A vertical layering method for numerical forecasting models adapted to terrain height tracking coordinates

Abstract

The present application relates to the technical field of computer systems based on specific calculation models, and specifically to a method for vertical stratification of numerical forecast models that adapts to terrain height tracking coordinates. The method is executed by a processor and includes: obtaining an initial data set; determining a vertical stratification parameter set; inputting the vertical stratification parameter set into the original prediction model and adjusting the parameters of the stratification function to obtain a primary prediction model; obtaining an original data set, a primary data set, and a primary prediction parameter set; obtaining comparison parameters; obtaining a preferred prediction parameter set; obtaining observation data; obtaining at least one optimal prediction parameter; determining the number of layers parameter corresponding to the optimal prediction parameter; and determining the optimal prediction model based on the stratification function corresponding to the number of layers parameter. The present application solves the problem in the prior art that the first layer height cannot be adjusted according to the actual predicted terrain height while ensuring the consistency of the vertical stratification characteristics, resulting in large errors in the prediction results.

Description

Technical Field

[0001] The present application relates to the technical field of computer systems based on specific computing models, and in particular to a vertical layering method of a numerical forecasting model adapted to terrain height tracking coordinates. Background Art

[0002] In recent years, the horizontal resolution of numerical forecast models has been continuously improved, and kilometer-level or even hundred-meter-level models have been gradually applied. However, in order to adapt to the improvement of horizontal resolution, the vertical resolution also needs to be improved, otherwise it will lead to the growth of false gravity waves, as well as false growth of instability and amplitude, which will increase the forecast error. Improving the vertical resolution of the model can also bring many benefits. For example, when the vertical gradient of temperature, humidity or wind is large, the lower vertical resolution cannot accurately capture these features. Increasing the vertical resolution can improve these problems. For example, in the test under denser vertical stratification, the simulation of fog is improved because the night jet stream and the wind shear at the top of the boundary layer are better characterized.

[0003] Increasing the vertical resolution may also provide a better presentation of the intensity and distribution of extreme weather. Existing technologies include experiments using the MM5 model to simulate tropical cyclones, which have shown that higher vertical resolution can more realistically reproduce the intensity, structure, etc. of tropical cyclones.

[0004] Among them, in the model layering, the height of the lowest layer of the model is very important. In the calculation of the model, the latent heat and sensible heat flux of the near-ground layer are diagnosed through the state variable diagnosis of the lowest layer height of the model. The layer of the model has a significant impact on the intensity and structure of tropical cyclones. If the lowest layer of the model is placed too high, the near-ground flux will be seriously underestimated. Therefore, in order to accurately calculate the near-ground momentum and flux of the model, the lowest layer of the model must be placed within the near-ground layer. In addition, the 2m temperature, 10m wind, etc. of the model are further obtained through the aforementioned flux diagnosis, and when the lowest layer height of the model is reduced, it helps to reduce the error of the 2m temperature and 2m humidity at night relative to the observation.

[0005] The CMA-MESO model is a regional model version of the numerical forecast system. The CMA-MESO 6.0 version currently in technical application has a horizontal resolution of up to 1 km. With the continuous development of the horizontal resolution of the CMA-MESO model, the natural vertical resolution also needs to adapt to the changes in the horizontal resolution as soon as possible. However, the commonly used layering schemes in the CMA-MESO model are all fixed-layer schemes, and the height of the first layer cannot be adjusted according to the height of the predicted terrain, resulting in large errors and poor accuracy in the prediction results. Summary of the invention

[0006] The purpose of this application is to provide a vertical stratification method for a numerical forecast model that is adapted to terrain height tracking coordinates, which solves the problem in the prior art that the height of the first layer cannot be adjusted according to the actual predicted terrain height while ensuring the consistency of vertical stratification characteristics, resulting in large errors in the prediction results.

