Large-scale driving field reconstruction adjustment method for convective scale ensemble forecast

Abstract

The invention discloses a flow scale ensemble forecasting large-scale driving field reconstruction adjustment method, which comprises the following steps of: performing dynamic downscaling deterministic forecasting on a global ensemble forecasting system driving field and an EC fine grid to obtain a driving disturbance field and control forecasting of a flow scale ensemble forecasting system; acquiring forecasting element values of the two modes in forecasting time efficiency, performing multi-sphere screening, calculating a forecasting statistical value, and performing polynomial fitting according to the forecasting statistical value to obtain a disturbance adjustment coefficient of the flow scale ensemble forecasting system at each moment; and adjusting the driving disturbance field according to the disturbance adjustment coefficient at each moment to obtain a reconstructed driving disturbance field, and superposing the reconstructed driving disturbance field to control forecast of the convective scale ensemble forecasting system to obtain a convective scale ensemble reconstructed driving field. The method not only can improve the efficiency and accuracy of flow scale ensemble forecast large-scale driving field reconstruction adjustment, but also plays an important role in relieving the problem of divergence of large-scale information in a global ensemble forecast system.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of atmospheric prediction, and in particular to a convection scale ensemble prediction large-scale driving field reconstruction adjustment method. BACKGROUND

[0002] With the intensification of global climate change, extreme weather events such as heavy rain and strong convection occur frequently, causing serious impact on human society and natural environment. Accurate prediction of these extreme weather events is of great significance for disaster prevention and mitigation and protection of people's life and property safety. The convection scale ensemble prediction system, as an advanced weather prediction tool, can provide high spatiotemporal resolution prediction information, which helps to improve the prediction accuracy of extreme weather events.

[0003] The traditional convection scale ensemble prediction system usually relies on the driving field provided by the global ensemble prediction system, including initial value field and lateral boundary condition. Although this method can provide certain large-scale information, due to the uncertainty and resolution limitation of the global ensemble prediction, the quality of large-scale prediction uncertainty information in the convection scale ensemble prediction system is limited, and the dispersion is large, which cannot effectively focus on the prediction uncertainty of local convection. In addition, the traditional dynamic downscaling method does not fully consider the uncertainty difference between different prediction models when processing the driving disturbance field, resulting in insufficient reliability of the prediction results. In order to overcome these shortcomings, the present application proposes a convection scale ensemble prediction large-scale driving field reconstruction adjustment method, which combines the global ensemble driving field with higher quality global deterministic prediction, reconstructs the background field and lateral boundary condition of the convection scale ensemble prediction system, and adjusts the original large-scale ensemble disturbance according to the changes of the updated background field and lateral boundary prediction quality, so as to further focus on the mesoscale and small scale prediction uncertainty of the convection scale ensemble prediction system. This method not only provides a reference for improving the background field and lateral boundary condition of the convection scale ensemble prediction system, but also alleviates the problem of divergence of large-scale information in the global ensemble prediction system. SUMMARY

[0004] The purpose of the present application is to provide a convection scale ensemble prediction large-scale driving field reconstruction adjustment method.

[0005] To achieve the above purpose, the present application is implemented according to the following technical scheme: The present application comprises the following steps: Performing dynamic downscaling to determine the driving disturbance field of the global ensemble prediction system, and performing dynamic downscaling to obtain the control prediction of the convection scale ensemble prediction system from the EC fine grid deterministic model prediction field; the driving field includes initial value field and lateral boundary; Obtaining the prediction element values of the EC fine grid deterministic model and the global ensemble prediction system control prediction within the prediction time, performing multi-sphere screening on the prediction element values, and calculating the prediction statistical values; According to the forecast statistical value, polynomial fitting is performed to obtain a disturbance adjustment coefficient at an analysis time of the convection scale ensemble prediction system and a disturbance adjustment coefficient at each prediction time; the disturbance adjustment coefficient at the analysis time is related to an initial value field; and the disturbance adjustment coefficient at the prediction time is related to a side boundary; According to the disturbance adjustment coefficient at the analysis time and the disturbance adjustment coefficient at each prediction time, the driving disturbance field is adjusted to obtain a reconstructed driving disturbance field, and the reconstructed driving disturbance field is superimposed on a control prediction of the convection scale ensemble prediction system to obtain a convection scale ensemble reconstructed driving disturbance field.

