Global focus on the Yellow River Basin's sub-seasonal-seasonal refined prediction method and system

Abstract

The embodiments of this specification provide a sub-seasonal-seasonal refined prediction method and system focusing on the Yellow River Basin globally, wherein the method comprises: processing the original CRA40 data, GFS data and OSTIA data into a data format that can be used for model input, driving the FGOALS-f2 model to run, using the FGOALS-f2 model to perform numerical simulation using dynamic ensemble prediction technology to obtain numerical simulation results; inputting the numerical simulation results into the FGOALS-UFS model, using the FGOALS-UFS model to perform regional zoom encryption prediction to generate prediction data, and converting the prediction data output by the FGOALS-UFS model into longitude-latitude grid prediction data that is convenient for subsequent analysis; using the NetCDF data format as the standard, using the longitude-latitude grid prediction data to generate common data required for business prediction, and interpolating to the resolution of a predetermined reading, and outputting multiple meteorological elements including potential height field, temperature field, wind field, specific humidity, ground temperature and precipitation.

Description

Technical Field

[0001] This document relates to the field of computer technology, and in particular to a sub-seasonal-seasonal refined prediction method and system focusing globally on the Yellow River Basin. Background Art

[0002] In multi-timescale forecasting, from weather to climate, subseasonal-to-seasonal (S2S) forecasts play a key interface role. They primarily focus on the potential predictability of weather events within a two-week to two-month timeframe, building a bridge between numerical weather forecasts and short-term climate predictions, bridging the gap between medium- and long-term weather forecasts and seasonal forecasts. The development and implementation of S2S-scale forecasts holds significant scientific significance and societal application value, both from the perspective of professional development in seamless weather-climate forecasting and from the perspective of societal needs. Subseasonal-scale forecasts are crucial for proactive disaster reduction and policy formulation by management decision-makers in the areas of extreme weather disaster prevention and mitigation, and are also crucial for agricultural production.

[0003] Subseasonal-seasonal forecasts are of great significance to the Yellow River Basin, particularly in agriculture, water resources management, disaster prevention and mitigation, and ecological protection. In agriculture, the Yellow River Basin is one of China's important agricultural production areas, and subseasonal-seasonal forecasts can help farmers make more accurate agricultural plans. Understanding weather trends over the coming weeks and months can help them choose appropriate crop planting times, fertilization, and irrigation strategies to maximize crop yields. Regarding water resources management, the Yellow River is China's second largest river, making water resource management crucial for the sustainable development of the basin. Subseasonal-seasonal forecasts can help water conservancy departments predict precipitation over the coming weeks and months, enabling them to better regulate reservoir storage, flood discharge, and irrigation, ensuring water supply in the basin. In terms of disaster prevention and mitigation, the Yellow River Basin is often threatened by floods, and subseasonal-seasonal forecasts can help governments and relief agencies prepare in advance. By predicting precipitation in the next few weeks to months, flood prevention and rescue measures can be taken in a timely manner to reduce the impact of floods on people's lives and property. In terms of ecological and environmental protection, the ecological environment of the Yellow River Basin is fragile. Subseasonal-seasonal forecasts can provide information on the temperature and precipitation in the next few weeks to months, helping relevant departments to formulate reasonable ecological protection strategies, take effective ecological restoration measures, and promote the improvement of the ecological environment in the Yellow River Basin.

[0004] The parameterization of physical processes is the greatest source of uncertainty in model simulations and predictions. The parameterization of the convection-cloud-radiation process is particularly complex, and the overall performance of general circulation models is largely limited by the parameterization of these wet physical processes. The current Resolved Convective Precipitation (RCP) scheme can make cumulus convective precipitation explicit. This scheme rewrites the traditional cumulus convective precipitation equation using a single-parameter cloud microphysics equation and then calculates their cloud microphysical properties separately. This reduces the errors introduced by traditional cumulus convective parameterization schemes due to the averaging of convective effects and their strong reliance on the accuracy of convective parameters.

[0005] The results show that the convective RCP scheme significantly improves the simulation errors of the tropical atmospheric intraseasonal oscillation (MJO) and the intertropical convergence zone (ITCZ) in the model. The weak MJO and the double equatorial convergence zone (Double_ITCZ) are internationally recognized difficulties in climate simulation. The FGOALS-f model using the RCP scheme simulates the MJO eastward with enhanced intensity and reasonable speed, with significantly improved performance.

[0006] Taking into account the demand for refined predictions in the Yellow River Basin, the variable grid downscaling technology not only ensures the stability and calculation accuracy of the forecast integral with fewer computing resources, but also removes the singularity of the poles through the cubic sphere grid, overcoming the problem of forecast calculation in the polar region. In addition, it also realizes the local refined dynamic downscaling calculation technology approach, which can better reflect the physical essential characteristics of atmospheric motion than the statistical downscaling method, and provides perfect technical support for the realization of refined forecasts, global-regional and integrated weather and climate forecasts.

[0007] For a long time, the numerical algorithms for solving atmospheric models have mainly been spectral methods and finite difference methods. Spectral methods have high computational accuracy and efficiency, and are therefore widely used in global models. Grid-type difference algorithms generally have second-order accuracy, are computationally simple, and can better perform horizontal region decomposition to adapt to large-scale parallel computing (except for extreme point problems). With the development of science and technology, finite volume, finite element and other algorithms have been introduced more and more into atmospheric models. These algorithms can be suitable for more general grid shapes and are suitable for developing large numerical models based on quasi-uniform grids to avoid the computational impact of polar region problems.

[0008] For spectral models, the main challenges associated with increasing resolution lie in their computational efficiency and the effective characterization of small-scale information. The Legendre transform in spectral methods has an O(N³) complexity, which leads to a sharp increase in computational complexity as resolution increases. The global nature of communication also affects the parallel efficiency of spectral methods. The European Forecast Center has made significant improvements in this regard. The fast Legendre transform significantly increases the computational and parallel efficiency of spectral methods, enabling global T7999 (approximately 2.5 km) spectral model calculations. Recently, the European Center has even conducted global seasonal simulations at 1 km resolution. In addition to computational efficiency, spectral methods also have limitations in characterizing small and medium-scale processes dominated by divergence modes due to the Gibbs effect, making it difficult to characterize fine small and medium-scale structures. In this regard, grid models are relatively more suitable.

[0009] For grid point models, regardless of the type of solution algorithm used (finite difference, finite volume or finite element), the design and use of the discretization algorithm should generally follow certain guiding principles. Discretization methods based on physical constraints have good theoretical and practical value. This is different from discretization algorithms that simply pursue computational accuracy (numerical convergence rate). Previous studies have shown that simply using computational accuracy as an indicator is difficult to meet the requirements of establishing efficient atmospheric numerical simulations (balance between performance and cost). Discretization methods that comply with physical constraints include some algorithms that satisfy the conservation of integral constraints, such as energy conservation, pseudo-energy conservation, etc., as well as discretization algorithms based on the principle of physical consistency. Of course, computational accuracy is still an important aspect. Summary of the Invention

[0010] The purpose of the present invention is to provide a sub-seasonal to seasonal refined prediction method and system focusing on the Yellow River Basin globally, aiming to solve the above-mentioned problems in the prior art.

