A method, device and equipment for generating marine weather data

Abstract

The application provides a marine meteorological data generation method, device and equipment, and relates to the technical field of data processing. The marine meteorological data generation method comprises the following steps: obtaining identified observation data and numerical simulation data of marine meteorological observation data; interpolating the numerical simulation data to a preset target grid to generate interpolated wind field data; generating corrected wind field data according to the interpolated wind field data and the identified observation data; generating simulated sea wave data according to the corrected wind field data; generating corrected sea wave data according to the simulated sea wave data and the identified observation data; and encoding the corrected wind field data and the corrected sea wave data into target marine meteorological data. The scheme of the application realizes the fusion of autonomous observation data and numerical simulation data, and efficiently generates high-precision auxiliary meteorological data meeting the input requirements of marine satellites.

Description

Technical Field

[0001] This invention relates to the field of data processing technology, and specifically to a method, apparatus, and equipment for generating marine meteorological data. Background Technology

[0002] The CFOSAT ocean satellite, equipped with a spectrometer, can acquire global sea surface wind field and wave direction spectrum information. When generating L2-level standard products, the satellite ground application system requires input of auxiliary meteorological data in a specific format. Specifically, the CMTO meteorological auxiliary data component in the processing program requires two independent GRIB format files, one containing meteorological parameters and the other containing wave parameters. However, data acquisition is constrained by international data policies, resulting in uncertainty and lag, making it difficult to guarantee operational autonomy, controllability, and long-term stability. Secondly, foreign data from the target country has limited accuracy in reflecting the unique marine environmental characteristics of the target country's nearshore waters, failing to fully utilize the increasingly sophisticated marine three-dimensional observation network (such as buoys and oceanographic stations) acquiring on-site observation data. Furthermore, existing data fusion and interpolation methods are insufficient in processing efficiency and accuracy. For example, when interpolating data to the N640 Gaussian reduced grid required by CFOSAT, traditional scatter interpolation algorithms are computationally inefficient, making it difficult to meet the high-frequency, near-real-time operational requirements. Summary of the Invention

[0003] This invention provides a method, apparatus, and equipment for generating marine meteorological data, so as to realize the fusion of autonomous observation data and numerical simulation data, and efficiently generate high-precision auxiliary meteorological data that meets the input requirements of marine satellites.

[0004] To solve the above-mentioned technical problems, the technical solution of the present invention is as follows: Embodiments of the present invention provide a method for generating marine meteorological data, comprising: Acquire marine meteorological observation data and numerical simulation data; The numerical simulation data is interpolated onto a preset target grid to generate interpolated wind field data; Based on the interpolated wind field data and the identified observation data, the corrected wind field data is generated; Based on the corrected wind field data, simulated ocean wave data is generated; Based on simulated wave data and identified observation data, corrected wave data is generated. The corrected wind field data and corrected wave data are encoded into target marine meteorological data.

[0005] The above-described solution of the present invention has at least the following beneficial effects: The method for generating marine meteorological data according to this invention involves acquiring identification observation data and numerical simulation data from marine meteorological observation data; interpolating the numerical simulation data onto a preset target grid to generate interpolated wind field data; generating corrected wind field data based on the interpolated wind field data and the identification observation data; generating simulated ocean wave data based on the corrected wind field data; generating corrected ocean wave data based on the simulated ocean wave data and the identification observation data; and encoding the corrected wind field data and the corrected ocean wave data into target marine meteorological data. This method achieves the fusion of autonomous observation data and numerical simulation data, and efficiently generates high-precision auxiliary meteorological data that meets the input requirements of marine satellites. Attached Figure Description

[0006] Figure 1 This is a flowchart illustrating the method for generating marine meteorological data according to the present invention.

[0007] Figure 2 This is a schematic diagram of the module block of the marine meteorological data generation device of the present invention. Detailed Implementation

[0008] Exemplary embodiments of the invention will now be described in more detail with reference to the accompanying drawings. While exemplary embodiments of the invention are shown in the drawings, it should be understood that the invention may be implemented in various forms and should not be limited to the embodiments set forth herein. Rather, these embodiments are provided so that this invention will be thorough and complete, and will fully convey the scope of the invention to those skilled in the art.

[0009] like Figure 1 As shown, an embodiment of the present invention proposes a method for generating marine meteorological data, comprising: Step 1: Acquire marine meteorological observation data and numerical simulation data; the observation data includes field observation data and satellite remote sensing data, wherein the field observation data includes at least wind speed and significant wave height, and the numerical simulation data includes at least wind field data and ocean wave data; Step 2: Interpolate the numerical simulation data onto a preset target grid to generate interpolated wind field data; Step 3: Generate corrected wind field data based on the interpolated wind field data and the identified observation data; Step 4: Generate simulated ocean wave data based on the corrected wind field data; Step 5: Generate corrected wave data based on simulated wave data and identified observation data; Step 6: Encode the corrected wind field data and corrected wave data into target marine meteorological data.

[0010] In this embodiment, a progressive, closed-loop technology chain integrating wind field driving, wave simulation, and observation fusion is constructed, achieving complementary advantages from data from different sources. Observational data is used to correct the background wind field, providing more accurate driving conditions for wave simulation. Subsequently, the observational data is again integrated into the simulated wave field to further correct deviations in the wave product. This phased, targeted, and gradual correction mechanism effectively integrates the authenticity of observational data with the spatiotemporal continuity of numerical models. The resulting analytical product possesses both high precision and high spatiotemporal consistency, significantly enhancing reliability. This solution can effectively integrate data from acquired ocean three-dimensional observation networks (such as buoys and ocean stations). Together with domestically produced numerical model products (such as GRAPES wind field and NEMO sea surface temperature), we have formed an independent and controllable auxiliary meteorological data production technology system. Addressing the computational bottleneck of interpolating data into the complex N640 Gaussian grid required by CFOSAT, this method abandons the traditional, less efficient scattered interpolation algorithm. Through batch processing and data parallelism, it significantly improves the interpolation calculation speed, making rapid gridding of massive amounts of data possible, thus meeting the near real-time operational requirements of satellite data processing. It also achieves the fusion of autonomous observation data and numerical simulation data, and efficiently generates high-precision auxiliary meteorological data, i.e., target marine meteorological data, that meets the input requirements of marine satellites.

[0011] In an optional embodiment of the present invention, step 1, acquiring identification observation data of marine meteorological observation data, includes: step 11, acquiring marine meteorological observation data from multiple sources; specifically, acquiring field data from different observation platforms, including observation data from large buoys (such as 10-meter disc buoys), ocean stations, and observation vessels, from the operational network and observation data transmission system of the National Marine Environmental Forecasting Center; wherein the observation equipment is divided into two categories: automatic instrument measurement equipment and manual observation equipment; extracting core elements such as wind speed (10-meter height wind speed), wind direction, and significant wave height from the above-mentioned field observation data, these elements are the basis for subsequent data fusion; acquiring remote sensing data from CFOSAT satellite and other satellites through external network downloads, wherein satellite remote sensing observation is divided into scatterometers, altimeters, etc. according to sensor type; acquiring sea surface wind field related elements from scatterometers and sea wave related elements from altimeters, ensuring that the acquired remote sensing data can cover the observation information of sea surface wind and waves along satellite orbits globally; acquiring operational analysis wind field data produced by the National Marine Environmental Forecasting Center, which... The data serves as the primary foundation for wind field numerical simulation. GFS wind field data from NCEP (National Centre for Environmental Forecasting) is acquired as a backup to ensure the availability of primary wind field data. Wind field data generated by the China Meteorological Administration's GRAPES (Global / Regional Assimilation Forecast Enhancement System) system is also acquired, providing global or regional wind field simulation results and enriching the sources of wind field data. Significant wave height field data generated by the wave numerical post-reporting subsystem (based on the WAVEWATCHIII model, using the ST4 source function scheme, optimizing key parameters for wave growth, and employing SMC variable resolution grids and hot-start technology) is acquired, serving as the core analysis field data for wave numerical simulation. Operational numerical analysis field data from ECMWF (European Centre for Medium-Range Weather Forecasts) and NCEP, including sea surface temperature and sea ice concentration, are also acquired. Additionally, sea surface temperature-related data generated by the National Marine Environmental Forecasting Center's high-resolution global marine dynamic environment numerical forecasting system (based on the 1 / 12° global NEMO model and employing ensemble filtering assimilation techniques) is obtained.

[0012] Step 12 involves automatic quality control of the marine meteorological observation data to identify and mark problematic data, thereby obtaining the results of automatic quality control. Specifically, this includes: pre-setting a control radius based on the buoy's normal drift range; determining whether the buoy exceeds its normal drift range by real-time monitoring of the distance between its current and initial coordinates; if so, automatically marking the observation data corresponding to the buoy's movement as problematic data (this strategy can completely identify this type of problematic data); pre-storing land information of the observation area; determining whether the buoy is in a shore maintenance state by performing mask calculations between the buoy's real-time coordinates and the pre-stored land information (if the coordinates fall within the land area, it is determined to be in a maintenance state); if determined to be in a maintenance state, automatically marking the observation data sent by the buoy in this state as useless information-type problematic data (this strategy can completely identify this type of problematic data); performing integrity verification on the acquired observation data files by verifying the file's checksum, format identifiers, etc., to determine whether the file is corrupted; if corrupted, automatically marking the observation data corresponding to that file as problematic data (this strategy can completely identify this type of problematic data); and checking the integrity of the data files within the data files. The system iterates through the data elements to check for missing core elements such as wind speed and significant wave height. If any elements are missing, the corresponding observation data for that file is automatically marked as problematic data. This strategy can completely identify this type of problematic data. The system compares the Q&A results obtained during the data download process (such as file information returned by the server) with the pre-stored file list information (including the expected server address, file path, file name, etc.). If inconsistencies are found (such as the actual server address not matching the pre-stored address, or the actual file name differing from the pre-stored file name), the observation data in this case is automatically marked as problematic data, and only an error message is displayed; the data is not directly deleted. The system compares the values ​​of elements such as wind speed and significant wave height in the observation data with the corresponding element values ​​in the same period's numerical simulation data (such as the significant wave height field of the numerical analysis wind field and the numerical wave reporting subsystem of the ocean wave). A weighted average method is used to determine whether the observation data is an extreme large or small value. If it is determined to be an outlier, the observation data is automatically marked as problematic data. However, this strategy can only partially identify this type of problematic data (it cannot completely distinguish between extreme true values ​​and outliers).

