GRAPES_GEPS global ensemble forecast method and system

Abstract

The present invention belongs to the technical field of weather forecasting, and discloses a GRAPES_GEPS global ensemble forecast method and system, the system comprising: a preprocessing module, a singular vector ground state initial value generation module, an initial singular vector calculation module, a control forecast initial value generation module, a disturbance field generation and ensemble forecast member disturbance initial value generation module, a control forecast integration module, a model disturbance parameter generation and 30 ensemble forecast member model integration modules, a post-processing module and an evolution singular vector calculation module. The present invention realizes the application of the initial singular vector method in the southern hemisphere, the northern hemisphere and the tropical cyclone region in the global GRAPES_GEPS system, uses the model disturbance SPPT scheme and the SKEB scheme to reflect the uncertainty of the model itself, can provide a wealth of probabilistic forecast guidance products based on the GRAPES global model for business services, and also provides more forecast reference information for forecasters.

Description

Technical Field

[0001] The invention belongs to the technical field of weather forecasting, and in particular relates to a GRAPES_GEPS global ensemble forecasting method and system. Background Art

[0002] At present, numerical weather forecast has become one of the most core tools in operational weather forecasting, but the inherent chaotic properties of the atmosphere and the many uncertainties in the initial field and physical process of the numerical model limit the predictability of a single model. Since ensemble forecasting provides a method to estimate the probability density function of the atmospheric state, it essentially conforms to the chaotic characteristics of atmospheric evolution, can provide a higher forecast skill score than a single deterministic forecast, and can also quantitatively estimate forecast uncertainty and provide probabilistic guidance for operational forecasting. Therefore, it has played a very important role in the numerical forecasting business system. However, in the process of researching and establishing the ensemble forecasting system, the generation technology of the initial value of the disturbance is one of the core technical issues. The ensemble forecasting system that only uses the initial value disturbance has the defects of insufficient divergence of the ensemble members and insufficient reliability of the ensemble forecasting system.

[0003] Global medium-range ensemble forecast is the main business component of the international advanced numerical forecast center. The National Center for Environmental Prediction of the United States and the European Center for Medium-Range Weather Forecasts established their own global ensemble forecast systems as early as 1992. The China Meteorological Administration established a global ensemble forecast system with the value-added vector method (BGM) as the initial value perturbation method at the end of 2006; in 2008, the T213 global ensemble forecast system based on the BGM initial value perturbation method was established, with 15 ensemble forecast members and a forecast time of 10 days; in 2014, the system was upgraded to the T639 global ensemble forecast, with the initial value perturbation still using the BGM method, and the SPPT method that takes into account model uncertainty, with a forecast time of 15 days. Due to the limitation of the perturbation method itself, the number of ensemble members is still 15. The generation technology of initial perturbations is one of the core technical issues in the research and establishment of ensemble forecasting systems. The singular vector method is currently the most advanced initial perturbation generation technology in the world. The SVs initial perturbation technology has been successfully applied in the ECMWF global ensemble forecasting system. In addition, Canada, Japan Meteorological Agency, etc. have also realized ensemble forecasting systems based on singular vectors, and some other countries are investing a lot of research in this regard. In 2018, the GRAPES Global Ensemble Prediction System (GRAPES_GEPS) independently developed by the Numerical Prediction Center of the China Meteorological Administration passed the business acceptance review, marking that my country has built a global deterministic and ensemble forecasting system with the independently developed GRAPES as the core, and a complete numerical weather forecasting system that provides support for professional models such as typhoons, environmental responses, and waves.

[0004] The existing technical solutions mainly include the following:

[0005] The global ensemble forecasting experimental method is constructed by using the singular vector (SVs) initial value perturbation method. However, the ensemble forecasting system that only uses the initial value perturbation has the defects of insufficient divergence of ensemble members and insufficient reliability of the ensemble forecasting system.

[0006] The experiment of initial singular vector perturbation (INISV experiment) was used, but it could not reduce the systematic negative deviation characteristics of the GRAPES-GEPS height field to a certain extent, and the ensemble discreteness of the system was insufficient.

[0007] In order to further increase the ensemble discreteness and improve the reliability of the system, only the random physical tendency perturbation scheme (SPPT scheme) or only the random kinetic energy compensation scheme (SKEB scheme) is used; however, it cannot further increase the ensemble discreteness of different forecast variables and reduce the ensemble mean error to a certain extent, especially in tropical regions.

[0008] Through the above analysis, the problems and defects of the prior art are as follows:

[0009] (1) The global ensemble forecasting experiment method is constructed using the singular vector (SVs) initial value perturbation method. The ensemble forecasting system that only uses initial value perturbations has the defects of insufficient divergence of ensemble members and insufficient reliability of the ensemble forecasting system.

[0010] (2) The experiments using initial singular vector perturbations cannot reduce the systematic negative deviation characteristics of the GRAPES-GEPS height field to a certain extent, and the ensemble discreteness of the system is insufficient.

[0011] (3) Using only the random physical tendency perturbation scheme (SPPT scheme) or only the random kinetic energy compensation scheme (SKEB scheme) cannot further increase the ensemble dispersion of different forecast variables and reduce the ensemble mean error to a certain extent, especially in the tropical region.

[0012] The difficulty of solving the above problems and defects is:

[0013] It is necessary to be able to find a perturbation method that meets the above conditions.

[0014] The significance of solving the above problems and defects is:

[0015] 1. In order to make the ensemble forecast system have ensemble member divergence and the ensemble forecast system reliable.

[0016] 2. It can reduce the systematic negative deviation characteristics of the GRAPES-GEPS height field to a certain extent, and the collective discreteness of the system increases.