[0007] The technical solution of this application:

[0008] The present application provides a method for vertical layering of a numerical prediction model adapted to terrain height tracking coordinates, the method being executed by a processor and comprising:

[0009] Obtain meteorological parameters to form an initial data set;

[0010] determining a vertical stratification parameter set based on an initial data set;

[0011] The vertical stratification parameter set is input into the original prediction model and the parameters of the stratification function are adjusted to obtain the primary prediction model, where the stratification function is:

[0012]

[0013]

[0014] in, is the total number of layers, is the model top height, The thickness of each layer of the model when it is evenly layered, To control the underlying parameters of the model, is the height of the first layer;

[0015] The initial data set is input into the original prediction model to obtain the original data set;

[0016] The original data set is input into the primary prediction model to obtain the primary data set;

[0017] The primary data set is input into the simulation model to obtain the primary prediction parameter set;

[0018] The original data set is input into the simulation model to obtain the comparison parameters;

[0019] Comparing the parameters in the primary prediction parameter set with the comparison parameters to obtain a preferred prediction parameter set;

[0020] Obtain observation data;

[0021] Comparing the parameters in the preferred prediction parameter set with the observed data to obtain at least one optimal prediction parameter;

[0022] Determine the layer number parameter corresponding to the best prediction parameter;

[0023] The best prediction model is determined based on the hierarchical function corresponding to the number of layers parameter.

[0024] In some embodiments, in the layered function, The calculation formula is:

[0025]

[0026] In some embodiments, the layered function The stratification is sparse at the top and dense at the bottom, and its first-order derivative for:

[0027]

[0028] In some embodiments, when b>1, Always greater than 0, Keeping the single increase, the height of the first layer is , which satisfies the following formula:

[0029]

[0030] In some embodiments, the primary data set is input into the simulation model, and the simulation time is at least 3 days; the original data set is input into the simulation model, and the simulation time is at least 3 days.

[0031] In some embodiments, the primary prediction model has a horizontal resolution of 0.03°×0.03.

[0032] In some embodiments, the observation data includes at least precipitation products, temperature verification data, and pressure layer element forecast verification data; wherein, the precipitation product is a fused grid precipitation product with a horizontal resolution of 0.05°×0.05°; the temperature verification data is a 2m temperature verification data set, and it is fused with ground element data with a horizontal resolution of 0.05°×0.05°; the pressure layer element forecast verification data has a horizontal resolution of 0.05°×0.05° and 32 vertical levels.

[0033] The technical solution of the present application has at least the following advantages and beneficial effects:

[0034] The present application provides a method for vertical stratification of numerical forecast models that adapts to terrain height tracking coordinates, and provides a new primary forecast model. A primary data set is obtained by adding a stratification function, and then the parameters in the primary forecast parameter set are compared with the comparison parameters to obtain a preferred forecast parameter set, and then the parameters in the preferred forecast parameter set are compared with the observation data to obtain at least one optimal forecast parameter; the vertical stratification parameters corresponding to the optimal forecast parameter are determined, and finally the optimal forecast model is determined. Compared with the stratification results of the prior art, the stratification results obtained by the stratification method provided by the present application have higher accuracy under the conditions of the same number of layers, and can realize the free adjustment of the height of the first layer, so that the forecast error of certain diagnostic variables is reduced, and the effect of the forecast of each isobaric surface element field and the precipitation forecast is more stable and the error is smaller, which improves the accuracy of the forecast. The present application solves the problem in the prior art that the first layer height cannot be adjusted according to the actual predicted terrain height under the premise of ensuring the consistency of the vertical stratification characteristics, resulting in a large error in the forecast result. BRIEF DESCRIPTION OF THE DRAWINGS

[0035] Figure 1 A schematic diagram of the vertical layering method of this application;

[0036] Figure 2 A schematic diagram of the first-order derivative of the stratification function in the vertical stratification method of this application;

[0037] Figure 3 The vertical layered comparison diagram of meso_71 and xu_71 in full-layer vertical distribution in Example 2 of the present application (a) and its thickness-height diagram below 500 m (b);

[0038] Figure 4 The layer number-height diagram (a), thickness-height diagram (b) and thickness-height diagram below 200 meters (c) of different vertical layered feature distributions in Example 2 of the present application;