[0013] Step 13: Based on the results of the automatic quality control, correct the problematic data to obtain the intervention-ready observation data. Specifically, a dedicated manual intervention window is set up. Forecasters use control files to perform targeted corrections on the following problematic data to obtain the intervention-ready observation data: For the problematic data marked in Step 12 due to changes in data server, file path, or file name, the forecaster, based on the actual situation, enters new server addresses, file paths, file names, and other naming rules in the manual intervention window, reconfigures data transmission parameters, and ensures that the system can correctly acquire and parse this part of the observation data to complete the data correction; For the abnormally large / small values ​​marked in Step 12, the forecaster combines other hydrological and meteorological data from the same period (such as surrounding buoy observation data, satellite remote sensing data, and numerical simulation results) to... Based on comprehensive judgment, if data anomalies are confirmed, the abnormal data is masked through a manual intervention window, and the observations for the time period and region containing the abnormal data are supplemented using reasonable interpolation or averaging methods to fill the data gaps. For sensor zero-point drift issues not identified in step 12 (for which there is no automatic quality control strategy), forecasters monitor the long-term trend of the observation data. If a regular zero-point shift is found in the data within a short period of time, the observation data for that period is marked as abnormal data, and the data is corrected through supplementary calculations to ensure data accuracy. If missing reports are found in step 12 due to sensor damage, forecasters mark the missing reporting period through a manual intervention window, and simultaneously use available observation data from the surrounding area (such as data from nearby buoys and oceanographic stations) to supplement the data, generating alternative observation data for the missing reporting period to complete data correction.

[0014] Step 14: Use the results of automatic quality control and the post-intervention observation data as the identification observation data. Specifically, this includes: integrating the obtained automatic quality control results (including the labeling information of various problem data, and the results of complete and partial identification) with the post-intervention observation data obtained after manual intervention correction (including the effective data obtained after reconfiguring the transmission parameters, the alternative data generated by supplementary calculation, and the effective data retained after masking anomalies) to form the final identification observation data.

[0015] In an optional embodiment of the present invention, step 2, interpolating the numerical simulation data onto a preset target grid to generate interpolated wind field data, includes: step 21, determining the four surrounding grid points corresponding to each target grid point in the numerical simulation data grid based on the latitude and longitude coordinates of the preset target grid. Specifically, the preset target grid is a GaussReduce N640 Gaussian grid, which has the structural characteristics of equidistant longitude distribution, non-equidistant (quasi-equidistant) latitude distribution, and no grid points at the poles, totaling 2,140,702 grid points; by calling the pre-stored N640 Gaussian grid latitude and longitude configuration file, extracting the unique identifier (such as grid number) of each target grid point in the grid and its corresponding longitude and latitude values, establishing the unique identifier and longitude of the target grid point. Establish a one-to-one correspondence between the longitude and latitude values ​​to ensure that the spatial location of each target grid point can be clearly traced. Obtain the structural parameters of the grid to which the numerical simulation wind field data to be interpolated (including operational analysis wind field data produced by the National Marine Environmental Forecasting Center, GFS wind field data from NCEP, or wind field data generated by the GRAPES system of the China Meteorological Administration). Specifically, this includes the latitudinal step size and longitudinal step size of the numerical simulation data grid, as well as the longitude range (from the minimum longitude value to the maximum longitude value) and latitude range (from the minimum latitude value to the maximum latitude value) covered by the grid. At the same time, extract the longitude value, latitude value, and corresponding grid index (used to identify the position of the grid point in the numerical simulation grid) of each grid point in the numerical simulation data grid to form a related dataset of numerical simulation grid point index, longitude value, and latitude value. For each target grid point, its location is determined within the latitude and longitude range of the numerical simulation data grid based on its latitude and longitude coordinates: Longitude location involves comparing the longitude value of the target grid point with the longitude sequence of the numerical simulation data grid to determine which two adjacent longitude values ​​the target grid point falls between, thus identifying the indices and latitude and longitude coordinates of the two numerical simulation grid points corresponding to these two adjacent longitude values; Latitude location involves comparing the latitude value of the target grid point with the latitude sequence of the numerical simulation data grid to determine which two adjacent latitude values ​​the target grid point falls between. Then, the indices and latitude and longitude coordinates of the two numerical simulation grid points corresponding to the two adjacent latitude values ​​are locked; the four surrounding grid points are determined, and the two numerical simulation grid points locked in the longitude direction are combined with the two numerical simulation grid points locked in the latitude direction to form a numerical simulation data grid cell composed of these four grid points. These four grid points are the four surrounding grid points corresponding to the current target grid point; the indices and latitude and longitude coordinates of these four grid points are recorded and associated with the unique identifier of the current target grid point; the above operation is repeated until the four surrounding numerical simulation grid points corresponding to all 2,140,702 target grid points are determined.

[0016] Step 22: Based on the latitude and longitude distance relationship between the target grid point and the four surrounding grid points, calculate the interpolation coefficients required for bilinear interpolation. Specifically, this includes: based on the established association relationship between the unique identifier of the target grid point and the information of the four surrounding numerical simulation grid points, extracting the longitude and latitude values ​​of each target grid point, as well as the longitude and latitude values ​​of the four surrounding numerical simulation grid points; labeling the four surrounding grid points as the first grid point (X1, Y1), the second grid point (X2, Y1), the third grid point (X2, Y2), and the fourth grid point (X1, Y2), where X represents longitude, Y represents latitude, and X2 > X1, Y2 > Y1; calculating the longitude and latitude distances between the target grid point and the four surrounding grid points. Distance Ratio: The meridional distance ratio is calculated as follows: The difference between the longitude value P1 of the target grid point and the longitude value X1 of the first grid point is calculated, and then divided by the difference between the longitude value X2 of the second grid point and the longitude value X1 of the first grid point. This yields the meridional distance ratio, denoted as Xscale. This ratio reflects the positional proportion of the target grid point relative to the unit formed by the four surrounding grid points in the longitude direction. The latitudinal distance ratio is calculated as follows: The difference between the latitude value P2 of the target grid point and the latitude value Y1 of the first grid point is calculated, and then divided by the difference between the latitude value Y2 of the fourth grid point and the latitude value Y1 of the first grid point. This yields the latitudinal distance ratio, denoted as Yscale. This ratio reflects the positional proportion of the target grid point relative to the unit formed by the four surrounding grid points in the latitudinal direction.

[0017] Derivation of Bilinear Interpolation Coefficients: Based on the aforementioned meridional distance scale Xscale and zonal distance scale Yscale, and following the calculation logic of bilinear interpolation, the interpolation coefficients for the four surrounding grid points are derived: The interpolation coefficient for the first grid point (X1, Y1) is obtained by multiplying 1-Xscale and 1-Yscale. This coefficient reflects the contribution weight of the first grid point to the interpolation result of the target grid point. The interpolation coefficient for the second grid point (X2, Y1) is obtained by multiplying Xscale and 1-Yscale. This coefficient reflects the contribution weight of the second grid point to the interpolation result of the target grid point. The interpolation coefficients for the third grid point (X2, Y2) are obtained by multiplying Xscale and Yscale. These coefficients reflect the contribution weight of the third grid point to the interpolation result of the target grid point. The interpolation coefficients for the fourth grid point (X1, Y2) are obtained by multiplying 1-Xscale and Yscale. These coefficients reflect the contribution weight of the fourth grid point to the interpolation result of the target grid point. Vectorized calculation and storage of the interpolation coefficients are implemented: a vectorized processing method is adopted, and the above-mentioned calculation of the meridional and latitudinal distance ratios and derivation of interpolation coefficients are performed synchronously on all target grid points to avoid the inefficiency caused by point-by-point iterative calculations. The four interpolation coefficients corresponding to each target grid point are associated with and stored with the unique identifier of that target grid point, forming a dataset of the unique identifier of the target grid point and the four interpolation coefficients.

[0018] Step 23: Using the interpolation coefficients, weighted fuse the wind field data of the four surrounding grid points in the numerical simulation data to obtain the interpolated wind field data for each target grid point. Specifically, this includes: for each target grid point, based on the indices of the four surrounding numerical simulation grid points recorded in Step 21, extracting the core wind field data, zonal wind speed component (U component) data, and meridional wind speed component (V component) data for each of the four grid points from the numerical simulation wind field dataset; ensuring that the extracted U component and V component data are accurately associated with the corresponding grid point index and the unique identifier of the target grid point, avoiding… Data mismatch; weighted fusion of U-component data based on interpolation coefficients: For the U-component interpolation result of the target grid point, the U-component data of the first grid point is multiplied by its corresponding interpolation coefficient to obtain the weighted value of the U-component of the first grid point; similarly, the U-component data of the second, third, and fourth grid points are multiplied by their corresponding interpolation coefficients to obtain their respective U-component weighted values; the weighted values ​​of the U-components of the above four grid points are summed to obtain the zonal wind speed component (U-component) interpolation result of the target grid point, which integrates the contributions of the U-component data of the four surrounding grid points.

[0019] The V component data is weighted and fused based on interpolation coefficients: using the same logic as the U component weighted fusion, the V component data of the first grid point is multiplied by its corresponding interpolation coefficient to obtain the weighted value of the V component of the first grid point; the V component data of the second, third, and fourth grid points are calculated in sequence and multiplied by their corresponding interpolation coefficients to obtain their respective V component weighted values; the V component weighted values ​​of the four grid points are summed to obtain the meridional wind speed component (V component) interpolation result of the target grid point, which integrates the contributions of the V component data of the four surrounding grid points.

[0020] Verify the rationality of the interpolated wind field data: For the interpolation results of the U and V components obtained for each target grid point, verify their rationality by combining them with the conventional numerical range of wind field data in the marine meteorological field (such as excluding abnormal results that exceed the reasonable wind speed threshold); if the interpolation results are within a reasonable range, then use the interpolation results of the U and V components as the interpolated wind field data of the target grid point and store them in association with the unique identifier of the target grid point; if the interpolation results are abnormal, return to step 21 to redetermine the four surrounding grid points of the target grid point, or return to step 22 to recalculate the interpolation coefficients until reasonable interpolated wind field data is obtained.