[0017] 3. It can further increase the ensemble dispersion of different forecast variables and reduce the ensemble mean error to a certain extent, especially in tropical regions. Summary of the invention

[0018] In view of the problems existing in the prior art, the present invention provides a GRAPES_GEPS global ensemble forecast method and system, and specifically relates to a global ensemble forecast method, system and medium for systematic negative deviation characteristics (GRAPES_GEPS) of height fields.

[0019] The present invention is achieved by a global ensemble prediction method of systematic negative deviation characteristics of height field, comprising:

[0020] Step 1: Preprocess the SST data; use the dynamic upscaling technique to interpolate the 0.25-degree resolution initial values ​​generated by the GRAPES-4DVAR assimilation analysis to 2.5-degree resolution to form the singular vector ground state initial values;

[0021] Step 2: Calculate the initial singular vectors corresponding to the northern and southern hemispheres according to the Lanczos algorithm; interpolate the 0.25-degree resolution initial values ​​generated by the GRAPES Global Four-Dimensional Variational Data Assimilation System (GRAPES-4DVAR) assimilation analysis to 0.5-degree resolution using dynamic upscaling technology to obtain the GRAPES global ensemble forecast control forecast initial values;

[0022] Step 3: Use the initial singular vectors and the evolved singular vectors to construct 15 disturbance fields through linear combination, and add / subtract 15 disturbance fields to the initial control forecast values ​​to construct the initial disturbance values ​​of the 1st to 30th ensemble forecast members;

[0023] Step 4: Use the GRAPES_GFS_2-1-2-2 model with a resolution of 0.5 degrees and the initial value of the control forecast to integrate for 360 hours and obtain the control forecast integration result; generate the parameters required for the perturbation schemes of the SPPT and SKEB models of the 30 ensemble forecast members;

[0024] Step 5: Using the model run with the added model perturbations SPPT and SKEB schemes, perform 360-hour integration of 30 ensemble forecast members to obtain the forecast results of 30 perturbation members; perform the global ensemble forecast of the systematic negative deviation characteristics (GRAPES_GEPS) of the height field;

[0025] Step six, based on the initial singular vectors corresponding to the northern and southern hemispheres, integrate and tangent the linear model to obtain the evolved singular vectors and perform forecast result detection.

[0026] Furthermore, in step 2, the calculation of the initial singular vectors corresponding to the northern hemisphere and the southern hemisphere according to the Lanczos algorithm also includes: when a tropical cyclone exists at the starting moment, calculating the initial singular vectors with the tropical cyclone area as the target area.

[0027] Further, the step three includes:

[0028] The initial disturbance field is generated based on the formula, and the GRAPES model analysis field is added and subtracted in pairs (,) to construct the initial disturbance values ​​of 2*M ensemble forecasts.

[0029] Further, the step three includes the following steps:

[0030] (1) Read SV configuration information from the configuration file namelist.svd;

[0031] (2) Read NH SVs, SH SVs, and TCSVs separately, and merge the read NH SVs, SH SVs, and TCSVs into global SVs perturbations;

[0032] (3) Scaling the global SVs disturbance field; performing Gaussian linear combination on the scaled SVs disturbance field to generate m initial value disturbance fields;

[0033] (4) Perform horizontal interpolation and smoothing processing on the m initial value disturbance fields, and perform addition and subtraction processing on the model analysis field to obtain 2*m ensemble forecast initial value fields.

[0034] Furthermore, the parameters include: random seed, disturbance magnitude, random disturbance wave number, loss time correlation coefficient and other parameters.

[0035] Furthermore, the business ensemble forecast products include: box plot data and typhoon path ensemble forecast data.

[0036] Another object of the present invention is to provide a global ensemble prediction system for systematic negative deviation characteristics of height fields, comprising:

[0037] Preprocessing module, used to preprocess the sea temperature data;

[0038] The singular vector ground state initial value generation module is used to interpolate the 0.25 degree resolution initial values ​​generated by the GRAPES-4DVAR (GRAPES Global Four-Dimensional Variational Data Assimilation System) assimilation analysis to 2.5 degree resolution using dynamic upscaling technology to form the singular vector ground state initial values;

[0039] The initial singular vector calculation module is used to calculate the initial singular vectors corresponding to the northern hemisphere and the southern hemisphere according to the Lanczos algorithm; it is also used to calculate the initial singular vectors with the tropical cyclone area as the target area when there is a tropical cyclone at the starting time;

[0040] The control forecast initial value generation module is used to interpolate the 0.25 degree resolution initial values ​​generated by the GRAPES-4DVAR (GRAPES global four-dimensional variational data assimilation system) assimilation analysis to 0.5 degree resolution using dynamic upscaling technology to obtain the GRAPES global ensemble forecast control forecast initial values;

[0041] The disturbance field generation and ensemble forecast member disturbance initial value generation module is used to construct 15 disturbance fields through linear combination using the initial singular vectors and the evolved singular vectors, and add / subtract 15 disturbance fields from the control forecast initial value to construct the disturbance initial values ​​of the 1st to 30th ensemble forecast members;

[0042] The control forecast integration module is used to integrate for 360 h using the GRAPES_GFS_2-1-2-2 model with a resolution of 0.5 degrees and the initial value of the control forecast of 0.5 degrees to obtain the control forecast integration result;

[0043] The model perturbation parameter generation and 30 ensemble forecast member model integration module is used to generate the parameters required by the SPPT and SKEB model perturbation schemes of the 30 ensemble forecast members; the model operation with the addition of the model perturbation SPPT and SKEB schemes is used to perform 360h integration of the 30 ensemble forecast members to obtain the forecast results of the 30 perturbation members;

[0044] Post-processing module for GRAPES_GEPS global ensemble forecast;

[0045] The evolved singular vector calculation module is used to obtain the evolved singular vectors according to the initial singular vectors corresponding to the northern hemisphere and the southern hemisphere and the integral tangent linear model.

