Meteorological ocean live simulation method and system based on satellite data

Abstract

The invention relates to the technical field of meteorological simulation, in particular to a meteorological ocean live simulation method and system based on satellite data. The method comprises the following steps: acquiring satellite multi-element connection data through a data connection system; constructing a numerical weather forecast product correction model, and correcting satellite multi-element connection data by using the numerical weather forecast product correction model; according to the corrected satellite multi-element connection data, utilizing a simulation algorithm to obtain a grid point meteorological element simulation result; based on the grid point meteorological element simulation result, a site meteorological element simulation result is obtained through an intelligent algorithm; and evaluating the grid point meteorological element simulation result and the site meteorological element simulation result to obtain a qualified meteorological ocean live simulation result. According to the method, grid point simulation is carried out after data correction, the site simulation result is obtained through an intelligent algorithm, the simulation result is evaluated, a qualified meteorological element simulation result is obtained, and the problem of obtaining the real condition of regions where meteorological ocean real conditions such as oceans and islands are difficult to obtain is solved.

Description

Technical Field

[0001] The present invention relates to the technical field of meteorological simulation, and particularly to a method and system for meteorological and oceanic real-time simulation based on satellite data. Background Art

[0002] With the progress of society, countries around the world have paid increasing attention to the ocean and ocean resources. Maritime transportation and economic production activities have been increasing. Against the backdrop of global warming, extreme weather events have been occurring continuously, and marine meteorological disasters have become more frequent and intense, posing a serious threat to the development of the marine economy. This urgently requires continuously enhancing the meteorological and oceanic simulation and early warning capabilities, and at the same time providing more high-quality and efficient meteorological guarantee services.

[0003] Currently, the technical equipment and capabilities of marine meteorological monitoring are insufficient. Conventional ocean surface wind data mostly come from traditional observation means such as buoy stations, islands, oil platforms, and ships, and are mostly distributed near the coastline. The observation in the central part of the ocean is basically in a vacuum state, making it difficult to meet the scientific research and business needs. At the same time, there are many blind spots in marine meteorological observations, the means of collecting, collating, and integrating and analyzing marine meteorological data are outdated, the capabilities of professional numerical models for marine meteorology are relatively weak, and the evaluation of numerical model simulations for marine meteorology is insufficient. Therefore, how to solve the problem of obtaining real-time conditions in areas where it is difficult to obtain meteorological and oceanic real-time conditions such as the ocean and isolated islands is an urgent problem to be solved currently. Summary of the Invention

[0004] In view of the deficiencies of existing methods and the requirements of practical applications, in order to improve the accuracy of meteorological and oceanic simulations and solve the problem of obtaining real-time conditions in areas where it is difficult to obtain meteorological and oceanic real-time conditions such as the ocean and isolated islands. On the one hand, the present invention provides a method for meteorological and oceanic real-time simulation based on satellite data, including the following steps: Obtain satellite multi-source connection data through a data connection system; construct a numerical weather prediction product correction model, and use the numerical weather prediction product correction model to correct the satellite multi-source connection data; according to the corrected satellite multi-source connection data, obtain grid point meteorological element simulation results using a simulation algorithm; based on the grid point meteorological element simulation results, obtain station meteorological element simulation results through an intelligent algorithm; evaluate the grid point meteorological element simulation results and the station meteorological element simulation results to obtain qualified meteorological and oceanic real-time simulation results. The present invention performs grid point simulation after correcting the data, then obtains station simulation results through an intelligent algorithm, and evaluates the simulation results to obtain qualified meteorological element simulation results, and solves the problem of obtaining real-time conditions in areas where it is difficult to obtain meteorological and oceanic real-time conditions such as the ocean and isolated islands through real-time simulation.

[0005] Optionally, after obtaining the satellite multi-source connection data through the data connection system, the satellite multi-source connection data is also preprocessed; the preprocessing includes data deduplication, data selection, data filtering, abnormal file elimination, and standardization processing. By preprocessing the satellite multi-source connection data, the present invention effectively improves the data quality, which is further beneficial to improving the simulation effect of the present invention.

[0006] Optionally, the numerical weather prediction product correction model includes a meteorological element feature extraction coding structure and a meteorological element feature restoration decoding structure; the meteorological element feature restoration decoding structure includes a sub-pixel convolution sub-module and a replication splicing sub-module. The sub-pixel convolution sub-module is used for expanding the meteorological element feature map, and the replication splicing sub-module is used for channel splicing of the feature maps of the meteorological element feature extraction coding structure and the meteorological element feature restoration decoding structure. By using the coding and decoding structures, the present invention improves the operation efficiency while reducing the loss of effective information during the image expansion process, which is beneficial to improving the efficiency of the present invention.

[0007] Optionally, correcting the satellite multi-source connection data by using the numerical weather prediction product correction model includes the following steps: Constructing a training and validation data set by using historical satellite multi-source connection data; optimizing the parameters of the numerical weather prediction product correction model, training and validating the numerical weather prediction product correction model with optimized parameters through the training and validation data set; correcting the satellite multi-source connection data according to the trained numerical weather prediction product correction model. By sorting out the historical data to construct a data set and then optimizing the parameters of the correction model, the present invention is beneficial to effectively revising the data, and further improves the simulation accuracy of the present invention.

[0008] Optionally, obtaining the grid point meteorological element simulation result by using the simulation algorithm according to the corrected satellite multi-source connection data includes the following steps: Constructing a grid point simulation model by using the simulation algorithm; performing initialization processing on the grid point simulation model; performing data assimilation processing based on the corrected satellite multi-source connection data; obtaining the grid point meteorological element simulation result through the data after data assimilation processing and the initialized grid point simulation model. By initializing the simulation model and then performing data assimilation processing according to the actual situation, the present invention is beneficial to eliminating the noise problem caused by the imbalance between the wind field and the mass field during cyclic assimilation.

[0009] Optionally, the data assimilation processing based on the corrected satellite multi-source connection data includes occultation data assimilation; The occultation data assimilation satisfies the following formula: Where, represents the refractive index value of the neutral atmosphere, represents the empirical coefficient related to the dry air pressure, represents the partial pressure of dry air, represents the absolute temperature, represents the water vapor pressure, represents the empirical coefficient of the temperature square term related to the water vapor pressure, represents the empirical coefficient of the temperature first term related to the water vapor pressure. The present invention inserts the refractive index value calculated for the background field geopotential height layer into the required observation height, which is conducive to directly comparing with the observed quantity, thereby improving the data assimilation processing efficiency.

[0010] Optionally, obtaining the meteorological element simulation result through the data after data assimilation processing and the grid simulation model after initialization processing includes the following steps: Interpolate the model layer to 45 geometric height layers; use diagnostic analysis technology to generate analysis elements such as visibility, cloud amount, lightning, and cloud base height. By interpolating the model layer to more geometric height layers, the present invention can significantly improve the atmospheric vertical resolution. At the same time, interpolating to more geometric height layers can enhance the prediction ability of the numerical model. The unconventional meteorological elements obtained by using diagnostic analysis technology are conducive to further improving the simulation elements of the present invention.

[0011] Optionally, obtaining the station meteorological element simulation result through the intelligent algorithm based on the grid meteorological element simulation result includes the following steps: Based on the grid meteorological element simulation result, obtain the station meteorological element; construct a station meteorological element correction model; use the historical measured station meteorological elements to establish a station meteorological element correction data set, and train and verify the station meteorological element correction model through the station meteorological element correction data set; according to the trained station meteorological element correction model and the station meteorological element, obtain the station meteorological element simulation result. The present invention obtains the station simulation result from the grid simulation result and revises it based on historical monitoring data, which is conducive to improving the accuracy of the station simulation result of the present invention.

