A temperature prediction method based on dynamic approximate vertical variation under complex terrain

Abstract

The application belongs to air temperature prediction and provides an air temperature prediction method based on dynamic approximate vertical variation under complex terrain, wherein the method dynamically calculates the approximate vertical variation by comprehensively considering the relationship between the coarse resolution mode air temperature prediction of the observation site, the coarse resolution mode terrain height, the air temperature real value and the real terrain height, and obtains the high resolution grid air temperature prediction product and the high resolution site air temperature prediction product of the observation site through the terrain downscaling technology, so that the fitting degree of the air temperature prediction product and the real terrain change is improved, and the accuracy of the air temperature prediction under the complex terrain is further improved.

Description

Technical Field

[0001] This invention relates to the field of temperature prediction technology, specifically to a temperature forecasting method based on dynamic approximate vertical variability under complex terrain. Background Technology

[0002] Temperature forecasting is a crucial component of meteorological services, playing a vital role in power dispatching, forest fire prevention, and daily life. Influenced by global warming and human activities, extreme heat, cold, rain, snow, and freezing weather—meteorological disasters directly related to temperature—are becoming increasingly frequent. This has led to higher demands for refined temperature forecasts, shifting from point-based to area-based forecasting, specifically from station-based forecasts to grid-based forecasts. This is particularly challenging in regions with complex underlying surfaces, where topography significantly impacts temperature, making accurate temperature forecasting extremely difficult.

[0003] In current technologies, temperature forecasting typically relies on numerical weather prediction models, which predict future weather conditions by simulating atmospheric physical and chemical processes. However, the application of numerical weather prediction models in areas with complex terrain has certain limitations. First, due to the relatively low resolution of numerical models, they cannot accurately capture the impact of terrain on local weather. Second, numerical models are usually based on average terrain height, while in reality, subtle differences in terrain height can lead to significant temperature variations. Furthermore, existing temperature forecasting models also show inadequacy in handling complex weather systems, such as cold waves and temperature fluctuations caused by severe convective weather.

[0004] For temperature forecasting in complex terrain, numerical models struggle to accurately depict such intricate underlying surface conditions, resulting in coarse-resolution forecasts with poor alignment to topographic changes and low accuracy. Therefore, improving the precision and accuracy of temperature forecasts for complex terrain regions is particularly urgent. Summary of the Invention

[0005] To address the shortcomings of existing technologies, this invention provides a temperature forecasting method based on dynamic approximate vertical variability under complex terrain, thereby improving the accuracy of temperature forecasting under complex terrain.

[0006] A temperature forecasting method based on dynamic approximate vertical variability under complex terrain includes:

[0007] Using bilinear interpolation, the coarse-resolution model temperature forecast and model terrain height of sample j in the target area during the training period are interpolated to the location of the observation station, respectively, to obtain the coarse-resolution model temperature forecast and coarse-resolution model terrain height of sample j at the observation station during the training period.

[0008] Based on the coarse-resolution model temperature forecast and coarse-resolution model terrain height of sample j at the observation station during the training period, as well as the actual temperature and actual terrain height of sample j at the observation station during the training period, the coarse-resolution approximate vertical variability of sample j at the observation station during the training period is calculated.

[0009] Based on the coarse-resolution approximate vertical variability of sample j at the observation station during the training period and all samples during the training period, the average coarse-resolution approximate vertical variability of the observation station is calculated.

[0010] The inverse distance weighted interpolation method is used to interpolate the average value of the coarse resolution approximate vertical variability of each observation station onto the high resolution grid, so as to obtain the high resolution approximate vertical variability of each grid point on the high resolution grid.

[0011] The coarse-resolution model temperature forecast and coarse-resolution model terrain height of the target area are interpolated onto the high-resolution grid using the bilinear interpolation method to obtain the high-resolution model temperature forecast and high-resolution model terrain height on the high-resolution grid.

[0012] Based on the high-resolution model temperature forecast, high-resolution model terrain height, high-resolution approximate vertical variability, and actual grid terrain height on the high-resolution grid, the high-resolution grid temperature forecast product for each grid point on the high-resolution grid is calculated.