[0021] Step 24: Combine the interpolated wind field data of all target grid points to form the interpolated wind field data. Specifically, this includes: based on the association relationship between the unique identifier of the target grid point and the interpolated wind field data (U component, V component) formed in Step 23, further improve the mapping system to ensure that the unique identifier of each target grid point corresponds to only one unique set of U component and V component interpolation results, and that the mapping relationship can be quickly queried through the unique identifier of the target grid point, providing a data association basis for subsequent integration. make Represents grid points in numerical simulation data grid The corresponding core wind field data (including zonal wind speed component U and meridional wind speed component V) has zonal and meridional step sizes for the numerical simulation data grid, respectively. and To achieve finer mesh adaptation, zero-fill wind speed is introduced. Its definition is revised as follows: ; in, For any positive integer, Z is defined as the set of integers; the wind field data from the original numerical simulation data grid... Interpolate to target grid (step size is) The process can be represented as a two-dimensional convolution, and the corrected formula is as follows: ; in, This is called the impulse response function, which is uniquely determined by the interpolation method. In this method... The double singer function is used, and it is defined as follows: ;in, , Let be the singer function, satisfying , ;in, This represents the wind field data after zero-filling. It is the core variable on the transition grid, and its unit is consistent with the original wind field data (e.g., m / s, corresponding to U / V components); (i, j) represents the integer index; M and N represent any positive integers, which represent the refinement factor of the original grid in the latitudinal (longitudinal) and longitudinal (latitude) directions, respectively. This represents the final interpolated wind field data on the target grid; p and q represent integer summation variables used to traverse all edge grid points in the transition grid that may affect the current target grid point.

[0022] Sort the interpolated wind field data according to the target grid structure rules: Referencing the standard data arrangement rules of the GaussReduce N640 Gaussian grid (which follow the spatial distribution logic of grid longitude from west to east and latitude from low to high latitude, and conform to the format requirements of subsequent data encoding and application), sort the associated dataset of unique identifiers of target grid points and interpolated wind field data. The sorted data must accurately reflect the spatial continuity of the N640 Gaussian grid, ensuring that the interpolated wind field data of adjacent target grid points maintain spatial adjacency in the dataset, facilitating spatial analysis during subsequent data application. Check the completeness of the interpolated wind field data: Count the number of unique identifiers of target grid points in the sorted dataset and compare it with the total number of grid points in the N640 Gaussian grid (2,140,702) to confirm whether there are any missing or duplicate interpolated wind field data for grid points. If data is missing, return to the above steps to recalculate the interpolated wind field data for the missing grid points. Field data; if duplicate data exists, remove duplicate data and retain the only valid interpolated wind field data until the dataset contains interpolated wind field data for all target grid points; verify the overall validity of the interpolated wind field data: perform overall validity verification on the sorted complete interpolated wind field dataset, including: batch verifying whether the interpolation results of the U and V components of all grid points are within the reasonable range of marine meteorological wind field data, excluding overall data offset or anomalies; check whether there are abrupt changes in the interpolated wind field data (U and V components) of adjacent grid points (such as the wind speed difference between adjacent grid points exceeding the reasonable fluctuation range under normal marine meteorological scenarios). If there are abrupt changes, locate the abrupt grid points and return to the previous steps for reprocessing to ensure spatial consistency of the data; if the integrity check and overall validity verification both pass, the sorted complete interpolated wind field dataset is the final interpolated wind field data; if it fails, repeat the above check and verification process until complete and valid interpolated wind field data is generated.

[0023] In an optional embodiment of the present invention, step 3, generating corrected wind field data based on the interpolated wind field data and the identified observation data, includes: Step 31 involves decomposing the interpolated wind field data into wind speed and wind direction components. Specifically, this includes: acquiring the interpolated wind field data, which uses a GaussReduce N640 grid as its carrier and contains the zonal wind speed component (denoted as the U component) and meridional wind speed component (denoted as the V component) for each target grid point; secondly, using conventional wind field parameter conversion methods in marine meteorology, calculating the wind speed (i.e., the magnitude of the wind speed, reflecting the strength of the wind) for each grid point based on the U and V components; simultaneously, determining the wind direction component (i.e., the direction of wind propagation, reflecting the direction of wind direction) for the same grid point based on the meteorological data direction calculation rules; finally, ensuring that the wind direction component remains unchanged from its initial calculation in subsequent wind field correction processes, serving only as a basic element for subsequent data integration, thus completing the decomposition of the interpolated wind field data into wind speed and wind direction components.

[0024] Step 32, using the wind speed rate as the first background field, extracts wind speed rate observation values ​​from the identification observation data. Specifically, this includes: determining the composition of the identification observation data, including on-site observation data (10-meter height wind speed observation data from large buoys, ocean stations, and observation vessels) and satellite remote sensing observation data (sea surface wind speed observation data from scatterometers); secondly, for the on-site observation data, extracting the 10-meter height wind speed observation values ​​(i.e., wind speed rate observation values) at the corresponding observation time and location from the observation records of each observation platform (buoy, ocean station, observation vessel), and recording the observation time and latitude / longitude coordinates corresponding to each observation value; for the satellite remote sensing data... The wind speed observation data (i.e., wind rate observation values) are extracted from the wind speed data along the track obtained by the scatterometer at the corresponding observation time and at each observation point within the track coverage area. The observation time and latitude and longitude coordinates are also recorded. Finally, all extracted wind rate observation values ​​are spatiotemporally matched to ensure that the observation time of each observation value is consistent with the time dimension of the interpolated wind field data (00, 06, 12, and 18:00 UTC every day), and that the latitude and longitude coordinates of the observation location are within the spatial coverage of the GaussReduceN640 Gaussian grid. The spatiotemporally matched wind rate observation values ​​are selected as the basic observation data for subsequent calculations.

[0025] Step 33: For each observation point, calculate the difference between the observed wind speed rate and the estimated background field value at that observation point to obtain the first observation increment. Specifically, this includes: determining the estimated background field value corresponding to each wind speed rate observation; for each spatiotemporally matched wind speed rate observation (including the latitude and longitude coordinates of the observation point), using a spatial interpolation method (consistent with the bilinear interpolation logic in Step 2 to ensure interpolation accuracy), interpolating the wind speed rate background field obtained in Step 31 (i.e., the interpolated wind speed rate grid data) to the latitude and longitude coordinates of that observation point to obtain the estimated wind speed rate value of that observation point in the background field; secondly, The first observation increment is calculated for each observation point: For each observation point, the estimated background wind speed rate for that observation point is subtracted from its corresponding observed wind speed rate. The difference is the first observation increment for that observation point. If the first observation increment is positive, it indicates that the observed value is greater than the estimated background field value, suggesting that the background field may have underestimated the wind speed at that location. If the first observation increment is negative, it indicates that the observed value is less than the estimated background field value, suggesting that the background field may have overestimated the wind speed at that location. Finally, the latitude and longitude coordinates of the observation point corresponding to each first observation increment are recorded to form a correlated dataset of observation point coordinates and first observation increments.

[0026] Step 34: For each grid point, collect all first observation increments within a preset influence radius, and perform a weighted average based on a weighting function to obtain the first analysis increment for that grid point. Specifically, this includes: First, pre-setting the influence radius for calculating the analysis increment: Based on empirical values ​​from marine meteorological data fusion, the influence radius for each grid point is set to 100 kilometers, meaning only the first observation increments corresponding to observation points within a 100-kilometer radius of that grid point are considered; Second, collecting the first observation increments grid by grid point: For each target grid point in the GaussReduceN640 Gaussian grid, based on its latitude and longitude coordinates, in the observation point coordinates-first observation increment association dataset, select all first observation increments whose latitude and longitude coordinates are within the 100-kilometer influence radius of that grid point. First, the local observation increment set for each grid point is formed. If there are no observation points within 100 kilometers of a grid point, the local observation increment set for that grid point is empty, and the subsequent first analysis increment is treated as 0. Second, the weighting function is determined and the weighted average is calculated: a weighting function related to the distance from the observation point to the grid point is used (the closer the distance, the greater the weight, to highlight the correction effect of the neighboring observation data on the grid point), and all the first observation increments in the local observation increment set of each grid point are assigned corresponding weight values. Finally, based on each first observation increment and its corresponding weight value, the weighted average of all the first observation increments of that grid point is calculated. This weighted average is the first analysis increment of that grid point, reflecting the comprehensive correction magnitude of the wind speed rate of that grid point by the surrounding observation data.

[0027] Step 35: Add the first analysis increment to the wind speed rate of the first background field to obtain the updated wind speed rate. Specifically, this includes: establishing the correlation between grid points, background field wind speed rate, and the first analysis increment: matching the background field wind speed rate data of each grid point obtained in step 31 with the first analysis increment of that grid point obtained in step 34, ensuring that the background field data of each grid point matches the corresponding correction increment; secondly, performing update calculation: for each grid point, summing its first analysis increment with the value of the wind speed rate background field to obtain the updated wind speed rate of that grid point (i.e., the initially corrected wind speed); if a grid point has no first observation increment (i.e., the local observation increment set is empty), then its first analysis increment is 0, and the updated wind speed rate is equal to the original wind speed rate background field value; finally, recording the updated wind speed rates of all grid points to form the initially updated wind speed rate grid data, completing the correction of the background field wind speed rate for a single operation.

[0028] Step 36: Using the updated wind speed rate as the new background field, perform iterative correction until the preset convergence condition is met to obtain the corrected wind speed rate. Specifically, this includes: setting the convergence condition for iterative correction: adopting the conventional convergence criteria in the field of marine meteorological data fusion, and presetting two convergence conditions (iteration stops when either condition is met). Condition one is that the average difference in wind speed rate among all grid points in the 'updated wind speed rate grid data' obtained from two adjacent iterations is less than a preset accuracy threshold (e.g., 0.1 m / s, this threshold is set based on the accuracy requirements of CFOSAT satellite spectrometer auxiliary data). Condition two is that the number of iterations reaches a preset maximum number (e.g., 3 times, set based on a balance between computational efficiency and accuracy). Next, start the iteration process: [The text abruptly ends here, likely due to an incomplete translation or a missing section.] The initially updated wind speed rate grid data is used as the new wind speed rate background field. Steps 32 to 35 are repeated, namely, re-extracting the matching wind speed rate observations from the identified observation data, calculating the observation increment for each observation point, calculating the analysis increment for each grid point, and updating the wind speed rate. Next, the iteration is judged and terminated: after each iteration, the wind speed rate grid data obtained in the current iteration is compared with that obtained in the previous iteration to check whether the preset convergence condition is met. If the convergence condition is met, the iteration is stopped, and the updated wind speed rate grid data obtained in the current iteration is determined as the corrected wind speed rate. If the convergence condition is not met, the updated wind speed rate grid data is used as the new background field, and the above iteration operation is repeated until the convergence condition is met.

