal trend decomposition method, the observed and simulated average precipitation is decomposed, the relative deviations of each sequence are calculated, and the processing performance of the preset hydrological model on the precipitation deviation of the basin is analyzed in combination with the hydrological process according to the relative deviations of each sequence. Finally, the contribution of the precipitation deviation to the simulated runoff deviation is quantitatively evaluated by a panel regression model. The present application can quantify and analyze the influence of precipitation deviation on the runoff simulation performance of a hydrological model, and provide important scientific basis and targeted guidance for subsequent model parameter optimization and performance improvement. BRIEF DESCRIPTION OF DRAWINGS

[0025] Figure 1 FIG. 1 shows a flowchart of the method for evaluating the performance of a hydrological model based on the combination of panel regression and hydrological process analysis according to an embodiment of the present application; Figure 2 FIG. 2 shows a schematic diagram of the time sequences of observed precipitation, simulated precipitation, and their respective decomposed components according to an embodiment of the present application; Figure 3 FIG. 3 shows a cumulative distribution function (CDF) diagram of the relative deviations of the precipitation sequences according to an embodiment of the present application; Figure 4 FIG. 4 shows a correlation analysis diagram of intermediate variables in a hydrological model driven by precipitation according to an embodiment of the present application; Figure 5 FIG. 5 shows a parameter analysis diagram of a hydrological model driven by precipitation according to an embodiment of the present application; Figure 6 This diagram illustrates the contribution of precipitation deviation to relative runoff deviation as proposed in the embodiments of the present invention. Figure 7 This diagram illustrates the components of the hydrological model performance evaluation system based on a combination of panel regression and hydrological process analysis proposed in this embodiment of the invention. Figure 8 This diagram illustrates the structure of the electronic device proposed in the embodiments of the present invention. Detailed Implementation

[0026] The accompanying drawings are for illustrative purposes only and should not be construed as limiting the scope of this patent. To better illustrate this embodiment, some parts of the accompanying drawings may be omitted, enlarged, or reduced, and do not represent the actual dimensions; It is understandable to those skilled in the art that some well-known details may be omitted from the accompanying drawings.

[0027] The technical solution of the present invention will be further described below with reference to the accompanying drawings and embodiments.

[0028] The positional relationships depicted in the accompanying drawings are for illustrative purposes only and should not be construed as limiting this patent. Example 1 This embodiment provides a method for evaluating the performance of a hydrological model based on a combination of panel regression and hydrological process analysis. The flowchart of this method can be found here. Figure 1 This includes the following steps: S10: Based on the acquired meteorological data and watershed attributes of each watershed, runoff simulation is performed using a preset hydrological model to obtain the simulated runoff of each watershed; the meteorological data includes the observed and simulated average precipitation of each watershed. S20: Decompose the surface average precipitation observations to obtain the observed precipitation trend sequence, observed precipitation seasonal sequence, and observed precipitation residual sequence for each watershed; decompose the surface average precipitation simulation for each watershed to obtain the simulated precipitation trend sequence, simulated precipitation seasonal sequence, and simulated precipitation residual sequence for each watershed. S30: Based on the sequence obtained in S20, calculate the relative deviation of the simulated precipitation trend sequence, the relative deviation of the simulated precipitation seasonal sequence, the relative deviation of the simulated precipitation residual sequence, and the relative deviation of the simulated runoff for each watershed. Using the relative deviations, combine the hydrological model to simulate runoff and analyze the performance of the preset hydrological model in handling watershed precipitation deviations. S40: Based on the relative deviation, construct a panel regression model to quantitatively evaluate the contribution of the relative deviation to the runoff deviation.

[0029] In the present embodiment, the preset hydrological model can be a commonly used hydrological model, such as a SWAT model, a VIC model, a HEC-HMS model, a HSPF model, etc. In the present embodiment, the hydrological model adopted is HBV a hydrological model. Meanwhile, in the present embodiment, the actual observed runoff of each basin is obtained by an official hydrological monitoring agency.