[0046] Further, the post-processing module includes:

[0047] The integrated data unified post-processing unit is used to convert the binary forecast results of all ensemble members into GRIB2 (computer-independent compressed binary code) format data;

[0048] The conventional ensemble forecast product post-processing unit is used to generate operational ensemble forecast products using the ensemble forecast results in the unified post-processed GRIB2 (computer-independent compressed binary code) format;

[0049] The extreme weather ensemble forecast product post-processing unit is used to calculate and generate the extreme forecast index of ground elements, ensemble forecast member deviation correction data and anomaly products, and 2m temperature anomaly probability forecast products using ensemble forecast member data, model climate data and historical climate data.

[0050] Another object of the present invention is to provide a storage medium for receiving a program input by a user, wherein the stored computer program enables an electronic device to execute the global ensemble prediction method for the systematic negative deviation characteristics of the height field, comprising the following steps:

[0051] Step 1: Preprocess the SST data; use the dynamic upscaling technique to interpolate the 0.25 degree resolution initial values ​​generated by the GRAPES-4DVAR (GRAPES Global Four-Dimensional Variational Data Assimilation System) assimilation analysis to 2.5 degree resolution to form the singular vector ground state initial values;

[0052] Step 2: Calculate the initial singular vectors corresponding to the northern and southern hemispheres according to the Lanczos algorithm; interpolate the 0.25-degree resolution initial values ​​generated by the GRAPES-4DVAR (GRAPES Global Four-Dimensional Variational Data Assimilation System) assimilation analysis to 0.5-degree resolution using dynamic upscaling technology to obtain the GRAPES global ensemble forecast control forecast initial values;

[0053] Step 3: Use the initial singular vectors and the evolved singular vectors to construct 15 disturbance fields through linear combination, and add / subtract 15 disturbance fields to the initial control forecast values ​​to construct the initial disturbance values ​​of the 1st to 30th ensemble forecast members;

[0054] Step 4: Use the GRAPES_GFS_2-1-2-2 model with a resolution of 0.5 degrees and the initial value of the control forecast to integrate for 360 hours and obtain the control forecast integration result; generate the parameters required for the perturbation schemes of the SPPT and SKEB models of the 30 ensemble forecast members;

[0055] Step 5: Run the model with the added model perturbations SPPT and SKEB schemes, perform 360-hour integration of 30 ensemble forecast members, and obtain the forecast results of 30 perturbation members; perform the GRAPES_GEPS global ensemble forecast;

[0056] Step six, based on the initial singular vectors corresponding to the northern and southern hemispheres, integrate and tangent the linear model to obtain the evolved singular vectors and perform forecast result detection.

[0057] Another object of the present invention is to provide an information data processing terminal, which includes a memory and a processor. The memory stores a computer program. When the computer program is executed by the processor, the processor executes the global ensemble forecasting method for the systematic negative deviation characteristics of the height field.

[0058] Combining all the above technical solutions, the advantages and positive effects of the present invention are as follows:

[0059] In the technical research and development of the GRAPES global ensemble forecasting business operation system, the present invention improves the defects of insufficient divergence of ensemble members and insufficient reliability of ensemble forecasting systems that only use initial value perturbations, thereby further increasing the ensemble discreteness of different forecast variables and reducing the ensemble average error to a certain extent, especially in tropical regions. The application of the initial singular vector method in three target areas, such as the southern hemisphere, the northern hemisphere and the tropical cyclone area, in the global GRAPES_GEPS system is realized, and the model perturbation SPPT scheme and SKEB scheme are used to reflect the uncertainty of the model itself, and an extreme weather ensemble forecast product is developed. From the perspective of meeting the needs of different time and space scales from the output of the global ensemble forecasting system, the forecasting elements are discussed, and their predictability can provide reference for model developers, front-line forecasters and users. In particular, the in-depth study of atmospheric predictability is very important for comprehensively improving the level of numerical forecasting systems from the perspective of science and technology. When the true source of predictability can be determined, model developers can modify and improve it in a targeted manner. Combining traditional predictability research with improved ensemble forecasts that can more objectively express forecast uncertainty will provide another valuable reference for predictability. It can provide users with confidence indexes of these probabilistic forecasts and provide a wealth of probabilistic forecast guidance products based on the GRAPES global model for business services. It is very beneficial to drive forecasters to use probabilistic forecast products in forecasting business and serve the society, and also provide forecasters with more forecast reference information. BRIEF DESCRIPTION OF THE DRAWINGS

[0060] Figure 1 It is a flow chart of the GRAPES_GEPS global ensemble forecast method provided by an embodiment of the present invention.

[0061] Figure 2 It is a flow chart of the operation of the GRAPES_GEPS global ensemble forecast system provided by an embodiment of the present invention.

[0062] Figure 3 It is a schematic diagram of the structure of the GRAPES_GEPS global ensemble forecast system provided by an embodiment of the present invention;

[0063] In the figure: 1. Preprocessing module; 2. Singular vector ground state initial value generation module; 3. Initial singular vector calculation module; 4. Control forecast initial value generation module; 5. Disturbance field generation and ensemble forecast member disturbance initial value generation module; 6. Control forecast integration module; 7. Model disturbance parameter generation and 30 ensemble forecast member model integration module; 8. Post-processing module; 9. Evolved singular vector calculation module.

[0064] Figure 4 It is a flow chart of generating ensemble forecast initial value disturbance based on GRAEPS singular vectors provided in an embodiment of the present invention.

[0065] Figure 5(a) is a schematic diagram of the evolution of the ensemble dispersion (dashed line) and ensemble mean error (solid line) of the 850 hPa zonal wind field u(a) in the tropical region with the forecast time in the SV_SPPT (blue line) and SV_SPPT_SKEB (red line) experiments provided by an embodiment of the present invention.

