Manufacturing method of high-precision high-altitude wind field reanalysis data set

Abstract

The invention relates to a manufacturing method of a high-precision high-altitude wind field reanalysis data set, and belongs to the technical field of meteorological high-altitude wind field fusion. According to the method, standard atmospheric pressure layer division with the interval of 250 meters of sounding second data and a standard atmospheric model, background field wind field processing based on ERA5, ERA5 and high-altitude detection wind field fusion adopting an optimal interpolation method are integrated, and a high-altitude wind field reanalysis data set is manufactured. According to the method, ERA5 reanalysis data and national high-altitude wind field detection data are fused, a high-precision high-altitude wind reanalysis data set of 80 layers below 20000 meters is made, and the data set can more finely and accurately reflect the wind field conditions of China and surrounding areas thereof.

Description

Technical Field

[0001] This invention belongs to the field of meteorological upper-air wind field fusion technology, specifically involving a method for creating a high-precision upper-air wind field reanalysis dataset. Background Technology

[0002] With the continued and in-depth application of extreme weather and climate analysis and artificial intelligence forecasting models, related industries have increasingly higher requirements for reanalysis data. Atmospheric reanalysis data refers to gridded data generated by combining numerical forecasting, numerical assimilation, various meteorological observations, and satellite radar remote sensing data. It generally includes atmospheric reanalysis products and land surface reanalysis products. Common examples include European reanalysis data and China Meteorological Administration reanalysis data.

[0003] In aerospace flight meteorological support, upper-level wind support is a crucial element. For upper-level wind forecasting and support review, the primary data sources for analysis are numerical models and reanalysis data. Currently, the numerical models and reanalysis data used in operational research mainly serve weather and climate forecasting, and their vertical resolution is insufficient for detailed upper-level wind forecasting and review. For example, the highest-quality European Centre for Medium-Range Reanalysis (ERA5) reanalysis data only has 29 layers from 1000 hPa to 50 hPa. Furthermore, the lack of high-resolution wind fields in the construction of AI-based upper-level wind forecasting models also hinders the development of intelligent models.

[0004] Research on high-precision upper-altitude wind field reanalysis datasets and related data fusion is relatively limited, mainly due to low demand and a small number of service industries involved. In recent years, with the increasingly sophisticated wind field requirements of aerospace flights, it is essential to conduct specialized research on wind field fusion technologies. Utilizing existing atmospheric reanalysis data and upper-altitude sounding data, wind field fusion can be performed through methods such as optimal interpolation, providing a data foundation for wind field analysis and intelligent modeling. Summary of the Invention

[0005] (a) Technical problems to be solved

[0006] The technical problem to be solved by this invention is how to provide a method for creating a high-precision upper-level wind field reanalysis dataset, so as to solve the problem of the lack of high vertical resolution wind fields in the construction of artificial intelligence forecasting models for upper-level winds.

[0007] (II) Technical Solution

[0008] To address the aforementioned technical problems, this invention proposes a method for creating a high-precision upper-level wind field reanalysis dataset, which includes the following steps:

[0009] S1, Standard pressure layer division with 250-meter intervals

[0010] The radiosonde data from upper-air sounding stations broadcast via satellite by the China Meteorological Administration were processed to obtain the observed wind field data of the isobaric layers at each station. Further processing yielded the second-by-second data for geopotential height and pressure at each station. A moving average was applied to the second-by-second data, and the averaged data was interpolated into geopotential height layer sequences and corresponding pressure layer sequences at 250-meter intervals. A horizontal average was then applied to the pressure layer sequences of each station within a specified range to obtain a pressure layer sequence ranging from 0 to 15000 meters, spaced 250 meters apart. Combined with the SA1976 model barosphere sequence used above 15,000 meters Subsequently, a sequence of pressure layers ranging from 0 to 20,000 meters is generated. ;

[0011] S2, Background Field Wind Field Processing

[0012] Read ERA5 reanalysis data to obtain , The wind field, totaling 29 layers, was used to obtain the pressure layer sequence based on wind field data from ERA5 reanalysis. The gridded background wind field data is used to obtain the background wind field corresponding to each upper-air sounding station, and the deviation between the observation and the background wind field data of each upper-air sounding station is calculated.