[0006] Further, the method for determining the driving disturbance field comprises: The driving disturbance field comprises a disturbance initial value field and a disturbance side boundary; The initial value field at the analysis time of the global ensemble prediction system is dynamically down-scaled to obtain an initial value field of the convection scale model, and the initial value field of the convection scale model is subtracted from a prediction field at the analysis time of the control prediction in the global ensemble prediction system to obtain a disturbance initial value field; The side boundary at each prediction time of the global ensemble prediction system is dynamically down-scaled to obtain a side boundary at each prediction time of the convection scale model, and the side boundary at each prediction time of the convection scale model is subtracted from a prediction field at each prediction time of the control prediction in the global ensemble prediction system to obtain a disturbance side boundary.

[0007] Further, the method for obtaining the control prediction of the convection scale ensemble prediction system comprises: The control prediction of the convection scale ensemble prediction system comprises a control prediction initial value field of the convection scale ensemble prediction system and a control prediction side boundary of the convection scale ensemble prediction system; The initial value field at the analysis time of the EC fine grid deterministic model is dynamically down-scaled to the same simulation region and resolution as the analysis time of the convection scale ensemble prediction to obtain a control prediction initial value field of the convection scale ensemble prediction system; The side boundary at each prediction time of the EC fine grid deterministic model is dynamically down-scaled to the simulation region and resolution at each prediction time of the convection scale ensemble prediction to obtain a control prediction side boundary of the convection scale ensemble prediction system.

[0008] Further, the method for calculating the forecast statistical value comprises: The forecast element values of the EC fine grid deterministic model and the control prediction of the convection scale ensemble prediction system within a prediction period are obtained; the forecast element values comprise zonal wind U, meridional wind V, temperature T, and height field H; the forecast element values comprise an atmospheric pressure level; A dynamic adaptive multi-sphere is constructed according to historical forecast element values, and the forecast element values within the prediction period are screened by the multi-sphere, and the expression is: in is the filtering condition, indicating time kind Hierarchical forecast element values With the The forecast factor vector of historical samples Euclidean distance Less than time kind Radius of forecast elements at each level , and the forecast element value and Prediction factor vector Euclidean distance Time function mapping With time decay weight The weighted cumulative time is greater than the empirical threshold , For time sensitivity, For the The historical time of the forecast element vector, is the dimension of forecast factor value, is the number of levels for the same forecast factor value, is the dimension weight, is the level weight, For the Prediction factor vector correspond kind The forecast element values ​​of the level, is the forecast element vector exist Standard deviation of forecast element values ​​of different categories at different levels, is the forecast element vector of The standard deviation of the forecast element values ​​at different levels of the class, for kind The standard forecast element radius of the level, is the regulating factor, for time Laplace operator of the atmospheric pressure field; The mean value of the upper level forecast element value and the lower level forecast element value of the same category is used to replace the screened forecast element value, and the root mean square error of different forecast elements of the EC fine grid deterministic model and the control prediction of the convective scale ensemble prediction system in each prediction period is calculated; The root mean square error ratio of different forecast elements of the EC fine grid deterministic model and the control prediction of the convective scale ensemble prediction system in each prediction period is calculated, the root mean square error ratio of each prediction time of the same forecast element is arranged in ascending order, the root mean square error confidence interval and the median of each prediction time of the same forecast element are obtained, and the median of the root mean square error of different prediction times of the same forecast element is taken as the prediction statistical value.

[0009] Further, the method for obtaining the disturbance adjustment coefficient of the analysis time of the convective scale ensemble prediction system and the disturbance adjustment coefficient of each prediction time comprises: According to the prediction statistical value of the same forecast element of different prediction times , polynomial fitting is performed to determine the adjustment coefficient, and the expression is: Among them is the adjustment coefficient of the forecast element , , , and the polynomial coefficients are respectively; The mean value of the adjustment coefficients of all forecast elements of the same prediction time is taken as the disturbance adjustment coefficient of the adjustment coefficient of the current prediction time , when , is the disturbance adjustment coefficient of the analysis time, when , is the disturbance adjustment coefficient of each prediction time .