[0012] Optionally, evaluating the grid meteorological element simulation result and the station meteorological element simulation result to obtain a qualified meteorological and oceanic actual situation simulation result includes: Conduct integrity check, correlation test, change range check, climatological boundary value check, internal consistency check, time consistency check, and test for lack of historical actual situation stations on the grid meteorological element simulation result and the station meteorological element simulation result; according to the inspection results, obtain a qualified meteorological and oceanic actual situation simulation result.. By evaluating the simulation result, the present invention further ensures the accuracy of the simulation result of the present invention.

[0013] Second aspect, to efficiently execute a meteorological and oceanic real - time simulation method based on satellite data provided by the present invention, the present invention further provides a meteorological and oceanic real - time simulation system based on satellite data, including a processor, an input device, an output device, and a memory. The processor, the input device, the output device, and the memory are interconnected. Among them, the memory is used to store a computer program, the computer program contains program instructions, and the processor is configured to call the program instructions to execute a meteorological and oceanic real - time simulation method as described in the first aspect of the present invention. The meteorological and oceanic real - time simulation system of the present invention has a compact structure and stable performance, and can stably execute the meteorological and oceanic real - time simulation method provided by the present invention, further enhancing the overall applicability and practical application ability of the present invention. Brief Description of the Drawings

[0014] Figure 1 It is a flowchart of a meteorological and oceanic real - time simulation method based on satellite data provided by an embodiment of the present invention; Figure 2 It is a framework diagram of a meteorological and oceanic real - time simulation system based on satellite data provided by an embodiment of the present invention. Detailed Embodiments

[0015] The specific embodiments of the present invention will be described in detail below. It should be noted that the embodiments described here are only for illustrative purposes and are not used to limit the present invention. In the following description, a large number of specific details are set forth in order to provide a thorough understanding of the present invention. However, it is obvious to those of ordinary skill in the art that the present invention does not have to be practiced with these specific details. In other instances, well - known circuits, software, or methods have not been specifically described in order to avoid obscuring the present invention.

[0016] Throughout the specification, the reference to "an embodiment", "embodiments", "an example", or "examples" means that a particular feature, structure, or characteristic described in connection with the embodiment or example is included in at least one embodiment of the present invention. Thus, the phrases "in an embodiment", "in embodiments", "an example", or "examples" appearing throughout the specification do not necessarily all refer to the same embodiment or example. In addition, the specific features, structures, or characteristics may be combined in any suitable combination and / or sub - combination in one or more embodiments or examples. Moreover, those of ordinary skill in the art should understand that the drawings provided herein are for illustrative purposes only and are not necessarily drawn to scale.

[0017] Please refer to Figure 1, in order to improve the accuracy of meteorological and oceanographic simulations and solve the problem of obtaining real-time conditions in difficult-to-reach areas such as the ocean and isolated islands. The present invention provides a method for simulating meteorological and oceanographic real-time conditions based on satellite data, as Figure 1 shown. In one embodiment, the method includes the following steps: S1. Obtain satellite multi-source connection data through a data connection system.

[0018] The data connection system is used to extract structured data, which generally comes from the aggregation of many sub-data. The way of referring to and aggregating sub-data for structured data is called connection. It realizes data collection, conversion, loading, and storage through specific technologies and methods to meet different application and business requirements. Numerical weather prediction is based on the basic laws of fluid mechanics and thermodynamics in physics. By applying these laws to the changes in the atmosphere, describing these changes with mathematical formulas, and then using a computer to solve these mathematical formulas to predict future weather conditions. In this process, the data connection system is responsible for extracting, integrating, and managing the required meteorological data from various data sources to provide accurate and timely data support for numerical weather prediction.

[0019] In an embodiment, according to the data collection strategy configurations such as the type of collected data, collection method, and collection time, various types of data and products such as satellite remote sensing and aircraft reports are regularly obtained from the data connection system through methods such as raw copying, virtual sharing, FTP acquisition, and data acquisition API interface calls.

[0020] Specifically, according to the collection strategies such as time, frequency, and method for various types of observation and detection data (meteorological conventional data, oceanographic conventional data, satellite remote sensing detection data, reanalysis data, and numerical prediction data), corresponding observation and detection meteorological data collection and processing programs are developed for the formats of various types of original observation and detection meteorological data, and data collection and warehousing are completed through the data collection and processing programs.

[0021] It should be understood that the real-time collection program needs to be specially customized according to the characteristics of each type of data source. For each type of data collection, an independent set of data parsing and processing programs needs to be customized, and it is necessary to ensure uninterrupted operation for 7×24 hours. When the data parsing and processing program for a certain type stops, it does not affect other data parsing programs.

[0022] Furthermore, the data parsing program includes: (1) Message identification Mainly realizes message identification for various types of data and product files such as conventional observation and detection data, oceanographic hydrology, satellite data, reanalysis data, and numerical prediction data.

[0023] File name pattern matching and recognition: According to the predefined naming rules for various types of meteorological data reports and product materials, perform file name pattern matching and recognition on the received meteorological data from the perspective of the file name naming rules. If the naming rule completely conforms to the naming format of a certain type of data, it proceeds to the next process.

[0024] File type recognition: File type recognition refers to performing file type recognition on the meteorological data set identified through file name pattern matching.

[0025] (2) Data decoding Decoding mainly realizes the original message decoding of conventional observation data, ocean hydrology, satellite data, reanalysis data, numerical forecast data, etc. using programs.

[0026] Decoding of meteorological standardized data sets: Use programs to decode messages such as surface, radiosonde, aircraft reports, etc.

[0027] HDF format decoding: Decode the data product materials stored in the database, convert them into the required data format and the file format that can be retrieved and processed by the display platform, and improve the retrieval and display speed. This data format is different from the binary files and ASCII files used in conventional data. It provides an overall directory structure and uses a binary tree method to establish an "index" of the file content, enabling quick access to information directly from nested files through the "index", and different types of data sources can be stored in one file, and at the same time, these data sources can contain their data information and other relevant information.

[0028] NETCDF format decoding: Decode real-time or numerical forecast and reanalysis products in the NETCDF format. Decode, splice fields, and encode the files to generate the format data files required by this system.

[0029] GRIB format decoding: Read and extract information according to time, spatial longitude and latitude range, element types, forecast lead time, etc. Decode, splice fields, and encode GRIB format (GRIB1, GRIB2) files to generate the required format data files.

[0030] Further, after obtaining satellite multi-source received data through the data acquisition system, preprocess the satellite multi-source received data.

[0031] The preprocessing includes data deduplication, data selection, data filtering, abnormal file elimination, and standardization processing.

[0032] Specifically, data deduplication includes data deduplication for the following five types of duplicate data: 1) The records are exactly the same, but there are deviations in the observation time and geographical location; 2) The observation time and geographical location are the same, but the records are different; 3) The position and observation time are basically consistent, but the recorded data show large deviations; 4) The observation location and record are exactly the same, but the observation time crosses the zero point; 5) The observation time and location are exactly the same, but the length of the significant digits retained in the decimal places of the data records is inconsistent.

[0033] Data selection means that based on the collected conventional meteorological data, conventional ocean data, satellite remote sensing data and other data, and based on the data fusion analysis business needs, the system supports the automatic selection of various types of data; it can also manually identify and select various types of data according to user needs; and perform format checks on the selected data. The content of the format check includes whether the element types are fully interpreted, whether the location, length, precision and other attributes of the elements are correctly interpreted, whether there are abnormal characters, whether the file is empty, etc., and eliminates various unusable data that cannot meet the format requirements.

[0034] Data filtering means that according to the collected conventional meteorological data, conventional ocean data, and satellite remote sensing data, based on the business needs of data fusion analysis, the system supports automatic filtering of various types of data; it can also manually identify and filter various types of data according to user needs.

