Applied to data assimilation and automatic station terrain height difference processing methods

Abstract

The present invention discloses a method for processing model and automatic station terrain height difference for data assimilation, comprising the following steps: (1) collecting individual cases of severe convection accompanied by heavy rain and strong wind, collecting automatic station data of the time of occurrence of the individual cases and performing preprocessing; (2) calculating the height difference between the observed and model terrain; and dividing the observation type based on the terrain height difference; (3) according to the height difference between the observed and model terrain, using an improved near-surface observation operator method based on M-O theory in three-dimensional variational assimilation to obtain the model background field of the observation position; the present invention adds the height difference between the observed and model terrain to the assimilation system and can maximize the use of all available ground observation data, thereby significantly improving the data utilization rate of the automatic weather station.

Description

Technical Field

[0001] The invention relates to the technical field of atmospheric science meteorology data assimilation, and in particular to a model applied to data assimilation and a method for processing terrain height differences of automatic stations. Background Art

[0002] In recent years, my country's automatic observation station network has become increasingly complete, ranking first in the world in terms of quantity and density. The automatic stations can provide high-frequency and high-precision meteorological observation data and capture the refined ground features of convective systems, which plays an important role in improving the accuracy of severe convective weather forecasts.

[0003] Assimilating automatic station data can improve the ground characteristics of convective systems and thus improve forecasts. However, due to the limitation of model resolution, terrain is often interpolated to the grid through high-resolution terrain data. This terrain height interpolation method causes the model to use one terrain height value to represent a larger grid area. In fact, the terrain heights at different locations in the grid are different, resulting in a deviation between the actual station height and the interpolated model height, and the deviation is widespread. Even with a high-resolution model, it is impossible to fully capture the details of the terrain, which makes it impossible for automatic station data to be fully used in numerical models. Summary of the invention

[0004] Purpose of the invention: The purpose of the present invention is to provide a model for data assimilation and a method for processing the terrain height difference of an automatic station, so as to solve the adaptability problem of the automatic station data under complex terrain and reduce the impact of the terrain height difference.

[0005] Technical solution: The model for data assimilation and the method for processing the terrain height difference of the automatic station described in the present invention include the following steps:

[0006] (1) Collect individual cases of severe convection accompanied by heavy rain and strong winds, collect automatic station data on the time of occurrence of individual cases, and perform preprocessing;

[0007] (2) Calculate the height difference between observed and model terrain; classify observation types based on the terrain height difference;

[0008] (3) According to the height difference between the observed and model terrain, the model background field at the observation location, namely temperature, specific humidity, and wind, is obtained using an improved near-surface observation operator method based on MO theory.

[0009] Furthermore, in step (1), the preprocessing is specifically as follows: quality control of the automatic station data includes: temporal and spatial consistency verification of the automatic station wind field observations, and outlier quality control based on GTS ground observations.

[0010] Furthermore, the spatiotemporal consistency test of the wind field observations of the automatic stations includes: eliminating the stations with missing measurements, eliminating the stations with abnormal high frequency in time, and eliminating the abnormal spatial observations; among them, the specific elimination of abnormal spatial observations is as follows: 0.25° is selected The spatial resolution is 0.25°, and the consistency is determined by comparing the observations of neighboring stations in the selected area. The formula is as follows:

[0011] ;

[0012] in, is 0.05. When the above formula is met, the current automatic station cannot pass the spatial consistency test.

[0013] Furthermore, outlier quality control based on GTS ground observations includes: eliminating problematic stations, identifying outliers in GTS and automatic station observations, and identifying and eliminating outlier stations; among them, identifying and eliminating outlier stations is as follows: the least squares method is used to calculate the fitted kernel density distribution line, and the value of 1‰ of the highest point where the number of matching stations on both sides of the fitted line is reduced to is used as the threshold. The area below 1‰ is the truncation interval. The inside and outside of the truncation interval distinguish the main part and the tail area in the data point set. Points outside the truncation threshold are eliminated as outliers.

