Method and system for determining parameterization scheme of regional climate model

Abstract

The invention discloses a parameterization scheme determination method and system for a regional climate model, and relates to the technical field of climate simulation. Comprising the following steps: determining all parameterization schemes corresponding to all physical processes of a target region by adopting a regional climate model; determining a variation coefficient of an average relative error and an absolute error of simulation results of all parameterization schemes of each physical process on different meteorological variables in the target area; dynamically generating a first weight value and a second weight value by adopting a Softmax function based on the two coefficients of each variable; calculating a comprehensive statistical index of each parameterization scheme, and selecting the optimal index as the optimal scheme of the physical process; and combining the optimal schemes of all the physical processes to obtain an overall optimal parameterization scheme combination of the target area. According to the method, through a double-layer weighting mechanism, key variables influencing the overall performance of the model and variables most sensitive to a specific physical process can be automatically identified.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of climate simulation, and in particular to a parameterization scheme determination method and system for a regional climate model. BACKGROUND

[0002] Earth System Models (ESM), especially Weather Research and Forecasting (WRF) and other Regional Climate Models (RCM), play a core role in climate change research, weather forecasting and related fields. One of the main features of the WRF model is that it provides a large number of parameterization scheme options for various physical processes (such as Microphysics_schemes, land surface processes, radiation, cumulus convection, etc.). In theory, these schemes can be combined into millions of different configurations. However, different scheme combinations have a significant impact on the simulation results of meteorological elements, and there is no universally optimal scheme combination that can be applied to all regions, all times and all variables. Therefore, how to select the best parameterization scheme combination for a specific research area is a key technical problem that must be solved before applying the WRF model, but it also introduces great uncertainty and computational challenges for model application.

[0003] To overcome the one-sidedness of single-variable evaluation, some researchers have proposed multi-variable comprehensive evaluation methods such as Model Climate Performance Index (MCPI) and Multi-Variable Integrated Evaluation (MVIE). These methods usually give equal weights to all variables involved in the evaluation or subjectively assign weights based on the researchers' experience.

[0004] Different meteorological variables have different sensitivities to different physical processes (such as microphysical processes and cumulus convection processes), and equal weight processing will dilute or hide the huge deviation of key constraint variables (such as precipitation and wind speed) that are simulated poorly but are crucial to the overall climate in the process of averaging, making the evaluation result distorted and leading to the selection of a scheme combination that is not the optimal solution in terms of physical mechanisms. SUMMARY

[0005] Therefore, it is necessary to provide a parameterization scheme determination method and system for a regional climate model to solve the above technical problems.

[0006] The parameterization scheme determination method for a regional climate model provided by the embodiments of the present application comprises the following steps. The regional climate model is used to determine all parameterization schemes corresponding to all physical processes of the target region, wherein the physical processes are used to describe the atmospheric physical mechanism existing in the target region, and the parameterization schemes are used to simulate the sub-grid scale process of the regional climate model; The coefficients of variation of the average relative errors and the absolute errors of the simulation results of different meteorological variables in the target region by all parameterization schemes of each physical process are determined, wherein the average relative errors represent the simulation effects of different meteorological variables by each parameterization scheme, and the coefficients of variation of the absolute errors represent the sensitivities of different meteorological variables to different parameterization schemes; Based on the coefficients of variation of the average relative errors and the absolute errors of the simulation results of each meteorological variable, a Softmax function is used as a weight conversion function to dynamically generate a first weight value and a second weight value for each meteorological variable, and the first weight value and the second weight value of each meteorological variable are used to determine the comprehensive statistical indicators of all parameterization schemes corresponding to each physical process; The parameterization scheme corresponding to the optimal comprehensive statistical indicator is used as the optimal parameterization scheme of each physical process, and all optimal parameterization schemes corresponding to all physical processes are combined to obtain the overall optimal parameterization scheme of the target region.

[0007] Optionally, the average relative errors of the simulation results of different meteorological variables in the target region by all parameterization schemes of each physical process are determined based on the following formula, and the average relative errors are used to describe the differences between the simulation values and the measured values of each meteorological variable by each parameterization scheme: ; The coefficients of variation of the absolute errors of the simulation results of different meteorological variables in the target region by all parameterization schemes of each physical process are determined based on the following formula, and the coefficients of variation of the absolute errors are used to describe the discrete degrees of the parameterization schemes in simulating each meteorological variable: ; wherein, the average relative error is, the coefficient of variation of the absolute error is, the simulation value of the meteorological variable of the parameterization scheme at the discrete point is, the measured value of the meteorological variable at the discrete point is, the parameterization scheme is, the discrete point is, the meteorological variable is, N the total number of parameterization schemes is, T the total number of discrete points is.