[0010] Further, the method for obtaining the convective scale ensemble reconstruction driving field comprises: The reconstruction driving disturbance field comprises a reconstruction disturbance initial value field and a reconstruction disturbance side boundary; the convective scale ensemble reconstruction driving field comprises a reconstruction initial value field and a reconstruction side boundary; The reconstruction disturbance initial value field is obtained according to the disturbance adjustment coefficient of the analysis time and the disturbance initial value field, and the reconstruction disturbance initial value field is superimposed on the control prediction initial value field of the convective scale ensemble prediction system to obtain the reconstruction initial value field, and the expression is: Among them is the reconstruction initial value field of the member in the convective scale ensemble prediction system at the analysis time, For the analysis time, the control initial value field of the convective scale ensemble prediction system is analyzed, For the analysis time, the initial value field of the convective scale ensemble prediction system is analyzed, The reconstruction disturbance initial value field is reconstructed; According to the disturbance adjustment coefficient and the disturbance side boundary of each prediction time, the reconstruction disturbance side boundary is obtained, and the reconstruction disturbance side boundary is superimposed on the control prediction side boundary of the convective scale ensemble prediction system to obtain the reconstruction side boundary, and the expression is: Among them For the analysis time, the initial value field of the convective scale ensemble prediction system is analyzed, For the analysis time, the initial value field of the convective scale ensemble prediction system is analyzed, For the analysis time, the initial value field of the convective scale ensemble prediction system is analyzed, For the analysis time, the initial value field of the convective scale ensemble prediction system is analyzed, For the analysis time, the initial value field of the convective scale ensemble prediction system is analyzed, For the analysis time, the initial value field of the convective scale ensemble prediction system is analyzed, For the analysis time, the initial value field of the convective scale ensemble prediction system is analyzed, For the analysis time, the initial value field of the convective scale ensemble prediction system is analyzed.

[0011] The beneficial effects of the present application are: The present application is a large-scale driving field reconstruction adjustment method for convective scale ensemble prediction, compared with the prior art, the present application has the following technical effects: The present application can improve the data preprocessing capability and enhance the model adaptability in the large-scale driving field reconstruction adjustment of the convective scale ensemble prediction through the steps of dynamic downscaling, multiple sphere screening, data statistics, polynomial fitting and disturbance adjustment and superposition, thereby improving the efficiency and precision of the large-scale driving field reconstruction adjustment of the convective scale ensemble prediction, optimizing the large-scale driving field reconstruction adjustment technology of the convective scale ensemble prediction, greatly saving resources and improving work efficiency, and realizing the reconstruction adjustment of the large-scale driving field of the convective scale ensemble prediction, providing a reference for the improvement of the background field and side boundary condition of the convective scale ensemble prediction, so as to alleviate the problem of divergence of large-scale information in the global ensemble prediction system. BRIEF DESCRIPTION OF DRAWINGS

[0012] Figure 1 The present application is a large-scale driving field reconstruction adjustment method for convective scale ensemble prediction, compared with the prior art, the present application has the following technical effects: Figure 2 The present application is a large-scale driving field reconstruction adjustment method for convective scale ensemble prediction, compared with the prior art, the present application has the following technical effects: Figure 3 The present application is a large-scale driving field reconstruction adjustment method for convective scale ensemble prediction, compared with the prior art, the present application has the following technical effects: Figure 4 The present application is a large-scale driving field reconstruction adjustment method for convective scale ensemble prediction, compared with the prior art, the present application has the following technical effects: DETAILED DESCRIPTION

[0013] The application will be further described in the following specific embodiments, which are illustrative of the application and are not intended to limit the application.