[0013] S3, Wind Farm Integration Processing

[0014] Calculate the observation field error cocorrelation matrix between upper-air sounding stations, the background field error cocorrelation matrix between upper-air sounding stations, and the background field error cocorrelation matrix between grid points and different upper-air sounding stations. Calculate weighting coefficients based on the above matrices. Combine these weighting coefficients with the deviation between the observations and background wind field data of each upper-air sounding station and the background wind field data of the grid points to obtain the optimal interpolated wind field data and create a reanalysis dataset of the wind field.

[0015] (III) Beneficial Effects

[0016] This invention proposes a method for creating a high-precision upper-air wind field reanalysis dataset. This invention fuses ERA5 reanalysis data with national upper-air wind field detection data to create a high-precision upper-air wind reanalysis dataset covering 80 layers below 20,000 meters. This dataset can more accurately reflect the wind field conditions in China and its surrounding areas. The high-precision upper-air wind reanalysis dataset proposed in this invention is based on ERA5 reanalysis data. ERA5 only shows wind fields at 29 layers below 20,000 meters, and its creation process did not fully integrate the dense vertical wind field profiles from various upper-air sounding stations in China. This dataset is stored in the standard NETCDF format and has a universal data reading interface. It is suitable for upper-air wind support in aerospace flights and the construction of corresponding artificial intelligence models. Attached Figure Description

[0017] Figure 1 This is a flowchart of the method of the present invention. Detailed Implementation

[0019] To make the objectives, contents, and advantages of the present invention clearer, the specific embodiments of the present invention will be described in further detail below with reference to the accompanying drawings and examples.

[0020] This invention pertains to the field of meteorological upper-air wind field fusion technology. It provides a method for creating a high-precision upper-air wind field reanalysis dataset. This method integrates radiosonde data and standard atmospheric model standard pressure layer division at 250-meter intervals, background wind field processing based on ERA5, fusion of ERA5 and upper-air sounding wind fields using the optimal interpolation method, and creates a reanalysis dataset of upper-air wind fields.

[0021] Specifically, the following steps are included:

[0022] S1, Standard pressure layer division with 250-meter intervals

[0023] S11. A sequence of pressure layers ranging from 0 to 20,000 meters, with intervals of 250 meters, was obtained using the US Standard Atmosphere (SA1976) atmospheric reference model. ;

[0024] S12. The sounding data from more than 120 upper-air sounding stations broadcast by the China Meteorological Administration via satellite were decoded station by station in BUFR format (Binary Universal Form for Representation of meteorological data) to obtain the observed wind field data of more than 200 isobaric layers at each station. , The decoding process continues to obtain the second-by-second data for geopotential height and air pressure, and a 30-second moving average is applied to the second-by-second data. The averaged data is then interpolated to form a geopotential height layer sequence with 250-meter intervals. and the corresponding pressure layer sequence .

[0025] S13. Select a geopotential height sequence of all upper-air sounding stations within the 20°N-30°N range, spaced 250 meters apart. and the corresponding pressure layer sequence And the sequence of barospheres spaced 250 meters apart at all upper-air sounding stations within the 20°N-30°N range. Horizontal averaging yields a sequence of 250-meter-level pressure layers ranging from 0 to 15000 meters. .

[0026] S14. Because high-altitude sounding stations use balloon sounding, the balloons typically burst and data is interrupted above 15,000 meters. Therefore, the SA1976 model's barosphere sequence is used above 15,000 meters. Subsequently, a sequence of pressure layers ranging from 0 to 20,000 meters was generated. .

[0027] (1)

[0028] S2, Background Field Wind Field Processing

[0029] S21. The China Meteorological Administration stipulates that all upper-air sounding stations nationwide shall conduct operational sounding at 08:00 and 20:00 daily, reading ERA5 reanalysis data from 1000hPa to 50hPa at 08:00 and 20:00, and obtaining... , The wind field has a total of 29 layers and a horizontal resolution of 0.25°.