[0013] Based on the high-resolution model temperature forecast, high-resolution model terrain height, high-resolution approximate vertical variability, and actual terrain height of each grid point on the high-resolution grid, the high-resolution station temperature forecast product of the observation station is calculated.

[0014] The high-resolution grid temperature forecast product and the high-resolution station temperature forecast product of the observation station are identified as the final high-resolution grid and observation station temperature forecast products.

[0015] Furthermore, based on the coarse-resolution model temperature forecast and coarse-resolution model terrain elevation of sample j at the observation station during the training period, as well as the actual temperature and actual terrain elevation of sample j at the observation station during the training period, the coarse-resolution approximate vertical variability of sample j at the observation station during the training period is calculated, including:

[0016] The coarse-resolution approximate vertical variability of sample j at the observation station during the training period is calculated using the following formula, based on the coarse-resolution model temperature forecast and coarse-resolution model terrain elevation of sample j at the observation station during the training period, as well as the actual temperature and actual terrain elevation of sample j at the observation station during the training period:

[0017]

[0018] In the formula, γi,j F represents the coarse-resolution approximate vertical variability of sample j at observation station i during the training period. i,j O i,j H represents the coarse-resolution model temperature forecast and the actual temperature of sample j at observation station i during the training period, respectively. i,m H i,o These represent the coarse-resolution model terrain height and the actual terrain height at observation station i, respectively.

[0019] Furthermore, based on the coarse-resolution approximate vertical variability of sample j at the observation station during the training period and all samples during the training period, the average coarse-resolution approximate vertical variability of the observation station is calculated, including:

[0020] The average coarse-resolution approximate vertical variability of the observation station is calculated using the method shown in the following formula, based on the coarse-resolution approximate vertical variability of sample j at the observation station during the training period and all samples during the training period:

[0021]

[0022] In the formula, γ represents the coarse-resolution approximate vertical variability average at observation station i, 3N represents all samples during the training period, and γ i,j This represents the coarse-resolution approximate vertical variability of sample j at observation station i during the training period.

[0023] Furthermore, using the inverse distance weighted interpolation method, the average value of the coarse-resolution approximate vertical variability of each observation station is interpolated onto the high-resolution grid to obtain the high-resolution approximate vertical variability of each grid point on the high-resolution grid, including:

[0024] The method shown in the following formula employs inverse distance weighted interpolation to interpolate the average value of the coarse-resolution approximate vertical variability of each observation station onto the high-resolution grid, thus obtaining the high-resolution grid approximate vertical variability at the k-th grid point:

[0025]

[0026] In the formula, λ i Here, Y represents the weighting coefficient, and Y represents the number of observation stations involved in the calculation. This represents the approximate vertical variation of the k-th grid point on the high-resolution grid.

[0027] Furthermore, based on the high-resolution model temperature forecast, high-resolution model terrain height, high-resolution approximate vertical variability, and actual grid terrain height for each grid point on the high-resolution grid, the high-resolution grid temperature forecast product for each grid point on the high-resolution grid is calculated, including:

[0028] The high-resolution grid temperature forecast product for the k-th grid point is calculated using the following formula, based on the high-resolution model temperature forecast, high-resolution model terrain height, high-resolution approximate vertical variability, and actual grid terrain height for each grid point on the high-resolution grid:

[0029]

[0030] In the formula, T k ′ represents the high-resolution grid temperature forecast product for the k-th grid point on the high-resolution grid, F k， For the high-resolution model temperature forecast at the k-th grid point on a high-resolution grid, H k,m H k,o These represent the high-resolution mode terrain height and the real grid terrain height at the k-th grid point on the high-resolution grid, respectively.