[0014] The application of the flow scale ensemble prediction large scale driving field reconstruction adjustment method and system includes the following steps: As shown in the embodiment, the method includes the following steps: Figure 1 The driving field of the global ensemble prediction system is dynamically downscaled to determine the driving disturbance field, and the EC fine grid deterministic model prediction field is dynamically downscaled to obtain the control prediction of the convective scale ensemble prediction system; the driving field includes initial value field and side boundary; The predicted element values of the EC fine grid deterministic model and the global ensemble prediction system control prediction within the prediction time are obtained, the predicted element values are screened by multiple spheres, and the prediction statistical values are calculated; According to the prediction statistical values, the disturbance adjustment coefficients at the analysis time and the disturbance adjustment coefficients at each prediction time of the convective scale ensemble prediction system are obtained by polynomial fitting; the disturbance adjustment coefficients at the analysis time are related to the initial value field; the disturbance adjustment coefficients at each prediction time are related to the side boundary; According to the disturbance adjustment coefficients at the analysis time and the disturbance adjustment coefficients at each prediction time, the driving disturbance field is adjusted to obtain a reconstructed driving disturbance field, and the reconstructed driving disturbance field is superimposed on the control prediction of the convective scale ensemble prediction system to obtain a convective scale ensemble reconstruction driving field.

[0015] In the embodiment, the method for determining the driving disturbance field includes: The driving disturbance field includes a disturbance initial value field and a disturbance side boundary; The initial value field of the global ensemble prediction system at the analysis time is dynamically downscaled to obtain the initial value field of the convective scale model, and the initial value field of the convective scale model is subtracted from the prediction field at the analysis time of the control prediction in the global ensemble prediction system to obtain the disturbance initial value field; The side boundary of the global ensemble prediction system at each prediction time is dynamically downscaled to obtain the side boundary of the convective scale model at each prediction time, and the side boundary of the convective scale model at each prediction time is subtracted from the prediction field at each prediction time of the control prediction in the global ensemble prediction system to obtain the disturbance side boundary.

[0016] In the embodiment, the method for obtaining the control prediction of the convective scale ensemble prediction system includes: The control prediction of the convective scale ensemble prediction system includes the control prediction initial value field of the convective scale ensemble prediction system and the control prediction side boundary of the convective scale ensemble prediction system; ​The initial forecast field at the time of EC fine-grid deterministic model analysis is dynamically downscaled to the same simulation area and resolution as that at the time of convective-scale ensemble forecast analysis, to obtain the initial forecast field of the convective-scale ensemble forecast system control; The lateral boundary forecast dynamics at each forecast time of the EC fine-grid deterministic model are downscaled to the simulation area and resolution of each forecast time of the convective-scale ensemble forecast to obtain the lateral boundary of the control forecast of the convective-scale ensemble forecast system.

[0017] In this embodiment, the method for calculating the forecast statistical value includes: Obtain the forecast element values ​​of the EC fine grid deterministic model and the convective scale ensemble forecast system control forecast within the forecast time; the forecast element values ​​include zonal wind U, meridional wind V, temperature T and height field H; the forecast element values ​​include atmospheric pressure levels; A dynamic adaptive multi-sphere is constructed based on the historical forecast element values, and the forecast element values ​​within the forecast time are screened by the multi-sphere. The expression is: in is the filtering condition, indicating time kind Hierarchical forecast element values With the The forecast factor vector of historical samples Euclidean distance Less than time kind Radius of forecast elements at each level , and the forecast element value and Prediction factor vector Euclidean distance Time function mapping With time decay weight The weighted cumulative time is greater than the empirical threshold , For time sensitivity, For the The historical time of the forecast element vector, is the dimension of forecast factor value, is the number of levels for the same forecast factor value, is the dimension weight, is the level weight, For the Prediction factor vector correspond kind The forecast element values ​​of the level, is the forecast element vector exist Standard deviation of forecast element values ​​of different categories at different levels, is the forecast element vector of The standard deviation of the forecast element values ​​at different levels of the class, for kind The standard forecast element radius of the level, is the regulating factor, for time Laplace operator of the atmospheric pressure field; The filtered forecast element values ​​are replaced by the mean of the forecast element values ​​of the previous level and the next level of the same category, and the root mean square error of different forecast elements of the EC fine grid deterministic model and the convective scale ensemble forecast system control forecast in each forecast period is calculated; The root mean square error ratios of different forecast elements of the EC fine-grid deterministic model and the convective-scale ensemble forecast system control forecast within each forecast period are calculated. The root mean square error ratios of the same forecast element at each forecast moment are arranged in ascending order to obtain the root mean square error confidence intervals and medians of the same forecast element at each forecast moment. The median of the root mean square errors of the same forecast element at different forecast moments is taken as the forecast statistic.