[0030] In the present embodiment, first, the meteorological data and basin attributes of each basin are obtained, wherein the basin attributes include: climatic indices: average precipitation, average evapotranspiration, precipitation seasonality, snowfall proportion, dryness index, high precipitation event frequency, high precipitation event duration, low precipitation event frequency, low precipitation event duration; topographic features: basin average elevation, basin average slope, basin area; vegetation features: forest proportion, maximum monthly average of leaf area index, difference between maximum and minimum monthly leaf area index, maximum monthly average of green vegetation, difference between maximum and minimum monthly green vegetation proportion, main land use cover type index, main land use cover type, 50% quantile of root depth; soil properties: bedrock depth, soil depth, soil bulk porosity, soil saturated hydraulic conductivity, maximum water content, sand proportion, silt proportion, clay proportion; geological features: most common stratigraphic landform type of the basin, proportion of the most common stratigraphic landform type of the basin, second most common stratigraphic landform type of the basin, proportion of the second most common stratigraphic landform type of the basin, proportion of carbonate sedimentary rock in the area of the basin, underground porosity, underground permeability.

[0031] The meteorological data and basin attributes of each basin are used to drive the preset HBV hydrological model to obtain the simulated runoff. Based on the seasonal trend decomposition method, the areal average precipitation observation and the areal average precipitation simulation are respectively decomposed. The areal average precipitation observation is decomposed to obtain an observed precipitation trend sequence, an observed precipitation seasonal sequence, and an observed precipitation residual sequence. The areal average precipitation simulation is decomposed to obtain a simulated precipitation trend sequence, a simulated precipitation seasonal sequence, and a simulated precipitation residual sequence. Then, the relative deviations of each sequence are respectively calculated, and according to the relative deviations of each sequence, the hydrological process, i.e., the parameters and intermediate variables of the preset hydrological model in the simulation process, are combined to analyze the processing performance of the preset hydrological model on the precipitation deviation of the basin. Finally, the relative deviations of each sequence are taken as panel data, and a fixed effect model is used for panel regression analysis to construct a panel regression model, so that the contributions of the relative deviations to the simulated runoff deviation can be respectively quantitatively evaluated.

[0032] Embodiment 2 In the present embodiment, meteorological data of 190 basins in a certain region are collected.

[0033] In S10, the meteorological data further includes: the surface average temperature and the potential evapotranspiration of each basin, and the surface average precipitation observation and the surface average precipitation simulation of each basin are obtained by: S101: divide the grids inside the basin according to the boundaries of each basin, and calculate the area of each grid; S102: obtain precipitation observation data and precipitation simulation data, and calculate the surface average precipitation observation of each basin according to the precipitation observation data and the surface average precipitation simulation of each basin according to the precipitation simulation data by using the weighted average method with the area of each grid as the weight.

[0034] Specifically, by using the Python language platform, the Geopandas library is used to read the basin boundary vector file, and the Xarray library is used to open the gridded precipitation dataset covering the study area; based on the boundaries of 190 basins, the precipitation grid is spatially cropped, and the weight coefficient of each grid cell is determined by calculating the area proportion of the intersection with the basin; the area weighted average method is used, and the NumPy library is used for matrix operation, the precipitation observation data and the precipitation simulation data of all effective grids in each basin are weighted and fused respectively, and the surface average precipitation observation and the surface average precipitation simulation of each basin are generated; the surface average precipitation observation and the surface average precipitation simulation of each basin are stored in Pandas DataFrame format, and the time resolution is daily, which is used as the standardized input data of the subsequent hydrological model.

[0035] In S10, based on the obtained meteorological data of each basin and the basin attribute of each basin, a preset hydrological model is used for simulation to obtain the simulated runoff of each basin, and the process is: C1: based on the meteorological data, the basin attribute, and the long short-term memory neural network, the static parameters and dynamic parameters of the hydrological model are obtained:

[0036] Wherein, is the static parameter of the basin b; is the dynamic parameter of the basin b; is the basin attribute of the basin b; is the long short-term memory neural network; represents from the first day to the t-th day; is the network parameter; is the meteorological data of the basin b, which includes the surface average precipitation simulation, temperature and potential evapotranspiration of the basin b; C2: driving the hydrological model according to the static parameters and the dynamic parameters, obtaining the simulated runoff of each basin:

[0037] wherein, is the simulated runoff of the basin b, is a preset HBV hydrological model.