[0008] Optionally, based on the average relative error and the coefficient of variation of the absolute error of the simulation result of each meteorological variable, a Softmax function is used as a weight conversion function to dynamically generate a first weight value and a second weight value for each meteorological variable, specifically including: The Softmax function is used to convert the average relative error and the coefficient of variation of the absolute error into the first weight value and the second weight value respectively, the value range of which is between 0 and 1 and the sum of all variable weights is 1; The average relative error is converted into the first weight value based on the following formula: ; The coefficient of variation of the absolute error is converted into the second weight value based on the following formula: ; Wherein, is the average relative error, is the coefficient of variation of the absolute error, is the first weight value, is the second weight value, N is the total number of parameterization schemes, is the parameterization scheme.

[0009] Optionally, the physical processes include but are not limited to microphysical processes, cumulus convection processes, planetary boundary layer processes, land surface processes and radiation processes.

[0010] Optionally, the comprehensive statistical indicators include but are not limited to root mean square error, Pearson correlation coefficient, standard deviation and average impact error index.

[0011] The embodiment of the application provides a parameterization scheme determination system of a regional climate model, comprising: A parameterization scheme module is configured to determine all parameterization schemes corresponding to all physical processes of a target region by using a regional climate model, wherein the physical processes are used to describe atmospheric physical mechanisms existing in the target region, and the parameterization schemes are used to simulate sub-grid scale processes of the regional climate model; An index determination module is configured to determine the average relative error and the coefficient of variation of the absolute error of simulation results of different meteorological variables in the target region by using all parameterization schemes of each physical process, wherein the average relative error represents the simulation effect of each parameterization scheme on different meteorological variables, and the coefficient of variation of the absolute error represents the sensitivity of each meteorological variable to different parameterization schemes; The index comprehensive module is configured to convert a Softmax function as a weight conversion function based on a coefficient of variation of the average relative error and the absolute error of the simulation result of each meteorological variable, dynamically generate a first weight value and a second weight value for each meteorological variable, respectively, and determine a comprehensive statistical index of all parameterization schemes corresponding to each physical process according to the first weight value and the second weight value of each meteorological variable. The summary module is configured to take the parameterization scheme corresponding to the optimal comprehensive statistical index as the optimal parameterization scheme of each physical process, and combine the optimal parameterization schemes corresponding to all physical processes to obtain the overall optimal parameterization scheme of the target region.

[0012] The above-mentioned parameterization scheme determination method and system of the regional climate model provided by the embodiments of the present application have the following beneficial effects compared with the prior art. The present application introduces a double-layer dynamic weighting mechanism, calculates the coefficient of variation of the average relative error and the absolute error of each meteorological variable, and converts the two indexes into weight values by using the Softmax function. This process is based on the weight calculation of the simulation data and the observation data of the model, automatically identifies the key variables affecting the overall performance of the model and the meteorological variables most sensitive to the specific physical process, and gives more attention, so as to occupy a more important position in the comprehensive evaluation, and avoids the problem that the key error is diluted or hidden in the existing equal weight method.

[0013] In addition, the weight of the meteorological variable is adjusted independently and dynamically in the parameterization process of each physical process, fully considering the response difference of different meteorological variables to different physical processes, enhancing the physical rationality and pertinence of the evaluation. This not only improves the simulation capability of the regional climate model on the key variables and extreme events, but also enhances the applicability and robustness of the parameterization in different target regions, realizing the leap from the average optimal to the mechanism optimal. BRIEF DESCRIPTION OF DRAWINGS

[0014] Figure 1 FIG. 1 is a flowchart of a parameterization scheme determination method of a regional climate model according to an embodiment of the present application. DETAILED DESCRIPTION

[0015] In order to make the purpose, technical scheme and advantages of the present application clearer, the present application will be further described in detail below in combination with the drawings and embodiments. It should be understood that the specific embodiments described herein are only used to explain the present application and do not limit the present application.