[0029] Step 37: Combine the corrected wind speed rate with the wind direction component to generate corrected wind field data, specifically including: Two types of core data are acquired: first, the corrected wind speed rate obtained in step 36 (covering all target grid points of the GaussReduceN640 Gaussian grid, reflecting the strength of the corrected wind); second, the wind direction component obtained from step 31 (covering all target grid points of the same Gaussian grid, while keeping the initial calculation results unchanged, reflecting the direction of the wind). Next, data associations at the grid point level are established: for each target grid point, the corrected wind speed rate is bound to the corresponding wind direction component, ensuring a one-to-one correspondence between wind speed and wind direction information for the same grid point. Third, wind field components are vectorized and restored: a conventional inverse transformation method for wind field parameters in marine meteorology is used, combined with vectorization calculation techniques to achieve efficient batch processing. Based on the corrected wind speed rate (denoted as V) and wind direction component (denoted as θ) bound to each grid point, the zonal wind speed component (U component) and meridional wind speed component (V component) are directly derived using vectorization calculation formulas. Specifically, utilizing the mapping relationship between GaussReduceN640 Gaussian grid numbers and latitude and longitude coordinates, the U and V components to be calculated are used as vectorization targets (denoted as ValueP, whose vector dimension is equal to the total number of target grid points), and are quickly calculated using the following vectorization formula: ValueP=V1×lmd1+V2×lmd2+V3×lmd3+V4×lmd4; Where V1, V2, V3, and V4 are the original wind field data near the point to be interpolated (the U and V components of the wind field are known values); lmd1-lmd4 are the vectorization coefficients during interpolation, which are functions of the latitude and longitude coordinates of reference points near ValueP (the target grid point). Their derivation is based on the latitude and longitude distance relationship between the point to be interpolated and the surrounding reference points. ; In the formula, P1 and P2 are the latitude and longitude of the target grid point, and X1, Y1, X2, and Y2 are the latitude and longitude of the surrounding reference points. , Since the GaussReduce N640 Gaussian grid has a spatial distribution pattern, the mapping relationship between grid number and latitude and longitude coordinates can be directly established through formulas. Therefore, the U and V components of all target grid points can be directly calculated in batches through the above vectorized calculation formulas, ensuring that the output wind field data format matches the input requirements of the subsequent wave numerical simulation subsystem (such as the subsystem based on the WAVEWATCHIII model). Moreover, the method has been tested and found to greatly improve data processing efficiency. Finally, the U and V components of all grid points are integrated to form a complete data set with the GaussReduce N640 Gaussian grid as the carrier. This set is the corrected wind field data.

[0030] The iterative correction process for the wind field rate described above can be quantitatively represented by a mathematical formula. The corrected iterative formula is as follows: ; in, Indicates the first After the nth iteration, the th The wind speed rate analysis field of each target grid point (i.e., the updated wind speed rate in step 36) is in m / s; For the first The update result after the next iteration; Indicates the first During the nth iteration, the 1st Each grid point has a preset influence radius (set to 100km in step 34 above) for the total number of valid observation points. If there are no valid observation points, then... At this point, the fractional term in the formula is 0. ; Indicates the first During the nth iteration, the 1st The observation point for the first The weight values ​​of each grid point (conforming to the rule in step 34 that "the closer the distance, the greater the weight", such as calculation based on a Gaussian function or inverse distance ratio); Indicates the first The wind speed rate observation values ​​at each observation point (i.e., the wind speed rate observation values ​​in step 32 based on spatiotemporal matching) are in m / s. Indicates the first In the nth iteration, based on the 1st... Wind speed rate analysis field at each grid point The first one obtained through spatial interpolation The wind speed rate estimate at each observation point (corresponding to the background field estimate at the observation point in step 33) is in m / s. Through the calculation and iterative correction of the observation increment and analysis increment, the advantages of the two types of data are complemented. The final corrected wind field data maintains the spatiotemporal continuity of the grid data and significantly improves the accuracy. The GaussReduceN640 Gaussian grid is used as the data carrier in the correction process.

[0031] In an optional embodiment of the present invention, step 4, generating simulated wave data based on the corrected wind field data, includes: step 41, determining the parameters of the wave numerical simulation model, including setting the source function scheme and optimizing key parameters controlling wave growth, specifically including: determining that the WAVEWATCHIII model is used as the core tool for wave numerical simulation in this step. This model is a mature operational model used for wave simulation in the field of marine meteorology, which can effectively support the simulation of wave generation, propagation and evolution processes at global and regional scales; setting the source function scheme, determining that the source function term of the model adopts the ST4 scheme, which has high simulation accuracy in characterizing the wave energy generation, transfer and dissipation process, and can more accurately reflect the energy change law of waves in the actual marine environment, and is compatible with CFOSAT satellite spectrometer-assisted meteorological data for wave simulation. The requirements for accuracy; optimization of key parameters controlling wave growth, taking into account the wave evolution characteristics of the target country's nearshore short-wind zone (the nearshore area of ​​the target country is relatively narrow, the wind field has a short range, and the wave growth process is unique), combined with on-site observation data of the target country's nearshore area (such as effective wave height and wind speed data observed by large buoys, and wave evolution process data recorded by ocean stations) and regional marine dynamic environment characteristics (such as nearshore water depth distribution, ocean current influence, etc.), the key parameters controlling wave energy generation efficiency, energy transfer rate at different wavenumber bands, and wave energy dissipation intensity in the WAVEWATCHIII model are adjusted and optimized; through multiple simulation experiments and comparative verification with observation data, parameter values ​​suitable for the wave growth law of the target country's nearshore short-wind zone are determined to ensure that the model can accurately capture the entire process characteristics of wave growth from generation to maturity in the target country's nearshore short-wind zone; Step 42: Based on the corrected wind field data and the wave numerical simulation model, perform wave numerical simulation to generate simulated wave data. Specifically, this includes: to balance the simulation of far-field swell generation and propagation with the simulation of nearshore short-wind areas, using an SMC variable-resolution grid as the model's computational grid; this grid has a relatively sparse resolution in the far-field open sea area to ensure simulation efficiency; and a higher resolution in the nearshore area to accurately capture the detailed evolution characteristics of nearshore short-wind areas; simultaneously, the grid configuration references ETOPO and GEBCO mixed topographic data to ensure the grid accurately reflects the actual sea depth distribution, providing a computational basis for wave simulation that conforms to real ocean topography; and to reduce the spin-up time during the model startup phase (i.e., the model reaching a stable state from its initial state). To ensure the spatiotemporal continuity of the wave simulation data, the WAVEWATCHIII mode's hot-start technology is enabled. The wave simulation results generated in the previous phase (such as the previous time) of the wave numerical simulation subsystem are used as the initial field for this simulation, replacing the initial conditions of the traditional cold start. This allows the model to quickly enter a stable simulation state, improving simulation efficiency and ensuring the spatiotemporal continuity of the simulation process. The corrected wind field data generated in step 2 (this data is GaussReduce N640 Gaussian grid data, which includes the zonal and meridional wind speed components of each grid point, combining realism and spatiotemporal continuity) is used as the core driving data of the model and input into the WAVEWATCHIII mode to provide high-precision wind field forcing conditions for the wave simulation.

[0032] Input auxiliary data includes sea surface temperature and sea ice concentration data provided by the operational numerical analysis fields of ECMWF and NCEP. This data characterizes the impact of the ocean surface environment on wave energy dissipation and propagation speed, providing a complete ocean environmental background for model operation. The WAVEWATCHIII model is launched. Based on the configured grid, hot-start initial field, corrected wind field driving data, and auxiliary data, the model simulates the generation, energy transfer, propagation, and dissipation of waves globally (focusing on the coastal areas of the target country) according to the ST4 source function scheme and optimized wind and wave growth parameters. After model operation, simulated wave data using a GaussReduce N640 Gaussian grid is output. This data core includes the significant wave height field (reflecting the strength of the waves) and also covers auxiliary wave parameters such as wave period and wave direction, forming a complete simulated wave dataset. Significant wave height data observed from nearshore buoys ( For example, the effective wave height observations recorded by the State Oceanic Administration's 10-meter disc buoy in the nearshore waters of the target country are compared point-to-point with the effective wave height field in the simulated wave data for verification. The mean absolute error and root mean square error of the simulated values ​​and observed values ​​are calculated. If the accuracy indicators meet the preset requirements (e.g., mean absolute error not greater than 0.2 meters, root mean square error not greater than 0.3 meters), the simulated wave data is determined to be valid data. If the accuracy requirements are not met, the process returns to step 41 to re-optimize the model parameters, or checks the validity of the corrected wind field data, adjusts and re-executes the simulation process until simulated wave data that meets the accuracy requirements is generated. By selecting the high-precision ST4 source function scheme and optimizing key parameters for the wave evolution characteristics of the nearshore short wind zone, the problem of insufficient simulation accuracy of the general model in the nearshore area is effectively solved. The SMC variable resolution grid ensures high-resolution simulation in the nearshore area while reducing the computational load in the far-field region and improving the model's operating efficiency.