[0038] Specifically, a hydrological model is built based on a PyTorch framework HBV In the simulation process, the face average precipitation simulation quantity, the air temperature, and the potential evapotranspiration are taken as the meteorological data input, and the meteorological data input and the basin attribute input jointly form a data input matrix; a long short-term memory neural network is used to establish a nonlinear mapping relationship among the basin attribute, the meteorological data, and the hydrological model parameters, to generate static parameters and dynamic parameters; one year of meteorological data input is added before the starting time of the meteorological data as a warm-up period, to ensure that the initial state of the hydrological model is stable; then the static parameters, the dynamic parameters, and the meteorological data input are used to drive the hydrological model to complete runoff simulation; and finally, the simulated runoff obtained by driving the hydrological model with the face average precipitation simulation quantity is generated.

[0039] In the embodiment, the face average precipitation observation quantity, the air temperature, and the potential evapotranspiration are also taken as the comparative meteorological data input, and the remaining steps are the same as the above process, which will not be described here again. Finally, the comparative simulated runoff obtained by driving the hydrological model with the face average precipitation observation quantity is generated, which is used for subsequent comparative analysis.

[0040] The network parameters of the long short-term memory neural network are optimized by using a gradient descent method, and the expression is as follows:

[0041] wherein, is the network parameter of the long short-term memory neural network after the i-th optimization; is the network parameter of the long short-term memory neural network after the (i+1)-th optimization; is a learning rate; is a loss function; is the gradient of the loss function with respect to the network parameter after the i-th optimization; is the meteorological data, including the face average precipitation simulation quantity, the air temperature, and the potential evapotranspiration; is the observed runoff.

[0042] In the embodiment, when the comparative simulated runoff is trained, the meteorological data used includes the face average precipitation observation quantity, the air temperature, and the potential evapotranspiration.

[0043] The loss function is a weighted combination of root mean square errors, and the expression of the loss function is as follows:

[0044] wherein, is a loss function, is the flow weight coefficient, and respectively represent the simulated and observed runoff of the basin b on the t-th day before transformation, and respectively represent the simulated and observed runoff of the basin b on the t-th day after transformation, , , is 10 -6 , B is the total number of basins, T is the total time length.

[0045] In S20, the observed areal average precipitation is decomposed into the observed precipitation trend sequence, the observed precipitation seasonal sequence, and the observed precipitation residual sequence of each basin, and the simulated areal average precipitation is decomposed into the simulated precipitation trend sequence, the simulated precipitation seasonal sequence, and the simulated precipitation residual sequence of each basin, and the expression is:

[0046] wherein, represents from the 1st day to the t-th day, and are the observed areal average precipitation and the simulated areal average precipitation, and are the observed precipitation trend sequence and the simulated precipitation trend sequence, and are the observed precipitation seasonal sequence and the simulated precipitation seasonal sequence, and are the observed precipitation residual sequence and the simulated precipitation residual sequence.

[0047] Specifically, the seasonal trend decomposition is realized based on the Statsmodels library of the Python platform, and the observed areal average precipitation and the simulated areal average precipitation are respectively decomposed into three components: the trend sequence reflecting the long-term change rule, the seasonal sequence representing the periodic fluctuation, and the residual sequence embodying the short-term fluctuation. As shown in Figure 2 , the horizontal coordinate is the year, and the vertical coordinate is the precipitation, which shows the change of the time sequence of the observed areal average precipitation and the simulated areal average precipitation and their decomposed components in different years in a basin, and the rBias value is used to quantify the deviation of the simulated areal average precipitation at different time scales, wherein the overall deviation of the simulated areal average precipitation sequence is extremely low, but the deviation of the seasonal sequence is as high as-195.14%, which indicates that the overall simulation effect is good, but the deviation of the seasonal sequence is large, which is the main source of the deviation of the simulated runoff of the basin.

[0048] The process of S30 is: S301: calculating the relative deviation of the simulated precipitation trend sequence of each basin The calculation expression is:

[0049] in, This represents the observed precipitation trend sequence of watershed b on day t. Let represent the simulated precipitation trend sequence of watershed b on day t, where b represents the watershed, y represents the hydrological year, and Y represents the length of the hydrological year. S302: Calculate the relative deviation of the simulated seasonal precipitation series for each watershed. The calculation expression is:

[0050] in, This represents the seasonal sequence of observed precipitation in watershed b on day t. Let represent the simulated seasonal precipitation sequence of watershed b on day t, where b represents the watershed, y represents the hydrological year, and Y represents the length of the hydrological year. S303: Calculate the relative deviation of the simulated precipitation residual sequence for each watershed. The calculation expression is:

[0051] in, This represents the residual sequence of observed precipitation in watershed b on day t. Let represent the simulated precipitation residual sequence of watershed b on day t, where b represents the watershed, y represents the hydrological year, and Y represents the length of the hydrological year. S304: Calculate the relative deviation of simulated runoff for each watershed. The calculation expression is:

[0052] in, This represents the observed runoff of watershed b on day t. Let represent the simulated runoff of watershed b on day t, where b represents the watershed, y represents the hydrological year, and Y represents the length of the hydrological year. S305: Based on relative deviation Relative deviation Relative deviation and relative deviation The process of runoff simulation combined with hydrological models is analyzed to examine the way and effect of the preset hydrological model on the precipitation deviation of the watershed.