[0014] Furthermore, step (2) includes the following steps:

[0015] (21) Select the four model grids closest to the observation location, namely the lower left, lower right, upper left, and upper right, and perform bilinear interpolation on the model terrain height values ​​at the four grids to obtain the model terrain height at the observation location. The formula is as follows:

[0016] ;

[0017] in, Represents the model terrain height after bilinear interpolation, represents the terrain height at the model grid location, , Represents the model grid point closest to the observation x , y The index in the direction, , , , Respectively represent the weights of the observation position to the lower left, lower right, upper left, and upper right grid points;

[0018] (22) Calculate the difference between the observed and model terrain heights: ;

[0019] (23) Threshold for classifying observation types for multiple cases The sensitivity test was conducted and the test results were objectively scored. The scoring parameters included the root mean square error RMSE of ground elements, the fair skill score ETS of precipitation, and the fractional skill score FSS. The optimal division threshold based on the actual effect was obtained. Greater than The automatic stations are divided into sounding types and assimilated according to the form of sounding data; Less than The automatic stations are classified as surface type and assimilated using the surface observation assimilation form;

[0020] (24) Eliminate observations with significant differences in terrain height, as follows: Observation elimination thresholds are set for multiple cases. The sensitivity test is carried out, and the test results are scored by objective scoring to obtain the optimal rejection threshold based on the actual effect. Greater than Automatic stations with different heights are excluded; in the ground observation type, the terrain height difference is excluded Less than Automatic station.

[0021] Furthermore, step (3) is as follows: at the original operator near ground height Introducing the height difference between actual observation and model terrain The new ground height is , that is, observe the actual near-ground height above the model terrain, update the temperature, specific humidity, wind operators involved in the near-ground forward observation operator, and the temperature, specific humidity, wind operators involved in the near-ground tangent operator; among them, the improved near-ground forward observation operator is as follows:

[0022] The forward observation operator formula for near-surface wind is as follows:

[0023] ;

[0024] ;

[0025] in, represents the difference between observed and model terrain height, , Indicates the background wind speed at the actual near-ground height position U , V Quantity, , Indicates the wind speed at the lowest level of the model U , V Quantity, is the actual height near the ground The profile function at Represents the profile function of the lowest level of the model;

[0026] The forward observation operator formulas for near-surface temperature and specific humidity are as follows:

[0027] ;

[0028] ;

[0029] in, , Indicates the background field temperature and humidity at the actual near-ground height. and They represent the bit temperature and mixing ratio of the lowest layer of the mode, and It represents the potential temperature and mixing ratio at the ground level. represents the ground pressure, the ideal gas constant , specific heat at constant pressure , and Indicates the actual height near the ground The stability function of temperature and specific humidity at and The stability function of the lowest layer temperature and specific humidity of the model;

[0030] The improved near-ground tangent operator is as follows:

[0031] Near-surface wind shear linear operator:

[0032] ;

[0033] ;

[0034] in, , Represents the disturbance of the background wind speed U and V components at the actual near-ground height position, , Indicates the wind speed at the lowest level of the model U , V The disturbance of the component, is the actual height near the ground The perturbation of the profile function at represents the perturbation of the profile function of the lowest level of the model;

[0035] Near-surface temperature and specific humidity tangent linear operator:

[0036] ;

[0037] ;

[0038] in, , Indicates the disturbance of background field temperature and specific humidity at the actual near-ground height position. and represent the disturbances of potential temperature and specific humidity in the lowest layer of the model, respectively. and It represents the disturbance of potential temperature and specific humidity at the ground. and Indicates the actual height near the ground The perturbation of the stability function at and represents the perturbation of the stability function of the lowest level of the model, Represents disturbances in surface air pressure.

[0039] Beneficial effects: Compared with the prior art, the present invention has the following significant advantages: the established improved near-surface observation operator scheme can apply the height difference between actual observation and model terrain to the assimilation system in real time, screen and modify the observation operator formula in real time, and can more accurately handle the difference between observation and model terrain; the improved height difference division scheme effectively avoids the information loss caused by eliminating too much observation data or not fully considering the terrain height difference in the previous scheme, and can maximize the use of all available ground observation data while adding the height difference between observation and model terrain to the assimilation system, significantly improving the data utilization rate of the automatic weather station. This improvement not only optimizes the comprehensive utilization of data, but also enhances the application performance of the assimilation system under complex terrain conditions; it can improve the forecast results of severe convective weather. BRIEF DESCRIPTION OF THE DRAWINGS

[0040] Figure 1 is a flow chart of the present invention;

[0041] Figure 2 is a schematic diagram of the present invention;

[0042] Figure 3 It is a precipitation forecast result diagram of the present invention. DETAILED DESCRIPTION

[0043] The technical solution of the present invention is further described below in conjunction with the accompanying drawings.