[0033] In an optional embodiment of the present invention, step 5, generating corrected wave data based on simulated wave data and identified observation data, includes: Step 51: Using the simulated wave data as the second background field, extract wave observation values ​​from the identified observation data. This specifically includes: confirming the specific source and type of the identified observation data. This data is valid data after automatic quality control and manual intervention correction in Step 1, containing two types of core data: field observation data and satellite remote sensing observation data. Among them, the field observation data comes from the observation records of the target country's nearshore meteorological data buoys (such as 10-meter disc buoys), ocean stations, and observation vessels of the State Oceanic Administration; the satellite remote sensing observation data comes from the orbital observation records of satellite altimeters (used to obtain wave-related elements). Extract core wave observation elements: For the field observation data, extract the effective wave height observation values ​​(effective wave height is the core element of wave numerical simulation and fusion, and is a key indicator of CFOSAT satellite spectrometer-assisted meteorological data) from the effective observation records of each observation platform (10-meter disc buoy, ocean station, observation vessel) at the corresponding observation time and location. Record the observation time (accurate to the hour) and latitude and longitude coordinates of the observation point for each significant wave height observation value; for satellite remote sensing observation data, extract the significant wave height observation values ​​of each observation point within the orbital coverage area from the altimeter-acquired ocean wave observation data, and record their observation time and latitude and longitude coordinates as well; perform spatiotemporal matching screening: using the simulated ocean wave data generated in step 4 as the matching benchmark, conduct spatiotemporal matching verification of the observation values; among them, the time dimension matching must ensure that the observation time of the ocean wave observation value is completely consistent with the output time of the simulated ocean wave data (00:00, 06:00, 12:00, and 18:00 UTC every day); the spatial dimension matching must ensure that the latitude and longitude coordinates of the observation point corresponding to the ocean wave observation value are within the spatial coverage area of ​​the GaussReduceN640 Gaussian grid to which the simulated ocean wave data belongs; filter out the significant wave height observation values ​​that simultaneously meet the spatiotemporal matching conditions to form a related dataset of observation point latitude and longitude - observation time - significant wave height observation value. Step 52: For each observation point, calculate the difference between the observed wave value and the estimated value of the background field at that observation point to obtain the second observation increment. Specifically, this includes: for each effective wave height observation value (including the corresponding latitude and longitude coordinates of the observation point) obtained in Step 51, using the same vectorized bilinear interpolation algorithm as in Step 2, interpolating the simulated wave data (background field, with the core being the effective wave height field of the GaussReduceN640 Gaussian grid) to the latitude and longitude coordinates of the observation point; during the interpolation process, by matching the latitude and longitude mapping relationship between the observation point coordinates and the background field grid, the estimated value of the effective wave height of the observation point in the background field is calculated to ensure that the spatial positioning accuracy of the estimated value is consistent with the accuracy of the background field grid.

[0034] Calculate the second observation increment for a single observation point: For each observation point, subtract the estimated significant wave height of the background field obtained through interpolation from the observed significant wave height of that point. The difference is the wave observation increment for that observation point, i.e., the second observation increment. If the second observation increment is positive, it indicates that the actual observed wave value at that observation point is greater than the simulated value of the background field, suggesting that the background field may underestimate the wave intensity in that area. If the second observation increment is negative, it indicates that the actual observed value is less than the simulated value of the background field, suggesting that the background field may overestimate the wave intensity in that area. Establish the association between the second observation increment and the observation point: Bind the wave observation increment of each observation point to its corresponding latitude and longitude coordinates and observation time to form a related dataset of observation point latitude and longitude - observation time - wave observation increment, ensuring that each observation increment can be traced back to a specific spatial location and time node.

[0035] Step 53: For each grid point, within a preset influence radius, collect all second observation increments and perform a weighted average based on a weighting function to obtain the second analysis increment for that grid point. Specifically, referring to the empirical parameters for calculating the wind field analysis increment in Step 3, and combining the spatial correlation characteristics of wave propagation, the influence radius of each grid point is preset to 100 kilometers; that is, only the wave observation increments corresponding to observation points within a 100-kilometer radius of the grid point are considered. Observation increments outside this range are not included in the calculation because they have little impact on the wave characteristics of the grid point, thus ensuring the spatial relevance of the second analysis increment. Sex; Collect the second observation increment point by grid point: For each target grid point in the GaussReduceN640 Gaussian grid, based on the latitude and longitude coordinates of the grid point, in the observation point latitude and longitude-observation time-wave observation increment association dataset formed in step 52, select all wave observation increments whose latitude and longitude coordinates are within the 100-kilometer influence radius of the grid point, forming a local wave observation increment set for the grid point; if there are no observation points that meet the conditions within 100 kilometers of a certain grid point (i.e., the local wave observation increment set is empty), then the second analysis increment of the grid point is treated as 0.

[0036] Determine the weighting function and calculate the weighted average: Use the inverse distance weighting function (i.e., the closer the observation point is to the target grid point, the greater the contribution weight of the observation increment of the observation point to the analysis increment of the grid point), and assign corresponding weight values ​​to all observation increments in the local wave observation increment set of each grid point; the weight calculation is an inverse distance weight, and the weight is the reciprocal of the distance. In the weight calculation process, the straight-line distance is calculated by the difference between the latitude and longitude of the observation point and the grid point. The smaller the distance, the larger the weight coefficient, and vice versa. Subsequently, each second observation increment is multiplied by its corresponding weight coefficient to obtain a weighted second observation increment. Then, all weighted second observation increments are summed and divided by the sum of the weight coefficients to obtain the wave analysis increment for that grid point. This increment reflects the comprehensive correction magnitude of the wave intensity at the grid point by the surrounding observation data. The second analysis increment is then linked to the grid point: the wave analysis increment (second analysis increment) of each grid point is bound to its corresponding grid point identifier (such as grid number) and latitude and longitude coordinates to form a linked dataset of grid point identifier-latitude and longitude-wave analysis increment, ensuring that the correction increment of each grid point can be accurately matched to the corresponding position in the background field.

[0037] Step 54: Add the second analysis increment to the second background field to obtain updated wave data. Specifically, this includes: acquiring the simulated wave data (background field) generated in Step 4, which uses a GaussReduce N640 Gaussian grid as its carrier, with the core being the simulated effective wave height value of each grid point; matching the simulated effective wave height value of each grid point in the background field with the wave analysis increment of the corresponding grid point in the grid point identifier-latitude-longitude wave analysis increment association dataset formed in Step 53, ensuring that the simulated value of each grid point accurately corresponds to the correction increment and avoiding data mismatch; and performing grid point wave data update: for each grid point, summing the wave analysis increment of that grid point with the simulated effective wave height value of that grid point in the background field to obtain the updated wave data. The effective wave height value after grid point update; if a grid point has no corresponding wave observation increment (i.e., the local wave observation increment set is empty and the analysis increment is 0), then the effective wave height value after grid point update is equal to the original effective wave height simulation value of the background field, maintaining the spatiotemporal continuity of the background field; forming a preliminary updated wave dataset: collect the effective wave height values ​​after grid point update, sort the data according to the standard spatial arrangement rules of GaussReduceN640 Gaussian grid (longitude from west longitude to east longitude, latitude from low latitude to high latitude), forming a preliminary updated wave dataset; this dataset still uses GaussReduceN640 Gaussian grid as the carrier, and the core contains the effective wave height value after grid point update, completing the first correction of the simulated wave data.

[0038] Step 55: Using the updated wave data as the new background field, perform iterative correction until the preset convergence conditions are met, and generate corrected wave data. Specifically, this includes: combining the accuracy requirements of CFOSAT satellite spectrometer-assisted meteorological data and referring to the conventional judgment standards in the field of marine meteorological data fusion, and preset two types of convergence conditions (the iteration will terminate when either condition is met): Condition 1: In the updated wave dataset generated between two consecutive iterations, the average difference in effective wave height of all grid points is less than a preset accuracy threshold (e.g., 0.1 meters). Condition 2: The number of iterations reaches the preset maximum number of iterations (e.g., 3 times, this number is set based on a balance between computational efficiency and accuracy to avoid wasting computational resources due to excessive iteration); Start the iterative correction process: Use the preliminarily updated wave dataset generated in step 54 as the new wave background field, and repeat the complete operation of steps 51 to 54: that is, re-extract the wave observation values ​​that match the spatiotemporal of the new background field from the identified observation data, calculate the wave observation increment for each observation point, calculate the wave analysis increment for each grid point, and update the effective wave height values ​​of the grid points to obtain the wave dataset after the second iteration; Iteration judgment and termination: After each iteration, compare the wave dataset generated in the current iteration with that generated in the previous iteration (the core comparison is the effective wave height values ​​of all grid points): If the average difference in effective wave height between the two datasets is less than the preset accuracy threshold (e.g., 0.1 meters), or the number of iterations has reached the preset maximum number (e.g., 3 times), then stop the iteration and determine the wave dataset generated in the current iteration as the corrected wave data; If any of the above convergence conditions are not met, continue to use the wave dataset generated in the current iteration as the new background field and repeat the operation of steps 51 to 54. The process continues until the convergence condition is met; the validity verification of the corrected wave data is performed: the effective wave height observation data of the target country's nearshore buoys (such as 10-meter disc buoys) is compared point-to-point with the effective wave height values ​​of the corresponding grid points in the corrected wave data; the mean absolute error, root mean square error, and dispersion index are calculated. If the accuracy index reaches the preset level (such as mean absolute error ≤ 0.18 meters, root mean square error ≤ 0.25 meters), the corrected wave data is confirmed to be valid; if the accuracy requirement is not met, the process returns to step 51 to re-check the extraction of observation data or... The interpolation calculation process involves adjusting and re-executing the iterative process until valid data is generated. The fused observation data (which has high fidelity) is effectively corrected through the calculation and iterative correction of observation increments and analysis increments, absorbing the advantages of both types of data. This process retains the spatiotemporal continuity of numerical simulation data while incorporating real wave information from field / remote sensing observation data. The final corrected wave data has significantly improved accuracy, effectively solving the problem of insufficient accuracy of general fusion methods in nearshore short-wind areas and complex topographic sea areas. This makes the corrected wave data more suitable for the nearshore marine environmental characteristics of the target country.