[0035] Abnormal file elimination refers to the function of judging and eliminating abnormalities based on the collected conventional meteorological data, conventional marine data, and satellite remote sensing detection data. When obtaining ground, high-altitude, meteorological and marine observation data, and meteorological satellite data, the files are stored in the local server through the transmission interface. In daily data observation, it is often the case that all or part of the data is lost or data anomalies occur due to equipment hardware and software failures or misoperation, causing certain losses to the accumulation of data. The causes of data anomalies include electromagnetic fields or pulse signal interference around the observation station, strong weather influences (such as heavy precipitation, etc.), etc. When the above reasons occur, the data in the collector often cannot be written, resulting in no data or data errors in the data file. At this time, such files will be eliminated as abnormal files, and such files will be restored by analyzing the reasons.

[0036] Standardization processing means that data normalization processing will standardize the collected conventional meteorological data, conventional ocean data, satellite data, reanalysis data and numerical forecast products according to metadata standards, including naming specifications, format conversion, normalization processing, etc., to generate standardized data that meets the requirements of the data assimilation subsystem.

[0037] Data standardization processing is carried out for the pre-processed conventional meteorological data, conventional ocean data, satellite data, reanalysis data and numerical forecast products. According to the relevant historical data compilation procedures, standardization processing, unified quality control and correction are carried out, and the coordinate system, time zone and measurement standards are normalized.

[0038] S2. Construct a numerical weather forecast product correction model, and use the numerical weather forecast product correction model to correct the satellite multivariate access data.

[0039] In an embodiment, the numerical weather forecast product correction model is constructed according to a deep learning model. In this embodiment, the deep learning model is a deep learning segmentation network. Further, the numerical weather forecast product correction model includes a meteorological element feature extraction coding structure and a meteorological element feature restoration decoding structure; The meteorological element feature restoration and decoding structure includes a sub-pixel convolution submodule and a copy and splicing submodule. The sub-pixel convolution submodule is used to expand the meteorological element feature map, and the copy and splicing submodule is used to channel-splice the feature map of the meteorological element feature extraction coding structure and the feature map of the meteorological element feature restoration and decoding structure.

[0040] Specifically, the encoding stage mainly realizes image dimension reduction and feature map extraction, and the core processing is the convolution layer and the pooling layer. The convolution layer uses two-dimensional convolution to extract features. If the input size is 2×48×48, the image size becomes 64×48×48 after the convolution kernel of 64×3×3 and the border filling is completed.

[0041] The role of the pooling layer in the network is to reduce the dimension of features, reduce parameters on the basis of retaining the main features of meteorological elements, and enhance the generalization ability of the model. For example, the size of the input model forecast meteorological element feature map is 64×48×48, and after pooling, the size is 64×24×24.

[0042] In the encoding stage, by stacking convolution and pooling operations, the input meteorological element grid map (2×48×48) can eventually be encoded into a meteorological element feature map (512×3×3).

[0043] The decoding stage mainly realizes image feature restoration, and the sub-pixel convolution sub-module is used to expand the feature map of meteorological elements. Compared with the deconvolution in the original deep learning segmentation network, it can improve the computational efficiency while reducing the loss of effective information in the image expansion process.

[0044] The copy and splicing submodule performs channel splicing on the feature maps of the meteorological element feature extraction coding structure and the meteorological element feature restoration decoding structure. For example, the size of the new feature map obtained by splicing the feature map of the encoder 512×6×6 and the feature map of the decoder 512×6×6 is 1024×6×6.

[0045] After decoding, the encoded feature map (512×3×3) can be decoded into the final output image (1×48×48).

[0046] Finally, the output image is restored to the output of meteorological element data to obtain the numerical weather prediction correction result.

[0047] Furthermore, the size of the convolution kernel depends on the feature distribution and discrimination. If the features to be extracted are small, a smaller convolution kernel should be selected. Selecting a larger convolution kernel will result in the loss of some local features. The size of the convolution kernel selected is determined according to the size of the training data, and the loss function is the mean square error (MSE).

[0048] In the formula, n represents the number of training samples, represents the observed values of the training set, represents the prediction correction results in the training set.

[0049] Furthermore, the correcting of the satellite multi-source data using the numerical weather prediction product correction model includes the following steps: Construct a training and validation dataset using historical satellite multi-source data; Optimize the parameters of the numerical weather prediction product correction model, and train and validate the numerical weather prediction product correction model with optimized parameters through the training and validation dataset; Correct the satellite multi-source data according to the trained numerical weather prediction product correction model.

[0050] Specifically, the labeled data is divided into a training set and a validation set, and the adaptive moment estimation optimization algorithm is used to optimize the parameters in the numerical weather prediction product correction model.

[0051] The adaptive moment estimation optimization algorithm means that by calculating the first-order moment estimation and second-order moment estimation of the gradient, independent adaptive learning rates are designed for different parameters, which can converge quickly and learn more efficiently. To further improve the calculation efficiency and the robustness of the model, batch training is performed on the network and the batch size is set to 32. The initial learning rate of the network is set to 0.001, and the learning rate decay factor is set to 0.5; if the scoring index of the validation set does not decrease for two consecutive times, the learning rate is decayed to 0.5×r; the total number of training iterations is 30 times, and the validation set score is calculated once for each iteration; finally, the model with the highest validation set score is selected and transferred to the test set for correction and scoring.

[0052] In the embodiments, based on the original data of the T799 numerical weather prediction product as the input information, the trained numerical weather prediction product correction model is applied to correct the corresponding prediction results and output the corrected prediction product for the corresponding prediction time. The corrected elements include the geopotential height, temperature, relative humidity, meridional wind U, and zonal wind V on the three-dimensional isobaric surface layer; the two-dimensional elements include surface pressure, sea-level pressure, surface temperature, 2-meter temperature, 2-meter relative humidity, 10-meter meridional wind U, and 10-meter zonal wind V.

[0053] The T799 numerical weather prediction product correction subsystem runs twice a day after the generation of the T799 numerical weather prediction product.

[0054] In the embodiments, under the condition of less information, due to the inability to obtain various types of observational data and the reduction of the assimilated data volume, the longitude of the T799 numerical weather prediction decreases, which in turn leads to the reduction of the high-resolution regional analysis and prediction capabilities with it as the background field. Using the T799 numerical weather prediction products for 1 year and the ERA5 reanalysis products for the corresponding time periods, based on artificial intelligence (deep learning) technology, a correction model for the T799 numerical weather prediction products is constructed; using this model, the rolling correction of the three-dimensional situation field and the two-dimensional surface element field of the T799 numerical weather prediction used to drive the WRF model in the grid point meteorological element simulation subsystem within the range of 65°E to 155°E in longitude and 15°S to 55°N in latitude for the next 0 - 24 hours at 3-hour intervals is realized, providing a background field with a relatively high accuracy for the grid point meteorological element simulation.

[0055] S3. According to the corrected satellite multi-source connection data, the grid point meteorological element simulation results are obtained using the simulation algorithm.

[0056] Specifically, the obtaining of the grid point meteorological element simulation results using the simulation algorithm according to the corrected satellite multi-source connection data includes the following steps: S31. Using the simulation algorithm, a grid point simulation model is constructed.

[0057] In the embodiments, a grid point simulation model is constructed according to the mesoscale meteorological model WRF and its three-dimensional variational assimilation system.

[0058] It should be understood that the observational data and assimilable data that can be obtained in different regions are different. Therefore, it is also necessary to first perform regional division according to the actual situation.