[0053] Specifically, such as Figure 3 As shown, there are three subplots. The vertical axis represents the cumulative distribution function (CDF) value, and the horizontal axis of subplot a represents the relative deviation. Value, subplot b, the horizontal axis represents the relative deviation. Value, the horizontal axis of subplot c represents the relative deviation. value.

[0054] In the present embodiment, the intermediate variables outputted in the simulation process of the preset hydrological model include evapotranspiration, soil moisture, and base flow index. Figure 4 The correlation analysis diagram of the intermediate variables of the hydrological model driven by precipitation is shown in FIG. 3, which includes three sub-diagrams. The horizontal coordinates of the sub-diagrams are the intermediate variables obtained by driving the hydrological model with the observed average precipitation, and the vertical coordinates are the intermediate variables obtained by driving the hydrological model with the simulated average precipitation. Specifically, the horizontal coordinates of sub-diagram a represent the evapotranspiration obtained by driving the hydrological model with the observed average precipitation, and the vertical coordinates represent the evapotranspiration obtained by driving the hydrological model with the simulated average precipitation; the horizontal coordinates of sub-diagram b represent the soil moisture obtained by driving the hydrological model with the observed average precipitation, and the vertical coordinates represent the soil moisture obtained by driving the hydrological model with the simulated average precipitation; and the horizontal coordinates of sub-diagram c represent the base flow index obtained by driving the hydrological model with the observed average precipitation, and the vertical coordinates represent the base flow index obtained by driving the hydrological model with the simulated average precipitation.

[0055] The parameter analysis diagram of the hydrological model driven by precipitation is shown in FIG. 4. The horizontal coordinates of the sub-diagrams are the results of driving the hydrological model with the observed average precipitation and the results of driving the hydrological model with the simulated average precipitation. The vertical coordinates of sub-diagram a are the ablation critical temperature; the vertical coordinates of sub-diagram b are the degree-day factor; the vertical coordinates of sub-diagram c are the re-freezing coefficient; the vertical coordinates of sub-diagram d are the snow water holding capacity; the vertical coordinates of sub-diagram e are the soil field water holding capacity; the vertical coordinates of sub-diagram f are the evaporation threshold coefficient; the vertical coordinates of sub-diagram g are the infiltration rate; the vertical coordinates of sub-diagram h are the surface runoff recession coefficient; the vertical coordinates of sub-diagram i are the soil flow recession coefficient; the vertical coordinates of sub-diagram j are the underground runoff recession coefficient; the vertical coordinates of sub-diagram k are the upper soil water storage threshold; the vertical coordinates of sub-diagram l are the gamma function concentration shape parameter; and the vertical coordinates of sub-diagram m are the gamma function concentration scale parameter. Through the joint comparison, the performance of the preset hydrological model in processing the precipitation deviation of the basin is analyzed. Figure 3 、 Figure 4 、 Figure 5

[0056] The process S40 is as follows: Based on the relative deviations , , and , the panel data is constructed, the panel regression analysis is performed through the fixed effect model, the panel regression model is constructed, and the fitting regression coefficients , and are obtained, so that the contributions , and of the relative deviations to the simulated runoff deviation are obtained.​ , meet the expression:

[0057] Wherein, is the fixed effect of the basin, is the time fixed effect, is the error term; is the contribution of the precipitation trend sequence, that is, the relative deviation 1% change causes the relative deviation change of the simulated runoff ; is the contribution of the precipitation seasonal sequence, that is, the relative deviation 1% change causes the relative deviation change of the simulated runoff ; is the contribution of the precipitation residual sequence, that is, the relative deviation 1% change causes the relative deviation change of the simulated runoff .