[0018] In this embodiment, the method for obtaining the disturbance adjustment coefficient of the convective-scale ensemble forecast system at the analysis moment and the disturbance adjustment coefficient at each forecast moment includes: Based on the same forecast factor Different forecast times The forecast statistics of Perform polynomial fitting to determine the adjustment coefficient. The expression is: in Forecast elements The adjustment factor, 、 、 are the polynomial coefficients respectively; Take all forecast elements at the same forecast time The mean of the adjustment coefficient is the current forecast time The adjustment coefficient of the disturbance adjustment coefficient is hour, To analyze the moment disturbance adjustment coefficient, when hour, For each forecast time The disturbance adjustment coefficient.

[0019] In this embodiment, the method for obtaining the convective scale set reconstructed driving field includes: The reconstructed driving disturbance field includes reconstructing the initial value field of the disturbance and reconstructing the side boundary of the disturbance; the reconstructed driving field of the flow scale set includes reconstructing the initial value field and reconstructing the side boundary; The reconstructed initial disturbance field is obtained based on the disturbance adjustment coefficient and the initial disturbance field at the analysis time. The reconstructed initial disturbance field is superimposed on the initial disturbance field of the convective scale ensemble forecast system control forecast, and the reconstructed initial field is obtained. The expression is: in For the analysis of the members of the convective scale ensemble forecast system The reconstruction initial value field, To analyze the initial forecast field of the convective scale ensemble forecast system, For analysis time members The reconstructed perturbation initial value field; The reconstructed disturbance lateral boundary is obtained according to the disturbance adjustment coefficient and the disturbance lateral boundary at each forecast moment. The reconstructed disturbance lateral boundary is superimposed on the control forecast lateral boundary of the convective-scale ensemble forecast system to obtain the reconstructed lateral boundary. The expression is: in for Members of the momentary convective scale ensemble forecast system The reconstructed side boundary of for The moment-by-moment convective scale ensemble forecast system controls the forecast side boundary. for Moment Member The reconstructed perturbation side boundary.

[0020] In the actual evaluation, the reconstruction and adjustment of the large-scale driving field of the convective-scale ensemble forecast in the XX region was taken as an example. The stream-scale ensemble forecast system uses the WRF model version XXX. The model area is set to a horizontal resolution of 3km. The simulation area ranges from 35.5° to 46.3°N and 105.2° to 122.4°E, with a total of 550×424 grid points, covering the XX region. The vertical layer is 59 model surfaces, and the model layer top is 10hPa. The system includes a control forecast and 20 disturbance member forecasts, a total of 21 ensemble members. The system forecasts twice a day, at 00:00 UTC and 12:00 UTC, with a forecast validity of 48h, and outputs forecast results hourly. All member physical process parameterization schemes are set as: Thompson microphysical scheme, Mellor-Yamada-Janjic (MYJ) boundary layer scheme and RRTMG long-short wave radiation scheme, and cumulus convection parameterization scheme is turned off; The global ensemble prediction system driving field is dynamically downscaled to determine a driving disturbance field, and the EC fine grid deterministic mode prediction field is dynamically downscaled to obtain a control prediction of the convective scale ensemble prediction system, wherein the horizontal resolution of the global ensemble prediction system is 0.5°*0.5°, and the horizontal resolution of the EC fine grid deterministic mode is 0.25°*0.25°; The prediction element values of the EC fine grid deterministic mode and the control prediction of the convective scale ensemble prediction system in the 12-60h prediction field are obtained, wherein the prediction element categories include zonal wind U, meridional wind V, temperature T and height field H, the statistical levels are divided into 10 / 50 / 100 / 200 / 250 / 500 / 700 / 850 / 925 / 1000hPa according to atmospheric pressure, the statistical period is 00UTC every 6h from June 21 to 28, 2022, the root mean square error ratio sample number at a certain prediction time (taking the 18h prediction field as an example) is 4 (prediction element categories)*10 (statistical levels)*8 (statistical cases) =320, after obtaining 320 root mean square error ratio samples, the root mean square error ratio confidence interval and the median at different prediction times are obtained by ascending from small to large, and the disturbance adjustment coefficients of the analysis time (initial condition) and each prediction time (side boundary condition) of the convective scale ensemble prediction system are obtained by polynomial fitting of the root mean square error ratio median as shown in Figure 2 ; The disturbance initial value field and the disturbance side boundary are multiplied by the adjustment coefficient corresponding to the time The adjusted reconstructed disturbance initial value field and the reconstructed disturbance side boundary are obtained, and the reconstructed disturbance initial value field and the reconstructed disturbance side boundary are superimposed on the initial value field and the side boundary of the control prediction of the convective scale ensemble prediction system to obtain the reconstructed initial value field and the reconstructed side boundary, and the Figure 3 precipitation prediction map, Figure 4 actual precipitation map are compared to show the accuracy of the reconstructed and adjusted precipitation probability prediction of the convective scale ensemble prediction large-scale driving field.