[0011] The present invention provides a sub-seasonal to seasonal refined prediction method focusing on the Yellow River Basin globally, comprising:

[0012] The original CRA40 data, GFS data and OSTIA data were processed into a data format that can be used for model input, driving the FGOALS-f2 model to run, and numerical simulations were performed using the FGOALS-f2 model using the dynamic ensemble prediction technology to obtain numerical simulation results.

[0013] Input the numerical simulation results into the FGOALS-UFS model, use the FGOALS-UFS model to perform regional zoom encryption prediction, generate prediction data, and convert the prediction data output by the FGOALS-UFS model into longitude and latitude grid prediction data that is convenient for subsequent analysis;

[0014] Using the NetCDF data format as the standard, the latitude and longitude grid prediction data is used to generate the common data required for business forecasting, and interpolated to the resolution of the predetermined reading, and the output includes multiple meteorological elements such as potential height field, temperature field, wind field, specific humidity, ground temperature and precipitation.

[0015] The present invention provides a sub-seasonal to seasonal refined prediction system focusing on the Yellow River Basin globally, comprising:

[0016] The FGOALS-f2 module is used to process the original CRA40 data, GFS data, and OSTIA data into a data format that can be used for model input, drive the FGOALS-f2 model, and use the FGOALS-f2 model to perform numerical simulations using dynamic ensemble prediction technology to obtain numerical simulation results;

[0017] The FGOALS-UFS module is used to input the numerical simulation results into the FGOALS-UFS model, use the FGOALS-UFS model to perform regional zoom encryption prediction, generate prediction data, and convert the prediction data output by the FGOALS-UFS model into longitude and latitude grid prediction data that is convenient for subsequent analysis;

[0018] The post-processing module is used to generate common data required for business forecasting using the latitude and longitude grid prediction data based on the NetCDF data format, and interpolate to the resolution of the predetermined reading, and output multiple meteorological elements including potential height field, temperature field, wind field, specific humidity, ground temperature and precipitation.

[0019] An embodiment of the present invention also provides an electronic device, comprising: a memory, a processor, and a computer program stored on the memory and executable on the processor. When the computer program is executed by the processor, the steps of the above-mentioned global sub-seasonal-seasonal refined prediction method focusing on the Yellow River Basin are implemented.

[0020] An embodiment of the present invention also provides a computer-readable storage medium, on which an implementation program for information transmission is stored. When the program is executed by a processor, the steps of the above-mentioned global sub-seasonal-seasonal refined prediction method focusing on the Yellow River Basin are implemented.

[0021] The adoption of the embodiments of the present invention is of great significance to agriculture, water resources management, disaster prevention and mitigation, and ecological and environmental protection in the Yellow River Basin. The implementation of 12.5 seasonal ensemble forecasts in the Yellow River Basin for the first time can help relevant departments and farmers make more informed decisions, improve production efficiency, reduce losses, and ensure public safety and the sustainable development of the ecological environment. BRIEF DESCRIPTION OF THE DRAWINGS

[0022] In order to more clearly illustrate one or more embodiments of this specification or the technical solutions in the prior art, the following briefly introduces the drawings required for use in the embodiments or the description of the prior art. Obviously, the drawings described below are only some embodiments recorded in this specification. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying any creative work.

[0023] Figure 1 This is a flow chart of a sub-seasonal to seasonal refined prediction method for the global Yellow River Basin according to an embodiment of the present invention;

[0024] Figure 2a is a schematic diagram of a cubic sphere grid according to an embodiment of the present invention;

[0025] Figure 2b 1. This is a schematic diagram of a cube sphere grid according to an embodiment of the present invention after Schmidt transformation to achieve a variable grid.

[0026] Figure 3a is a schematic diagram of the Yellow River Basin according to an embodiment of the present invention;

[0027] Figure 3b Schematic diagram of the Yellow River Basin grid change according to an embodiment of the present invention;

[0028] Figure 4 is a technical flow chart of the preprocessing module of an embodiment of the present invention;

[0029] Figure 5 Schematic diagram of an initialization scheme of the FGOALS-f2 prediction system according to an embodiment of the present invention;

[0030] Figure 6 Schematic diagram of the FGOALS-f2 prediction method according to an embodiment of the present invention;

[0031] Figure 7 is a schematic diagram of an execution script according to an embodiment of the present invention;

[0032] Figure 8 Schematic diagram of post-processing of FGOALS-UFS prediction test results according to an embodiment of the present invention;

[0033] Figure 9 This is a schematic diagram of a global sub-seasonal to seasonal refined prediction system focusing on the Yellow River Basin according to an embodiment of the present invention;

[0034] Figure 10 is a schematic diagram of an electronic device according to an embodiment of the present invention. DETAILED DESCRIPTION

[0035] In order to enable those skilled in the art to better understand the technical solutions in one or more embodiments of this specification, the technical solutions in one or more embodiments of this specification will be clearly and completely described below in conjunction with the drawings in one or more embodiments of this specification. Obviously, the described embodiments are only part of the embodiments of this specification, not all of the embodiments. Based on one or more embodiments of this specification, all other embodiments obtained by ordinary technicians in this field without making creative efforts should fall within the scope of protection of this document.

[0036] Method Example

[0037] According to an embodiment of the present invention, a sub-seasonal to seasonal refined prediction method focusing on the Yellow River Basin is provided. Figure 1 Flowchart of the sub-seasonal-seasonal refined prediction method for the global focus on the Yellow River Basin according to an embodiment of the present invention. Figure 1 As shown, the sub-seasonal-seasonal refined prediction method for the global focus on the Yellow River Basin according to an embodiment of the present invention specifically includes:

[0038] Step S101: Processing the original CRA40 data, GFS data, and OSTIA data into a data format that can be used for model input, driving the FGOALS-f2 model to run, and using the FGOALS-f2 model to perform numerical simulation using dynamic ensemble prediction technology to obtain numerical simulation results. Step S101 specifically includes:

[0039] Download the original CRA40 data, GFS data and OSTIA data to the server through the enhanced data download program;

[0040] Perform data quality control by checking data integrity and variable value rationality;

[0041] Process the original CRA40 data, GFS data and OSTIA data into NetCDF format;

[0042] Creating a prediction experiment includes case creation, case setting, and case compilation. Case creation creates a configuration folder for the simulation case in the corresponding directory; case setting is used to configure model components, external forcing files, model integration start and end times, and model output variables; case compilation generates an executable file cesm.exe in the compilation directory;

[0043] After successful compilation, an executable file of FGOALS-f2 is generated. After running the executable file, the FGOALS-f2 mode runs successfully.