[0066] Figure 5(b) is a schematic diagram of the evolution of the ensemble dispersion (dashed line) and ensemble mean error (solid line) of the 850hPa meridional wind field v(b) in the tropical region with the forecast time in the SV_SPPT (blue line) and SV_SPPT_SKEB (red line) experiments provided by an embodiment of the present invention.

[0067] Figure 5(c) is a schematic diagram of the evolution of the ensemble dispersion (dashed line) and ensemble mean error (solid line) of the 850hPa temperature t(c) in the tropical region in the SV_SPPT (blue line) and SV_SPPT_SKEB (red line) experiments with the forecast time provided by an embodiment of the present invention.

[0068] Figure 5(d) is a schematic diagram of the evolution of the ensemble dispersion (dashed line) and ensemble mean error (solid line) of the 850 hPa potential height h(d) in the tropical region with the forecast time in the SV_SPPT (blue line) and SV_SPPT_SKEB (red line) experiments provided by an embodiment of the present invention.

[0069] Figure 6 It is a flowchart of post-processing and product production of the GRAPES_GEPS system provided by an embodiment of the present invention.

[0070] Figure 7 It is a vertical section along 50 degrees north latitude of the set initial value disturbance member 4 provided in an embodiment of the present invention.

[0071] Figure 8 It is a comprehensive scorecard of forecast effects of the GRAPES_GEPS V1.0 global ensemble forecast system and the T639_EPS global ensemble forecast system during the parallel experiment provided by an embodiment of the present invention. DETAILED DESCRIPTION

[0072] In order to make the purpose, technical solution and advantages of the present invention more clearly understood, the present invention is further described in detail below in conjunction with the embodiments. It should be understood that the specific embodiments described herein are only used to explain the present invention and are not used to limit the present invention.

[0073] In view of the problems existing in the prior art, the present invention provides a global ensemble forecasting method, system and medium for systematic negative deviation characteristics of height field. The present invention is described in detail below in conjunction with the accompanying drawings.

[0074] like Figure 1 As shown, the global ensemble forecasting method for systematic negative deviation characteristics of height fields provided by the embodiment of the present invention includes:

[0075] S101, preprocessing the SST data; using the dynamic upscaling technique to interpolate the 0.25 degree resolution initial values ​​generated by the GRAPES-4DVAR (GRAPES Global Four-Dimensional Variational Data Assimilation System) assimilation analysis to 2.5 degree resolution, forming the singular vector ground state initial values;

[0076] S102, calculate the initial singular vectors corresponding to the northern and southern hemispheres according to the Lanczos algorithm; interpolate the 0.25 degree resolution initial values ​​generated by the GRAPES-4DVAR (GRAPES Global Four-Dimensional Variational Data Assimilation System) assimilation analysis to 0.5 degree resolution using the dynamic upscaling technique to obtain the GRAPES global ensemble forecast control forecast initial values;

[0077] S103, using the initial singular vector and the evolved singular vector, constructing 15 disturbance fields through linear combination, and adding / subtracting the 15 disturbance fields to the control forecast initial value to construct the disturbance initial values ​​of the 1st to 30th ensemble forecast members;

[0078] S104, using the GRAPES_GFS_2-1-2-2 model with a resolution of 0.5 degrees and the initial value of the control forecast of 0.5 degrees to integrate for 360 hours, and obtain the control forecast integration result; generate the parameters required for the perturbation scheme of the SPPT and SKEB models of 30 ensemble forecast members;

[0079] S105, using the model operation with the added model perturbations SPPT and SKEB scheme, perform 360-hour integration of 30 ensemble forecast members to obtain the forecast results of 30 perturbation members; perform GRAPES_GEPS global ensemble forecast;

[0080] S106, according to the initial singular vectors corresponding to the northern hemisphere and the southern hemisphere, the linear model is integrated and cut to obtain the evolved singular vectors, and the forecast result is tested.

[0081] The method provided in the embodiment of the present invention for calculating the initial singular vectors corresponding to the northern hemisphere and the southern hemisphere according to the Lanczos algorithm also includes: when a tropical cyclone exists at the starting moment, calculating the initial singular vectors with the tropical cyclone area as the target area.

[0082] Step S103 provided in the embodiment of the present invention includes:

[0083] The initial disturbance field is generated based on the formula, and the GRAPES model analysis field is added and subtracted in pairs (,) to construct the initial disturbance values ​​of 2*M ensemble forecasts.

[0084] Step S103 provided in the embodiment of the present invention includes the following steps:

[0085] (1) Read SV configuration information from the configuration file namelist.svd;

[0086] (2) Read NH SVs, SH SVs, and TCSVs separately, and merge the read NH SVs, SH SVs, and TCSVs into global SVs perturbations;

[0087] (3) Scaling the global SVs disturbance field; performing Gaussian linear combination on the scaled SVs disturbance field to generate m initial value disturbance fields;

[0088] (4) Perform horizontal interpolation and smoothing processing on the m initial value disturbance fields, and perform addition and subtraction processing on the model analysis field to obtain 2*m ensemble forecast initial value fields.

[0089] The parameters provided by the embodiment of the present invention include: random seed, disturbance magnitude, random disturbance wave number, loss time correlation coefficient and other parameters.

[0090] The business ensemble forecast products provided by the embodiment of the present invention include: box plot data and typhoon path ensemble forecast data.