[0044] like Figure 1-Figure 2 As shown, the present invention provides a model and an automatic station terrain height difference processing method for data assimilation, comprising the following steps:

[0045] (1) Collect individual cases of severe convection accompanied by heavy rain and strong winds, collect automatic station data on the time of occurrence of the individual cases and perform preprocessing; the preprocessing is as follows: select individual cases of severe convection accompanied by heavy rain and strong winds that occurred between 2020 and 2023, prepare automatic station data on the time of occurrence of the individual cases; perform quality control on the automatic station data, including: temporal and spatial consistency test of wind field observations at the automatic station, and outlier quality control based on GTS ground observations.

[0046] The spatiotemporal consistency test of wind field observations at automatic stations includes: eliminating stations with missing measurements at automatic stations, eliminating stations with high-frequency abnormal observations, and eliminating spatial abnormal observations; among them, the elimination of stations with high-frequency abnormal observations is as follows: statistical analysis of the three-year automatic station data is performed year by year, and the values ​​of 0.7% and 99.3% in the sorted list of multi-year statistical data are set as the maximum and minimum abnormal value thresholds for station elimination. Eliminate stations with abnormal value occurrence frequencies higher than 50%;

[0047] The specific elimination of spatial abnormal observations is as follows: Select 0.25° The spatial resolution is 0.25°, and the consistency is determined by comparing the observations of neighboring stations in the selected area. The formula is as follows:

[0048] ;

[0049] in, is 0.05. When the above formula is met, the current automatic station cannot pass the spatial consistency test.

[0050] Outlier quality control based on GTS ground observations includes: elimination of problematic stations, identification of outliers in GTS and automatic station observations, and identification and elimination of outlier stations;

[0051] Among them, the specific elimination of problematic stations is as follows: eliminate stations with missing meteorological elements in GTS and stations with missing longitude and latitude in automatic station data;

[0052] The outlier identification of GTS and automatic station observations is as follows: GTS stations are matched according to the latitude and longitude coordinates of automatic stations, and stations within 1 km of the two are selected to ensure the consistency of changes in meteorological elements; based on the selection of matching stations, kernel density estimation (KDE) is performed on the directly observed meteorological elements, including temperature, pressure and wind field, and the accuracy and deviation of automatic station observations are discussed through kernel probability density distribution.

[0053] The specific steps of identifying and removing outlier sites are as follows: the least squares method is used to calculate the fitted kernel density distribution line, and the value of 1‰, where the number of matching sites on both sides of the fitted line is reduced to the highest point, is used as the threshold. The area below 1‰ is the truncation interval, and the main part and tail area of ​​the data point set are distinguished inside and outside the truncation interval. Points outside the truncation threshold are considered outliers and are removed.

[0054] (2) Calculating the height difference between the observed and model terrains; and classifying the observation types based on the terrain height difference. Step (2) includes the following steps:

[0055] (21) Select the four model grids closest to the observation location, namely the lower left, lower right, upper left, and upper right, and perform bilinear interpolation on the model terrain height values ​​at the four grids to obtain the model terrain height at the observation location. The formula is as follows:

[0056] ;

[0057] in, Represents the model terrain height after bilinear interpolation, represents the terrain height at the model grid location, , Represents the model grid point closest to the observation x , y The index in the direction, , , , Respectively represent the weights of the observation position to the lower left, lower right, upper left, and upper right grid points;

[0058] (22) Calculate the difference between the observed and model terrain heights: ;

[0059] (23) Threshold for classifying observation types for multiple cases The sensitivity test was conducted and the test results were objectively scored. The scoring parameters included the root mean square error RMSE of ground elements, the fair skill score ETS of precipitation, and the fractional skill score FSS. The optimal division threshold based on the actual effect was obtained. Greater than The automatic stations are divided into sounding types and assimilated according to the form of sounding data; Less than The automatic stations are classified as surface type and assimilated using the surface observation assimilation form;

[0060] (24) Eliminate observations with significant differences in terrain height, as follows: Observation elimination thresholds are set for multiple cases. The sensitivity test is carried out, and the test results are scored by objective scoring to obtain the optimal rejection threshold based on the actual effect. Greater than Automatic stations with different heights are excluded; in the ground observation type, the terrain height difference is excluded Less than Automatic station.