[0021] The above only describes the preferred embodiments of the present application and is not intended to limit the present application, and any modification, equivalent replacement, improvement, etc. made within the spirit and principle of the present application shall be included in the protection scope of the present application.

Claims

1. A method for reconstructing and adjusting the large-scale driving field of convective-scale ensemble forecasts, characterized by: The following steps are involved: S1. Dynamically downscaling the global ensemble forecast system driving field to determine the driving disturbance field, and dynamically downscaling the EC fine-grid deterministic model forecast field to obtain the control forecast of the convective-scale ensemble forecast system; the driving field includes the initial value field and the lateral boundary; S2. Obtain forecast element values ​​within the forecast timeframe of the EC fine-grid deterministic model and the global ensemble forecast system control forecast, perform multi-sphere screening on the forecast element values, and calculate forecast statistics; S3. Performing polynomial fitting based on the forecast statistical value to obtain a disturbance adjustment coefficient at the analysis time of the convective-scale ensemble forecast system and a disturbance adjustment coefficient at each forecast time; the disturbance adjustment coefficient at the analysis time is related to the initial value field; and the disturbance adjustment coefficient at the forecast time is related to the lateral boundary; S4. Adjust the driving disturbance field according to the disturbance adjustment coefficient at the analysis moment and the disturbance adjustment coefficient at each forecast moment to obtain a reconstructed driving disturbance field, and superimpose the reconstructed driving disturbance field on the control forecast of the convective-scale ensemble forecast system to obtain a convective-scale ensemble reconstructed driving field.

2. The method for reconstructing and adjusting the large-scale driving field of the convective-scale ensemble forecast according to claim 1 is characterized in that: The method for determining the driving disturbance field comprises: The driving disturbance field includes a disturbance initial value field and a disturbance side boundary; The initial value field of the global ensemble forecast system at the analysis time is dynamically downscaled to obtain the initial value field of the convective scale model, and the initial value field of the convective scale model is subtracted from the forecast field at the control forecast analysis time in the global ensemble forecast system to obtain the perturbation initial value field; The lateral boundaries of each forecast moment of the global ensemble forecast system are dynamically downscaled to obtain the lateral boundaries of each forecast moment of the convective scale model, and the lateral boundaries of each forecast moment of the convective scale model are subtracted from the forecast fields of each forecast moment of the control forecast in the global ensemble forecast system to obtain the disturbance lateral boundaries.

3. The method for reconstructing and adjusting the large-scale driving field of the convective-scale ensemble forecast according to claim 1 is characterized in that: The method for obtaining a control forecast of a convective-scale ensemble forecast system comprises: The control forecast of the convective scale ensemble forecast system includes a convective scale ensemble forecast system control forecast initial value field and a convective scale ensemble forecast system control forecast lateral boundary; The initial forecast field at the time of EC fine-grid deterministic model analysis is dynamically downscaled to the same simulation area and resolution as that at the time of convective-scale ensemble forecast analysis, to obtain the initial forecast field of the convective-scale ensemble forecast system control; The lateral boundary forecast dynamics at each forecast time of the EC fine-grid deterministic model are downscaled to the simulation area and resolution of each forecast time of the convective-scale ensemble forecast to obtain the lateral boundary of the control forecast of the convective-scale ensemble forecast system.