[0031] Furthermore, based on the high-resolution model temperature forecast, high-resolution model terrain height, high-resolution approximate vertical variability, and actual terrain height of each grid point on the high-resolution grid, the high-resolution station temperature forecast product for the observation station is calculated, including:

[0032] Based on the high-resolution model temperature forecast, high-resolution model terrain height, high-resolution approximate vertical variability, and the actual terrain height of the observation station at each grid point on the high-resolution grid, the temperature forecasts for four high-resolution grid points on a plane with the same terrain height around observation station i are calculated. The formulas for calculating the temperature forecast values ​​for the four grid points are as follows:

[0033]

[0034] In the formula, g is the g-th grid point among four high-resolution grid points on a plane with the same terrain height around observation station i, and T g For the high-resolution temperature forecast at the g-th grid point, F g For the high-resolution model temperature forecast of the g-th grid point, H represents the high-resolution approximate vertical variability at the g-th grid point. g,m H represents the high-resolution terrain height at the g-th grid point. g,o The actual terrain height of station i corresponding to the g-th grid point;

[0035] Based on the calculated temperature forecasts at four high-resolution grid points on a plane with the same terrain elevation around observation station i, a high-resolution station temperature forecast product is calculated using the bilinear interpolation method. Attached Figure Description

[0036] To more clearly illustrate the specific embodiments of the present invention, the accompanying drawings used in the description of the specific embodiments or prior art will be briefly introduced below. In all the drawings, the elements or parts are not necessarily drawn to scale.

[0037] Figure 1 A flowchart illustrating a temperature forecasting method based on dynamic approximate vertical variability under complex terrain, provided in an embodiment of the present invention;

[0038] Figure 2 This is a schematic diagram of a bilinear interpolation method provided in an embodiment of the present invention;

[0039] Figure 3 This is a schematic diagram of an inverse distance weighted interpolation method provided in an embodiment of the present invention;

[0040] In the attached figure, (x,y) represents the coordinates of the observation station; (x1,y1), (x2,y1), (x1,y2), and (x2,y2) represent the coordinates of the four grid points near the observation station (lower left, lower right, upper left, and upper right, respectively); (x,y1) and (x,y2) represent the coordinates of the two middle grid points whose x-coordinates are the same as the x-coordinates of the observation station; k represents the k-th grid point; R is the interpolation radius; r1 represents the straight-line distance from the first observation station to the k-th grid point; r2 represents the straight-line distance from the second observation station to the k-th grid point; and r3 represents the straight-line distance from the third observation station to the k-th grid point. Detailed Implementation

[0041] The embodiments of the technical solution of the present invention will now be described in detail with reference to the accompanying drawings. These embodiments are merely illustrative of the technical solution of the present invention and are therefore intended to limit the scope of protection of the present invention.

[0042] In one embodiment, such as Figure 1 As shown, a temperature forecasting method based on dynamic approximate vertical variability is provided for complex terrain, including:

[0043] 1. Using bilinear interpolation, the coarse-resolution model temperature forecast and model topographic height of sample j in the target area during the training period are interpolated to the location of the observation station, respectively, to obtain the coarse-resolution model temperature forecast and coarse-resolution model topographic height of sample j at the observation station during the training period, including:

[0044] (1) Using the bilinear interpolation method, the coarse-resolution model temperature forecast of sample j in the target area during the training period is interpolated to the location of the observation station, so as to obtain the coarse-resolution model temperature forecast of sample j at the observation station during the training period.

[0045] Bilinear interpolation is a commonly used numerical analysis method used to interpolate data from one type of grid point to another type of grid point or non-grid point.

[0046] The training period includes multiple samples, and each sample in the training period can be regarded as a forecast day. The numerical weather prediction model will give a model temperature forecast result for each forecast day of the target area. In order to distinguish it from the high-resolution model temperature forecast in the following text, it is named coarse-resolution model temperature forecast here.

[0047] like Figure 2 As shown, the bilinear interpolation method is used to interpolate the coarse-resolution model temperature forecast of sample j in the target area during the training period to the location of the observation station, so as to obtain the coarse-resolution model temperature forecast of sample j at the observation station during the training period.

[0048]

[0049] In the formula, F(x,y) is the coarse-resolution model temperature forecast of sample j at the observation station during the training period, and (x,y) is the coordinate of the corresponding position of the observation station; F(x1,y1), F(x2,y1), F(x1,y2), and F(x2,y2) are the coarse-resolution model temperature forecasts of sample j at four grid points near the observation station (lower left, lower right, upper left, and upper right) during the training period, respectively, and (x1,y1), (x2,y1), (x1,y2), and (x2,y2) are the coordinates of the four grid points; F(x,y1) and F(x,y2) are the coarse-resolution model temperature forecasts of two middle grid points with the same abscissa as the abscissa of the observation station.