[0058] In this embodiment, the contribution diagram of the precipitation deviation to the runoff relative deviation is as shown in Figure 6 , wherein the abscissa is the simulated precipitation trend sequence, the simulated precipitation seasonal sequence, and the observed precipitation residual sequence, and the ordinate is the contribution degree of the precipitation deviation to the runoff deviation change.

[0059] The analysis of Figure 6 is carried out, the contribution of different precipitation sequence deviations to the runoff deviation is quantified and analyzed, and the influence of the precipitation deviation on the runoff simulation performance of the hydrological model is obtained.

[0060] Embodiment 3 The embodiment provides a hydrological model performance evaluation system based on the combination of panel regression and hydrological process analysis, which is used to realize the hydrological model performance evaluation method based on the combination of panel regression and hydrological process analysis, and the system is referred to Figure 7 , and the system comprises: A precipitation and runoff acquisition module is configured to simulate runoff of each basin by using a preset hydrological model according to meteorological data of each basin and basin attributes of each basin, wherein the meteorological data comprises a surface average precipitation observation and a surface average precipitation simulation of each basin. A precipitation decomposition module is configured to decompose the surface average precipitation observation to obtain an observed precipitation trend sequence, an observed precipitation seasonal sequence and an observed precipitation residual sequence of each basin, and decompose the surface average precipitation simulation of each basin to obtain a simulated precipitation trend sequence, a simulated precipitation seasonal sequence and a simulated precipitation residual sequence of each basin. The relative deviation analysis module is configured to calculate, based on the sequence output by the precipitation decomposition module, a relative deviation of the simulated precipitation trend sequence of each basin, a relative deviation of the simulated precipitation seasonal sequence of each basin, a relative deviation of the simulated precipitation residual sequence of each basin, and a relative deviation of the simulated runoff of each basin, and use the relative deviations to analyze the processing performance of the preset hydrological model on the precipitation deviation of the basin in combination with the process of runoff simulation by the hydrological model. The panel regression module is configured to construct a panel regression model based on the relative deviations and quantitatively evaluate the contribution of the relative deviations to the runoff deviation, respectively.

[0061] The embodiment also provides an electronic device, which includes a memory, a processor, and a program stored in the memory and capable of running on the processor.

[0062] The embodiment also provides an electronic device, which includes a processor, a memory, and a program stored in the memory and capable of running on the processor.

[0063] Specifically, referring to Figure 8 The embodiment also provides an electronic device, which includes a bus 401, a transceiver 402, an antenna 403, a bus interface 404, a processor 405, and a memory 406.

[0064] The transceiver 402 is configured to acquire at least one of meteorological data of each basin, basin attributes, and observed runoff. The processor 405 is configured to drive a preset hydrological model by using the meteorological data of each basin and the basin attributes to obtain simulated runoff.

[0065] In Figure 8In particular embodiments, bus architecture (represented generally by the bus 401) links various circuits including one or more processors represented by the processor 405 and memory represented by the memory 406. The bus 401 can also link various other circuits such as peripheral devices, voltage regulators, and power management circuitry, all of which are well known in the art and thus, not further described herein. The bus interface 404 provides an interface between the bus 401 and the transceiver 402. The transceiver 402, which can be a single element or multiple elements such as multiple receivers and transmitters, provides a means for communicating with various other apparatus over a transmission medium. Data processed by the processor 405 is transmitted over a wireless medium via the antenna 403, which further receives data and communicates the data to the processor 405.

[0066] The processor 405 is responsible for managing the bus 401 and general processing, and can also provide various functions including timing, peripheral interfaces, voltage regulation, power management, and other control functions. The memory 406 can be used to store data used by the processor 405 during execution of operations.

[0067] Optionally, the processor 405 can be a central processing unit (CPU), an application specific integrated circuit (ASIC), a field programmable gate array (FPGA) or complex programmable logic device (CPLD).

[0068] The embodiments are merely exemplary and are not to be construed as limiting the scope of the application. Those skilled in the art will be able to ascertain many modifications without departing from the scope of the underlying principles. All references, patents and patent applications cited herein are hereby incorporated by reference. The embodiments shown and described are only examples of the application and choice of equivalent techniques should be interpreted as possible applications of the principles of the application. Changes can be made in the function and arrangement of elements without departing from the scope of the application as recited in the claims.