[0016] Earth System Models (ESM) and Regional Climate Models (RCM) are indispensable scientific tools in the detection of climate change, weather forecasting, prediction of future climate, and adaptation research. These models simulate the interactions between the atmosphere, ocean, land, and cryosphere through complex mathematical and physical equations.

[0017] Among them, the Weather Research and Forecasting (WRF) model is a widely used new generation of mesoscale numerical model internationally. Due to its advanced dynamic core, rich physical process options, and good community support, the WRF model is not only widely used for operational weather forecasting, but also plays a core role in the Coordinated Regional Climate Downscaling Experiments (CORDEX) promoted by the World Climate Research Program (WCRP), providing high-resolution climate change predictions for various regions around the world.

[0018] A core technical challenge of numerical models is that, due to limitations in computing power, many key physical processes (such as cloud formation and dissipation, turbulence in the atmospheric boundary layer, cumulus convection, radiation transfer, etc.) have scales much smaller than the model's grid resolution, which cannot be directly resolved. To reflect the influence of these "sub-grid scale" processes in the model, scientists have developed parameterization schemes. Parameterization schemes use resolvable large-scale variables (such as temperature, humidity, wind field) to approximately and empirically express the overall effect of these small-scale processes.

[0019] One of the main features of the WRF model is its high flexibility and modularity, which provides users with multiple different parameterization schemes for each key physical process. For example, there are dozens of schemes for microphysical processes, and there are also multiple schemes for planetary boundary layer, cumulus convection, and land surface processes. Each scheme is based on different physical assumptions and mathematical expressions. Users can combine these schemes, theoretically generating millions or even tens of millions of different model configurations.

[0020] However, this flexibility also brings great uncertainty: (1) Sensitivity of results: choosing different combinations of parameterization schemes will produce significant, even fundamental differences in the simulation results of key meteorological elements such as temperature, precipitation, and wind speed.

[0021] (2) No universal optimal solution: There is no "universal" or generally optimal parameterization scheme combination that can be applied to all geographical regions, all time scales, and all weather and climate phenomena. Selecting the most suitable scheme combination for a specific region and research goal is a necessary prerequisite to ensure the reliability of the simulation results.

[0022] Currently, there are mainly the following technical solutions and their inherent limitations to solve the above problems: 1. Single variable evaluation method.

[0023] Researchers usually focus on the model's simulation ability for a single key variable (such as precipitation or temperature), and evaluate it by calculating indicators such as correlation coefficient (R) and root mean square error (RMSE). For example, Taylor Diagram can integrate multiple statistical indicators into a diagram for visual comparison.

[0024] Defects: This method ignores the complex coupling and interaction between variables in the climate system. Over-optimizing the simulation accuracy of a single variable often leads to a decline in the simulation performance of other key variables, thus failing to comprehensively and objectively reflect the overall performance of the model.

[0025] 2. Multi-variable equal weight or subjective weight evaluation method: To overcome the one-sidedness of single variable evaluation, researchers have proposed multi-variable comprehensive evaluation methods such as Model Climate Performance Index (MCPI) and Multi-Variable Integrated Evaluation (MVIE). These methods usually assign equal weights to all variables involved in the evaluation, or subjectively assign weights based on the researchers' experience.

[0026] Defects: (1) Covering up key error sources: Equal weight processing can "dilute" or "cover up" the large deviations of "key constraint variables" (such as precipitation and wind speed) that are simulated poorly but are crucial to the overall climate, resulting in distorted evaluation results and ultimately selecting a scheme combination that is not the optimal solution in terms of physical mechanisms.

[0027] (2) Ignoring sensitivity differences: Different meteorological variables have different sensitivities to different physical processes (such as microphysical processes and cumulus convection processes). For example, precipitation is much more sensitive to changes in microphysical schemes than temperature. Existing methods use the same set of fixed weights when evaluating all physical processes, which does not conform to physical laws and cannot objectively attribute the sources of model uncertainty, leading to the accumulation and transmission of errors in the sequential evaluation process.

[0028] 3. Sequential sensitivity analysis method: To reduce the huge computational cost of evaluating all combinations, researchers usually use the method of adjusting and selecting the optimal parameterization scheme for each physical process one by one.