[0039] The following specific examples illustrate the detailed implementation process of all steps in the marine meteorological data generation method described in this invention: Step 1 aims to acquire identification observation data and numerical simulation data from marine meteorological observations, providing a foundational data source for subsequent fusion. Specifically, this includes: Step 11: Obtain field data from different observation platforms, including large buoys (such as 10-meter disc buoys), ocean stations, and observation vessels, from the operational network and observation data transmission system of the National Marine Environmental Forecasting Center. Extract core elements such as wind speed, wind direction, and significant wave height at a height of 10 meters from the aforementioned field observation data. Then, acquire remote sensing data from CFOSAT satellites and other satellites via internal network downloads. Satellite remote sensing observations are categorized based on sensor type, such as scatterometers and altimeters. Ensure that the acquired remote sensing data covers observational information of sea surface winds and waves along satellite orbits within the target area; acquire operational analysis wind field data (spatial resolution 0.25°×0.25°) produced by the National Marine Environmental Forecasting Center, which serves as the primary foundational data for wind field numerical simulation; acquire GFS wind field data (resolution 0.25°×0.25°) from NCEP (National Centers for Environmental Forecasting), as a backup for the primary wind field data to ensure its availability; acquire wind field data (resolution 0.125°×0.125°) generated by the China Meteorological Administration's GRAPES system; and acquire wave numerical post-reports. The effective wave height field data (resolution 0.1°×0.1°) generated by the subsystem (based on the WAVEWATCHIII model, using the ST4 source function scheme, optimizing key parameters for wind and wave growth, and employing SMC variable resolution grids and hot-start technology) serves as the core analysis field data for ocean wave numerical simulation. Template files from ECMWF (European Centre for Medium-Range Weather Forecasts) are also acquired. Simultaneously, sea surface temperature data and global sea ice data generated by the National Marine Environmental Forecasting Center's high-resolution global ocean dynamic environment numerical prediction system (based on the 1 / 12° global NEMO model and employing ensemble filtering assimilation techniques) are also acquired.

[0040] Step 12, obtaining the results of automatic quality control: Specifically, buoy station movement control: Based on the normal drift range of the buoy, a preset control radius of 5km is established; by real-time monitoring of the distance between the buoy's current coordinates and initial coordinates, it is determined whether the buoy exceeds the normal drift range. If it does, the observation data corresponding to the buoy station movement is automatically marked as problematic data. This strategy can completely identify this type of problematic data; land information of the observation area is pre-stored; by performing mask calculations between the buoy's real-time coordinates and the pre-stored land information, it is determined whether the buoy is in a shore maintenance state (if the coordinates fall within the land area, it is determined to be in a maintenance state); if it is determined to be in a maintenance state, the observation data sent by the buoy in this state is automatically marked as useless information-type problematic data. This strategy can completely identify this type of problematic data; the integrity of the acquired observation data file is verified by checking the file's checksum, format identifier, etc., to determine whether the file is corrupted. If corrupted, the observation data corresponding to that file is automatically marked as problematic data; the elements within the data file are traversed and checked. The system checks for missing key elements such as wind speed and significant wave height. If any elements are missing, the corresponding observation data for that file is automatically marked as problematic data. This strategy can fully identify such problematic data. It compares the Q&A results obtained during the data download process (such as file information returned by the server) with the pre-stored file list information (including the expected server address, file path, and filename). If inconsistencies are found (such as the actual server address not matching the pre-stored address, or the actual filename differing from the pre-stored filename), the observation data in this case is automatically marked as problematic data, and only an error message is displayed; the data is not directly deleted. The system also compares the values ​​of elements such as wind speed and significant wave height in the observation data with the corresponding values ​​in the same period's numerical simulation data (such as the significant wave height field of the numerical analysis wind field and the numerical wave reporting subsystem). A weighted average method is used to determine whether the observation data is an extreme large or small value. If it is determined to be an outlier, the observation data is automatically marked as problematic data; however, this strategy can only partially identify such problematic data.

[0041] Step 13: For the problematic data marked in Step 12 due to changes in data server, file path, or filename, the forecaster, based on the actual situation, enters the new server address, file path, filename, and other naming rules in the manual intervention window, reconfigures the data transmission parameters, and ensures that the system can correctly acquire and parse this part of the observation data; for the abnormally large / small values ​​marked in Step 12, the forecaster makes a comprehensive judgment based on other hydrological and meteorological data of the same period (such as surrounding buoy observation data, satellite remote sensing data, and numerical simulation results). If the data is confirmed to be abnormal, the abnormal data is masked through the manual intervention window, and the observation values ​​of the time period and area where the abnormal data is located are supplemented using reasonable interpolation or averaging methods; for the sensor zero-point drift problem not identified in Step 12 ( Without an automatic quality control strategy, forecasters monitor the long-term trend of observation data. If a regular zero-point shift is found in the data within a short period, the observation data for that period is marked as abnormal and deleted. If missing reports are found in step 12 due to sensor damage, forecasters mark the missing report period through a manual intervention window and delete the abnormal data. The automatic quality control results obtained in step 12 (including the marking information of various types of problematic data, and the results of complete and partial identification) are integrated with the post-intervention observation data obtained after manual intervention correction in step 13 (including the effective data obtained after reconfiguring transmission parameters, the replacement data generated by supplementary calculation, and the effective data retained after masking anomalies) to form the final identified observation data, providing a high-quality data foundation for subsequent fusion.

[0042] Step 2: This step uses vectorized bilinear interpolation to interpolate the numerical simulation data onto a preset GaussReduce N640 Gaussian grid to obtain the interpolated wind field data. Specifically, this includes: Step 21: Obtain the target grid and numerical simulation grid parameters: The preset target grid is the GaussReduce N640 Gaussian grid, which has the structural characteristics of equidistant longitude distribution, non-equidistant (quasi-equidistant) latitude distribution, and no grid points at the poles, totaling 2,140,702 grid points; by calling the pre-stored N640 Gaussian grid latitude and longitude configuration file, extract the unique identifier (such as grid number) of each target grid point in the grid and its corresponding longitude and latitude values, and establish a one-to-one correspondence between the unique identifier of the target grid point, the longitude value, and the latitude value; obtain the structural parameters of the grid to which the numerical simulation wind field data to be interpolated (including the operational analysis wind field of the National Marine Environmental Forecasting Center, the GFS wind field of NCEP, or the GRAPES system wind field) belongs, including the latitudinal step size, the longitudinal step size, and the overall longitude range (from the minimum longitude value to the maximum longitude value) and latitude range (from the minimum latitude value to the maximum latitude value) of the grid; at the same time, extract the numerical simulation... The longitude and latitude values ​​of each grid point in the simulated grid, along with their corresponding grid indices, form a related dataset of numerical simulation grid point index-longitude-latitude values. The process involves matching each target grid point with its four surrounding grid points: For each target grid point, its location within the numerical simulation grid is determined based on its latitude and longitude coordinates: Longitude direction: The longitude value of the target grid point is compared with the longitude sequence of the numerical simulation grid to determine which two adjacent longitude values ​​it lies between, thus locking the indices and coordinates of the corresponding two numerical simulation grid points; Latitude direction: The latitude value of the target grid point is compared with the latitude sequence of the numerical simulation grid to determine which two adjacent latitude values ​​it lies between, thus locking the indices and coordinates of the corresponding two numerical simulation grid points; Combining to form the four surrounding grid points: The two grid points locked in the longitude and latitude directions are combined to obtain the four surrounding numerical simulation grid points corresponding to the target grid point. Their indices and coordinates are recorded and associated with the unique identifier of the target grid point, until all 2,140,702 target grid points are matched.

[0043] Step 22: Based on the correlation established in Step 21, for each target grid point, extract its own longitude and latitude values, as well as the longitude and latitude values ​​of the four surrounding numerical simulation grid points; label the four surrounding grid points as the first grid point (X1, Y1), the second grid point (X2, Y1), the third grid point (X2, Y2), and the fourth grid point (X1, Y2), where X represents longitude and Y represents latitude, and X2 > X1 and Y2 > Y1; calculate the distance ratios: Meridional distance ratio (Xscale): calculate the difference between the longitude value of the target grid point (P1) and the longitude value of the first grid point (X1), divide it by the difference between the longitude value of the second grid point (X2) and the longitude value of the first grid point (X1), reflecting the positional proportion of the target grid point in the longitude direction; Latitude distance ratio (Yscale): calculate the difference between the latitude value of the target grid point (P2) and the latitude value of the first grid point (X1, Y2). The difference between the latitude values ​​of the fourth grid point (Y1) and the first grid point (Y1) is divided by the difference between the latitude values ​​of the fourth grid point (Y2) and the first grid point (Y1) to reflect the positional proportion of the target grid point in the latitudinal direction. The interpolation coefficients are derived: based on Xscale and Yscale, the interpolation coefficients of the four surrounding grid points are derived using bilinear interpolation logic: First grid point coefficient: (1-Xscale) × (1-Yscale); Second grid point coefficient: Xscale × (1-Yscale); Third grid point coefficient: Xscale × Yscale; Fourth grid point coefficient: (1-Xscale) × Yscale. Vectorized processing is used to synchronously calculate the interpolation coefficients of all target grid points, avoiding the inefficiency of point-by-point iteration. The four interpolation coefficients of each target grid point are associated with its unique identifier and stored to form a target grid point unique identifier - four interpolation coefficients dataset.

[0044] Step 23: For each target grid point, based on the indices of the four surrounding numerical simulation grid points recorded in Step 21, extract the zonal wind speed components (U component) and meridional wind speed components (V component) of these four grid points from the numerical simulation wind field dataset, ensuring that the data is accurately associated with the grid point index and the unique identifier of the target grid point; multiply the U component data of each surrounding grid point by its corresponding interpolation coefficient to obtain the weighted value of the U component; sum the four weighted values ​​to obtain the U component interpolation result of the target grid point; using the same logic as the U component, multiply the V component data of each surrounding grid point by its corresponding interpolation coefficient, and sum them to obtain the V component interpolation result of the target grid point; compare the U component and V component interpolation results of the target grid point with the normal numerical range of the marine meteorological wind field data (excluding obvious outliers); if the result is reasonable, associate and store it with the unique identifier of the target grid point; if it is abnormal, return to Step 21 to rematch the surrounding grid points or recalculate the coefficients in Step 22 until the result is reasonable.

[0045] Step 24: Improve the mapping relationship between the unique identifier of the target grid point and the interpolated wind field data (U component, V component), ensuring that each unique identifier corresponds to only one set of interpolation results and can be quickly queried through the unique identifier; sort the associated dataset according to the standard arrangement rules of GaussReduce N640 Gaussian grid (longitude from west to east longitude, latitude from low latitude to high latitude) to ensure the spatial continuity of the data (adjacent grid point data remain spatially adjacent in the dataset); count the number of unique identifiers of the target grid points in the statistical dataset and compare it with the total number of N640 grid points (2,140,702). If any are missing, return to the previous steps to recalculate; if any are duplicated, remove the duplicate data; batch verify whether the U component and V component of all grid points are within a reasonable range, and check whether there are abrupt changes in adjacent grid point data (to exclude spatial consistency issues); if all pass, the sorted dataset is the interpolated wind field data; if it fails, repeat the verification process until the data is complete and valid.