[0061] (3) According to the height difference between the observed and model terrain, the improved near-surface observation operator method based on MO theory is used to obtain the model background field of the observation position, namely, temperature, specific humidity, and wind. Specifically, based on the Monin-Obukhov (MO) similarity theory, a near-surface observation operator combining terrain height difference is established. The specific process is as follows: Introduce the height difference between the actual observed and model terrain into the original operator near-surface height, and obtain the new near-surface height, that is, the actual near-surface height observed above the model terrain, and update the temperature, specific humidity, and wind operators involved in the near-surface forward observation operator, as well as the temperature, specific humidity, and wind operators involved in the near-surface tangent operator; Among them, the improved forward near-surface observation operator is as follows:

[0062] The forward observation operator formula for near-surface wind is as follows:

[0063] ;

[0064] ;

[0065] in, represents the difference in terrain height between observation and model, , Indicates the background wind speed at the actual near-ground height position U , V Quantity, , Indicates the wind speed at the lowest level of the model U , V Quantity, is the actual height near the ground The profile function at Represents the profile function of the lowest level of the model;

[0066] The forward observation operator formulas for near-surface temperature and specific humidity are as follows:

[0067] ;

[0068] ;

[0069] in, , Indicates the background field temperature and humidity at the actual near-ground height. and They represent the bit temperature and mixing ratio of the lowest layer of the mode, and It represents the potential temperature and mixing ratio at the ground level. represents the ground pressure, the ideal gas constant , specific heat at constant pressure , and Indicates the actual height near the ground The stability function of temperature and specific humidity at and The stability function of the lowest layer temperature and specific humidity of the model;

[0070] The improved near-ground tangent operator is as follows:

[0071] Near-surface wind shear linear operator:

[0072] ;

[0073] ;

[0074] in, , Represents the disturbance of the background wind speed U and V components at the actual near-ground height position, , Indicates the wind speed at the lowest level of the model U , V The disturbance of the component, is the actual height near the ground The perturbation of the profile function at represents the perturbation of the profile function of the lowest level of the model;

[0075] Near-surface temperature and specific humidity tangent linear operator:

[0076] ;

[0077] ;

[0078] in, , Indicates the disturbance of background field temperature and specific humidity at the actual near-ground height position. and represent the disturbances of potential temperature and specific humidity in the lowest layer of the model, respectively. and It represents the disturbance of potential temperature and specific humidity at the ground. and Indicates the actual height near the ground The perturbation of the stability function at and represents the perturbation of the stability function of the lowest level of the model, Represents disturbances in surface air pressure.

[0079] The improved observation operator of the present invention, which adds the actual terrain height, can more realistically reflect the influence of local terrain on meteorological variables (such as temperature, humidity, wind speed, etc.), and more accurately invert the near-ground conditions at the observation location. For near-ground observations located on the model terrain, the improved near-ground observation operator is applied to invert the model background field (temperature, specific humidity, wind) at the observation location, which is used to calculate the observation increment in the minimization cost function. Figure 3 As shown, the two height difference processing schemes of the present invention can improve precipitation forecast. The partitioning scheme improves the FSS score and ETS score of precipitation in the range of 3-12h, and the improved near-surface observation operator scheme improves the FSS score and ETS score of precipitation in the range of 1-12h.