[0050] Preferably, the numerical prediction models include the European Centre for Medium-Range Weather Forecasts (ECMWF) Global Numerical Prediction Model, the China Meteorological Administration (CMA) Global Numerical Prediction Model, the CMA Regional Numerical Prediction Model, and the Chongqing Mesoscale Regional Numerical Prediction Model.

[0051] (2) Using the bilinear interpolation method, the coarse resolution model terrain height on the target area is interpolated to the location of the observation station to obtain the coarse resolution model terrain height of the observation station.

[0052] The interpolation calculation method for the coarse-resolution model terrain height of the observation station is the same as that for the coarse-resolution model temperature forecast of the observation station. That is, a bilinear interpolation method is used to interpolate the coarse-resolution model terrain height of the target area to the location of the observation station, thus obtaining the coarse-resolution model terrain height of the observation station. However, the coarse-resolution model terrain height of the observation station does not need to be calculated once for each sample j (i.e., each forecast day) during the training period, because the coarse-resolution model terrain height of the target area is the same on each forecast day. Therefore, the coarse-resolution model terrain height of the observation station is calculated only once, and remains the same for all other forecast days.

[0053] 2. Based on the coarse-resolution model temperature forecast and coarse-resolution model terrain height of sample j at the observation station during the training period, as well as the actual temperature and actual terrain height of sample j at the observation station during the training period, the coarse-resolution approximate vertical variability of sample j at the observation station during the training period is calculated.

[0054] Preferably, the following formula is used to calculate the coarse-resolution approximate vertical variability of sample j at the observation station during the training period, based on the coarse-resolution model temperature forecast and coarse-resolution model terrain height of sample j at the observation station during the training period, as well as the actual temperature and actual terrain height of sample j at the observation station during the training period:

[0055]

[0056] In the formula, γ i,j F represents the coarse-resolution approximate vertical variability of sample j at observation station i during the training period. i,j O i,j H represents the coarse-resolution model temperature forecast and the actual temperature of sample j at observation station i during the training period, respectively. i,m H i,o These represent the coarse-resolution model terrain height and the actual terrain height at observation station i, respectively.

[0057] During the training period, the actual temperature and actual terrain elevation of sample j at observation station i were measured by instruments.

[0058] 3. Based on the coarse-resolution approximate vertical variability of sample j at the observation station during the training period and all samples during the training period, the average coarse-resolution approximate vertical variability of the observation station is calculated.

[0059] Preferably, the average coarse-resolution approximate vertical variability of the observation station is calculated using the method shown in the following formula, based on the coarse-resolution approximate vertical variability of sample j at the observation station during the training period and all samples during the training period:

[0060]

[0061] In the formula, γ represents the coarse-resolution approximate vertical variability average at observation station i, 3N represents all samples during the training period, and γ i,j This represents the coarse-resolution approximate vertical variability of sample j at observation station i during the training period.

[0062] Optionally, all samples during the training period are generated by sliding sampling and mixing data from N days before the forecast date, N days before the forecast date of the previous year, the forecast date, and N-1 days after the forecast date, resulting in a total of 3N days of samples during the training period.

[0063] Preferably, based on the principle of minimizing the mean absolute error, the optimal sliding training period N is determined month by month to obtain the approximate vertical variability at observation station i that most closely resembles the real atmosphere. The specific steps are as follows:

[0064] For each month, N ranges from 5 to 70 days, with 5-day intervals. The average approximate vertical variability of all samples at each observation station during the training period is calculated daily for each N value. The coarse-resolution model temperature forecasts for the observation stations are then corrected. The specific correction method is as follows:

[0065]

[0066] In the formula, F i，m T i,m These are the coarse-resolution model temperature forecast and temperature correction for observation station i on day m, respectively. Let be the approximate vertical variability average value of observation station i on day m.

[0067] Then, the daily temperature correction values ​​of the observation stations during this month were verified, with the mean absolute error as the verification index. The specific calculation method is as follows:

[0068]

[0069] In the formula, N1 is the number of days in the current month, and O i,m The actual temperature value of observation station i on day m, MAE i The mean absolute error for the current month is the MAE. i When the value of N is at its minimum, it is the optimal sliding training period. The average value of the approximate vertical variability of all samples at observation station i is calculated monthly under the optimal sliding training period.