[0029] Limitations: The accuracy of this method is highly dependent on the accuracy of each evaluation step. Since the "optimal" solution selected in the previous physical process becomes the basis for subsequent evaluations, any deviations in early evaluations will be propagated and amplified, affecting the reliability of the final combination.

[0030] This invention provides a method for determining the parameterization scheme of a regional climate model, the method comprising: All parameterization schemes corresponding to all physical processes in the target region are determined using a regional climate model. The physical processes describe the atmospheric physical mechanisms existing in the target region, while the parameterization schemes simulate the sub-grid-scale processes of the regional climate model.

[0031] The mean relative error and coefficient of variation of absolute error for all parameterization schemes of each physical process in the simulation results of different meteorological variables in the target area are determined. The mean relative error characterizes the simulation effect of each parameterization scheme on different meteorological variables, while the coefficient of variation of absolute error characterizes the sensitivity of each meteorological variable to different parameterization schemes.

[0032] Based on the coefficients of variation of the average relative error and absolute error of the simulation results for each meteorological variable, the Softmax function is used as the weight transformation function to dynamically generate a first weight value and a second weight value for each meteorological variable. Then, based on the first and second weight values ​​for each meteorological variable, the comprehensive statistical index of all parameterization schemes corresponding to each physical process is determined.

[0033] The parameterization scheme corresponding to the optimal comprehensive statistical index is used as the optimal parameterization scheme for each physical process, and the optimal parameterization schemes corresponding to all physical processes are combined to obtain the overall optimal parameterization scheme for the target area.

[0034] The core idea of ​​this method is that the weights are not fixed or subjectively set, but are dynamically calculated and generated from simulated and observed data based on the model's actual performance during the evaluation process. This method can be used in conjunction with various multivariate performance indicators (such as MIEI).

[0035] The specific implementation is as follows: I. Define the statistical basis for weight calculation.

[0036] The dynamic weighting method proposed in this study is mainly used to obtain the variable weights in the above evaluation indicators. (i.e., the first and second weight values), this is an objective weighting method, independent of the statistical indicators used, and applicable to all multi-objective evaluation methods. The foundation of this method mainly consists of two statistical indicators: the mean relative error (MRE). ) and the coefficient of variation of absolute error ( ).

[0037] Mean Relative Error (MRE): used to describe the difference between the simulated and observed values of each meteorological variable for each parameterization scheme, representing the simulation ability of a certain physical parameterization process for the meteorological variable.

[0038] (1) (2) (3) Coefficient of Variation of Absolute Error (CVAE): used to describe the dispersion of the parameterization scheme in simulating each meteorological variable, representing the sensitivity of a certain meteorological variable to the physical parameterization process.

[0039] (4) (5) where, is the simulated value of the meteorological variable at the discrete point for the parameterization scheme, is the observed value of the meteorological variable at the discrete point, indicates the absolute error of the meteorological variable in the parameterization scheme, is the parameterization scheme, is the discrete point, is the meteorological variable, is the total number of parameterization schemes, N is the total number of discrete points. T

[0040] II. Weight Transformation

[0041] The core is to give higher weight values to variables with larger mean relative errors (worse simulation effects) or higher coefficients of variation (more sensitive to the physical process) in the simulation process. Therefore, the function is used to convert the above two statistical indicators of the variable into the corresponding 0-1 weight value of each variable. The Softmax function is an activation function commonly used in classification tasks, which converts the original numerical output into a probability distribution, with each output value between 0 and 1 and the sum of all outputs being 1. In addition, it scales the input through the exponential function, making the impact of larger numerical values on the output more significant, which is very suitable for the requirement that the weight needs to be more sensitive to the difference between variables in this study. It is worth noting that the form of the exponential function is not fixed, and in actual application, any form of function can be used to scale the statistical results according to the actual data structure.

[0042] ​​​ wherein, is the mean relative error, is the coefficient of variation of absolute error, is the first weight value, is the second weight value.

[0043] III. Sequential evaluation and preferred implementation process.