Claims

1. A hydrological model performance evaluation method based on panel regression combined with hydrological process analysis, characterized in that, The method comprises the following steps: S10: According to the obtained meteorological data of each basin and the basin attributes of each basin, the runoff simulation is performed by using a preset hydrological model to obtain the simulated runoff of each basin; the meteorological data comprises a surface average precipitation observation and a surface average precipitation simulation of each basin; S20: The surface average precipitation observation is decomposed to obtain an observed precipitation trend sequence, an observed precipitation seasonal sequence and an observed precipitation residual sequence of each basin; the surface average precipitation simulation of each basin is decomposed to obtain a simulated precipitation trend sequence, a simulated precipitation seasonal sequence and a simulated precipitation residual sequence of each basin; S30: Based on the sequences obtained in S20, the relative deviation of the simulated precipitation trend sequence of each basin, the relative deviation of the simulated precipitation seasonal sequence of each basin, the relative deviation of the simulated precipitation residual sequence of each basin and the relative deviation of the simulated runoff of each basin are calculated respectively; the relative deviations are used to analyze the processing performance of the preset hydrological model on the precipitation deviation of the basin in combination with the process of runoff simulation by using the hydrological model; S40: Based on the relative deviations, a panel regression model is constructed to quantitatively evaluate the contribution of the relative deviations to the runoff deviation.

2. The method for evaluating the performance of a hydrological model based on the combination of panel regression and hydrological process analysis according to claim 1, characterized in that, In S10, the meteorological data further comprises a temperature of each basin and a potential evapotranspiration of each basin; and the surface average precipitation observation and the surface average precipitation simulation of each basin are obtained by: S101: Dividing a grid located in the interior of the basin according to the boundary of each basin; and calculating the area of each grid; S102: Obtaining precipitation observation data and precipitation simulation data; and calculating the surface average precipitation observation of each basin from the precipitation observation data by using a weighted average method with the area of each grid as a weight; and calculating the surface average precipitation simulation of each basin from the precipitation simulation data.

3. The method of claim 1, wherein the method is characterized by, In S10, the simulated runoff of each basin is obtained by using a preset hydrological model based on the obtained meteorological data of each basin and the basin attributes of each basin; and the process is as follows: C1: obtaining static parameters and dynamic parameters of a hydrological model based on the meteorological data, the basin attribute, and a long short-term memory neural network; wherein, is the static parameter of the basin b; is the dynamic parameter of the basin b; is the basin attribute of the basin b; is the long short-term memory neural network; represents from the 1st day to the tth day; is the network parameter; is the meteorological data of the basin b, including the areal mean precipitation simulation amount, air temperature and potential evapotranspiration of the basin b; C2: driving the hydrological model according to the static parameters and the dynamic parameters to obtain simulated runoff of each watershed: wherein, is the simulated runoff for the catchment b, is a predetermined HBV hydrological model.

4. The method for evaluating the performance of a hydrological model based on the combination of panel regression and hydrological process analysis according to claim 3, characterized in that, network parameters of the long short-term memory neural network The gradient descent method is applied for optimization, and the expression is: wherein, is the network parameter of the long short-term memory neural network after the i-th optimization; is the network parameter of the long short-term memory neural network after the i+1-th optimization; is the learning rate; is the loss function; is the gradient of the loss function with respect to the network parameter after the i-th optimization; is the meteorological data, including the surface mean precipitation simulation, air temperature and potential evapotranspiration; is the observed runoff.

5. The method for evaluating the performance of a hydrological model based on the combination of panel regression and hydrological process analysis according to claim 4, characterized in that, The loss function is a weighted combination of root mean square errors, and the loss function expression is: in, For loss function, For traffic weighting coefficients, and Let these represent the simulated runoff and observed runoff of watershed b on day t, respectively, before the transformation. and Let these represent the simulated runoff and observed runoff of the transformed watershed b on day t, respectively. , , 10 -6 , B The total number of watersheds, T This represents the total time duration.

6. The method of claim 1, wherein the method is characterized by, In S20, the areal average precipitation observation is decomposed to obtain an observed precipitation trend sequence, an observed precipitation seasonal sequence, and an observed precipitation residual sequence of each basin, and the areal average precipitation simulation quantity is decomposed to obtain a simulated precipitation trend sequence, a simulated precipitation seasonal sequence, and a simulated precipitation residual sequence of each basin, and the expression is as follows: wherein, denotes from day 1 to day t, and are the surface mean precipitation observation and the surface mean precipitation simulation, respectively, and are the observed precipitation trend series and the simulated precipitation trend series, respectively, and are the observed precipitation seasonal series and the simulated precipitation seasonal series, respectively, and are the observed precipitation residual series and the simulated precipitation residual series, respectively.