[0046] Step 3, generating corrected wind field data, specifically includes: Step 31, obtaining interpolated wind field data: extracting the interpolated wind field data generated in Step 2 (GaussReduce N640 grid, containing U and V components for each grid point); using conventional marine meteorological wind field parameter conversion methods, calculating wind speed (wind strength) based on the U and V components of each grid point; determining the wind direction component (wind direction) based on the same U and V components using meteorological data direction calculation rules; subsequent correction only applies to wind speed, while the wind direction component retains its initial calculation result, serving as the basic element for subsequent data integration, completing the decomposition of interpolated wind field data → wind speed + wind direction component.

[0047] Step 32: The identified observation data includes the field observation data processed in Step 1 and satellite remote sensing data (sea surface wind speed data from the scatterometer); from the observation records of each observation platform, extract the 10-meter height wind speed observation value (wind speed rate observation value) at the corresponding observation time and location, and record the observation time and latitude / longitude coordinates; from the scatterometer's orbital wind speed data, extract the wind speed observation value (wind speed rate observation value) at the observation point within the orbital coverage area at the corresponding observation time, and record the observation time and latitude / longitude coordinates; ensure that the observation time of the wind speed rate observation value is consistent with the time dimension of the interpolated wind field data (00:00, 06:00, 12:00, and 18:00 UTC daily), and that the latitude / longitude of the observation location is within the GaussReduceN640 grid spatial coverage area; select matching observation values ​​as the basis for subsequent calculations.

[0048] Step 33: For each spatiotemporally matched wind speed rate observation (including the latitude and longitude of the observation point), the same bilinear interpolation method as in Step 2 is used to interpolate the wind speed rate background field (interpolated wind speed rate grid) obtained in Step 31 to the coordinates of the observation point to obtain the estimated background field value; the difference between the wind speed rate observation value of the observation point and the estimated background field value is the observation increment of the observation point; if the increment is positive, it indicates that the background field underestimates the wind speed; if it is negative, it indicates that the background field overestimates the wind speed; record the latitude and longitude coordinates of the observation point corresponding to each observation increment to form the observation point coordinates-observation increment dataset.

[0049] Step 34: Based on experience in marine meteorological data fusion, the influence radius of each grid point is set to 100 kilometers (considering only the observation increments within a 100-kilometer radius). For each target grid point, based on its latitude and longitude coordinates, all observation increments within the 100-kilometer influence radius are selected from the observation point coordinates-observation increment dataset to form a local observation increment set for that grid point. If there is no observation point, the set is empty, and subsequent analysis increments are treated as 0. A distance-inverse weighting function (the closer the observation point is to the grid point, the greater the weight) is used to assign weights to the increments in the local observation increment set. The average value of all weighted increments is calculated, which is the analysis increment for that grid point (reflecting the comprehensive correction magnitude of the surrounding observation data).

[0050] Step 35: Match the background field data of wind speed rate for each grid point obtained in Step 31 with the analysis increment for that grid point obtained in Step 34 to ensure a correct match; perform update calculation: updated wind speed rate for each grid point = background field value of wind speed rate + analysis increment; if the analysis increment is 0 (no observation increment), the updated value is equal to the original background field value; record the updated wind speed rate for all grid points to form the preliminary updated wind speed rate grid data.

[0051] Step 36, set convergence conditions: Condition 1: The average difference in wind speed rate among all grid points in the updated wind speed rate grid data of two adjacent iterations is <0.1 m / s (set based on the accuracy requirements of CFOSAT auxiliary data); Condition 2: The number of iterations reaches 3 (balancing computational efficiency and accuracy); The iteration can be stopped when either condition is met; The initially updated wind speed rate grid data is used as the new background field, and steps 32-35 (extracting matching observations, calculating observation increments, calculating analysis increments, and updating wind speed rates) are repeated to obtain the updated data for the next iteration; After each iteration, the current wind speed rate grid data is compared with the previous iteration. If the convergence condition is met, the current data is the corrected wind speed rate; if not, the iteration continues until convergence.

[0052] Step 37: Extract the corrected wind speed rate (covering all target grid points) from Step 36 and the wind direction component from Step 31 (keeping the initial value); for each target grid point, bind the corrected wind speed rate with the corresponding wind direction component to ensure a one-to-one correspondence between wind speed and wind direction; using the conventional marine meteorological wind field parameter inverse transformation method, based on the bound wind speed rate and wind direction components, recalculate the U and V components of each grid point to ensure that the format matches the input requirements of the subsequent wave numerical simulation subsystem; integrate the U and V components of all grid points to form the corrected wind field data of the GaussReduceN640 grid carrier.

[0053] Step 4, generating simulated ocean wave data, specifically includes: Step 41, using the WAVEWATCHIII model as the core tool for ocean wave numerical simulation (supporting global and regional scale ocean wave evolution simulation); the model source function term adopts the ST4 scheme, which has high accuracy in characterizing wave energy generation, transfer, and dissipation, accurately reflecting actual ocean wave energy changes, and adapting to the accuracy requirements of CFOSAT auxiliary data; considering the wave evolution characteristics of the nearshore short wind zone in Area A (narrow sea area and short wind field influence distance), combined with the nearshore field observation data of Area A (significant wave height and wind speed data from buoys and ocean stations) and the regional ocean dynamic environment characteristics (water depth distribution and ocean current influence), the key parameters controlling wind and wave energy generation efficiency, energy transfer rate, and energy dissipation intensity in the WAVEWATCHIII model are adjusted and optimized; through multiple simulation experiments and comparative verification with observation data, the parameter values ​​suitable for the wind and wave growth law of the nearshore short wind zone in Area A are determined to ensure that the model accurately captures the entire process of wind and wave growth from generation to maturity in the short wind zone.

[0054] Step 42: Use SMC variable resolution grid and refine it near the shore; use ETOPO+GEBCO hybrid topographic data for water depth data to ensure accurate reflection of the actual sea depth distribution and provide a real topographic basis for simulation; use the simulation results of the previous time step of the wave numerical simulation subsystem as the initial field for this simulation (replacing the traditional cold start) to ensure the spatiotemporal continuity of simulation data and improve simulation efficiency; input the corrected wind field data generated in step 3 to provide high-precision wind field forcing conditions for wave simulation.

[0055] The WAVEWATCHIII mode is activated, and based on the configured grid, hot-start initial field, driving data, and auxiliary data, the generation, energy transfer, propagation, and dissipation of ocean waves on a global scale are simulated according to the ST4 source function scheme and optimized parameters. The global ocean wave significant wave height background field data are obtained by calculating through mode integration.

[0056] Step 5: Generate corrected wave data, specifically including: Step 51: Identify the observation data, which includes the field observation data processed in Step 1 (significant wave height records from nearshore buoys, ocean stations, and observation vessels) and satellite remote sensing data (significant wave height records along the orbit from satellite altimeters); extract core wave observation elements: extract the significant wave height observation values ​​(core wave fusion elements) corresponding to the observation time and location from the effective records of each observation platform, and record the observation time (accurate to the hour) and latitude and longitude coordinates; extract the significant wave height observation values ​​of observation points within the orbit coverage area corresponding to the observation time from the altimeter wave data along the orbit, and record the observation time and latitude and longitude coordinates; using the simulated wave data generated in Step 4 as a benchmark, ensure that the observation time of the wave observation values ​​is consistent with the output time of the simulated data (00:00, 06:00, 12:00, and 18:00 UTC daily), and that the latitude and longitude of the observation points are within the GaussReduceN640 grid coverage area of ​​the simulated data; filter out the matching significant wave height observation values ​​to form a dataset of observation point latitude and longitude - observation time - significant wave height observation values.

[0057] Step 52: For each matched valid wave height observation (including the latitude and longitude of the observation point), the same vectorized bilinear interpolation algorithm as in Step 2 is used to interpolate the simulated wave data (background field, with the core being the valid wave height field) to the coordinates of the observation point, obtaining the estimated value of the valid wave height of the background field (ensuring that the spatial positioning accuracy is consistent with the background field grid); the difference between the observed value of the valid wave height of the observation point and the estimated value of the background field is the wave observation increment of that observation point; if the increment is positive, it indicates that the background field underestimates the wave intensity; if it is negative, it indicates that the background field overestimates the wave intensity; the wave observation increment is bound to the latitude and longitude of the observation point and the observation time to form the observation point latitude and longitude-observation time-wave observation increment dataset, ensuring that the increment can be traced back to the specific spatiotemporal location.

[0058] Step 53: Referring to the experience gained from wind field analysis increments in Step 3, and considering the spatial correlation of wave propagation, the influence radius of each grid point is preset to 100 kilometers (considering only wave observation increments within a 100-kilometer radius). For each target grid point, based on its latitude and longitude coordinates, all wave observation increments within the 100-kilometer influence radius are selected from the observation point latitude / longitude-observation time-wave observation increment dataset, forming a local wave observation increment set for that grid point. If there is no observation point, the set is empty, and the analysis increment is treated as 0. A distance-inverse weighting function is used (the closer the observation point is to the grid point, the greater the weight), assigning a weight to each increment in the local increment set (the smaller the distance, the larger the weight coefficient). Each increment is multiplied by its corresponding weight coefficient, summed, and divided by the sum of the weight coefficients to obtain the wave analysis increment for that grid point (reflecting the comprehensive correction magnitude of the surrounding observation data to the wave intensity of the grid point). The wave analysis increment is bound to the grid point identifier (such as number) and latitude / longitude to form a grid point identifier-latitude / longitude-wave analysis increment dataset, ensuring that the correction increment accurately matches the background field location.