[0070] 4. Using the inverse distance weighted interpolation method, the average value of the coarse resolution approximate vertical variability of each observation station is interpolated onto the high resolution grid to obtain the high resolution approximate vertical variability of each grid point on the high resolution grid.

[0071] Preferably, the method shown in the following formula employs inverse distance weighted interpolation to interpolate the average value of the coarse-resolution approximate vertical variability of each observation station onto the high-resolution grid, thereby obtaining the high-resolution approximate vertical variability at the k-th grid point on the high-resolution grid:

[0072]

[0073] In the formula, λ i Here, Y represents the weighting coefficient, and Y represents the number of observation stations involved in the calculation. This represents the approximate vertical variation of the k-th grid point on the high-resolution grid.

[0074] Preferably, the weighting coefficient λ i The calculation formula is as follows:

[0075]

[0076] in, R is the interpolation radius, which is taken as 0.1° in this paper. i This represents the distance from the i-th observation station to the k-th grid point. For example... Figure 3 As shown, a circle with radius R is drawn with the k-th grid point on the high-resolution grid as its center. The number of observation stations within this circle and the straight-line distance from each observation station to the center (i.e., the k-th grid point) are then calculated. Weights are assigned based on these straight-line distances, resulting in the weighting coefficient λ. i .

[0077] 5. Using the bilinear interpolation method, the coarse-resolution model temperature forecast and coarse-resolution model terrain height of the target area are interpolated onto the high-resolution grid to obtain the high-resolution model temperature forecast and high-resolution model terrain height of each grid point on the high-resolution grid;

[0078] Here, the bilinear interpolation method is also used to interpolate the coarse-resolution model temperature forecast and coarse-resolution model terrain height of the target area onto the high-resolution grid, so as to obtain the high-resolution model temperature forecast and high-resolution model terrain height of each grid point on the high-resolution grid. The difference here is that the target area is located exactly on the high-resolution grid.

[0079] 6. Based on the high-resolution model temperature forecast, high-resolution model terrain height, high-resolution approximate vertical variability, and actual grid terrain height of each grid point on the high-resolution grid, calculate the high-resolution grid temperature forecast product for each grid point on the high-resolution grid.

[0080] Preferably, the high-resolution grid temperature forecast product for the k-th grid point is calculated using the method shown in the following formula, based on the high-resolution model temperature forecast, high-resolution model terrain height, high-resolution approximate vertical variability, and actual grid terrain height for each grid point on the high-resolution grid:

[0081]

[0082] In the formula, T k ′ represents the high-resolution grid temperature forecast product for the k-th grid point on the high-resolution grid, F k， For the high-resolution model temperature forecast at the k-th grid point on a high-resolution grid, H k,m H k,o These represent the high-resolution mode terrain height and the real grid terrain height at the k-th grid point on the high-resolution grid, respectively.

[0083] 7. Based on the high-resolution model temperature forecast, high-resolution model terrain height, high-resolution approximate vertical variability, and actual terrain height of each grid point on the high-resolution grid, the high-resolution station temperature forecast product of the observation station is calculated.

[0084] Preferably, the high-resolution station temperature forecast product of the observation station is calculated based on the high-resolution model temperature forecast, high-resolution model terrain height, high-resolution approximate vertical variability, and the actual terrain height of the observation station at each grid point on the high-resolution grid, which also uses the bilinear interpolation method.

[0085] Based on the high-resolution model temperature forecast, high-resolution model terrain height, high-resolution approximate vertical variability, and the actual terrain height of the observation station at each grid point on the high-resolution grid, the temperature forecasts for four high-resolution grid points on a plane with the same terrain height around observation station i are calculated. The formulas for calculating the temperature forecast values ​​for the four grid points are as follows:

[0086]

[0087] In the formula, g is the g-th grid point among four high-resolution grid points on a plane with the same terrain height around observation station i, and T g For the high-resolution temperature forecast at the g-th grid point, F g For the high-resolution model temperature forecast of the g-th grid point, H represents the high-resolution approximate vertical variability at the g-th grid point. g,m H represents the high-resolution terrain height at the g-th grid point. g,o The actual terrain height of station i corresponding to the g-th grid point;

[0088] Then, based on the calculated temperature forecasts at four high-resolution grid points on a plane with the same terrain elevation around observation station i, the following method is used: Figure 2 The bilinear interpolation method shown is used to calculate high-resolution station temperature forecast products for the observation stations.