[0044] The ensemble samples are generated by combining the lagged average forecast method and the time-lagged perturbation method;

[0045] Two FGOALS-f2 ensembles are obtained by combining the atmospheric initial values ​​at 00:00, 06:06, 12:02, and 18:00 and the ocean initial values ​​closest to the date using the LAF method. The time-lag perturbation method is used to increase the number of ensembles as needed, forming four ensemble prediction samples based on the start time of 00:00, 06:06, 12:02, and 18:00. Two ensemble sample initial values ​​are generated for each start time according to the length of the atmospheric forcing time, and the number of ensemble samples can be expanded to 8 as needed.

[0046] The FGOALS-f2 model is used to forecast variables for 65 days on a daily basis, with each day as the starting day. The forecast results on the starting day and the 20 days before it are included in the current ensemble members, and the next 65 ensemble forecasts based on the starting day are obtained in turn. The 40 ensemble members are averaged to obtain the syndicate mean forecast factor based on the starting day.

[0047] Step S102: input the numerical simulation results into the FGOALS-UFS model, use the FGOALS-UFS model to perform regional zoom encryption prediction, generate prediction data, and convert the prediction data output by the FGOALS-UFS model into longitude and latitude grid prediction data that is convenient for subsequent analysis; the FGOALS-UFS model specifically includes: adopting a unified forecast model UFS, wherein the dynamic framework of the UFS is the global grid FV3 divided by a cubic sphere, and the physical process of the UFS adopts the universal atmospheric physics package CCPP.

[0048] Step S103 uses the latitude and longitude grid forecast data to generate common data required for business forecasting, using the NetCDF data format as a standard. The data is interpolated to a predetermined reading resolution, and output includes multiple meteorological elements, including geopotential height, temperature, wind, specific humidity, surface temperature, and precipitation. The global resolution of these meteorological elements is 100 km, the regional focus is 12.5 km, the vertical resolution is 32 layers, the top model layer is 1 hPa, and the temporal resolution and assimilation time window are 6 hours.

[0049] The above technical solutions of the embodiments of the present invention are described in detail below with reference to the accompanying drawings.

[0050] The global sub-seasonal to seasonal refined prediction system focusing on the Yellow River Basin is based on the internationally advanced climate system model FGOALS-f2 with independent intellectual property rights, and carries out sub-seasonal to seasonal predictions focusing on the Yellow River Basin.

[0051] The forecast output of the climate system model FGOAS-f2 is used as external forcing to drive the global focused gridded downscaling system FGOALS-UFS, achieving a 65-day ensemble forecast covering the Yellow River Basin (1200 km x 1200 km) at a resolution of 12.5 km. This ultimately provides sub-seasonal to seasonal forecast products for multiple meteorological elements, including geopotential height, temperature, wind, specific humidity, temperature, precipitation, and more. The following describes this forecast system, focusing on the global climate model FGOAS-f2, the refined forecast model FGOALS-UFS, input data, data preprocessing and initialization techniques, dynamic ensemble forecasting technology, operational process, post-processing modules, and data products.

[0052] Global Climate Model FGOASL-f2

[0053] FGOALS-f2 is a climate system model designed for intraseasonal and seasonal climate prediction. It comprises four component models: atmosphere, ocean, land, and sea ice, and a coupler. The atmospheric model is the independently developed Finite Volume Atmosphere Model of IAP LASG Version 2 (FAMIL2), with a horizontal resolution of 100 kilometers and 32 vertical layers. FAMIL2 employs a independently developed Resolved Convective Precipitation (RCP) scheme to explicitly represent cumulus convective precipitation. This scheme rewrites traditional cumulus convection using single-parameter cloud microphysics equations and then calculates their cloud microphysical properties separately, mitigating errors in traditional cumulus convection parameterization schemes due to averaging of convective effects and a strong reliance on convective parameter accuracy. The RCP scheme achieves near-scale adaptation at horizontal resolutions ranging from 12.5 to 100 kilometers, requiring no parameter adjustments. It has excellent simulation capabilities for the tropical intraseasonal oscillation (MJO), the Double-ITCZ, the El Niño-Southern Oscillation (ENSO), and tropical cyclones. The land surface model is CLM4.0, with the same horizontal resolution as the atmosphere model. The ocean model is POP2, which uses a dispalced-pole grid, rotating the North Pole to Greenland, with a grid resolution of gx1v6 (equivalent to a 1°×1° resolution and 60 vertical layers). The sea ice model is CICE4, with the same horizontal resolution as the ocean model. Interpolation and data exchange between the component models are performed directly using the seventh-generation coupler, CPL7, developed by the National Center for Atmospheric Research (NCA), enabling coupled parallel computing. For ease of use, data are interpolated to a horizontal resolution of 1 degree and a vertical resolution of 17 standard isobars. Daily data are generated using the time-varying Newton relaxation method, assimilating initial atmospheric variables from the CRA40 reanalysis data, including temperature, humidity, surface pressure, sea level pressure, and surface wind.

[0054] FGOALS-f2 successfully addressed the issue of coupled model instability, enabling the coupled system model to operate stably for over 500 model years at C96 (100 km). The FGOALS-f2 climate system model is primarily used for weather and climate prediction on multiple timescales, including sub-seasonal, seasonal, and interannual.

[0055] Focus on the refined prediction model FGOALS-UFS in the Yellow River Basin.

[0056] The refined climate prediction model FGOALS-UFS, focusing on the Yellow River Basin, utilizes the Common Infrastructure for Modeling the Earth (CIME) Unified Forecast System (UFS), a unified forecasting model developed by the National Centers for Environmental Prediction (NCEP) and the National Center for Atmospheric Research (NCAR). The UFS model dynamical framework uses the FV3 global grid with cubic sphere subdivision. The UFS model physics utilizes the Global Model Testbed (GMTB) in collaboration with NOAA and NCAR, integrating physical parameters with the Common Community Physics Package (CCPP) software framework. The CCPP's physical process parameter schemes are highly interoperable, with automatic allocation of process variables, making it easy to use and error-free. A globally focused stretched grid is a key method for improving the horizontal resolution of the model system. Global focused stretching is a key feature of FGOALS-UFS.