[0091] like Figure 2-Figure 3 As shown, the global ensemble prediction system for systematic negative deviation characteristics of height fields provided by the embodiment of the present invention includes:

[0092] Preprocessing module 1, used to preprocess the sea temperature data;

[0093] Singular vector ground state initial value generation module 2 is used to interpolate the 0.25 degree resolution initial values ​​generated by GRAPES-4DVAR (GRAPES global four-dimensional variational data assimilation system) assimilation analysis to 2.5 degree resolution using dynamic upscaling technology to form the singular vector ground state initial values;

[0094] The initial singular vector calculation module 3 is used to calculate the initial singular vectors corresponding to the northern hemisphere and the southern hemisphere according to the Lanczos algorithm; and is also used to calculate the initial singular vectors with the tropical cyclone area as the target area when there is a tropical cyclone at the starting time;

[0095] The control forecast initial value generation module 4 is used to interpolate the 0.25 degree resolution initial values ​​generated by the GRAPES-4DVAR (GRAPES global four-dimensional variational data assimilation system) assimilation analysis to 0.5 degree resolution using the dynamic upscaling technology to obtain the GRAPES global ensemble forecast control forecast initial values;

[0096] The disturbance field generation and ensemble forecast member disturbance initial value generation module 5 is used to construct 15 disturbance fields by linear combination using the initial singular vector and the evolved singular vector, and add / subtract 15 disturbance fields to the control forecast initial value to construct the disturbance initial values ​​of the 1st to 30th ensemble forecast members;

[0097] The control forecast integration module 6 is used to integrate for 360 hours using the GRAPES_GFS_2-1-2-2 model with a resolution of 0.5 degrees and the initial value of the control forecast of 0.5 degrees to obtain the control forecast integration result;

[0098] The model perturbation parameter generation and 30 ensemble forecast member model integration module 7 is used to generate the parameters required by the SPPT and SKEB model perturbation schemes of the 30 ensemble forecast members; the model operation with the addition of the model perturbation SPPT and SKEB schemes is used to perform 360h integration of the 30 ensemble forecast members to obtain the forecast results of the 30 perturbation members;

[0099] Post-processing module 8, used for GRAPES_GEPS global ensemble forecast;

[0100] The evolved singular vector calculation module 9 is used to obtain the evolved singular vectors according to the initial singular vectors corresponding to the northern hemisphere and the southern hemisphere by integrating and tangent linear models.

[0101] The post-processing module 8 provided in the embodiment of the present invention includes:

[0102] The integrated data unified post-processing unit is used to convert the binary forecast results of all ensemble members into GRIB2 (computer-independent compressed binary code) format data;

[0103] The conventional ensemble forecast product post-processing unit is used to generate operational ensemble forecast products using the ensemble forecast results in the unified post-processed GRIB2 (computer-independent compressed binary code) format;

[0104] The extreme weather ensemble forecast product post-processing unit is used to calculate and generate the extreme forecast index of ground elements, ensemble forecast member deviation correction data and anomaly products, and 2m temperature anomaly probability forecast products using ensemble forecast member data, model climate data and historical climate data.

[0105] The technical solution of the present invention is further described below in conjunction with specific embodiments.

[0106] Embodiment 1:

[0107] 1. Initial value perturbation program design based on singular vectors

[0108] Based on the formula, the initial perturbation field is generated, and the GRAPES model analysis field is added and subtracted in pairs (,) to construct 2*M ensemble forecast perturbation initial values. The specific implementation scheme of constructing the initial perturbation value of the ensemble forecast by global GRAPES SVs is as follows: Figure 4 .

[0109] The calculation of global GRAPES SVs is carried out in three target areas: mid- and high-latitudes in the northern and southern hemispheres and near tropical cyclones. The number of SVs obtained in different target areas is different. The GRAEPS SVs of these three target areas are linearly combined into the initial perturbation field of the ensemble forecast through the following process:

[0110] (1) Read SV configuration information from the configuration file namelist.svd;

[0111] (2) read NH SVs, SH SVs, and TCSVs separately and merge them into global SVs disturbance;

[0112] (3) Scaling the global SVs disturbance field;

[0113] (4) Perform Gaussian linear combination on the scaled SVs disturbance field to generate m initial value disturbance fields;

[0114] (5) Perform horizontal interpolation and smoothing on the m initial value disturbance fields, and filter out those individual values ​​that exceed the analysis error by a large amount, so as to ensure the stability of the ensemble forecast and the rationality of the initial value disturbance;

[0115] (6) Finally, these m initial value disturbance fields are added and subtracted on the model analysis field to obtain 2*m ensemble forecast initial value fields.

[0116] 2. Analysis of the structure of the disturbance field of the initial value of the ensemble forecast

[0117] In order to further study the rationality of using GRAEPS SVs to construct the ensemble forecast initial value perturbation scheme, the structure analysis of the constructed ensemble forecast initial value perturbation field is carried out. The following is an example of the experimental results at 00UTC on May 7, 2013. Among them, the singular vector calculation scheme of the GRAPES global model is shown in Table 1. Here, the model initial value field uses FNL data, and the model resolution is 0.5°×0.5° horizontally and 60 layers vertically.

[0118] Table 1 Computational configuration of the GRAPES SVs experiment

[0119]

[0120] Figure 7 The horizontal structure of the potential temperature disturbance of member 4 generated by linear combination in the 00UTC experiment on May 7, 2013 at the 25th layer of the model surface (about 500hPa) and its vertical structure along 50 degrees north latitude are given. It can be seen that the horizontal distribution and magnitude of the potential temperature disturbance field are relatively reasonable, and the vertical profile also shows a typical baroclinic structure that tilts westward with height, which is consistent with the instability characteristics of the weather-scale atmosphere reflected by the singular vector disturbance itself.

[0121] 3. Mode perturbation technology combining SPPT and SKEB schemes

[0122] In the process of developing the model perturbation technology of the GRAPES_GEPS system, the SPPT scheme was first studied. When the SKEB scheme was developed, the SPPT scheme had already been maturely applied. Therefore, after the SKEB scheme was introduced into the GRAPES_GEPS system, more attention was paid to the impact of continuing to use the SKEB scheme on the basis of the model perturbation SPPT scheme. Therefore, after analyzing the impact of the SPPT scheme and the SKEB scheme separately, the following two groups of ensemble forecast experiments were conducted, as shown in Table 2: SV_SPPT experiment using only the model perturbation SPPT scheme, and SV_SPPT_SKEB experiment using both the model perturbation SPPT and SKEB schemes. The results of these two groups of experiments were compared and analyzed to determine whether the application of the SKEB scheme can further improve the GRAPES_GEPS system.