[0059] In the embodiments, the settings of the grid point simulation area are as follows: Regional range: The longitude is between 70°E and 150°E, and the latitude is between 10°S and 50°N; Coordinates of the regional center point: (110°E, 20°N); Horizontal resolution: 9 Km; Horizontal grid points: 900 grid points in the meridional direction and 700 grid points in the zonal direction, totaling 900×700 grid points; Vertical structure: 45 layers in total, and the top of the model layer is located at 30 hPa; Update frequency: Update once every hour; Types of assimilated data: Surface, radiosonde, aircraft reports and other observational data within the regional scope, FY-4B cloud wind data and multi-channel scanning imaging radiometer ARGI data, FY-3D microwave humidity sounding data, FY-3E microwave temperature sounding data and microwave humidity sounding data, OceanSat-2 scatterometer sea surface wind field data and buoy station data. Among them, for surface, radiosonde, aircraft reports, and FY-4B, data within half an hour before and after the selected assimilation time point are assimilated; since FY-3D / 3E and OceanSat-2 are polar-orbiting satellites and cross the area twice a day, when assimilating operational data, the data is selected within a half-hour window before and after the analysis time. If there is no such polar-orbiting satellite data within the assimilation window, the assimilation analysis of the above polar-orbiting satellite data will not be carried out.

[0060] S32. Initialize the grid point simulation model.

[0061] For the monthly average vegetation cover data in the multi-year climatology of the WRF model, update it based on the latest remote sensing observation data to form vegetation cover data with higher resolution; at the same time, based on the latest original data of soil distribution and soil profiles in China, optimize the update of soil types to improve the prediction accuracy of numerical weather prediction.

[0062] Based on satellite remote sensing observation data in the past decade, establish a static dataset for updated soil, vegetation, etc. that conforms to the actual development situation, and update the vegetation cover data, soil types, and the new soil hydraulics parameter table of NOAH.

[0063] Filter initialization is to eliminate the noise caused by the imbalance between the wind field and the mass field. This kind of noise is likely to cause calculation instability during the integration process, reduce the prediction level, and prevent assimilation from being carried out due to background field noise during the data assimilation cycle; initialization can construct a continuous field at the initial moment for some variables that are not analyzed at the initial moment (such as cloud water content, etc.).

[0064] Using the analysis initial value and the prediction model, integrate forward and backward respectively from the given analysis moment to form a set of time series. Process this time series with a digital filter to filter out short-period oscillations and obtain the adjusted initial field.

[0065] According to different ways of introducing non-adiabatic processes during the model integration process, the implementation methods of digital filtering can be divided into three types. They include the ADFI method, the DDFI method, and the TDFI method.

[0066] The ADFI method analyzes the data by integrating it forward and backward for T hours from the moment t = 0, obtaining a time series centered on the moment t = 0. Both the forward and backward integrations of the model are adiabatic processes; the DDFI method is similar to the ADFI method, but the forward integration of the model uses a non-adiabatic process; the TDFI method is that the model first integrates adiabatically backward from the moment t = 0 to -T, and then integrates forward from -T to T using a non-adiabatic process, but only the time series obtained from the non-adiabatic process within the time period [-T, T] is used for filtering.

[0067] S33. Perform data assimilation processing based on the corrected satellite multi-source retrieved data.

[0068] Utilize the WRFDA three-dimensional data assimilation technology to achieve the assimilation of various conventional observation data, autonomous meteorological and oceanographic satellites, and occultation data.

[0069] Occultation atmospheric sounding can detect and obtain the vertical profiles of parameters such as temperature, humidity, and pressure in the middle and low latitudes of the global atmosphere, the vertical profile of ionospheric electron density, and the horizontal distribution of peak electron density, enriching the observational data that can be assimilated in the numerical prediction system and having important potential for improving the simulation accuracy of grid three-dimensional meteorological elements.

[0070] The error of occultation data has a direct impact on the assimilation effect. Therefore, effective quality control is required first. Quality control needs to find out the outlier data that differ greatly from the data average. The outlier data is usually determined by calculating the standard deviation from the average of the data sample. However, the outlier data itself has a great impact on the average and standard deviation of the data, thus affecting the effective identification of the outlier data. To reduce this possibility, a reasonable scheme will be designed to effectively identify the outlier data, perform quality control on the occultation data, and improve the rationality of the assimilation input data; The quality control of occultation data is completed by the following 4 steps: (a) Range check Since the ionospheric error is opposite to that of the upper troposphere, a range check is first performed, that is, the data of each vertical profile is checked as follows. If or , then this data is excluded.

[0071] (b) Data consistency test Calculate the Z-score of each data: In the formula represents the bending angle i or the refractive index , the subscript "i" represents the i-th observational data, is the biweight average, It is the double-weighted mean square deviation. The standard of Z-score greater than 3, 4, and 5 is used to determine the abnormal data (i.e., outlier data). After comparing the amount of suspicious abnormal data and the corresponding observed values with the model simulation values calculated from the large-scale analysis field temperature and water vapor, it is found that. This step of quality control is performed on the data itself without inputting any model simulation values.

[0072] (c) Compatibility test with the background field Calculate the Z-score of the following variables: , The data with Z-score greater than 4 is regarded as suspicious abnormal data.

[0073] (d) Symmetry test There may be negative systematic errors in the occultation observation data in the lower troposphere. Therefore, the symmetry test of the data error is carried out for the data passing through the above three steps. The symmetry test is only carried out for the data below 4 km. Denote the set , According to the set and , construct the following two data sets: , Then the data set and the data set are both symmetric. Calculate the double-weighted standard variances and of these two data sets respectively, and take: as the standard variance of the original data set. Finally, calculate the Z-score of each data. The data with Z-score greater than 3 is defined as the possible error data.

[0074] Since the vertical resolution of the occultation data is relatively high (more than 1000 layers in the vertical direction), how to effectively sparsify and reduce the computational amount of assimilation is also a problem that needs to be studied. In this embodiment, the height logarithmic linear interpolation method is used to interpolate the occultation data to the model layer.

[0075] The direct assimilation method is used to assimilate the occultation sounding data. Constructing a reasonable observation operator is an important part of direct assimilation. In this embodiment, the direct assimilation of the autonomous occultation data is realized by constructing a refractive index observation operator.

[0076] The amount of bending of the radio ray through the atmosphere along its path depends on the refractive index n of the air along the path. For convenience, the refractive index N is used to replace the refractive index n, and their relationship is as follows: For radio occultation sounding, the scattering term can be ignored, and the ionospheric effect can usually be removed during the preprocessing, such as through ionospheric bending angle correction. The forward model only deals with the refractive index of the neutral atmosphere.

[0077] Furthermore, the occultation data assimilation satisfies the following formula: where, represents the refractive index value of the neutral atmosphere, represents the empirical coefficient related to the dry air pressure, represents the partial pressure of dry air, represents the absolute temperature, represents the water vapor pressure, represents the empirical coefficient of the temperature square term related to the water vapor pressure, represents the empirical coefficient of the temperature first term related to the water vapor pressure.

[0078] It can be understood that in each background field layer, the water vapor pressure e comes from the specific humidity q: r is the dry mixing ratio of water vapor, defined as: The water vapor pressure e, the background field specific humidity q, and the pressure p: .

[0079] Furthermore, the vertical sounding data assimilation of Fengyun-3D / 3E satellites means that the microwave temperature sounding data obtained by MWTS-II of Fengyun-3D / 3E satellites and the microwave humidity sounding data obtained by MWHS-II can conduct refined sounding of the atmospheric temperature and humidity states. Carrying out the construction of the application of the vertical sounding data assimilation of Fengyun 3D / 3E satellites has important potential for improving the simulation ability of regional refined grid meteorological elements.

[0080] In satellite observations, there are many factors that can lead to large errors, such as weather conditions (clear sky, cloudy or full-day coverage), ground conditions (sea surface, land, sea ice, etc.), geographical location (e.g., mid-latitudes or tropics), observation geometry (sub-satellite point or limb measurement), response characteristics and accuracy of the sensor during orbital operation, errors in radiation transfer models and background fields, etc. To ensure the consistency between the quality of surrounding data and analysis results, quality control must be carried out first. This includes extreme value detection, limb detection, cloud detection, etc.