[0057] FGOALS-UFS is a system based on the FGOALS-f2 global focused fine-grained forecast system (resolution ~100 km) to provide variable-grid downscaling forecasts focused on the Yellow River Basin. The Unified Forecast System (UFS), originally a weather forecasting system used by the United States National Oceanic and Atmospheric Administration, is driven by the FGOALS-f2 model to generate forecasts, forming the FGOALS-UFS model. Based on the six-hour three-dimensional temperature, geopotential height, humidity, meridional wind, and zonal wind output from the FGOALS-f2 global forecast system (resolution ~100 km), the FGOALS-UFS system is driven by the Newton relaxation method (nudging) to provide variable-grid fine-grained forecasts focused on the Yellow River Basin. The aggregation center is Yan'an (36°N, 110°E). Nudging analysis and forecast nudging time lag methods are combined to achieve ensemble fine-grained forecasts with a focal area resolution of 12.5 km. For the first time, 12.5 seasonal ensemble forecasts were achieved in the Yellow River Basin. The input data for the global aggregated variable grid FGOALS-UFS C96_r4 are shown in Table 1:

[0058] Table 1

[0059]

[0060] FGOALS-UFS utilizes the cubic-grid finite volume dynamics kernel FV3, which offers outstanding performance in terms of computational efficiency, accuracy, local refinement, and scalability. The FV3 dynamics kernel implements a variable grid stretching function through the Schmidt transformation. FGOALS-UFS uses a spherical cubic grid and applies the Schmidt (1977) transformation to "pull" grid intersections toward a "target" point (corresponding to the center point of focused high resolution), thereby achieving a globally clustered stretched variable grid. Clustered stretching is accomplished in two steps: the grid is stretched toward the South Pole to achieve the desired level of refinement, and then a rigid body rotation is used to rotate the South Pole to the target point. Stretching toward the South Pole means that the longitude of the point remains unchanged, only the latitude is changed, greatly simplifying the transformation process.

[0061] The transformation from latitude to θ is given by:

[0062]

[0063] The distortion D is a function of the stretch factor c, which can be any positive number. If c = 1, no stretching occurs, which is the setting for a uniform mesh. The Schmidt transform is similar to other forms of transforms and does not require any changes to the solver of the dynamical framework.

[0064]

[0065] At C 96 resolution, the uniform grid resolution is 100 km globally, and the focus area resolution under 8x focusing can reach 12.5 km, covering an area of ​​1200 km*1200 km centered on Yan'an.

[0066] FV3 supports global focus variable grid, Figure 2a and Figure 2b The grid diagram before and after the grid change is shown. The grid-changing FV3 can smoothly transition to a higher resolution in the focal area, achieving the encryption function of a specific area. This grid-changing system is also called the Little Apple fine-grained downscaling system (the world is regarded as an apple, and the focal area is the apple core at the center). At C96 resolution (a focal area has 96 uniform grid points), the resolution of the focal area reaches 12.5 kilometers. The outermost area is the background area provided by FGOALS-f2 (as shown in Figure 3, the resolution gradually transitions from 100 kilometers at the periphery to 12.5 kilometers inward). The resolution is 100 kilometers, and the resolution gradually increases towards the focal area.

[0067] The atmospheric data uses weather forecast data from the Global Forecast System (GFS). GFS is a global numerical weather forecast model system developed and maintained by the National Centers for Environmental Information (NCEP) of the United States. It is widely used to generate weather forecasts worldwide and provide meteorological information to meteorologists, weather forecasters, researchers and the public.

[0068] The GFS model runs multiple times daily, generating global weather forecasts for the next several days. This data includes various meteorological elements, such as temperature, wind speed and direction, precipitation, and cloud cover. Forecasts range from hours to days and are primarily used for short- and medium-term weather forecasts.

[0069] The ocean data used is the global sea surface temperature observation dataset (METOFFICE-GLO-SST-L4-NRT-OBS-SST-V2, hereinafter referred to as OSTIA) released by the UK Met Office. This dataset contains global sea surface temperature observations obtained from data sources such as satellite observations and buoys.

[0070] OSTIA is a non-real-time data that has undergone a certain level of processing. Currently in version V2, it offers shorter latency and higher spatial resolution and accuracy than version V1. OSTIA data is of great significance in many fields. Due to its significant influence on processes such as ocean circulation and air-sea interactions, it plays a vital role in weather forecasting, climate research, and marine ecosystem monitoring.

[0071] GFS data is stored in a standard meteorological data format, with the original data in GRIB (Gridded Binary) format. This system downloads the raw data and preprocesses it into NetCDF format, which effectively stores multidimensional grid data and facilitates data processing and analysis. The preprocessed GFS data has a horizontal spatial resolution of 1°×1°, a vertical spatial resolution of 32 layers for isobaric surface elements, and a temporal resolution of 6 hours. Specific element information is shown in Table 2:

[0072] Table 2

[0073]

[0074]

[0075] OST data is stored in a standard meteorological data format. The original data is in GRIB format. After downloading the original data, this system preprocesses it into NetCDF format, which can effectively store multi-dimensional grid data and facilitate data processing and analysis.

[0076] Data preprocessing technology mainly includes three parts: data download, data quality control and data format preprocessing. The data download part mainly provides guarantees for timely and complete download and transmission of the initial field of the prediction system to the server where the prediction system is located by enhancing the robustness and automatic error correction function of the data download program, performing dual-server backup download, and expanding the download function of multi-source data, etc. from the aspects of software and hardware; the data quality control part mainly provides guarantees for the correct initialization of the prediction system by detecting the integrity of the data file and the rationality of the variable values; the data format preprocessing part mainly converts various formats of data into data formats that can be directly read by the atmospheric and ocean component models of the prediction system through functions such as data space interpolation, missing variable calculation, and data unit unification, ensuring that the prediction system can correctly identify each data variable. The main function of the preprocessing module is to provide the prediction system with real-time, stable, and accurate sea temperature and atmospheric initial fields to ensure that the system can start smoothly at any time. The technical process of the preprocessing module is as follows: Figure 4 .

[0077] The initialization scheme is the key to the prediction system. The FGOALS-f2 model initializes atmospheric and oceanic conditions based on the nudging technique described by Jeuken et al. (1996). Specifically, the atmospheric initialization scheme of the FGOALS-f2 prediction system utilizes a nudging scheme with a weighting factor that varies with time. The assimilation scheme is divided into two parts: reanalysis data (Reanalysis) nudging and forecast data (Forecast) nudging. The Global Analysis Dataset (CRA-40) of the China Meteorological Administration is used as the true value in the reanalysis data nudging. CRA40 was developed by the China Meteorological Administration and uses the Global Spectral Model (GSM) and Gridpoint Statistical Interpolation (GSI) of the National Centers for Environmental Prediction (NCEP) of the United States for atmospheric reanalysis (Liu et al., 2023). The process is shown in the following formula:

[0078] x(t)=x modsl (t)+N(t)[x obs (t)-x modsl (t)]

[0079] Where t represents a certain moment, x represents the value after nudging, and x model are the values ​​calculated by the model’s own physical and dynamical processes. obsrepresents the "true" value at each time step, calculated using bilinear interpolation from the 6-hour reanalysis data. N is the nudging coefficient (Newton relaxation coefficient) with a time-varying weighting factor calculated using the following two equations:

[0080]

[0081]

[0082] Δt is the model integration step, which is 1800 seconds at the KJ-FGOALS-f2 C96 resolution. T is the nudging time window. To avoid initial shock, reanalysis data nudging began on January 1, 1976, and the model values ​​were forced with "true" values ​​every 6 hours. Figure 5 Figure (a) shows the weighting factor factor0(t) as it varies cosine over time. At the beginning and end of the time window, a larger nudging coefficient causes the model to approach the true value more quickly. However, in the middle of the time window, the confidence that the reanalysis represents the true value decreases. To maintain the model's performance, a smaller nudging coefficient is chosen.