4. The method for reconstructing and adjusting the large-scale driving field of the convective-scale ensemble forecast according to claim 1 is characterized in that: The method for calculating the forecast statistical value comprises: Obtain the forecast element values ​​of the EC fine grid deterministic model and the convective scale ensemble forecast system control forecast within the forecast time; the forecast element values ​​include zonal wind U, meridional wind V, temperature T and height field H; the forecast element values ​​include atmospheric pressure levels; A dynamic adaptive multi-sphere is constructed based on the historical forecast element values, and the forecast element values ​​within the forecast time are screened by the multi-sphere. The expression is: in is the filtering condition, indicating time kind Hierarchical forecast element values With the The forecast factor vector of historical samples Euclidean distance Less than time kind Radius of forecast elements at each level , and the forecast element value and Prediction factor vector Euclidean distance Time function mapping With time decay weight The weighted cumulative time is greater than the empirical threshold , For time sensitivity, For the The historical time of the forecast element vector, is the dimension of forecast factor value, is the number of levels for the same forecast factor value, is the dimension weight, is the level weight, For the Prediction factor vector correspond kind The forecast element values ​​of the level, is the forecast element vector exist Standard deviation of forecast element values ​​of different categories at different levels, is the forecast element vector of The standard deviation of the forecast element values ​​at different levels of the class, for kind The standard forecast element radius of the level, is the regulating factor, for time Laplace operator of the atmospheric pressure field; The filtered forecast element values ​​are replaced by the mean of the forecast element values ​​of the previous level and the next level of the same category, and the root mean square error of different forecast elements of the EC fine grid deterministic model and the convective scale ensemble forecast system control forecast in each forecast period is calculated; The root mean square error ratios of different forecast elements of the EC fine-grid deterministic model and the convective-scale ensemble forecast system control forecast within each forecast period are calculated. The root mean square error ratios of the same forecast element at each forecast moment are arranged in ascending order to obtain the root mean square error confidence intervals and medians of the same forecast element at each forecast moment. The median of the root mean square errors of the same forecast element at different forecast moments is taken as the forecast statistic.

5. The method for reconstructing and adjusting the large-scale driving field of the convective-scale ensemble forecast according to claim 1 is characterized in that: The method for obtaining the disturbance adjustment coefficient of the convective-scale ensemble forecast system at the analysis moment and the disturbance adjustment coefficient at each forecast moment includes: Based on the same forecast factor Different forecast times The predicted statistical value of Perform polynomial fitting to determine the adjustment coefficient. The expression is: in Forecast elements The adjustment factor, 、 、 are the polynomial coefficients respectively; Take all forecast elements at the same forecast time The mean of the adjustment coefficient is the current forecast time The adjustment coefficient of the disturbance adjustment coefficient is hour, To analyze the moment disturbance adjustment coefficient, when hour, For each forecast time The disturbance adjustment coefficient.

6. The method for reconstructing and adjusting the large-scale driving field of the convective-scale ensemble forecast according to claim 1 is characterized in that: The method for obtaining a convective-scale ensemble reconstructed driving field comprises: The reconstructed driving disturbance field includes reconstructing the initial value field of the disturbance and reconstructing the side boundary of the disturbance; the reconstructed driving field of the flow scale set includes reconstructing the initial value field and reconstructing the side boundary; The reconstructed initial disturbance field is obtained based on the disturbance adjustment coefficient and the initial disturbance field at the analysis time. The reconstructed initial disturbance field is superimposed on the initial disturbance field of the convective-scale ensemble forecast system control forecast. The expression is: in For the analysis of the members of the convective scale ensemble forecast system The reconstruction initial value field of To analyze the initial forecast field of the convective scale ensemble forecast system, For analysis time members The reconstructed perturbation initial value field; The reconstructed disturbance lateral boundary is obtained according to the disturbance adjustment coefficient and the disturbance lateral boundary at each forecast moment. The reconstructed disturbance lateral boundary is superimposed on the control forecast lateral boundary of the convective-scale ensemble forecast system to obtain the reconstructed lateral boundary. The expression is: in for Members of the momentary convective scale ensemble forecast system The reconstructed side boundary of for The moment-by-moment convective scale ensemble forecast system controls the forecast side boundary. for Moment Member The reconstructed perturbation side boundary.

Images