[0081] Extreme value detection means that for the microwave temperature (MWTS II) detection radiation value between 50K and 350K, and the microwave humidity (MWHSII) detection radiation value between 80 and 340K, those exceeding the range will be excluded.

[0082] Limb detection means that during the scanning and detection process of the MWTS II and MWHS II detectors, for the detection points at the edge of the scanning line, due to the severe tilt of the detection angle, the atmospheric radiation path passed by the scanning point to the detector channel is longer than that from the sub-satellite point to the detector channel, resulting in a decrease in the observed radiation amount and generating a limb effect. Some scanning edge points can be removed according to the scanning line length and the number of scanning points to reduce the limb effect.

[0083] Cloud detection means that although microwave detectors can penetrate clouds to detect the temperature and humidity of the atmosphere, the water droplets and ice crystals in precipitation clouds are larger than the radiation wavelength, so the scattering result will weaken the signal between clouds. This will affect the detection, so precipitation clouds must be detected. According to the probability (%) statistically calculated from ATOVS 1D data, in the current assimilation, the following are usually used: Where and are the observed brightness temperatures of channel 1 and channel 15 respectively. When the radiation brightness temperature must be excluded.

[0084] Based on the radiation transfer model RTTOV, using a high-precision spectral line model, introducing the spectral response function of the FY-3D / 3E microwave channels, radiation transfer simulations are carried out for the typical atmospheric profile distributions in different regions and seasons, and a fast radiation transfer model capable of processing various channel data is established.

[0085] Furthermore, there is also the data assimilation of the AGRI on the FY-4B multi-channel scanning imaging radiometer.

[0086] The AGRI imager detector on Fengyun-4B satellite consists of 14 channels in total. Water vapor channels 09 and 10 mainly obtain atmospheric water vapor information centered at 200 hPa and 400 hPa, and can detect atmospheric humidity information at high frequencies.

[0087] Assimilating the AGRI data of Fengyun-4 requires establishing an observation operator for the infrared imager based on a radiative transfer model. In the embodiment, instrument parameterization coefficients applicable to the radiative transfer model are constructed to realize the addition of the observation operator for assimilating the infrared imager data in the regional prediction model WRF.

[0088] Statistical analysis and evaluation are carried out for the AGRI data of Fengyun-4B. This includes statistics for ocean regions and land regions, etc., to make preliminary preparations for further data assimilation.

[0089] In the embodiment, there is also the assimilation of Fengyun-4B cloud motion wind data. Since the calculation of cloud motion wind vectors is based on the recognition of image modules (scales of dozens to hundreds of kilometers). The temporal and spatial changes in cloud areas, as well as possible errors or limitations in each stage such as image processing, calculation, and recognition, may make individual wind vectors unreasonable or make the group of cloud motion wind vectors show a certain degree of disorder. It is necessary to perform a certain quality control on this data before assimilation. According to the principle of continuity, the fluid motion speeds of adjacent points will not have large differences in magnitude and direction, that is, there is correlation and consistency. Therefore, correcting the cloud motion wind vectors at the same longitude and latitude points with differences in adjacent time steps, and filtering out the extreme cloud motion wind vectors within the adjacent range at the same time step can improve the quality of cloud motion wind data.

[0090] Specifically, error characteristic statistics refers to determining the error distribution characteristics and structural characteristics of cloud motion wind data, and establishing the observation error and its error covariance matrix of cloud motion wind data based on the comparison and verification of the characteristics of cloud motion wind data, the numerical prediction background field, and conventional meteorological sounding observations, as well as error statistical analysis.

[0091] The screening of data refers to taking the median value among the data half an hour before and after the regular data as a reference and comparing the three at the same longitude and latitude points.

[0092] The selection of single-point channels at the same longitude and latitude refers to the situation that when combining data from two channels, there may be data from different channels at the same longitude and latitude points. According to error statistics, the errors of wind speed and direction data in the water vapor channels are mostly smaller than those in the infrared channels compared with the sounding winds at each layer. Therefore, for points at the same longitude and latitude, the data values of the water vapor channels are taken.

[0093] Data deviation correction means that since the high-level cloud-trace winds have certain systematic deviation characteristics, the systematic deviation can be well removed through deviation correction, reducing the cloud-trace wind error. Considering the continuity of the wind distribution with height in the atmosphere. Therefore, for the situation where the wind speed and wind direction errors in the middle and lower layers are large and the error patterns are not obvious, the thermal wind principle is adopted for control. That is, the vertical shear of the wind and the background temperature field should approximately satisfy the thermal wind principle and not deviate too much. Through the extreme value control of the wind speed and the wind direction control, the data with unreliable quality is excluded.

[0094] The embodiment also includes the HY-2 scatterometer data assimilation. Since the scatterometer can not only provide sea surface wind speed data but also sea surface wind direction data, the satellite microwave scatterometer data can be used to well track and monitor the tropical cyclone circulation, which is of great significance for studying the atmospheric circulation over the ocean. The scatterometer has become an ideal remote sensor for obtaining the ocean surface wind field over a large area due to its advantages such as wide coverage, all-weather, all-day, and long-term continuous observation. With the development of the scatterometer wind field inversion technology, the accuracy of the scatterometer sea surface wind field has met the operational requirements.

[0095] The sea surface wind field is often affected by rainfall. At the same time, due to the large inversion errors of low wind speeds and high wind speeds, the data affected by rain and with wind speeds below 5 m / s and above 20 m / s need to be excluded when in use.

[0096] S34. Obtain the grid meteorological element simulation result through the data after data assimilation processing and the grid simulation model after initialization processing.

[0097] Specifically, obtaining the meteorological element simulation result through the data after data assimilation processing and the grid simulation model after initialization processing includes the following steps: S341. Interpolate the model layer to 45 geometric height layers.

[0098] In the embodiment, the vertical layer interpolation processing of the simulation data to 45 geometric height layers (height above the ground) includes 2, 10, 30, 50, 70, 100, 150, 200, 250, 300, 350, 400, 450, 500, 800, 1000, 1500, 2000, 2500, 3000, 3500, 4000, 4500, 5000, 5500, 6000, 7000, 8000, 9000, 10000, 11000, 12000, 13000, 14000, 15000, 16000, 17000, 18000, 19000, 20000, 21000, 22000, 23000, 24000, 25000 (meters).

[0099] S342. Use diagnostic analysis techniques to generate visibility, cloud amount, lightning, and cloud base height analysis elements.

[0100] Visibility is closely related to flight activities. Poor visibility is a serious visual range obstacle in flight activities, directly causing difficulties for visual flight and even endangering flight safety. The visibility usually forecasted by general meteorological stations is generally in several levels such as 1, 2, 4, 6, 8, 10, and above 10 kilometers. When the visibility is below 4 kilometers, it is called complex weather; when the visibility is greater than 10 kilometers, the visibility is considered good. Visibility is also the most critical determining factor for whether this airport is open or closed. At airports with relatively poor ground navigation equipment, it needs to be closed when the visibility is less than 0.8 kilometers. Visibility is closely related to the weather conditions at that time. When weather processes such as rainfall, fog, haze, and sandstorms occur, the atmospheric transparency is relatively low, so the visibility is poor.

[0101] Visibility diagnosis refers to calculating visibility using water vapor mixing ratio, cloud water mixing ratio, cloud ice mixing ratio, rain water mixing ratio, and snow mixing ratio. The expression is: Where , is the overall effect of cloud water mixing ratio, cloud ice mixing ratio, rain water mixing ratio, and snow mixing ratio.

[0102] Cloud water: ; Rain water: ; Cloud ice: ; Snow: ; C represents the mixing ratio in g / kg.

[0103] Cloud amount diagnosis refers to converting relative humidity into the corresponding layer cloud amount through an empirical exponential function using the quadratic relationship between the difference between relative humidity and its critical value: Where, represents the cloud amount of the k-th layer, with values ranging from 0 to 1; is the relative humidity of the k-th layer; is the relative humidity threshold of the k-th layer.