[0030] S22, to , The wind field is interpolated horizontally along the latitudinal and longitudinal directions to obtain wind field data with a horizontal resolution of 0.125° and a vertical layer of 29.

[0031] S23. For wind field data at 0.125°, if the pressure layer sequence is in the vertical direction... If the isobaric layer within the ERA5 overlaps with the existing isobaric layer in ERA5, then the wind field corresponding to the overlapping layer in ERA5 will be used as... Background wind field at corresponding levels , For non-overlapping layers, logarithmic pressure interpolation is used to interpolate each grid point in the vertical direction, obtaining the grid background wind field data of the corresponding isobaric layer from the existing upper and lower layers in ERA5. , For example, ERA5 already has two levels, 500 hPa and 550 hPa, but not... 525 hPa, here we take 500 hPa and 550 hPa ERA5 The wind field interpolation reached 525 hPa. For example, see Equation 2.

[0032] (2)

[0033] S24. Obtain the longitude and latitude positions of each upper-air sounding station, based on... , The background wind field data from the gridded data was used to calculate the background wind field corresponding to each upper-air sounding station using bilinear interpolation. , And calculate the deviation between the observations from each upper-air sounding station and the background wind field data. , .

[0034]

[0035] (3)

[0036] S3, Wind Farm Integration Processing

[0037] S31. Calculate the observation field error cocorrelation matrix between upper-air sounding stations. Starting from south to north and west to east along the horizontal direction, check each grid point to determine if there is an observation station within a 400 km effective radius of that grid point. If there is more than one upper-air sounding station within the effective radius, calculate the observation field error cocorrelation matrix between the upper-air sounding stations. Assuming that the observation field errors of each upper-air sounding station do not affect each other, the cocorrelation of the observation field error is only related to the location, and is determined according to the standard that "two upper-air sounding stations at the same location take a value of 1, and those at different locations take a value of 0".

[0038] S32. Calculate the background field error cocorrelation matrix between upper-air sounding stations, and the background field error cocorrelation matrix between grid points and different upper-air sounding stations.

[0039] Calculate the background field error cocorrelation between different high-altitude sounding stations within an effective grid radius of 400 km. Background field error cocorrelation Treating it as a Gaussian function, and assuming that the cocorrelation only involves the effective available range of upper-air sounding, where the characteristic distance scale in the meridional direction is 200 km and the characteristic distance scale in the zonal direction is 400 km, the meridional distance between two different upper-air sounding stations is... Latitudinal distance is The background field error cocorrelation values ​​of the two upper-air sounding stations are: The calculation formula is shown in Equation 4.

[0040] And using the same method, the meridional distance between the grid points and the upper-air sounding station is defined as follows: Latitudinal distance is Then, according to Equation 5, the background field error cocorrelation between the grid points and the high-altitude sounding station is calculated. .

[0041] (4)

[0042] (5)

[0043] S33. Calculate the weighting coefficients using LU decomposition according to Equation 6. Based on this, Equations 7 and 8 are used for calculation. , The bias-weighted values ​​are then summed with the gridded background wind field data to obtain the optimal interpolated wind field data. , .

[0044] (6)

[0045] (7)

[0046] (8)

[0047] S34, Wind Field , The data was written in the NETCDF (network Common Data Format) format according to four dimensions: time, pressure layer, latitude, and longitude, and a reanalysis dataset of the wind field was created.

[0048] Beneficial effects:

[0049] This invention fuses ERA5 reanalysis data with nationwide upper-air wind field detection data to create a high-precision upper-air wind reanalysis dataset covering 80 layers below 20,000 meters. This dataset provides a more detailed and accurate reflection of the wind field conditions in China and its surrounding areas. The high-precision upper-air wind reanalysis dataset proposed in this invention is based on ERA5 reanalysis data. ERA5 only shows wind fields at 29 layers below 20,000 meters, and its creation process did not fully integrate the dense vertical wind field profiles from various upper-air sounding stations across China. This dataset is stored in the standard NETCDF format and has a universal data reading interface. It is suitable for upper-air wind support in aerospace flights and the construction of corresponding artificial intelligence models.