[0083] The forecast data nudging algorithm is the same as the reanalysis data nudging algorithm. In this part, a specific day is selected from the reanalysis data nudging as the initial condition for the forecast. To ensure a certain degree of credibility for the forecast fields from the previous few days, weather forecast data is also nudged into the model, and new weighting factors that vary over time are added to the forecast period. N is calculated as follows:

[0084] N1(t)=N0(t)*factor1(t)

[0085]

[0086] The parameter m is the number of days for nudging the forecast data, and the duration should be expected to be less than 10 days. Figure 5 In the figure, (b) shows how factor0(t) varies with the cosine function.

[0087] like Figure 5 Figure 2. Temporal variation of weighting factors in (a) reanalysis nudging and (b) forecast nudging. In (a), reanalysis nudging uses CRA40 reanalysis data and starts integration based on the CMIP6 historical experiment. In (b), forecast nudging uses GFS weather forecast data. Dashed lines indicate the four ensemble members using time-lagged perturbations.

[0088] Monthly and seasonal climate model forecasts, due to their long forward integration times, have greater uncertainty than short- to medium-term numerical weather forecasts of less than two weeks. Therefore, the use of ensemble forecasting techniques is essential (Luan Yihua et al., 2016; Ma Leiming et al., 2017). By probabilistically analyzing the forecasts of dozens of members, we can obtain probabilistic information on various possible climate change outcomes, providing more scientific technical support for climate assurance decision-making (Zhou Tianjun et al., 2014). Similarly, due to the high computational cost associated with the large number of forecast members, the long forecast timeframe, and the refinement of forecast products, this system utilizes time-delay ensemble forecasting techniques (Wang Lei, 2020).

[0089] Specifically, the ensemble samples are generated by combining the lagged average forecast method and the time-lag perturbation method. Among them, the lagged average forecast method (LAF) was proposed by Hoffman and Kalnay in the 1980s. For example, the introduction of Dalcher et al. (1988) Figure 5 As shown in the figure, a 5-day forecast based on LAF is obtained by taking the arithmetic or weighted average of the forecast results for the same day (the 5th day of the forecast when LT=0, the 6th day of the forecast when LT=1, …, the n+5th day of the forecast when LT=n) under different lead times (LT; LT=0 day for the current day, LT=1 day for the previous day, …, LT=n day for the previous n days).

[0090] FGOALS-f2 generates two ensemble samples per wall-clock day, derived from the LAF method by combining the atmospheric initial values ​​at 00:00, 06:00, 12:00, and 18:00 with the nearest ocean initial values. In addition to the LAF method, FGOALS-f2 also incorporates a time-lagged perturbation method, allowing for the temporary increase of the number of ensembles as needed. Four ensemble prediction samples are generated based on the start time (00:00, 06:00, 12:00, and 18:00). Two ensemble sample initial values ​​are generated for each start time, depending on the duration of the atmospheric forcing. This number can be expanded to eight ensemble samples as needed.

[0091] The FGOALS-f2 model provides daily forecasts for 65 days (including the current day, two ensemble members). Each day is designated as the start date. The forecast results for the start date and the 20 days preceding it (a total of 40 ensemble members) are incorporated into the current ensemble member (a total of 40 ensemble members), and an ensemble forecast for the next 65 days based on the start date is generated. The ensemble average of the 40 ensemble members is then used to generate the ensemble mean forecast factor based on the start date.

[0092] like Figure 6As shown in Figure 3, the 12-pentad ensemble mean forecasts output by the model were processed using the above method to test the feasibility and forecasting skills of the comparative statistical downscaling method.

[0093] FGOALS-f2 prediction run: Data preprocessing converts the raw CRA40, GFS, and OSTIA data into a data format suitable for model input, driving the FGOALS-f2 model run. To perform numerical simulations using the FGOALS-f2 model, you first need to create a prediction experiment, which includes three steps: case creation, case settings, and case compilation. Case creation creates a configuration folder for the simulation case in the corresponding directory; case settings are used to configure model components, external forcing files, model integration start and end times, model output variables, etc.; case compilation generates the executable file cesm.exe in the compilation directory. The FGOALS-f2 executable program generates an executable file after successful compilation. Running the executable file successfully executes the model.

[0094] FGOALS-f2 data preprocessing:

[0095] (1) Run the GFS data preprocessing program gfs_process.sh

[0096] This shell script obtains the original GFS data on the run date and uses the methods provided by ncl to process the original data into netCDF data of the specifications required for model running.

[0097] (2) Run the OST data preprocessing program

[0098] This shell script obtains the raw OST data of the run date and calls the methods provided by ncl and cdo to process the raw data into netCDF data of the specifications required for model running.

[0099] FGOALS-f2 creates a prediction test:

[0100] (1) FGOALS-f2 case creation

[0101] Enter the scripts folder in the main directory of the FGOALS-f2 model and create a simulation case using the system script create_newcase. The location of the script is:

[0102] $HOME / FGOALS / scripts / create_newcase

[0103] Create a simulation case named BQ_job1 and use the following command:

[0104] $create_newcase-case BQ_job1-res C96_f09_g16-compset FAMIP-mach jyun-user_grid_file user_config_grid.xml.

[0105] Among them, the -case parameter sets the case name (the name of the simulation case); the -res parameter sets the model resolution, where C96 represents the atmospheric model resolution; the -compset parameter selects the composition of each component model; -mach specifies the computer where the model runs, and -user_grid_file sets the grid of each component model.

[0106] (2) FGOALS-f2 case setting

[0107] After completing the creation of the simulation case, it is necessary to further configure the simulation case, including model component configuration, external forcing file configuration, model integration start and end time, model output variables, etc. The case settings are mainly performed in the simulation case configuration folder, the specific location of which is described in (1).

[0108] The configuration steps are as follows:

[0109] First, run the cesm_setup_famil script as follows: $HOME / FGOALS / scripts / BQ_job1 / cesm_setup_famil. Then, in the simulation case configuration folder, open the $HOME / FGOALS / scripts / BQ_job1 / env_run.xml file. According to the various variable annotations in the file, set the model operation mode (such as startup, branch, and hybrid), reference case, integration start and end time, restart file output frequency, etc.