[0089] 8. The high-resolution grid temperature forecast product and the high-resolution station temperature forecast product from the observation stations are identified as the final high-resolution grid and observation station temperature forecast products.

[0090] This invention comprehensively considers the relationship between coarse-resolution model temperature forecasts, coarse-resolution model terrain height, actual temperature, and actual terrain height at observation stations, dynamically calculates the approximate vertical variability, and obtains high-resolution grids and high-resolution temperature forecast products for observation stations through terrain downscaling technology. This improves the fit between temperature forecast products and actual terrain changes, and further enhances the accuracy of temperature forecasts under complex terrain.

[0091] In the description of this specification, the references to terms such as "one embodiment," "some embodiments," "example," "specific example," or "some examples," etc., indicate that a specific feature, structure, material, or characteristic described in connection with that embodiment or example is included in at least one embodiment or example of the present invention. In this specification, the illustrative expressions of the above terms do not necessarily refer to the same embodiment or example. Furthermore, the specific features, structures, materials, or characteristics described may be combined in any suitable manner in one or more embodiments or examples. Moreover, without contradiction, those skilled in the art can combine and integrate the different embodiments or examples described in this specification, as well as the features of different embodiments or examples.

[0092] The above embodiments are only used to illustrate the technical solutions of the present invention, and are not intended to limit it. Although the present invention has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some or all of the technical features therein. Such modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the scope of the technical solutions of the embodiments of the present invention, and they should all be covered within the scope of the claims and specification of the present invention.

Claims

1. A temperature forecasting method based on dynamic approximate vertical variability under complex terrain, characterized in that, include: Using bilinear interpolation, the coarse-resolution model temperature forecast and model terrain height of sample j in the target area during the training period are interpolated to the location of the observation station, respectively, to obtain the coarse-resolution model temperature forecast and coarse-resolution model terrain height of sample j at the observation station during the training period. Based on the coarse-resolution model temperature forecast and coarse-resolution model terrain height of sample j at the observation station during the training period, as well as the actual temperature and actual terrain height of sample j at the observation station during the training period, the coarse-resolution approximate vertical variability of sample j at the observation station during the training period is calculated. Based on the coarse-resolution approximate vertical variability of sample j at the observation station during the training period and all samples during the training period, the average coarse-resolution approximate vertical variability of the observation station is calculated. The inverse distance weighted interpolation method is used to interpolate the coarse-resolution approximate vertical variability of each observation station onto the high-resolution grid, thereby obtaining the high-resolution approximate vertical variability of each grid point on the high-resolution grid. The coarse-resolution model temperature forecast and coarse-resolution model terrain height of the target area are interpolated onto the high-resolution grid using the bilinear interpolation method to obtain the high-resolution model temperature forecast and high-resolution model terrain height on the high-resolution grid. Based on the high-resolution model temperature forecast, high-resolution model terrain height, high-resolution approximate vertical variability, and actual grid terrain height on the high-resolution grid, the high-resolution grid temperature forecast product for each grid point on the high-resolution grid is calculated. Based on the high-resolution model temperature forecast, high-resolution model terrain height, high-resolution approximate vertical variability, and actual terrain height of each grid point on the high-resolution grid, the high-resolution station temperature forecast product of the observation station is calculated. The high-resolution grid temperature forecast product and the high-resolution station temperature forecast product of the observation station are identified as the final high-resolution grid and observation station temperature forecast products. Based on the coarse-resolution model temperature forecast and coarse-resolution model terrain elevation of sample j at the observation station during the training period, as well as the actual temperature and actual terrain elevation of sample j at the observation station during the training period, the coarse-resolution approximate vertical variability of sample j at the observation station during the training period is calculated, including: The coarse-resolution approximate vertical variability of sample j at the observation station during the training period is calculated using the following formula, based on the coarse-resolution model temperature forecast and coarse-resolution model terrain elevation of sample j at the observation station during the training period, as well as the actual temperature and actual terrain elevation of sample j at the observation station during the training period: In the formula, This represents the coarse-resolution approximate vertical variability of sample j at observation station i during the training period. , These represent the coarse-resolution model temperature forecast and the actual temperature of sample j at observation station i during the training period, respectively. , These represent the coarse-resolution model terrain height and the actual terrain height at observation station i, respectively. Based on the coarse-resolution approximate vertical variability of sample j at the observation station during the training period and all samples during the training period, the average coarse-resolution approximate vertical variability of the observation station is calculated, including: The average coarse-resolution approximate vertical variability of the observation station is calculated using the method shown in the following formula, based on the coarse-resolution approximate vertical variability of sample j at the observation station during the training period and all samples during the training period: In the formula, Let be the coarse-resolution approximate average vertical variability of observation station i, and 3N represent all samples during the training period. This represents the coarse-resolution approximate vertical variability of sample j at observation station i during the training period.