[0104] Integrated low clouds , medium clouds and high clouds to obtain the total cloud amount .

[0105] To be consistent with the customary values in cloud observations, the output products will , , and all the values of will be multiplied by 10.

[0106] Lightning diagnosis refers to statistically calculating the simulation factor values during lightning occurrences based on the historical actual situation of lightning observations, calculating the correlations between various forecast factors, selecting the factors with better correlations with thunderstorm occurrences, using the statistical method of Naive Bayes classification and the index nesting method to calculate the probability of lightning occurrence or non-occurrence, judging whether lightning occurs, and using the grid simulation results to delineate the lightning occurrence areas, thus establishing a lightning weather diagnosis model.

[0107] The specific algorithm implementation is as follows: Lightning is a type of severe convective weather, which is the result of the development of mesoscale convective activities. The occurrence, development, and maintenance of lightning generally require three conditions: moisture condition, instability condition, and lifting condition. For the target area, select the forecast factors that have good indicative effects on lightning occurrence in the target area from the severe convective parameters such as the Showalter index and SI index. Among them, the factors describing the lifting condition are relatively complex because the lifting force can come from synoptic-scale systems (such as upward motions in fronts, cyclones, etc.) or mesoscale systems (such as strong upward motions in mesoscale shear lines, convergence lines, etc.) or upward motions caused by terrain lifting and local heating unevenness.

[0108] It is described by the large-scale vertical velocity, and for the moisture condition and instability condition, extended factors will be selected to characterize.

[0109] By calculating the correlation coefficients between the historical cases of thunderstorms occurring in the target area and the above-listed simulation factors, and considering the differences between various indices, select the simulation factors for judging thunderstorm weather. The selection criteria for the simulation factors are to include as much as possible the elements characterizing the moisture condition, as well as the instability index and energy index; The specific steps are as follows: (a) Calculate the correlation coefficients between the lightning cases in the historical case library and the above simulation factors; (b) Sort the correlation coefficients and select the simulation factors with obvious correlations; (c) Calculate the correlations between various factors and select the relatively independent simulation factors as the lightning simulation factors.

[0110] (d) Simulate lightning using the index nesting method The index nesting method first defines the thresholds, then uses the simulation factors for nesting to establish a diagnostic simulation equation. And finally, perform the simulation by substituting the simulation factors into the equation.

[0111] Defining the thresholds of each simulation factor includes: Use the historical thunderstorm cases and non - thunderstorm weather cases within the specified area as the simulation factor library, statistically calculate the thresholds of each simulation factor when thunderstorms occur, and store them.

[0112] By judging whether each forecast factor of the diagnostic case reaches the threshold, each convective index reaching the standard threshold is recorded as 1, and not reaching is recorded as 0. The comprehensive convective index value is an integer from 0 to N. The larger the value, the greater the possibility of severe convection, and the smaller the value, the smaller the possibility of severe convection. Calculate the comprehensive index to judge whether lightning occurs.

[0113] Cloud - base height diagnosis includes the diagnosis method of the lifting condensation level and the relative humidity threshold method. Use the hindcast experiment to test the applicability of the above two methods and adopt the method with better effect.

[0114] Specifically, for the diagnosis of the lifting condensation level, when an unsaturated moist air parcel is lifted, as the air parcel rises, the temperature decreases according to the dry adiabatic lapse rate, and the saturation water vapor pressure corresponding to its temperature also decreases accordingly. There must be one and only one height at which the saturation water vapor pressure is equal to the water vapor pressure of the air parcel, that is, the temperature and dew point are equal, and water vapor begins to condense. This height is called the lifting condensation level. In the real atmosphere, unsaturated air does not always start to condense immediately when it reaches the lifting condensation level because there can be supersaturated states in the atmosphere. Nevertheless, introducing the concept of the lifting condensation level can help people understand the possibility of condensation at this height when moist air rises adiabatically. The lifting condensation level is generally close to the lower bound of the dynamically triggered convective cloud. The calculation formula is: In the formula, T and respectively represent the average temperature and average dew - point temperature of the air parcel from the starting lifting height to the lifting condensation level.

[0115] In another embodiment, the relative humidity threshold method (WR95 algorithm) is adopted. From the ground upwards, when the relative humidity ≥ 87%, it is considered to enter the cloud, and this height is regarded as the cloud - base height.

[0116] In the embodiment, both the large area and the three key areas will adopt the hourly cycling assimilation method to achieve the simulation of grid meteorological elements per hour. During cold start, the background field of the large area is generated by the global numerical forecast correction product, and the analysis and forecast fields of the large area are used as the background fields for the three key areas. During warm start, the 1-hour forecast field of each area is used as the background field for the next analysis moment. The delay time of the observational data of the statistical deployment unit is counted, and the start time of each analysis moment is set. First, the data assimilation of the large area is carried out, and then the assimilation analysis field of the large area is used as the background field. At the same time, the assimilation analysis of the three key areas is carried out. After the assimilation analysis of each area is completed, the assimilation analysis field is used as the initial field for a 1-hour forecast to provide the background field for the next moment. For example, if the surface report and Fengyun-4B satellite data are delayed by 20 minutes, the start time of the large area is set at the 20th minute of each hour; the assimilation analysis of the large area is completed within 8 minutes; the assimilation analysis of the three key areas is carried out after the assimilation analysis of the large area is completed, and the time is controlled within 8 minutes.

[0117] The cold start and warm start methods are jointly used to carry out rapid cycling assimilation. Since the T799 numerical weather forecast product has two forecasts per day, which are generated at about 02:00 and 14:00 Beijing time respectively, providing forecasts for the next 10 days with a time interval of 3 hours. In order to apply the latest global numerical forecast correction product as the background field and meet the timeliness requirements, cold starts are planned at 02:00 and 14:00 every day respectively to generate the initial boundary values of the large area; the warm start uses the one-hour forecast result after assimilation analysis as the background field. Furthermore, a block processing parallel method can also be adopted. Specifically, that is, according to the computing resources and processing time, multiple regions are divided into multiple sub-regions respectively, and model initialization, data assimilation, and diagnostic analysis are executed in sequence. Subsequently, the assimilation analysis fields of all sub-regions are integrated to form grid meteorological element simulation data.

[0118] S4. Based on the grid meteorological element simulation results, obtain the site meteorological element simulation results through an intelligent algorithm.

[0119] Specifically, the obtaining of the site meteorological element simulation results through an intelligent algorithm based on the grid meteorological element simulation results includes the following steps: S41. Based on the grid meteorological element simulation results, obtain the site meteorological elements.

[0120] Specifically, based on the grid simulation results of ground grid temperature, pressure, relative humidity, wind field, etc. output by the grid simulation model, using time and space algorithms, the interpolation of conventional and unconventional meteorological elements from grid points to sites is realized to obtain the site meteorological elements.

[0121] S42. Construct a site meteorological element correction model.

[0122] In the embodiment, the site meteorological element correction model is constructed based on a convolutional neural network including a residual network.

[0123] S43. Establish a site meteorological element correction data set using historical measured site meteorological elements, train and verify the site meteorological element correction model through the site meteorological element correction data set, and obtain the site meteorological element simulation result according to the trained site meteorological element correction model and the site meteorological elements.

[0124] Construct a data set using historical measured site meteorological elements, set hyperparameters, and train the corresponding site actual situation correction model for each site. The historical measured site meteorological elements include site historical observation data, historical grid simulation data, and historical Fengsi satellite precipitation and lightning data.

[0125] In the embodiment, the main component of the site meteorological element correction model is the convolutional layer. Each convolutional layer obtains multiple non-linear output mappings by sliding many convolutional kernels on the input mapping, and can extract abstract spatial features from the original spatial vector.