[0110] Finally, configure the atmospheric component model for the FGOALS-f2 model separately. This configuration file consists of two files: input.nml and diag_table. Both files are located in the simulation run directory: $HOME / FGOALS / runs / BQ_job1. input.nml is used to configure the storage location of the atmospheric model's external forcing fields, such as aerosols and greenhouse gases, as well as the data assimilation method. diag_table is used to set the type and frequency of atmospheric output variables.

[0111] (3) FGOALS-f2 example compilation

[0112] The simulation case compilation is also performed in the case configuration directory $HOME / FGOALS / scripts / BQ_job1 / . After entering the case configuration directory, run the compilation script BQ_job1.build_famil as follows to start the simulation case compilation. $. / BQ_job1.build_famil The compilation process not only compiles the model execution program, but also performs a series of configuration checks such as the model run initial field detection based on the simulation case settings in (2).

[0113] To run the FGOALS-f2 prediction test: In the $HOME / FGOALS / scripts / directory, run $. / BQ_job1.run to submit the job. After the run, results are generated in the $HOME / FGOALS / runs / BQ_job1 / run / directory. View the corresponding output files. Results are also recorded in $HOME / FGOALS / scripts / BQ_job1 / CaseStatus.

[0114] Post-processing of FGOALS-f2 prediction test results:

[0115] The specific process of post-processing of FGOALS-f2 prediction test results is as follows:

[0116] (1) Soft link data to a temporary / dedicated directory;

[0117] (2) Copy postp.csh to a temporary / dedicated directory;

[0118] (3) Check whether relevant parameters need to be modified;

[0119] (4) Execute the script, such as Figure 7 As shown in the following example, --input_mosaic and --input_file are required parameters, and the remaining parameters are optional. To post-process multiple pattern output data files, use the postp.csh script, $postp.csh$HOME / FGOALS / runs / BQ_job1 / run.

[0120] FGOALS-UFS Refined Forecasting Experiment: Using the FGOALS-UFS model to perform regional zoom and intensified forecasting, we first create a case. Then, we run FGOALS-UFS to generate forecast data. Finally, we convert the tile forecast data output by FGOALS-UFS into longitude and latitude grid forecast data for subsequent analysis.

[0121] FGOALS-UFS creates a prediction test:

[0122] (1) Create a new prediction test folder

[0123] Enter the scripts folder in the FGOALS-UFS main directory and copy the new_case folder to the new test case folder. The new_case folder is the prediction test run template. The folder path is:

[0124] $HOME / FGOALS_UFS / scripts

[0125] Taking the creation of a test case named BQ_job1 as an example, the command to copy the folder is:

[0126] $cp -rp new_case BQ_job1

[0127] (2) Related settings for the experiment

[0128] After creating the prediction test case, you need to set up the test case in detail. The steps are as follows:

[0129] First enter the BQ_job1 directory:

[0130] $cd$HOME / FGOALS-UFS / scripts / BQ_job1

[0131] Use the VIM editor to modify the task name in setup_run.sh in this directory to be consistent with the task name customized in the previous operation, which is BQ_job1 here:

[0132] $vi setup_run.sh

[0133] Change case_name to the name of the newly created case:

[0134] case_name=new_case->case_name=BQ_job1

[0135] Similarly, modify submit.sh accordingly:

[0136] $vi submit.sh

[0137] Change all new_case to the corresponding new case name:

[0138] new_case->BQ_job1

[0139] Run the setup_run.sh script:

[0140] $setup_run.sh

[0141] The script will be run in the directory

[0142] $HOME / FGOALS_UFS / runs

[0143] Create a new instance run directory

[0144] $HOME / FGOALS_UFS / runs / BQ_job1

[0145] Then, in the running directory $HOME / FGOALS_UFS / runs / BQ_job1, use the VIM editor to set the variables in the model_configure file, including the integration duration, integration start time, etc. The common settings are shown in Table 3:

[0146] Table 3

[0147] nhours_fcst Prediction of points duration (hours) start_day: Forecast start date: value range 1-31 start_month Forecast start month: value range 1-12 start_year Forecast start year: value range 0000-

[0148] Finally, the atmospheric component of FGOALS-UFS is configured separately. The configuration file is atm_in and the path is: $HOME / FGOALS_UFS / runs

[0149] Common configuration modifications are shown in Table 4:

[0150] Table 4

[0151]

[0152]

[0153] (3) Complete environment configuration

[0154] The environment variables of FGOALS-UFS are basically the same as those of FGOALS. The environment variables that need to be modified are stored in the env_jyun-hebian file. Before running FGOALS-UFS, users only need to execute this file to complete the environment configuration of the mode.

[0155] FGOALS-UFS prediction test run:

[0156] (1) Run directory: There are seven folders in the $$HOME / FGOALS-UFS directory, namely bld, inputs, libs, postp, runs, scripts, and tools. The bld folder is used to store environment variable configuration files and pattern executable files, the inputs folder is used to store pattern input data, the libs folder is the library that the pattern depends on, the runs folder stores newly created cases, the scripts folder contains templates for creating new cases, etc., and the postp and tools folders are used to store data processing tools.

[0157] All the input data required by FGOALS-UFS is located in the $HOME / FGOALS-UFS / inputs directory, which contains folders such as forcing, global, init, and redapp. The forcing folder is used to store the processed data output by FGOALS that is suitable for driving the core focus area of ​​FGOALS-UFS. The global folder stores the data used to drive the world. The init folder is used to store the initial conditions. The redapp folder is used to store the grid information of FGOALS-UFS. The redapp folder is different for different focus areas. In this project, the FGOALS-UFS grid information is suitable for Party A.

[0158] Submitting a job: After successfully compiling, the executable program of FGOALS-UFS is located in $HOME / FGOALS_UFS / runs / BQ_job1. In this directory, you can use mpirun to process jobs in parallel, or you can use submit.sh to submit jobs.

[0159] Post-processing of FGOALS-UFS prediction test results: The FGOALS-XAPP data post-processing script is fregrid_parallel, which is located in the directory $HOME / FGOALS_UFS / postp. Figure 8 As shown in the following example, --input_mosaic and --input_file are required parameters, and the remaining parameters are optional. To post-process multiple pattern output data files, use the postp.csh script.

[0160] Data post-processing is based on the NetCDF data format. The original output data of all ensemble members are used to generate the common data required for business forecasting and interpolated to 1 degree resolution.

[0161] Same resolution:

[0162] (1) Atmosphere: Horizontal resolution is 1°×1°, vertically there are 17 standard isobaric surfaces (1000, 925, 850, 700, 600, 500, 400, 300, 250, 200, 150, 100, 70, 50, 30, 20, 10 hPa)

[0163] (2) Ocean: Horizontal resolution is 1°×1°.