[0039] Figure 5 This is an information table of the exemplary experiment in Example 2 of the present application;

[0040] Figure 6 The wind speed comparison chart (a), temperature comparison chart (b), RMSE comparison chart of potential height field (c), and TS score comparison chart (d) of the average isobaric surface of each experimental group in Example 2 of the present application for 36 hours from August 20 to 26, 2023 are as follows;

[0041] Figure 7 This is a comparison chart of the 2m temperature root mean square error of each group in Example 2 of the present application for 12 hours (a), 24 hours (b), and 36 hours (c) forecasts from August 20 to 26, 2023. DETAILED DESCRIPTION

[0042] The following will be combined with the attached examples of the present application Figure 1-7 , the technical solutions in the embodiments of the present application are clearly and completely described. Obviously, the described embodiments are only part of the embodiments of the present application, not all of the embodiments. Based on the embodiments in the present application, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present application.

[0043] Example

[0044] Please refer to Figure 1-Figure 7 The present application provides a method for vertical stratification of a numerical forecast model that is adapted to terrain height tracking coordinates. The method is executed by a processor and includes:

[0045] Obtain meteorological parameters to form an initial data set;

[0046] determining a vertical stratification parameter set based on an initial data set;

[0047] The vertical stratification parameter set is input into the original prediction model and the parameters of the stratification function are adjusted to obtain the primary prediction model, where the stratification function is:

[0048]

[0049]

[0050] in, is the total number of layers, is the model top height, The thickness of each layer of the model when it is evenly layered, To control the underlying parameters of the model, is the height of the first layer;

[0051] The initial data set is input into the original prediction model to obtain the original data set;

[0052] The original data set is input into the primary prediction model to obtain the primary data set;

[0053] The primary data set is input into the simulation model to obtain the primary prediction parameter set;

[0054] The original data set is input into the simulation model to obtain the comparison parameters;

[0055] Comparing the parameters in the primary prediction parameter set with the comparison parameters to obtain a preferred prediction parameter set;

[0056] Obtain observation data;

[0057] Comparing the parameters in the preferred prediction parameter set with the observed data to obtain at least one optimal prediction parameter;

[0058] Determine the layer number parameter corresponding to the best prediction parameter;

[0059] The best prediction model is determined based on the hierarchical function corresponding to the number of layers parameter.

[0060] It is worth noting that in order to keep up with the horizontal resolution, the vertical resolution needs to be improved. In the process of improving the vertical resolution, the number of vertical layers is continuously adjusted and improved to achieve the accuracy of weather forecasting. In the prior art, under the background of the continuous development of the horizontal resolution of the CMA-MESO model, the vertical resolution also needs to adapt to the change of the horizontal resolution. However, the commonly used layering schemes in the CMA-MESO model are fixed layer schemes, which cannot adapt to the change of the vertical layer number of the highly following coordinates, and the height of the first layer cannot be adjusted, resulting in a large error in the results of the existing numerical prediction. This embodiment provides a new primary prediction model, which obtains a primary data set through the added layering function, and then compares the parameters in the primary prediction parameter set with the comparison parameters to obtain a preferred prediction parameter set, and then compares the parameters in the preferred prediction parameter set with the observed data to obtain at least one optimal prediction parameter; the number of layers corresponding to the optimal prediction parameter is determined, and finally the optimal prediction model is determined. It is worth noting that, compared with the stratification results of the prior art, the stratification results obtained by the stratification method provided in this embodiment are more accurate when implemented under the same number of layers. On the other hand, the stratification function provided in this method can realize the free adjustment of the height of the first layer, thereby making the numerical prediction results more accurate. This embodiment solves the problem that the prior art cannot adjust the height of the first layer according to the actual terrain height while ensuring the consistency of vertical stratification characteristics, resulting in large errors in the prediction results. It should be noted that the vertical stratification method provided in this embodiment can provide more adjacent layers with similar thicknesses, which can improve the calculation accuracy of the model and make it closer to the second-order accuracy.