[0059] Step 54: Obtain the simulated wave data (background field, including the simulated effective wave height value of each grid point) generated in Step 4. Match the simulated effective wave height value of each grid point with the wave analysis increment in Step 53 to avoid data mismatch. The updated effective wave height value of each grid point = simulated effective wave height value of the background field + wave analysis increment. If the analysis increment is 0 (no observation increment), the updated value is equal to the original simulated value to maintain the spatiotemporal continuity of the background field. Collect the updated effective wave height values ​​of all grid points and sort them according to the GaussReduceN640 grid standard arrangement rules (longitude from west to east, latitude from low latitude to high latitude) to form the preliminary updated wave dataset.

[0060] Step 55, set convergence conditions: Condition 1: The average difference in effective wave height among all grid points in the updated wave dataset after two consecutive iterations is <0.1m; Condition 2: The number of iterations reaches 3; The iteration can be stopped when either condition is met; Start the iteration process: Use the initially updated wave dataset as the new background field, repeat steps 51-54 (extract matching observations, calculate observation increments, calculate analysis increments and update effective wave heights) to obtain the wave dataset for the next iteration.

[0061] After each iteration, compare the current wave dataset with the previous iteration (core comparison: significant wave height). If the convergence condition is met, the current dataset is the corrected wave data; otherwise, continue iterating until convergence. Use the significant wave height observation data of nearshore buoys in area A and compare it point-to-point with the significant wave height values ​​of the corresponding grid points in the corrected wave data. Calculate the mean absolute error, root mean square error, and dispersion index (requirements: mean absolute error ≤ 0.18m, root mean square error ≤ 0.25m, comparable to ERA5 accuracy). If these conditions are met, the data is valid; otherwise, return to step 51 to re-examine the observation data extraction or interpolation process, adjust, and iterate again until the data is valid.

[0062] Step 6: Using the ECcodes program set developed by ECMWF, the corrected wind field data (including U and V components) generated in Step 3 is encoded into a GRIB2 format meteorological parameter file, and the corrected ocean wave data (including significant wave height, wave period, and wave direction) generated in Step 5 is encoded into a GRIB2 format ocean wave parameter file. Ensure that the grid structure of both files is GaussReduceN640, and the data time is 00:00, 06:00, 12:00, and 18:00 (UTC) daily. Finally, a product that meets the requirements of CFOSAT satellite spectrometer-assisted meteorological data is formed and delivered for operational applications.

[0063] The marine meteorological data generation method described in this invention uses a progressive correction based on wind field and ocean waves, fusing observational data (high realism) with numerical simulation data (good spatiotemporal continuity). The final product has an average absolute error of ≤0.18m for significant wave height and ≤0.15m for wind speed, achieving both high precision and high spatiotemporal consistency, significantly improving reliability. The method optimizes model parameters for the wave evolution characteristics of the nearshore short-wind zone in Area A, combining them with the SMC variable-resolution grid to effectively improve the simulation accuracy of nearshore waves in Area A, overcoming the accuracy limitations of general models in short-wind zones and complex topographical sea areas. 80% of the data uses domestically sourced data (GRAPES wind field, NEMO sea surface temperature, and nearshore buoy observations in Area A). The interpolation and fusion algorithms are independently developed, forming an independent and controllable auxiliary data production system, reducing reliance on external data such as ECMWF and NCEP.

[0064] like Figure 2 As shown, embodiments of the present invention also provide a marine meteorological data generation apparatus 20, comprising: The acquisition module 201 is used to acquire identification observation data and numerical simulation data of marine meteorological observation data; The processing module 202 is used to interpolate the numerical simulation data onto a preset target grid to generate interpolated wind field data; generate corrected wind field data based on the interpolated wind field data and the identified observation data; generate simulated ocean wave data based on the corrected wind field data; generate corrected ocean wave data based on the simulated ocean wave data and the identified observation data; and encode the corrected wind field data and the corrected ocean wave data into target marine meteorological data.

[0065] It should be noted that this device is the same as the method described above. All implementations of the method described above are applicable to the embodiments of this device and can achieve the same technical effect.

[0066] Embodiments of the present invention also provide a computing device, including: a processor and a memory storing a computer program, wherein the computer program, when executed by the processor, performs the above-described method. All implementations in the above method embodiments are applicable to this embodiment and can achieve the same technical effects.

[0067] Embodiments of the present invention also provide a computer-readable storage medium, comprising: stored instructions, which, when executed on a computer, cause the computer to perform the above-described method. All implementations in the above method embodiments are applicable to this embodiment and can achieve the same technical effects.

[0068] The above description represents the preferred embodiments of the present invention. It should be noted that those skilled in the art can make various improvements and modifications without departing from the principles of the present invention, and these improvements and modifications should also be considered within the scope of protection of the present invention.

Claims

1. A method for generating marine meteorological data, characterized in that, include: Acquire marine meteorological observation data and numerical simulation data; The numerical simulation data is interpolated onto a preset target grid to generate interpolated wind field data, including: Based on the latitude and longitude coordinates of the preset target grid, determine the four surrounding grid points of each target grid point in the numerical simulation data grid; Based on the latitude and longitude distance relationship between the target grid point and the four surrounding grid points, the interpolation coefficients required for bilinear interpolation are generated. The wind field data of the four surrounding grid points in the numerical simulation data are weighted and fused using the interpolation coefficients to obtain the interpolated wind field data of each target grid point. The interpolated wind field data of all target grid points are combined to form the interpolated wind field data. Based on the interpolated wind field data and the identified observation data, corrected wind field data is generated, including: The interpolated wind field data is decomposed into wind speed and wind direction components; Using the wind speed rate as the first background field, wind speed rate observation values ​​are extracted from the identification observation data; For each observation point, the difference between the observed wind speed rate and the estimated value of the background field at that observation point is calculated to obtain the first observation increment; For each grid point, within a preset radius of influence, all first observation increments are collected and weighted averaged based on a weighting function to obtain the first analytical increment for that grid point. The first analysis increment is added to the wind speed rate of the first background field to obtain the updated wind speed rate; The updated wind speed rate is used as the new background field, and iterative correction is performed until the preset convergence condition is met to obtain the corrected wind speed rate. The corrected wind speed rate is combined with the wind direction component to generate corrected wind field data; Based on the corrected wind field data, simulated ocean wave data is generated; Based on simulated wave data and identified observation data, corrected wave data is generated, including: Using the simulated wave data as a second background field, wave observation values ​​are extracted from the identification observation data; For each observation point, the difference between the observed wave value and the estimated value of the background field at that observation point is calculated to obtain the second observation increment; For each grid point, within a preset influence radius, all second observation increments are collected and weighted averaged based on a weighting function to obtain the second analysis increment for that grid point; The second analytical increment is added to the second background field to obtain the updated wave data; The updated wave data is used as the new background field for iterative correction until the preset convergence condition is met, and the corrected wave data is generated. The corrected wind field data and corrected wave data are encoded into target marine meteorological data.

2. The method for generating marine meteorological data according to claim 1, characterized in that, The marine meteorological observation data includes field observation data and satellite remote sensing data, wherein the field observation data includes at least wind speed and significant wave height, and the numerical simulation data includes at least wind field data and ocean wave data.

3. The method for generating marine meteorological data according to claim 1, characterized in that, Identification observation data for acquiring marine meteorological observation data includes: Acquire marine meteorological observation data from multiple sources; Automatic quality control is performed on the marine meteorological observation data to identify and label problematic data, and the results of automatic quality control are obtained. Based on the results of the automatic quality control, the problematic data is corrected to obtain the post-intervention observation data. The results of automated quality control and the observation data after intervention are used as the observation data for identification.

4. The method for generating marine meteorological data according to claim 1, characterized in that, Based on the corrected wind field data, simulated ocean wave data is generated, including: Determine the parameters of the wave numerical simulation model, including setting the source function scheme and optimizing the key parameters for controlling wave growth; Based on the corrected wind field data and the wave numerical simulation model, wave numerical simulation is performed to generate simulated wave data.

5. A device for generating marine meteorological data, characterized in that, include: The acquisition module is used to acquire identification observation data and numerical simulation data from marine meteorological observations. The processing module is used to interpolate the numerical simulation data onto a preset target grid to generate interpolated wind field data. This includes: determining the four surrounding grid points of each target grid point in the numerical simulation data grid based on the latitude and longitude coordinates of the preset target grid; generating interpolation coefficients required for bilinear interpolation based on the latitude and longitude distance relationship between the target grid point and the four surrounding grid points; weighting and fusing the wind field data of the four surrounding grid points in the numerical simulation data using the interpolation coefficients to obtain the interpolated wind field data for each target grid point; and combining the interpolated wind field data of all target grid points to form the... Interpolated wind field data; based on the interpolated wind field data and identified observation data, corrected wind field data is generated, including: decomposing the interpolated wind field data into wind speed rate and wind direction components; using the wind speed rate as a first background field, extracting wind speed rate observation values ​​from the identified observation data; for each observation point, calculating the difference between the wind speed rate observation value and the estimated value of the background field at that observation point to obtain a first observation increment; for each grid point, collecting all first observation increments within a preset influence radius, and performing a weighted average based on a weighting function to obtain the first analysis increment of that grid point; and then... The first analytical increment is added to the wind speed rate of the first background field to obtain an updated wind speed rate; the updated wind speed rate is used as a new background field, and iterative correction is performed until a preset convergence condition is met to obtain a corrected wind speed rate; the corrected wind speed rate is combined with the wind direction component to generate corrected wind field data; simulated wave data is generated based on the corrected wind field data; corrected wave data is generated based on the simulated wave data and the identified observation data, including: using the simulated wave data as a second background field, extracting wave observation values ​​from the identified observation data; for each At each observation point, the difference between the observed wave value and the estimated value of the background field at that observation point is calculated to obtain the second observation increment. For each grid point, within a preset influence radius, all second observation increments are collected and weighted averaged based on a weighting function to obtain the second analysis increment for that grid point. The second analysis increment is added to the second background field to obtain updated wave data. The updated wave data is used as the new background field for iterative correction until a preset convergence condition is met, generating corrected wave data. The corrected wind field data and the corrected wave data are encoded into target marine meteorological data.

6. A computing device, characterized in that, include: A processor, a memory storing a computer program, wherein the computer program, when executed by the processor, performs the method as described in any one of claims 1 to 4.

7. A computer-readable storage medium, characterized in that, A storage instruction that, when executed on a computer, causes the computer to perform the method as described in any one of claims 1 to 4.

Images