[0164] The interpolation method used is bilinear interpolation. The unknown value f(x, y) at the grid point P(x, y) can be obtained by using the known point Q 11 (x1,y1),Q 12 (x1,y2),Q 21 (x2,y1),Q 22 (x2,y2), the calculation formula is:

[0165]

[0166] The output data of the global sub-seasonal-seasonal refined forecast system focused on the Yellow River Basin includes multiple meteorological elements, including geopotential height, temperature, wind, specific humidity, surface temperature, and precipitation. The global resolution is 100 km, the regional focus is 12.5 km, the model has 32 vertical layers, the top layer is 1 hPa, and the temporal resolution and assimilation window is 6 hours. This system achieved a 12.5-sub-seasonal ensemble forecast for the Yellow River Basin for the first time. The output variable information is shown in Table 5:

[0167] Table 5

[0168]

[0169]

[0170]

[0171] From the above description, it can be seen that the independently developed climate model and RCP solution can calculate the refined focus area of ​​the Yellow River Basin at low cost and with ultra-high resolution.

[0172] 1. Independently developed climate model and RCP solution

[0173] FGOALS-f2 is a global climate model that provides software source code and has the potential for further improvement. Its independently developed Resolved Convective Precipitation (RCP) scheme makes cumulus convective precipitation explicit, reducing the errors caused by the averaging of convective effects and the strong reliance on convective parameter accuracy in traditional cumulus convective parameterization schemes. This has enhanced the influence of my country's climate model in the international academic community.

[0174] 2. Low-cost ultra-high-performance computing

[0175] The global sub-seasonal-seasonal refined forecast system focused on the Yellow River Basin utilizes FV3 variable grid downscaling technology, ensuring the stability and accuracy of forecast integrals with minimal computational resources. The regionally focused refined forecast system, FGOALS-UFS, far outperforms the WRF regional climate model. Using the same number of cores and hardware, it completed flood season forecast tests in just one-tenth the time of the WRF model.

[0176] 3. Yellow River Basin Refined Focus Area

[0177] The first sub-seasonal-seasonal grid prediction system with a global resolution of 100km and a focal area resolution of 12.5km, with the Yellow River Basin as the focal area, realizes refined forecasts of the Yellow River Basin and can provide refined forecasts of various future sub-seasonal-seasonal meteorological elements to meet the needs of agriculture, water resources management, disaster prevention and mitigation, and ecological and environmental protection in the Yellow River Basin.

[0178] Device Example 1

[0179] According to an embodiment of the present invention, a sub-seasonal to seasonal refined prediction system focusing on the Yellow River Basin is provided. Figure 9 Schematic diagram of a sub-seasonal to seasonal refined prediction system focusing on the Yellow River Basin in accordance with an embodiment of the present invention. Figure 9 As shown, the sub-seasonal-seasonal refined prediction system for the global focus on the Yellow River Basin according to an embodiment of the present invention specifically includes:

[0180] The FGOALS-f2 module 90 is used to process the original CRA40 data, GFS data, and OSTIA data into a data format that can be used for model input, drive the FGOALS-f2 model, and use the FGOALS-f2 model to perform numerical simulation using dynamic ensemble prediction technology to obtain numerical simulation results. The FGOALS-f2 module 90 is specifically used to:

[0181] Download the original CRA40 data, GFS data and OSTIA data to the server through the enhanced data download program;

[0182] Perform data quality control by checking data integrity and variable value rationality;

[0183] Process the original CRA40 data, GFS data and OSTIA data into NetCDF format;

[0184] Creating a prediction experiment includes case creation, case setting, and case compilation. Case creation creates a configuration folder for the simulation case in the corresponding directory; case setting is used to configure model components, external forcing files, model integration start and end times, and model output variables; case compilation generates an executable file cesm.exe in the compilation directory;

[0185] After successful compilation, an executable file of FGOALS-f2 is generated. After running the executable file, the FGOALS-f2 mode runs successfully;

[0186] The ensemble samples are generated by combining the lagged average forecast method and the time-lagged perturbation method;

[0187] Two FGOALS-f2 ensembles are obtained by combining the atmospheric initial values ​​at 00:00, 06:06, 12:02, and 18:00 and the ocean initial values ​​closest to the date using the LAF method. The time-lag perturbation method is used to increase the number of ensembles as needed, forming four ensemble prediction samples based on the start time of 00:00, 06:06, 12:02, and 18:00. Two ensemble sample initial values ​​are generated for each start time according to the length of the atmospheric forcing time, and the number of ensemble samples can be expanded to 8 as needed.

[0188] The FGOALS-f2 model is used to forecast variables for 65 days on a daily basis, with each day as the starting day. The forecast results on the starting day and the 20 days before it are included in the current ensemble members, and the next 65 ensemble forecasts based on the starting day are obtained in turn. The 40 ensemble members are averaged to obtain the syndicate mean forecast factor based on the starting day.

[0189] The FGOALS-UFS module 92 is used to input the numerical simulation results into the FGOALS-UFS model, use the FGOALS-UFS model to perform regional zoom encryption prediction, generate prediction data, and convert the prediction data output by the FGOALS-UFS model into longitude and latitude grid prediction data that is convenient for subsequent analysis; the FGOALS-UFS model specifically includes: adopting a unified forecast model UFS, wherein the dynamic framework of the UFS is the global grid FV3 divided by a cubic sphere, and the physical process of the UFS adopts the universal atmospheric physics package CCPP.

[0190] The post-processing module 94 is used to generate common data required for business forecasting using the latitude and longitude grid prediction data based on the NetCDF data format, and interpolate to the resolution of the predetermined reading, and output multiple meteorological elements including potential height field, temperature field, wind field, specific humidity, ground temperature and precipitation.

[0191] The embodiment of the present invention is an apparatus embodiment corresponding to the above-mentioned method embodiment. The specific operations of each module can be understood by referring to the description of the method embodiment, which will not be repeated here.

[0192] Device Example 2

[0193] An embodiment of the present invention provides an electronic device, such as Figure 10 As shown, it includes: a memory 100, a processor 102 and a computer program stored in the memory 100 and capable of running on the processor 102, and when the computer program is executed by the processor 102, the steps described in the method embodiment are implemented.

[0194] Device Example 3

[0195] An embodiment of the present invention provides a computer-readable storage medium, on which a program for implementing information transmission is stored. When the program is executed by the processor 102, the steps described in the method embodiment are implemented.

[0196] The computer-readable storage medium in this embodiment includes, but is not limited to, ROM, RAM, magnetic disk, or optical disk.

[0197] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention, rather than to limit it. Although the present invention has been described in detail with reference to the above embodiments, those skilled in the art should understand that they can still modify the technical solutions described in the above embodiments, or replace some or all of the technical features therein with equivalents. However, these modifications or replacements do not cause the essence of the corresponding technical solutions to deviate from the scope of the technical solutions of the embodiments of the present invention.