7. The method of claim 6, wherein the method is characterized by, The process of S30 is as follows: S301: Calculate the relative deviation of the simulated precipitation trend sequence of each basin The calculation expression is: wherein, represents the observed precipitation trend series of the basin b on the t-th day, represents the simulated precipitation trend series of the basin b on the t-th day, b represents a basin, y represents a hydrological year, and Y represents the time length of a hydrological year; S302: Calculate the relative deviation of the seasonal sequence of simulated precipitation of each basin The calculation expression is: wherein, represents the observed precipitation seasonal sequence for basin b on day t, represents the simulated precipitation seasonal sequence for basin b on day t, b represents a basin, y represents a hydrological year, and Y represents the length of a hydrological year; S303: Calculate the relative deviation of the simulation precipitation residual sequence of each basin The calculation expression is: wherein, represents the observed precipitation residual series of basin b on day t, represents the simulated precipitation residual series of basin b on day t, b represents a basin, y represents a hydrological year, and Y represents the length of a hydrological year; S304: Calculate the relative deviation of the simulated runoff of each watershed The calculation expression is: wherein, Qb,t represents the observed runoff of the catchment b on day t, Qb,t represents the simulated runoff of the catchment b on day t, b represents the catchment, y represents the hydrological year, and Y represents the length of the hydrological year; S305: based on the relative deviation , the relative deviation , the relative deviation and the relative deviation , combined with the hydrological model to simulate the runoff process, analyze the processing method and effect of the preset hydrological model on the basin precipitation deviation.

8. The method of claim 7, wherein the method is characterized by, The process of S40 is as follows: based on the relative deviation , the relative deviation , the relative deviation and the relative deviation constructing panel data, performing panel regression analysis by fixed effect model, constructing panel regression model, fitting regression coefficients , and , the contribution of the relative deviation to the simulated runoff deviation , and , satisfy the expression: ​ where, is the basin fixed effect, is the time fixed effect, is the error term; is the contribution of the trend of the precipitation series, i.e. the relative bias change in the simulated runoff relative bias for a 1% change ; is the contribution of the seasonal of the precipitation series, i.e. the relative bias change in the simulated runoff relative bias for a 1% change ; is the contribution of the residual of the precipitation series, i.e. the relative bias change in the simulated runoff relative bias for a 1% change .

9. A hydrological model performance evaluation system based on panel regression combined with hydrological process analysis, for implementing the hydrological model performance evaluation method based on panel regression combined with hydrological process analysis according to any one of claims 1-8, characterized in that, The system comprises: a precipitation and runoff obtaining module, configured to obtain the simulated runoff of each basin by using a preset hydrological model according to the obtained meteorological data of each basin and the basin attributes of each basin; the meteorological data comprises a surface average precipitation observation and a surface average precipitation simulation of each basin; a precipitation decomposing module, configured to decompose the surface average precipitation observation to obtain an observed precipitation trend sequence, an observed precipitation seasonal sequence and an observed precipitation residual sequence of each basin; and decompose the surface average precipitation simulation of each basin to obtain a simulated precipitation trend sequence, a simulated precipitation seasonal sequence and a simulated precipitation residual sequence of each basin; and a panel regression model is constructed to quantitatively evaluate the contribution of the relative deviations to the runoff deviation. The relative deviation analysis module is configured to calculate, based on the sequence output by the precipitation decomposition module, a relative deviation of the simulated precipitation trend sequence of each basin, a relative deviation of the simulated precipitation seasonal sequence of each basin, a relative deviation of the simulated precipitation residual sequence of each basin, and a relative deviation of the simulated runoff of each basin, and to analyze, by using the relative deviations, a processing performance of a preset hydrological model on precipitation deviation of a basin in combination with a runoff simulation process of the hydrological model. The panel regression module is configured to construct a panel regression model based on the relative deviations, and to quantitatively evaluate a contribution of the relative deviations to runoff deviation.

10. An electronic device comprising: The memory, the processor, and a program stored in the memory and capable of running on the processor; and the processor is configured to read the program in the memory to implement the steps in the hydrological model performance evaluation method based on panel regression and hydrological process analysis combination according to any one of claims 1 to 8.

Images