[0123] Table 2 GRAPES_GEPS system ensemble forecast test plan

[0124]

[0125] The impact of the SPPT and SKEB schemes is mainly concentrated in the tropical region. As can be seen from Figure 5, the use of the SKEB scheme can further increase the ensemble dispersion of different forecast variables based on the SPPT scheme and reduce the ensemble mean error to a certain extent. This shows that the SKEB scheme is a good supplement to the existing SPPT scheme, and the two model perturbation schemes can be applied to the GRAPES_GEPS system at the same time.

[0126] 4. Construction of the GRAPES Global Ensemble Prediction System

[0127] Based on the above key technologies: (1) GRAPES global singular vector technology suitable for ensemble forecasting applications; (2) GRAPES_GEPS initial value perturbation generation technology based on singular vectors; (3) SPPT model perturbation technology that reflects the uncertainty of the physical process of the GRAPES global model; (4) SKEB model perturbation technology that compensates for the kinetic energy dissipation of the GRAPES global model; (5) Tropical cyclone singular vector perturbation technology; (6) Post-processing technology for conventional ensemble forecast products and extreme weather ensemble forecast products. On the PAI-SUGON high-performance computer of the China Meteorological Administration, the scheduling of the system operation process based on the ECFLOW technology was established through concurrent scheduling and concurrent number control mechanisms. The technical parameters of the system are detailed in Table 3.

[0128] Table 3 GRAPES_GEPS global ensemble prediction system parameters

[0129]

[0130] GRAPE_GEPS system operation process is as follows Figure 3 The system integrates multiple modules such as data preprocessing, singular vector solution, ensemble forecast member integration mode, and product post-processing technology. The post-processing of GRAPES_GEPS global ensemble forecast mainly integrates three post-processing modules: unified post-processing of 31 ensemble forecast member integration data, post-processing of conventional ensemble forecast products, and post-processing of extreme weather forecast products. The specific production process of post-processing and products is shown in Figure 6 The ensemble forecast verification module is installed on the business verification platform, and the forecast results are verified after the product post-processing is completed. The main functions of each module are briefly described as follows:

[0131] (1) SST data preprocessing: SST data are preprocessed. The output results are stored in the file system of Sugon high-performance computers, providing input data for the calculation of singular vectors and model integration.

[0132] (2) Generation of singular vector ground state initial values. The 0.25-degree resolution initial values ​​generated by the GRAPES-4DVAR (GRAPES global four-dimensional variational data assimilation system) assimilation analysis are interpolated to 2.5-degree resolution using dynamic upscaling technology to form the singular vector ground state initial values ​​for the calculation of singular vectors. The input is the 0.25-degree GRAPES-4DVAR (GRAPES global four-dimensional variational data assimilation system) assimilation analysis initial field, and the output is the 2.5-degree initial value. The output results are stored in the file system of the Sugon high-performance computer to provide input data for the calculation of singular vectors.

[0133] (3) Solving the initial singular vectors. Solve the initial singular vectors corresponding to the northern hemisphere and the southern hemisphere according to the Lanczos algorithm. If there are tropical cyclones (up to 6) at the start time, it is also necessary to calculate the initial singular vectors with the tropical cyclone area as the target area. The input is the 2.5-degree singular vector ground state initial value, and the output is the 2.5-degree initial singular vector. The output results are stored in the file system of the Sugon high-performance computer, providing input data for the generation of the initial value of the perturbation.

[0134] (4) Generation of initial values ​​of control forecasts. The initial values ​​of 0.25 degree resolution generated by the assimilation analysis of GRAPES-4DVAR (GRAPES global four-dimensional variational data assimilation system) are interpolated to 0.5 degree resolution using dynamic upscaling technology to obtain the initial values ​​of GRAPES global ensemble forecast control forecasts. The input is the 0.25 degree GRAPES-4DVAR assimilation analysis field, and the output is the 0.5 degree initial values ​​of control forecasts. The output results are stored in the file system of Sugon high-performance computers to provide input data for model integration.

[0135] (5) Generation of disturbance field and initial disturbance values ​​of ensemble forecast members. Using the initial singular vectors and the evolved singular vectors, 15 disturbance fields are constructed through linear combination, and the initial control forecast values ​​are added / subtracted by 15 disturbance fields to construct the initial disturbance values ​​of the 1st to 30th ensemble forecast members. The input of the disturbance field is the 2.5-degree initial singular vector and the evolved singular vector, and the output is the input and the output is the 0.5-degree disturbance field. The input of the initial disturbance value of the ensemble forecast member is the 0.5-degree control forecast initial field and the 0.5 disturbance field. The output results are stored in the file system of the Sugon high-performance computer, providing the initial disturbance value data of the ensemble forecast members for the model integration.

[0136] (6) Control forecast integration. The GRAPES_GFS_2-1-2-2 model with a resolution of 0.5 degrees and the initial value of the control forecast of 0.5 degrees are used to integrate for 360 hours to obtain the control forecast integration results. The input data is the initial value of the control forecast of 0.5 degrees, and the output is the 6-hour forecast results within the 360-hour forecast period. The output results are stored in the file system of the Sugon high-performance computer.

[0137] (7) Model perturbation parameter generation and model integration of 30 ensemble forecast members. Generate the random seeds, perturbation magnitude, random perturbation wave number, loss time correlation coefficient and other parameters required by the SPPT and SKEB model perturbation schemes for 30 ensemble forecast members. Use the model running program with the addition of model perturbations SPPT and SKEB schemes to complete the 360-hour integration of 30 ensemble forecast members and obtain the forecast results of 30 perturbation members. The input data are 30 perturbation initial values, random seeds, perturbation magnitude, random perturbation wave number, loss time correlation coefficient and other parameters. The output is the 6-hour forecast results within the 360-hour forecast period. The output results are stored in the file system of the Sugon high-performance computer.