Claims

1. A sub-seasonal to seasonal refined forecasting method focusing on the Yellow River Basin, characterized by: include: The original CRA40 data, GFS data, and OSTIA data were processed into a data format that can be used for model input, driving the FGOALS-f2 model to run. The FGOALS-f2 model was used to perform numerical simulations using the dynamical ensemble prediction technology to obtain numerical simulation results. The numerical simulations using the FGOALS-f2 model using the dynamical ensemble prediction technology specifically include: The ensemble samples are generated by combining the lagged average forecast method and the time-lagged perturbation method; Two FGOALS-f2 ensembles are obtained by combining the atmospheric initial values ​​at 00:00, 06:06, 12:02, and 18:00 and the ocean initial values ​​closest to the date using the LAF method. The time-lag perturbation method is used to increase the number of ensembles as needed, forming four ensemble prediction samples based on the start time of 00:00, 06:06, 12:02, and 18:

00. Two ensemble sample initial values ​​are generated for each start time according to the length of the atmospheric forcing time, and the number of ensemble samples can be expanded to 8 as needed. The FGOALS-f2 model is used to forecast variables daily for 65 days, with each day being the starting day. The forecast results for the starting day and the 20 days before it are incorporated into the current ensemble members. The ensemble forecast for the next 65 days based on the starting day is then obtained. The ensemble average of the 40 ensemble members is then used to obtain the synoptic ensemble mean forecast factor based on the starting day. Inputting the numerical simulation results into the FGOALS-UFS model, using the FGOALS-UFS model to perform regional zoom encryption prediction to generate prediction data, and converting the prediction data output by the FGOALS-UFS model into latitude and longitude grid prediction data that is convenient for subsequent analysis; wherein, the FGOALS-UFS model specifically includes: adopting the unified forecast model UFS, wherein the dynamic framework of the UFS is the global grid FV3 of the cubic sphere subdivision, and the physical process of the UFS adopts the common atmospheric physics package CCPP; Using the NetCDF data format as the standard, the latitude and longitude grid prediction data is used to generate the common data required for business forecasting, and interpolated to the resolution of the predetermined reading, and the output includes multiple meteorological elements such as potential height field, temperature field, wind field, specific humidity, ground temperature and precipitation.

2. The method according to claim 1, characterized in that The original CRA40 data, GFS data, and OSTIA data are processed into a data format that can be used for model input. The specific steps to drive the FGOALS-f2 model include: Download the original CRA40 data, GFS data and OSTIA data to the server through the enhanced data download program; Perform data quality control by checking data integrity and variable value rationality; Process the original CRA40 data, GFS data and OSTIA data into NetCDF format; Creating a prediction experiment includes case creation, case setting, and case compilation. Case creation creates a configuration folder for the simulation case in the corresponding directory; case setting is used to configure model components, external forcing files, model integration start and end times, and model output variables; case compilation generates an executable file cesm.exe in the compilation directory; After successful compilation, an executable file of FGOALS-f2 is generated. After running the executable file, the FGOALS-f2 mode runs successfully.

3. The method according to claim 1, characterized in that The global resolution of the multiple meteorological elements is 100 kilometers, the regional focus area is 12.5 kilometers, there are 32 vertical layers, the top layer of the model is 1hPa, and the time resolution and assimilation time window are 6 hours.

4. A global sub-seasonal-seasonal refined forecast system focusing on the Yellow River Basin, characterized by: include: The FGOALS-f2 module is used to process the original CRA40 data, GFS data, and OSTIA data into a data format that can be used for model input, drive the FGOALS-f2 model, and use the FGOALS-f2 model to perform numerical simulations using dynamic ensemble prediction technology to obtain numerical simulation results. Specifically, it is used to generate ensemble samples using a combination of the lagged average prediction method and the time-lagged perturbation method. Two FGOALS-f2 ensembles are obtained by combining the atmospheric initial values ​​at 00:00, 06:06, 12:02, and 18:00 and the ocean initial values ​​closest to the date using the LAF method. The time-lag perturbation method is used to increase the number of ensembles as needed, forming four ensemble prediction samples based on the start time of 00:00, 06:06, 12:02, and 18:

00. Two ensemble sample initial values ​​are generated for each start time according to the length of the atmospheric forcing time, and the number of ensemble samples can be expanded to 8 as needed. The FGOALS-f2 model is used to forecast variables daily for 65 days, with each day being the starting day. The forecast results for the starting day and the 20 days before it are incorporated into the current ensemble members. The ensemble forecast for the next 65 days based on the starting day is then obtained. The ensemble average of the 40 ensemble members is then used to obtain the synoptic ensemble mean forecast factor based on the starting day. The FGOALS-UFS module is used to input the numerical simulation results into the FGOALS-UFS model, use the FGOALS-UFS model to perform regional zoom encryption prediction, generate prediction data, and convert the prediction data output by the FGOALS-UFS model into longitude and latitude grid prediction data that is convenient for subsequent analysis; The FGOALS-UFS model specifically includes: adopting the unified forecast model UFS, wherein the dynamic framework of the UFS is the global grid FV3 of the cubic sphere division, and the physical process of the UFS adopts the common atmospheric physics package CCPP; The post-processing module is used to generate common data required for business forecasting using the latitude and longitude grid prediction data based on the NetCDF data format, and interpolate to the resolution of the predetermined reading, and output multiple meteorological elements including potential height field, temperature field, wind field, specific humidity, ground temperature and precipitation.

5. The system according to claim 4, characterized in that The FGOALS-f2 module is specifically used for: Download the original CRA40 data, GFS data and OSTIA data to the server through the enhanced data download program; Perform data quality control by checking data integrity and variable value rationality; Process the original CRA40 data, GFS data and OSTIA data into NetCDF format; Create a prediction experiment, including case creation, case setting, and case compilation. Case creation creates a configuration folder for the simulation case in the corresponding directory; case setting is used to configure model components, external forcing file configuration, model integration start and end time, and model output variables; The individual compilation will generate the executable file cesm.exe in the compilation directory; After successful compilation, an executable file of FGOALS-f2 is generated. After running the executable file, the FGOALS-f2 mode runs successfully.

6. An electronic device, characterized in that: include: A memory, a processor, and a computer program stored in the memory and executable on the processor, wherein when the computer program is executed by the processor, the steps of the sub-seasonal-seasonal refined prediction method for the global focus on the Yellow River Basin are implemented as described in any one of claims 1 to 3.

7. A computer-readable storage medium, characterized in that The computer-readable storage medium stores an implementation program for information transmission, and when the program is executed by a processor, the steps of the sub-seasonal-seasonal refined prediction method for the global focus on the Yellow River Basin are implemented as described in any one of claims 1 to 3.

Images