[0044] In the sequential evaluation mode, the optimal scheme of each physical process of the model is screened one by one. The specific application process in the WRF simulation evaluation includes the following steps: Step 1: Run all parameterization schemes in the first physical parameterization process (for example, Microphysics_schemes), and calculate the absolute error and relative error of the simulation results of all schemes for different variables and the observed values; Step 2: Calculate the mean relative error and the coefficient of variation of absolute error of the simulation results of different variables to quantify the error degree and sensitivity of different variables; Step 3: Use the Softmax function to convert the two statistical indicators calculated above into two sets of weights as the importance measure of each variable in the simulation process; Step 4: According to the obtained weights, further calculate the comprehensive statistical indicators of each scheme, such as Root Mean Square Error (RMSE), Pearson Correlation Coefficient (R), Standard Deviation (SD), Mean Integral Square Error (MIEI), etc. (select according to the research area and research target, for example, select different comprehensive statistical indicators if you want to be closer in value or more consistent in frequency), and then determine the best scheme in the first physical parameterization process.

[0045] Step 5: Use the best scheme determined by Step 4 to continue the simulation of the next physical parameterization process, and repeat the above steps until the best scheme of all physical parameterization processes is determined.

[0046] Finally, the scheme set composed of the optimal schemes of each physical process is the overall optimal parameterization scheme combination of the model screened by the method of the present application.

[0047] IV. Beneficial effects.

[0048] 1. Objectivity and scientificity.

[0049] The application calculates the weight based on model simulation data and observation data, eliminates subjectivity and randomness of human setting weight in the evaluation process, and makes the evaluation result more objective, reproducible and scientific.

[0050] 2. Strong pertinence, can grasp the main contradiction.

[0051] Through the double-layer weighting mechanism, the method can automatically identify the key variables affecting the overall performance of the model and the variables most sensitive to the specific physical process, and give higher attention. This makes the optimization process more targeted and can effectively grasp the main source of uncertainty.

[0052] 3. Better performance, more robust results.

[0053] Through the application in the arid area of northwest China and the coastal area of southeast China, and compared with the traditional equal weight method, the parameterization scheme combination screened out by the application performs more stably and better in the simulation effect in the independent extreme climate year verification, proving the universality and effectiveness of the application in different climate backgrounds.

[0054] 4. Flexibility and universality.

[0055] The dynamic weight calculation core of the application is an independent and objective weighting method, which can be flexibly combined with any multivariate comprehensive statistical index (such as MIEI, RMSE, etc.), is not limited to a specific evaluation framework, and can be applied to the evaluation work of WRF and other complex numerical models with multiple processes, multiple scheme options and multiple output variables.

[0056] Based on the same inventive concept, an embodiment of the application provides a parameterization scheme determination system of a regional climate model, which comprises: A parameterization scheme module is configured to determine all parameterization schemes corresponding to all physical processes of a target region by using a regional climate model. The physical processes are used to describe the atmospheric physical mechanisms existing in the target region, and the parameterization schemes are used to simulate the sub-grid scale processes of the regional climate model.

[0057] An index determination module is configured to determine the variation coefficients of the average relative errors and the absolute errors of the simulation results of all parameterization schemes of each physical process on different meteorological variables in the target region. The average relative errors represent the simulation effects of different parameterization schemes on different meteorological variables, and the variation coefficients of the absolute errors represent the sensitivities of different meteorological variables to different parameterization schemes.

[0058] The index comprehensive module is configured to convert the Softmax function as a weight conversion function based on the average relative error and the coefficient of variation of the absolute error of the simulation result of each meteorological variable, and dynamically generate a first weight value and a second weight value for each meteorological variable respectively. According to the first weight value and the second weight value of each meteorological variable, the comprehensive statistical index of all parameterization schemes corresponding to each physical process is determined.

[0059] The summary module is configured to take the parameterization scheme corresponding to the optimal comprehensive statistical index as the optimal parameterization scheme of each physical process, and combine the optimal parameterization schemes corresponding to all physical processes to obtain the overall optimal parameterization scheme of the target region.

[0060] The above embodiments only express several embodiments of the present application, and the description is relatively specific and detailed, but it cannot be understood as a limitation on the scope of the patent. It should be noted that for ordinary skilled persons in the art, without departing from the concept of the present application, a number of modifications and improvements can be made, which are within the scope of protection of the present application.