[0138] (8) Unified post-processing of ensemble forecast member integral data. Unified post-processing converts all ensemble member binary forecast results into GRIB2 (Computer Independent Compressed Binary Code) format data for product post-processing and result archiving. This conversion process is started immediately after the ensemble forecast results of a certain time are generated, providing timely GRIB2 (Computer Independent Compressed Binary Code) data for product post-processing.

[0139] (9) Post-processing of conventional ensemble forecast products. Generate operational ensemble forecast products using the uniformly post-processed GRIB2 (Computer Independent Compressed Binary Code) format ensemble forecast results. The conventional ensemble forecast product module will generate ensemble forecast box plot data and typhoon path ensemble forecast data.

[0140] (10) Post-processing of extreme weather ensemble forecast products. Using ensemble forecast member data, model climate data, and historical climate data, etc., calculate and generate extreme forecast indexes for ground elements, ensemble forecast member deviation correction data and anomaly products, and 2-meter temperature anomaly probability forecast products.

[0141] (11) Solution of evolved singular vectors. According to the initial singular vectors corresponding to the northern and southern hemispheres, the linear model is integrated and cut to obtain the evolved singular vectors. The input is the 2.5 degree singular vector ground state initial value and the initial singular vector, and the output is the 2.5 degree evolved singular vector. The output results are stored in the file system of the Sugon high-performance computer.

[0142] Proof section (specific examples / experiments / simulations / pharmacological analysis / positive experimental data that can prove the inventiveness of the invention, etc.)

[0143] 1. Parallel test inspection and evaluation

[0144] (1) Comprehensive evaluation of forecasting capabilities

[0145] Figure 8A comprehensive score card for the parallel test period (61 days from September 1, 2018 to October 31, 2018) is given. It can be seen that the northern hemisphere wind field, except for the tropical wind field and the northern hemisphere 850hpa temperature field, are generally improved or equivalent compared with GRAPES_EPS and T639_EPS; all elements in the tropical region are improved overall.

[0146] In summary, in the technical research and development of the GRAPES global ensemble forecasting operational operation system provided by the present invention, the defects of the ensemble forecasting system that only uses initial value perturbations, such as insufficient divergence of ensemble members and insufficient reliability of the ensemble forecasting system, have been improved, so that the ensemble discreteness of different forecast variables can be further increased, and the ensemble average error can be reduced to a certain extent, especially in tropical regions. The application of the initial singular vector method in three target areas, namely the southern hemisphere, the northern hemisphere and the tropical cyclone area, in the global GRAPES_GEPS system is realized, and the model perturbation SPPT scheme and the SKEB scheme are used to reflect the uncertainty of the model itself, and an extreme weather ensemble forecast product is developed. It can provide a wealth of probabilistic forecast guidance products based on the GRAPES global model for business services, which is very beneficial to driving forecasters to use probabilistic forecasting products in depth in forecasting business and serve the society, and also provides forecasters with more forecast reference information.

[0147] It should be noted that the embodiments of the present invention can be implemented by hardware, software, or a combination of software and hardware. The hardware part can be implemented using dedicated logic; the software part can be stored in a memory and executed by an appropriate instruction execution system, such as a microprocessor or dedicated design hardware. It can be understood by a person of ordinary skill in the art that the above-mentioned devices and methods can be implemented using computer executable instructions and / or contained in a processor control code, such as a carrier medium such as a disk, CD or DVD-ROM, a programmable memory such as a read-only memory (firmware), or a data carrier such as an optical or electronic signal carrier. Such code is provided on the carrier medium. The device and its modules of the present invention can be implemented by hardware circuits such as very large-scale integrated circuits or gate arrays, semiconductors such as logic chips, transistors, etc., or programmable hardware devices such as field programmable gate arrays, programmable logic devices, etc., can also be implemented by software executed by various types of processors, and can also be implemented by a combination of the above-mentioned hardware circuits and software, such as firmware.

[0148] The above description is only a specific implementation mode of the present invention, but the protection scope of the present invention is not limited thereto. Any modifications, equivalent substitutions and improvements made by any technician familiar with the technical field within the technical scope disclosed by the present invention and within the spirit and principle of the present invention should be covered by the protection scope of the present invention.

Claims

1. A global ensemble prediction method for systematic negative deviation characteristics of height fields, characterized by: The global ensemble prediction method for the systematic negative deviation characteristics of the height field comprises: preprocessing the sea temperature data; Calculate the initial singular vectors corresponding to the northern and southern hemispheres; construct the initial perturbation values ​​of the ensemble forecast members; integrate for 360 hours to obtain the control forecast integration results; Conduct GRAPES_GEPS global ensemble forecast; According to the initial singular vectors corresponding to the northern and southern hemispheres, the linear model is integrated and tangent to obtain the evolving singular vectors for the prediction result detection; The global ensemble prediction method for the systematic negative deviation characteristics of the height field specifically includes: Step 1: The 0.25-degree resolution initial values ​​generated by the GRAPES global four-dimensional variational data assimilation system are interpolated to 2.5-degree resolution using dynamic upscaling technology to form the singular vector ground state initial values; Step 2: Calculate the initial singular vectors corresponding to the northern and southern hemispheres according to the Lanczos algorithm; interpolate the 0.25-degree resolution initial values ​​generated by the GRAPES global four-dimensional variational data assimilation system to 0.5-degree resolution using dynamic upscaling technology to obtain the GRAPES global ensemble forecast control forecast initial values; Step 3: Use the initial singular vectors and the evolved singular vectors to construct 15 disturbance fields through linear combination, and add / subtract 15 disturbance fields to the initial control forecast values ​​to construct the initial disturbance values ​​of the 1st to 30th ensemble forecast members; Step 4: Use the GRAPES_GFS_2-1-2-2 model with a resolution of 0.5 degrees and the initial value of the control forecast to integrate for 360 hours and obtain the control forecast integration result; generate the parameters required for the perturbation schemes of the SPPT and SKEB models of the 30 ensemble forecast members; Step 5: Using the model operation with the added model perturbations SPPT and SKEB schemes, perform 360-hour integration of 30 ensemble forecast members to obtain the forecast results of 30 perturbation members; perform global ensemble forecast of the systematic negative deviation characteristics of the height field; Step six, based on the initial singular vectors corresponding to the northern and southern hemispheres, integrate and tangent the linear model to obtain the evolved singular vectors and perform forecast result detection.