[0050] The above description is only a preferred embodiment of the present invention. It should be noted that for those skilled in the art, several improvements and modifications can be made without departing from the technical 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 creating a high-precision upper-level wind field reanalysis dataset, characterized in that, The method includes the following steps: S1, Standard pressure layer division with 250-meter intervals The radiosonde data from upper-air sounding stations broadcast via satellite by the China Meteorological Administration were processed to obtain the observed wind field data of the isobaric layers at each station. Further processing yielded the second-by-second data for geopotential height and pressure at each station. A moving average was applied to the second-by-second data, and the averaged data was interpolated into geopotential height layer sequences and corresponding pressure layer sequences at 250-meter intervals. A horizontal average was then applied to the pressure layer sequences of each station within a specified range to obtain a pressure layer sequence ranging from 0 to 15000 meters, spaced 250 meters apart. Combined with the SA1976 model barosphere sequence used above 15,000 meters Subsequently, a sequence of pressure layers ranging from 0 to 20,000 meters is generated. ; S2, Background Field Wind Field Processing Read ERA5 reanalysis data to obtain , The wind field, totaling 29 layers, was used to obtain the pressure layer sequence based on wind field data from ERA5 reanalysis. The gridded background wind field data is used to obtain the background wind field corresponding to each upper-air sounding station, and the deviation between the observation and the background wind field data of each upper-air sounding station is calculated. S3, Wind Farm Integration Processing Calculate the observation field error cocorrelation matrix between upper-air sounding stations, the background field error cocorrelation matrix between upper-air sounding stations, and the background field error cocorrelation matrix between grid points and different upper-air sounding stations. Calculate weighting coefficients based on the above matrices. Combine these weighting coefficients with the deviation between the observations and background wind field data of each upper-air sounding station and the background wind field data of the grid points to obtain the optimal interpolated wind field data and create a reanalysis dataset of the wind field.

2. The method for creating a high-precision upper-level wind field reanalysis dataset as described in claim 1, characterized in that, S1 includes: S11. A sequence of pressure layers ranging from 0 to 20,000 meters, with intervals of 250 meters, was obtained using the American Standard Reference Atmospheric Model SA1976. ; S12. Decode the sounding data from multiple upper-air sounding stations broadcast by the China Meteorological Administration via satellite broadcast into BUFR format for each station to obtain the wind field data of the isobaric layer at each station. , The process continues, decoding the second-by-second data for geopotential height and air pressure, and then performing a moving average of the second-by-second data over N seconds. Finally, the averaged data is interpolated to form a geopotential height layer sequence with 250-meter intervals. and the corresponding pressure layer sequence ; S13. Select a sequence of geopotential height layers at 250-meter intervals from all upper-air sounding stations within a specified range. and the corresponding pressure layer sequence And for all upper-air sounding stations within the specified range, a 250-meter interval sequence of barospheres was established. Horizontal averaging yields a sequence of 250-meter-level pressure layers ranging from 0 to 15000 meters. ; S14. Because the high-altitude sounding stations use balloon sounding, the balloons will explode and cause data interruption above 15,000 meters. Therefore, the barosphere sequence of the SA1976 model is used above 15,000 meters. Subsequently, a sequence of pressure layers ranging from 0 to 20,000 meters was generated. ; (1)。 3. The method for creating a high-precision upper-level wind field reanalysis dataset as described in claim 2, characterized in that, In S12, N is 30.

4. The method for creating a high-precision upper-level wind field reanalysis dataset as described in claim 2, characterized in that, In S13, the specified range is: 20°N-30°N.