[0126] The model receives an input of 3×3×n, where n represents the number of input feature factors (number of channels); the model introduces a residual block for residual learning, and by introducing skip connections across modules at the beginning and end, the feature learning process will not degenerate, and the gradient problem in the deep network structure is alleviated; the model finally outputs the actual meteorological element values (conventional meteorological elements, precipitation, visibility, and lightning) at the correction moment, calculates the loss with the true value at the input moment, and updates the model parameters through backpropagation.

[0127] From input to output, the model sequentially includes a convolutional layer (including 64 convolutional kernels of 3×3×n), three residual modules (each residual module sequentially includes a convolutional layer, a BN layer, a non-linear activation function, and a convolutional layer, and the number of convolutional kernels in the convolutional layer is 64), and a convolutional layer (including 1 convolutional kernel of 3×3×64), and the value at the center point is used as the output. Adam optimizer is used to update the gradient during training, and the learning rate decreases exponentially with the number of learning times. Set a fixed number of learning times to obtain the optimal model.

[0128] Adopt the loss function: where n represents the number of samples, represents the model input value, represents the true observed value.

[0129] During the training process, in order to reflect the correlation between elements, the temperature, humidity, wind, air pressure, cloud, precipitation, visibility, and lightning values from grid simulations are used as inputs, while the actual temperature, humidity, wind, air pressure, cloud, precipitation, visibility, and lightning are used as labels for training to generate a site actual situation correction model based on grid simulations.

[0130] Furthermore, according to the trained site meteorological element correction model and the site meteorological elements, the simulation results of the site meteorological elements are obtained.

[0131] S5. Evaluate the grid meteorological element simulation results and the site meteorological element simulation results to obtain qualified meteorological and oceanic actual situation simulation results.

[0132] In the embodiment, the evaluating the grid meteorological element simulation results and the site meteorological element simulation results to obtain qualified meteorological and oceanic actual situation simulation results includes: performing integrity check, correlation test, change range check, climatological threshold value check, internal consistency check, time consistency check, and test for missing historical actual situation sites on the grid meteorological element simulation results and the site meteorological element simulation results; and obtaining qualified meteorological and oceanic actual situation simulation results according to the check results.

[0133] Specifically, the integrity check refers to checking the file naming, format, and data integrity of the observation data. Check the integrity and non - missingness of the data files in sequence, as well as whether the product file names and formats meet the specification requirements. Ensure the correct data format to meet the needs of subsequent dataset production by checking whether the file classification code is accurate, whether the file is empty, and whether the data storage format is unified and standard. When data is missing, write the missing information into the data management log and provide relevant information on replaceable data. The test conclusions are divided into: 0: not tested; 1: conforms; 2: suspicious; 3: does not conform; 9: missing value, and are identified with different identification codes.

[0134] Correlation test means that, according to the physical properties of meteorological variables and the relationships of climate characteristics, it is standardized and restricted so that the elements themselves or their changing trends maintain the consistency they should have. The test is carried out according to the mutual relationships among meteorological data, such as: the relationship between precipitation, cloud type and cloud amount; the relationship between air temperature, dew point and humidity; the relationship between air temperature and weather phenomena. This test calculates and discriminates the suspicious data that appears. For example, when the wind speed > 4 m / s and there is no precipitation, the general visibility > 1000 m; when heavy rain occurs, the general visibility < 1000 m; when rainstorm occurs, the general visibility < 100 m; when the dew point value ≤ the temperature value and the temperature dew point difference > 3°C, there is generally no obvious precipitation; there is generally no precipitation when it is clear and few clouds, and the air temperature is not too high when it is cloudy; there is generally no precipitation when the cloud base height > 5000 m. The test conclusions are divided into: 0: not tested; 1: conforms; 2: suspicious; 3: does not conform; 9: missing value, and are identified with different identification codes.

[0135] Range check means that for the data within the specified area and time domain, range check is carried out according to elements. The data beyond the element range is suspicious data and should be further checked to determine whether the data is correct. Generally, the difference between the two-time observation values of air temperature ≤ 10.0°C, the difference between the two-time observation values of air pressure ≤ 10.0 hPa, and the difference between the two-time observation values of dew point ≤ 7.0°C.

[0136] Climatological limit value check means that for the element values that cannot occur from the climatological perspective, the observation records should be within the climatological limit values. Climatological extreme value check means the check whether the meteorological records are beyond the climatological extreme values. Climatological extreme values refer to the meteorological records with a very small occurrence probability at a fixed meteorological station within a certain time range. When the climatological limit values cannot be determined, the extended values of the extreme values (maximum and minimum values) of historical records can be used instead. The data beyond the climatological limit value range are error data and will not participate in the subsequent checks.

[0137] Internal consistency check means that the relationships among the meteorological element records observed at the same time must conform to certain rules. For ground observation data, internal consistency is the consistency among elements. It is based on the fact that there is more or less a correlation among the elements measured at the same moment within an observation point, and it detects whether the meteorological elements with physical characteristic correlations are consistent. For example: the consistency among vapor pressure, dew point temperature, air temperature and relative humidity, and the consistency among sea level air pressure, station air pressure and air temperature. If the relationship Td ≤ T between the air temperature T and the dew point temperature Td on the same isobaric surface is not satisfied, it is considered that there is an error in the temperature or dew point temperature.

[0138] For the inspection of wind direction and wind speed, the following methods are adopted: If the wind direction degree is greater than 360 degrees or less than 0 degrees, the wind component is marked as incorrect data; if the wind direction degree is greater than 1 degree and the wind speed is 0, the wind component is marked as incorrect data; the extreme value of the wind speed is set at 80 m / s, and wind observations greater than this extreme value are considered incorrect observations.

[0139] For the consistency inspection between temperature and humidity elements, the dew point temperature is first calculated from the temperature and relative humidity. If the difference between the dew point and the temperature is greater than 30 or less than 0 degrees, the temperature and relative humidity data are marked as suspicious data.

[0140] The time consistency inspection refers to the inspection of whether the changes in meteorological records follow specific rules within a certain time range. Most meteorological elements (except wind, precipitation, and evaporation) change continuously, and their changes over time should be continuous. Within a certain time interval, the fluctuations of the same element before and after should be within a certain range. The time consistency inspection is responsible for checking the time consistency of the data, that is, checking whether the recorded values of the same element at the same time of adjacent two days are within a certain value range. The recorded value of the element on the previous day is used as the reference value for the inspection, and the recorded value of the adjacent next day is checked. If it exceeds the given threshold, the data to be inspected is considered suspicious data.

[0141] The time consistency inspection methods include: 1) If the data of consecutive 24-hour regular observations are the same, the relevant data are suspicious (except for missing observations).

[0142] 2) The boundary value test of the inter-daily change distribution of elements. For elements with a certain diurnal variation pattern, the inter-daily changes of the element values at each time have similar distribution characteristics. The method adopted is: all the data within one time step before and after a certain time step within a month are used to form a sequence, and the mean and standard deviation of the statistical sequence are calculated. If the absolute difference between the data to be inspected and the mean of the sequence is greater than or equal to 3 times the standard deviation of the sequence, the data is considered suspicious.

[0143] For the inspection of the lack of historical actual situation stations, when the target station does not have sufficient long-term observational data, the verification is carried out by selecting the simulated values of the 4 adjacent grid points around the target station.

[0144] Taking the station observational data as the reference sequence, calculate the correlation between the data sequences of the 4 grid points and the reference sequence in turn, and select the grid point with the best correlation as the reference station.

[0145] Then, calculate the mean absolute error, maximum error, minimum error between the data sequence of the most relevant grid point and the reference sequence, and the percentage of the absolute error greater than the mean absolute error ; at the same time, perform the above calculations on the data sequences of the other 3 grid points and the reference sequence, and retain the mean absolute error, maximum error, and minimum error.