2. The global ensemble forecasting method for systematic negative deviation characteristics of height field according to claim 1, characterized in that: In step 2, the calculation of the initial singular vectors corresponding to the northern hemisphere and the southern hemisphere according to the Lanczos algorithm also includes: when a tropical cyclone exists at the starting time, calculating the initial singular vectors with the tropical cyclone area as the target area.

3. The global ensemble forecasting method for systematic negative deviation characteristics of height field according to claim 1, characterized in that: The step three includes: generating an initial disturbance field based on a formula, performing pairwise (,) addition and subtraction calculations on the GRAPES model analysis field, and constructing 2*M initial disturbance values ​​of the ensemble forecast; The specific steps include: (1) Read SV configuration information from the configuration file namelist.svd; (2) Read NH SVs, SH SVs, and TCSVs separately, and merge the read NH SVs, SH SVs, and TCSVs into global SVs perturbations; (3) Scaling the global SVs disturbance field; performing Gaussian linear combination on the scaled SVs disturbance field to generate m initial value disturbance fields; (4) Perform horizontal interpolation and smoothing processing on the m initial value disturbance fields, and perform addition and subtraction processing on the model analysis field to obtain 2*m ensemble forecast initial value fields.

4. The global ensemble forecasting method for systematic negative deviation characteristics of height field according to claim 1, characterized in that: The parameters include: random seed, disturbance magnitude, random disturbance wave number, loss time correlation coefficient and other parameters.

5. The global ensemble prediction method for systematic negative deviation characteristics of height field according to claim 1, characterized in that: Business ensemble forecast products include: box plot data and typhoon path ensemble forecast data.

6. A global ensemble prediction system for systematic negative deviation characteristics of height fields implementing the global ensemble prediction method for systematic negative deviation characteristics of height fields as claimed in any one of claims 1 to 5, characterized in that: The systematic negative deviation characteristics of the height field global ensemble forecast system include: Preprocessing module, used to preprocess the sea temperature data; The singular vector ground state initial value generation module is used to interpolate the 0.25 degree resolution initial values ​​generated by the GRAPES-4DVAR assimilation analysis to 2.5 degree resolution using dynamic upscaling technology to form the singular vector ground state initial values; The initial singular vector calculation module is used to calculate the initial singular vectors corresponding to the northern hemisphere and the southern hemisphere according to the Lanczos algorithm; it is also used to calculate the initial singular vectors with the tropical cyclone area as the target area when there is a tropical cyclone at the starting time; The control forecast initial value generation module is used to interpolate the 0.25 degree resolution initial values ​​generated by the GRAPES-4DVAR assimilation analysis to 0.5 degree resolution using the dynamic upscaling technique to obtain the GRAPES global ensemble forecast control forecast initial values; The disturbance field generation and ensemble forecast member disturbance initial value generation module is used to construct 15 disturbance fields through linear combination using the initial singular vectors and the evolved singular vectors, and add / subtract 15 disturbance fields from the control forecast initial value to construct the disturbance initial values ​​of the 1st to 30th ensemble forecast members; The control forecast integration module is used to integrate for 360 h using the GRAPES_GFS_2-1-2-2 model with a resolution of 0.5 degrees and the initial value of the control forecast of 0.5 degrees to obtain the control forecast integration result; The model perturbation parameter generation and 30 ensemble forecast member model integration module is used to generate the parameters required by the SPPT and SKEB model perturbation schemes of the 30 ensemble forecast members; the model operation with the addition of the model perturbation SPPT and SKEB schemes is used to perform 360h integration of the 30 ensemble forecast members to obtain the forecast results of the 30 perturbation members; Post-processing module for GRAPES_GEPS global ensemble forecast; The evolved singular vector calculation module is used to obtain the evolved singular vectors according to the initial singular vectors corresponding to the northern hemisphere and the southern hemisphere and the integral tangent linear model.

7. The global ensemble prediction system for systematic negative deviation characteristics of height fields according to claim 6, characterized in that: The post-processing module comprises: The integrated data unified post-processing unit is used to convert the binary forecast results of all ensemble members into GRIB2 format data; Conventional ensemble forecast product post-processing unit, used to generate operational ensemble forecast products using the ensemble forecast results in the unified post-processed GRIB2 format; The extreme weather ensemble forecast product post-processing unit is used to calculate and generate the extreme forecast index of ground elements, ensemble forecast member deviation correction data and anomaly products, and 2m temperature anomaly probability forecast products using ensemble forecast member data, model climate data and historical climate data.

8. A storage medium for receiving a program input by a user, wherein the stored computer program enables an electronic device to execute the global ensemble forecasting method for systematic negative deviation characteristics of a height field as claimed in any one of claims 1 to 5.

9. An information data processing terminal, characterized in that: The information data processing terminal includes a memory and a processor, wherein the memory stores a computer program, and when the computer program is executed by the processor, the processor executes the global ensemble forecasting method for systematic negative deviation characteristics of the height field as described in any one of claims 1 to 5.

Images