Claims

1. A model for data assimilation and a method for processing terrain height differences of automatic stations, characterized in that: The following steps are involved: (1) Collect individual cases of severe convection accompanied by heavy rain and strong winds, collect automatic station data on the time of occurrence of individual cases, and perform preprocessing; (2) Calculate the height difference between observed and model terrain; classify observation types based on the terrain height difference; (3) According to the height difference between the observed and model terrain, the improved near-surface observation operator method based on MO theory is used to obtain the model background field at the observation location, namely, temperature, specific humidity, and wind; the details are as follows: Introducing the height difference between actual observation and model terrain The new ground height is , that is, observe the actual near-ground height above the model terrain, update the temperature, specific humidity, wind operators involved in the near-ground forward observation operator, and the temperature, specific humidity, wind operators involved in the near-ground tangent operator; among them, the improved near-ground forward observation operator is as follows: The forward observation operator formula for near-surface wind is as follows: ； ； in, represents the difference in terrain height between observation and model, , Indicates the background wind speed at the actual near-ground height position U , V Quantity, , Indicates the wind speed at the lowest level of the model U , V Quantity, is the actual height near the ground The profile function at Represents the profile function of the lowest level of the model; The forward observation operator formulas for near-surface temperature and specific humidity are as follows: ； ； in, , Indicates the background temperature and humidity at the actual near-ground height. and They represent the bit temperature and mixing ratio of the lowest layer of the mode, and It represents the potential temperature and mixing ratio at the ground level. represents the ground pressure, the ideal gas constant , specific heat at constant pressure , and Indicates the actual height near the ground The stability function of temperature and specific humidity at and The stability function of the lowest layer temperature and humidity of the model; The improved near-ground tangent operator is as follows: Near-surface wind shear linear operator: ； ； in, , Represents the disturbance of the background wind speed U and V components at the actual near-ground height position, , Indicates the wind speed at the lowest level of the model U , V The disturbance of the component, is the actual height near the ground The perturbation of the profile function at represents the perturbation of the profile function of the lowest level of the model; Near-surface temperature and specific humidity tangent linear operator: ； ； in, , Indicates the disturbance of background field temperature and specific humidity at the actual near-ground height position. and represent the disturbances of potential temperature and specific humidity in the lowest layer of the model, respectively. and It represents the disturbance of potential temperature and specific humidity at the ground. and Indicates the actual height near the ground The perturbation of the stability function at and represents the perturbation of the stability function of the lowest level of the model, Represents disturbances in surface air pressure.

2. The method for processing the model and automatic station terrain height difference used in data assimilation according to claim 1, characterized in that: In step (1), the preprocessing is as follows: Quality control of the automatic station data includes: spatiotemporal consistency verification of the automatic station wind field observations, and outlier quality control based on GTS ground observations.

3. The method for processing the model and automatic station terrain height difference used for data assimilation according to claim 2, characterized in that: The spatiotemporal consistency test of wind field observations at automatic stations includes: eliminating stations with missing measurements at automatic stations, eliminating stations with high-frequency abnormalities in time, and eliminating spatial abnormal observations; among them, the specific elimination of spatial abnormal observations is as follows: 0.25° is selected The spatial resolution is 0.25°, and the consistency is determined by comparing the observations of neighboring stations in the selected area. The formula is as follows: ； in, is 0.

05. When the above formula is met, the current automatic station cannot pass the spatial consistency test.

4. The method for processing model and automatic station terrain height difference for data assimilation according to claim 2, characterized in that: The outlier quality control based on GTS ground observations includes: eliminating problematic stations, identifying outliers in GTS and automatic station observations, and identifying and eliminating outlier stations; among them, identifying and eliminating outlier stations is as follows: the least squares method is used to calculate the fitted kernel density distribution line, and the value of 1‰ of the highest point at which the number of matching stations on both sides of the fitted line is reduced is used as the threshold, and the area below 1‰ is the truncation interval. The inside and outside of the truncation interval distinguish the main part and the tail area in the data point set, and the points outside the truncation threshold are eliminated as outliers.

5. The method for processing model and automatic station terrain height difference for data assimilation according to claim 1, characterized in that: Step (2) includes the following steps: (21) Select the four model grids closest to the observation location, namely the lower left, lower right, upper left, and upper right, and perform bilinear interpolation on the model terrain height values ​​at the four grids to obtain the model terrain height at the observation location. The formula is as follows: ； in, Represents the model terrain height after bilinear interpolation, represents the terrain height at the model grid location, , Represents the model grid point closest to the observation x , y The index in the direction, , , , Respectively represent the weights of the observation position to the lower left, lower right, upper left, and upper right grid points; (22) Calculate the difference between the observed and model terrain heights: ; (23) Threshold for classifying observation types for multiple cases The sensitivity test was conducted and the test results were objectively scored. The scoring parameters included the root mean square error RMSE of ground elements, the fair skill score ETS of precipitation, and the fractional skill score FSS. The optimal division threshold based on the actual effect was obtained. Greater than The automatic stations are divided into sounding types and assimilated according to the form of sounding data; Less than The automatic stations are classified as surface type and assimilated using the surface observation assimilation form; (24) Eliminate observations with significant differences in terrain height, as follows: Observation elimination thresholds are set for multiple cases. The sensitivity test was conducted by objectively scoring the test results to obtain the optimal rejection threshold based on the actual effect; in the classification of sounding observation types, the terrain height difference Greater than Automatic stations with different heights are excluded; in the ground observation type, terrain height differences are excluded Less than Automatic station.

Images