Claims

1. A method for determining the parameterization scheme of a regional climate model, characterized in that, include: All parameterization schemes corresponding to all physical processes in the target region are determined using a regional climate model; wherein, the physical processes are used to describe the atmospheric physical mechanisms existing in the target region, and the parameterization schemes are used to simulate the subgrid-scale processes of the regional climate model; Determine the average relative error and the coefficient of variation of the absolute error of all parameterization schemes for each physical process for the simulation results of different meteorological variables in the target area; wherein, the average relative error characterizes the simulation effect of each parameterization scheme on different meteorological variables, and the coefficient of variation of the absolute error characterizes the sensitivity of each meteorological variable to different parameterization schemes. Based on the coefficient of variation of the average relative error and absolute error of the simulation results for each meteorological variable, the Softmax function is used as the weight transformation function to dynamically generate the first weight value and the second weight value for each meteorological variable; and based on the first weight value and the second weight value for each meteorological variable, the comprehensive statistical index of all parameterization schemes corresponding to each physical process is determined. The parameterization scheme corresponding to the optimal comprehensive statistical index is used as the optimal parameterization scheme for each physical process, and the optimal parameterization schemes corresponding to all physical processes are combined to obtain the overall optimal parameterization scheme for the target area.

2. The method for determining the parameterization scheme of a regional climate model as described in claim 1, characterized in that, The average relative error of the simulation results of all parameterization schemes for each physical process on different meteorological variables in the target area is determined based on the following formula, whereby the average relative error is used to describe the difference between the simulated and measured values ​​of each meteorological variable for each parameterization scheme: ； The coefficient of variation of the absolute error of the simulation results of all parameterization schemes for each physical process on different meteorological variables in the target area is determined based on the following formula, wherein the coefficient of variation of the absolute error is used to describe the degree of dispersion of the parameterization scheme when simulating each meteorological variable: ； in, The average relative error, The coefficient of variation is the absolute error. The parameterized scheme represents the simulated values ​​of meteorological variables at discrete points. These are the measured values ​​of meteorological variables at discrete points. For parameterization schemes, For discrete points, For meteorological variables, N The total number of parameterization schemes, T The total number of discrete points.

3. The method for determining the parameterization scheme of a regional climate model as described in claim 1, characterized in that, The coefficients of variation of the average relative error and absolute error of the simulation results based on each meteorological variable are used. The Softmax function is used as the weighting transformation function to dynamically generate a first weight value and a second weight value for each meteorological variable, specifically including: The Softmax function is used to convert the coefficients of variation of the average relative error and the absolute error into first weight values ​​and second weight values, respectively, which are in the range of 0 to 1 and the sum of the weights of all variables is 1. The average relative error is converted into a first weight value based on the following formula: ； The coefficient of variation of the absolute error is converted into a second weight value based on the following formula; ； in, The average relative error, The coefficient of variation is the absolute error. The first weight value, This is the second weight value. N The total number of parameterization schemes, This is a parameterized scheme.

4. The method for determining the parameterization scheme of a regional climate model as described in claim 1, characterized in that, The physical processes include, but are not limited to: microphysical processes, cumulus convection processes, planetary boundary layer processes, land surface processes, and radiation processes.

5. The method for determining the parameterization scheme of a regional climate model as described in claim 1, characterized in that, The comprehensive statistical indicators include, but are not limited to: root mean square error, Pearson correlation coefficient, standard deviation, and average impact error index.

6. A system for determining the parameterization scheme of a regional climate model, characterized in that, include: The parameterization scheme module is used to determine all parameterization schemes corresponding to all physical processes in the target region using a regional climate model; wherein, the physical processes are used to describe the atmospheric physical mechanisms existing in the target region, and the parameterization schemes are used to simulate the subgrid-scale processes of the regional climate model; The index determination module is used to determine the average relative error and the coefficient of variation of the absolute error of the simulation results of all parameterization schemes for each physical process on different meteorological variables in the target area; wherein, the average relative error characterizes the simulation effect of each parameterization scheme on different meteorological variables, and the coefficient of variation of the absolute error characterizes the sensitivity of each meteorological variable to different parameterization schemes. The index synthesis module is used to dynamically generate a first weight value and a second weight value for each meteorological variable based on the coefficient of variation of the average relative error and absolute error of the simulation results of each meteorological variable, using the Softmax function as the weight transformation function; and to determine the comprehensive statistical index of all parameterization schemes corresponding to each physical process based on the first weight value and the second weight value of each meteorological variable. The summary module is used to take the parameterization scheme corresponding to the best comprehensive statistical index as the best parameterization scheme for each physical process, and combine the best parameterization schemes corresponding to all physical processes to obtain the overall best parameterization scheme for the target area.

Images