2. The temperature forecasting method based on dynamic approximate vertical variability under complex terrain as described in claim 1, characterized in that, Using the inverse distance weighted interpolation method, the average value of the coarse-resolution approximate vertical variability of each observation station is interpolated onto the high-resolution grid to obtain the high-resolution approximate vertical variability of each grid point on the high-resolution grid, including: The method shown in the following formula employs inverse distance weighted interpolation to interpolate the average value of the coarse-resolution approximate vertical variability of each observation station onto the high-resolution grid, thus obtaining the high-resolution grid approximate vertical variability of the k-th grid point on the high-resolution grid: In the formula, Here, Y represents the weighting coefficient, and Y represents the number of observation stations involved in the calculation. This represents the approximate vertical variation of the k-th grid point on the high-resolution grid.

3. The temperature forecasting method based on dynamic approximate vertical variability under complex terrain as described in claim 2, characterized in that, Based on the high-resolution model temperature forecast, high-resolution model terrain height, high-resolution approximate vertical variability, and actual grid terrain height for each grid point on the high-resolution grid, the high-resolution grid temperature forecast product for each grid point on the high-resolution grid is calculated, including: The high-resolution grid temperature forecast product for the k-th grid point is calculated using the following formula, based on the high-resolution model temperature forecast, high-resolution model terrain height, high-resolution approximate vertical variability, and actual grid terrain height for each grid point on the high-resolution grid: In the formula, This is a high-resolution grid temperature forecast product for the k-th grid point on a high-resolution grid. For the high-resolution model temperature forecast of the k-th grid point on the high-resolution grid, These represent the high-resolution mode terrain height and the actual grid terrain height, respectively, for the k-th grid point on the high-resolution grid.

4. The temperature forecasting method based on dynamic approximate vertical variability under complex terrain as described in claim 1, characterized in that, Based on the high-resolution model temperature forecast, high-resolution model terrain height, high-resolution approximate vertical variability, and actual terrain height of each grid point on the high-resolution grid, the high-resolution station temperature forecast products for the observation stations are calculated, including: Based on the high-resolution model temperature forecast, high-resolution model terrain height, high-resolution approximate vertical variability, and the actual terrain height of the observation station at each grid point on the high-resolution grid, the temperature forecasts for four high-resolution grid points on a plane with the same terrain height around observation station i are calculated. The formulas for calculating the temperature forecast values ​​for the four grid points are as follows: In the formula, g is the g-th grid point among four high-resolution grid points on a plane with the same terrain height around observation station i. For high-resolution temperature forecasts at the g-th grid point, For the high-resolution model temperature forecast of the g-th grid point, For the high-resolution approximate vertical variability at the g-th grid point, Let g be the high-resolution model terrain height at the g-th grid point. The actual terrain height of station i corresponding to the g-th grid point; Based on the calculated temperature forecasts at four high-resolution grid points on a plane with the same terrain elevation around observation station i, a high-resolution station temperature forecast product is calculated using the bilinear interpolation method.

Images