[0146] Finally, calculate the error between the site data to be verified and the simulation values of the adjacent 4 grid points, and complete the verification according to the anomaly discrimination principle.

[0147] Please refer to Figure 2 , in the embodiment, in order to efficiently execute a meteorological and oceanic real-time simulation method provided by the present invention, the present invention also provides a meteorological and oceanic real-time simulation system based on satellite data, including: an input device, an output device, a processor, and a memory. The input device, the output device, the processor, and the memory are interconnected. The memory contains program instructions, and the program instructions are used for the steps of the meteorological and oceanic real-time simulation method based on satellite data. The meteorological and oceanic real-time simulation system based on satellite data of the present invention has a compact structure and stable performance, and can stably execute the meteorological and oceanic real-time simulation method based on satellite data of the present invention, further improving the overall applicability and practical application ability of the present invention.

[0148] In the embodiment, the so-called processor may be a central processing unit (CPU), and the processor may also be other general-purpose processors, digital signal processors (DSPs), application specific integrated circuits (ASICs), field-programmable gate arrays (FPGAs) or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components, etc. The general-purpose processor may be a microprocessor or the processor may also be any conventional processor, etc. The input device may be used to obtain data information. The output device may be used to output the result obtained from the program instructions included in the computer program stored in the memory provided by the present invention. The memory may include a read-only memory and a random access memory, and provide instructions and data to the processor. A part of the memory may also include a non-volatile random access memory.

[0149] In a possible implementation, the memory may include a program storage area and a data storage area. The program storage area may store an operating system and application programs required for at least one function, etc.; the data storage area may store data created during use. In addition, the memory may include a read-only memory and a random access memory, and provide instructions and data to the processor. A part of the memory may also include NVRAM. The memory stores an operating system, operation instructions, executable modules or data structures, or subsets thereof, or extended sets thereof. The operation instructions may include various operation instructions for implementing various operations. The operating system may include various system programs for implementing various basic tasks and processing hardware-based tasks.

[0150] The embodiments also provide a storage medium having a computer program stored thereon. When the computer program is executed by a processor, the steps of the above-mentioned meteorological and oceanic actual situation simulation method based on satellite data are implemented.

[0151] The storage medium may include: various media that can store program codes, such as USB flash drives, external hard drives, read-only memories (ROM), random access memories (RAM), magnetic disks, or optical discs.

[0152] In summary, the present invention performs grid simulation after correcting data, then obtains station simulation results through an intelligent algorithm, evaluates the simulation results, and obtains qualified meteorological element simulation results. By means of actual situation simulation, the problem of obtaining actual situations in difficult-to-reach areas such as the ocean and isolated islands is solved.

[0153] Therefore, the present invention effectively overcomes various disadvantages in the prior art and has high industrial utilization value.

[0154] 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 them; although the present invention has been described in detail with reference to the foregoing embodiments, those of ordinary skill in the art should understand that: they can still modify the technical solutions described in the foregoing embodiments, or perform equivalent replacements on some or all of the technical features; and 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, and they should all be covered by the scope described in the present invention.

Claims

1. A meteorological and oceanic actual situation simulation method based on satellite data, characterized in that, The meteorological and oceanic actual situation simulation method based on satellite data includes the following steps: Obtain satellite multi-source connection data through a data connection system; Construct a numerical weather prediction product correction model, and use the numerical weather prediction product correction model to correct the satellite multi-source connection data; According to the corrected satellite multi-source connection data, obtain the grid meteorological element simulation result by using a simulation algorithm; Based on the grid meteorological element simulation result, obtain the station meteorological element simulation result through an intelligent algorithm; Evaluate the grid meteorological element simulation result and the station meteorological element simulation result to obtain a qualified meteorological and oceanic actual situation simulation result.

2. The meteorological and oceanic actuality simulation method based on satellite data according to claim 1, characterized in that, After obtaining the satellite multi-source connection data through the data connection system, preprocess the satellite multi-source connection data; The preprocessing includes data deduplication, data selection, data filtering, abnormal file elimination, and standardization processing.

3. The meteorological and oceanic actual situation simulation method based on satellite data according to claim 1, characterized in that The numerical weather prediction product correction model includes a meteorological element feature extraction coding structure and a meteorological element feature restoration decoding structure; The meteorological element feature restoration decoding structure includes a sub-pixel convolution sub-module and a replication splicing sub-module. The sub-pixel convolution sub-module is used for expanding the meteorological element feature map, and the replication splicing sub-module is used for channel splicing of the feature maps of the meteorological element feature extraction coding structure and the meteorological element feature restoration decoding structure.

4. The meteorological and oceanic actuality simulation method based on satellite data according to claim 1, characterized in that, The step of using the numerical weather prediction product correction model to correct the satellite multi-source connection data includes the following steps: Construct a training and validation data set by using historical satellite multi-source connection data; Optimize the parameters of the numerical weather prediction product correction model, and train and validate the optimized numerical weather prediction product correction model through the training and validation data set; According to the trained numerical weather prediction product correction model, correct the satellite multi-source connection data.

5. The meteorological and oceanic real-time simulation method based on satellite data according to claim 1, characterized in that, The step of obtaining the grid meteorological element simulation result by using a simulation algorithm according to the corrected satellite multi-source connection data includes the following steps: Use a simulation algorithm to construct a grid simulation model; Perform initialization processing on the grid simulation model; Perform data assimilation processing based on the corrected satellite multi-source connection data; Obtain the grid meteorological element simulation result through the data after data assimilation processing and the initialized grid simulation model.

6. The meteorological and oceanic actuality simulation method based on satellite data according to claim 5, characterized in that The data assimilation processing based on the corrected satellite multi-source connection data includes occultation data assimilation; The occultation data assimilation satisfies the following formula: Among them, represents the refractive index value of the neutral atmosphere, represents the empirical coefficient related to the dry air pressure, represents the partial pressure of dry air, represents the absolute temperature, represents the water vapor pressure, represents the empirical coefficient of the temperature square term related to the water vapor pressure, represents the empirical coefficient of the temperature first term related to the water vapor pressure.

7. The meteorological and oceanic actual situation simulation method based on satellite data according to claim 5, characterized in that, The step of obtaining the meteorological element simulation result through the data after data assimilation processing and the initialized grid simulation model includes the following steps: Interpolate the model layer to 45 geometric height layers; Use diagnostic analysis technology to generate analysis elements such as visibility, cloud amount, lightning, and cloud base height.

8. The meteorological and oceanic actual situation simulation method based on satellite data according to claim 1, wherein The step of obtaining the station meteorological element simulation result through an intelligent algorithm based on the grid meteorological element simulation result includes the following steps: Obtain the station meteorological element based on the grid meteorological element simulation result; Construct a station meteorological element correction model; Use historical measured station meteorological elements to establish a station meteorological element correction data set, and train and validate the station meteorological element correction model through the station meteorological element correction data set; Obtain the simulation result of the station meteorological elements according to the trained correction model of the station meteorological elements and the station meteorological elements.

9. The meteorological and oceanic real-time simulation method based on satellite data according to claim 1, wherein Evaluate the simulation result of the grid meteorological elements and the simulation result of the station meteorological elements to obtain a qualified meteorological and oceanic actual situation simulation result, including: Conduct integrity check, correlation test, range of change check, climatological threshold value check, internal consistency check, time consistency check, and test for lack of historical actual situation stations on the simulation result of the grid meteorological elements and the simulation result of the station meteorological elements; Obtain a qualified meteorological and oceanic actual situation simulation result according to the inspection result.

10. A meteorological and oceanic real-time simulation system based on satellite data, characterized in that, The meteorological and oceanic actual situation simulation system based on satellite data includes: an input device, an output device, a processor, and a memory. The input device, the output device, the processor, and the memory are interconnected. The memory includes program instructions, and the program instructions are used to execute the meteorological and oceanic actual situation simulation method according to any one of claims 1-9.

Images