5. The method for creating a high-precision upper-level wind field reanalysis dataset as described in claim 2, characterized in that, S2 includes: S21. The China Meteorological Administration stipulates that all upper-air sounding stations nationwide shall conduct operational sounding at 08:00 and 20:00 daily, reading ERA5 reanalysis data from 1000hPa to 50hPa at 08:00 and 20:00, and obtaining... , The wind field has a total of 29 layers and a horizontal resolution of 0.25°. S22, to , The wind field is interpolated horizontally along the latitudinal and longitudinal directions to obtain wind field data with a horizontal resolution of 0.125° and a vertical layer of 29. S23. For wind field data at 0.125°, if the pressure layer sequence is in the vertical direction... If the isobaric layer within the ERA5 overlaps with the existing isobaric layer in ERA5, then the wind field corresponding to the overlapping layer in ERA5 will be used as... Background wind field at corresponding levels , For non-overlapping layers, logarithmic pressure interpolation is used to interpolate each grid point in the vertical direction for all horizontal grid points. The grid background wind field data of the corresponding isobaric layer is calculated from the existing upper and lower layers in ERA5. ; S24. Obtain the longitude and latitude positions of each upper-air sounding station, based on... , The background wind field data from the gridded data was used to calculate the background wind field corresponding to each upper-air sounding station using bilinear interpolation. , And calculate the deviation between the observations from each upper-air sounding station and the background wind field data. , ; (3)。 6. The method for creating a high-precision upper-level wind field reanalysis dataset as described in claim 5, characterized in that, In step S3, calculating the observation field error cocorrelation matrix between upper-air sounding stations includes: checking each grid point horizontally from south to north and from west to east to determine whether there is an observation station within an effective radius of 400 km at that grid point; if there is more than one upper-air sounding station within the effective radius, calculating the observation field error cocorrelation matrix between the upper-air sounding stations. .

7. The method for creating a high-precision upper-level wind field reanalysis dataset as described in claim 6, characterized in that, When calculating the cocorrelation of observation field errors between upper-air sounding stations, it is assumed that the observation field errors of each upper-air sounding station do not affect each other. Then the cocorrelation of observation field errors is only related to the location, and the value is taken according to the standard that "two upper-air sounding stations at the same location take a value of 1, and those at different locations take a value of 0".

8. The method for creating a high-precision upper-level wind field reanalysis dataset as described in claim 7, characterized in that, In step S3, calculating the background field error cocorrelation matrix between upper-air sounding stations, and the background field error cocorrelation matrix between grid points and different upper-air sounding stations, includes: Calculate the background field error cocorrelation between different high-altitude sounding stations within an effective grid radius of 400 km. ; Cocorrelate the background field error Treating it as a Gaussian function, assuming the cocorrelation only involves the effective available range of upper-air sounding, where the characteristic distance scale in the meridional direction is 200 km and the characteristic distance scale in the zonal direction is 400 km; the meridional distance between two different upper-air sounding stations is... Latitudinal distance is The background field error cocorrelation values ​​of the two upper-air sounding stations are: The calculation formula is shown in Equation 4; And using the same method, the meridional distance between the grid points and the upper-air sounding station is defined as follows: Latitudinal distance is Then, according to Equation 5, the background field error cocorrelation between the grid points and the high-altitude sounding station is calculated. ; （4） （5）。 9. The method for creating a high-precision upper-level wind field reanalysis dataset as described in claim 8, characterized in that, In step S3, weighting coefficients are calculated based on the aforementioned matrix. Combining these weighting coefficients with the deviation between the observations from each upper-air sounding station and the background wind field data, and the gridded background wind field data, the optimal interpolated wind field data is obtained, including: The weighting coefficients are calculated using LU decomposition according to Equation 6. Based on this, Equations 7 and 8 are used for calculation. , The bias-weighted values ​​are then summed with the gridded background wind field data to obtain the optimal interpolated wind field data. , : (6) (7) (8)。 10. The method for creating a high-precision upper-level wind field reanalysis dataset as described in claim 9, characterized in that, In step S3, creating the wind field reanalysis dataset includes: [containing the wind field data]. , The data was written in the NETCDF network general data format according to four dimensions: time, pressure layer, latitude, and longitude, and a reanalysis dataset of the wind field was created.

Images