[0061] It is worth noting that the vertical layering method of the numerical forecast model used in the prior art can only use fixed values ​​in the process. The vertical layering method of the numerical forecast model provided in this embodiment can adjust two parameters, specifically the total number of layers. and control the underlying parameters of the model , and then adjust the entire model to make the entire numerical forecast model more flexible.

[0062] Furthermore, the primary data set is input into the simulation model, and the simulation time is at least 3 days; the original data set is input into the simulation model, and the simulation time is at least 3 days.

[0063] Furthermore, in the primary prediction model, the horizontal resolution is 0.03°×0.03°.

[0064] It should be noted that the horizontal resolution of the model is 0.03°×0.03°, and the simulation range is 90°E-117°E in longitude and 20°N-38°N in latitude. The physical process is the same as the options of the existing technology. The microphysical scheme adopts the dual-parameter microphysical scheme LIUMA, the longwave radiation is the RRTM scheme, the shortwave radiation scheme is the Dudhia scheme, the near-surface scheme is the sfclay scheme, the land surface process is the improved Noah scheme, the boundary layer parameterization scheme is a combination of the EDMF and NMRF schemes, the turbulent terrain dragging scheme is added, and the convection parameterization scheme is not enabled. At the initial moment, the cloud analysis system is used to introduce the initial cloud information field.

[0065] Furthermore, the observation data at least includes precipitation products, temperature verification data, and pressure layer element forecast verification data; wherein the precipitation product is a fused grid precipitation product with a horizontal resolution of 0.05°×0.05°; the temperature verification data is a 2m temperature verification data set, and it is fused with ground element data with a horizontal resolution of 0.05°×0.05°; the pressure layer element forecast verification data has a horizontal resolution of 0.05°×0.05° and 32 vertical levels. It should be noted that the verification data used is the atmospheric analysis data recognized by the prior art, and the vertical level is set to 32 levels, which are publicly shared and consistent with the number of layers of the model input and output data.

[0066] It should be noted that the initial field and boundary conditions use the forecast field data of the Global Forecast System GFS 0.05°×0.05° and 32 vertical layers. The data used for cloud analysis in the model include: the blackbody radiation brightness temperature and total cloud amount of the Fengyun-2G satellite, and the radar reflectivity factor of the weather radar network.

[0067] Furthermore, in the hierarchical function, The calculation formula is:

[0068]

[0069] It should be noted that is the height of the first layer.

[0070] Furthermore, the hierarchical function The stratification is sparse at the top and dense at the bottom, and its first-order derivative for:

[0071]

[0072] Furthermore, when b>1, Always greater than 0, Keeping the single increase, the height of the first layer is , which satisfies the following formula:

[0073]

[0074] It should be noted that when >1 hour, is always greater than 0, then Keeping the monotonic increase, the stratification function can be reasonably stratified, so When the above formula is satisfied, Keep the single increase. The above formula is recorded as formula A.

[0075] For details, please refer to Figure 2 , Figure 2 is the first derivative of the stratification function in the vertical stratification method Schematic diagram of Ability to display layered functions in more detail properties, which demonstrates Down It should be noted that The settings need to meet Requirements. Figure 2 The gray solid line shows the result. When k=1 Very close to 0, if It may cause Reduce in a small range. The value of The reduction and increases with the increase of and In the range of The maximum value of does not exceed 2.5 m. Due to the integral stability of the model, the height of the first layer of the model (i.e. the lowest layer) should not be set too low. Therefore, the requirements of formula A can be fully met in the model layer setting, which can ensure It is monotonically increasing, satisfying the characteristics of sparse upper part and dense lower part.

[0076] Depend on Figure 2 and Figure 4 As shown, exist and There are few changes nearby. The variation is the largest near the bottom layer and the top layer, so it can be seen that the thickness of adjacent layers is close to each other in the layered results, while the thickness variation between adjacent layers in the middle layer is large. However, in the CMA-MESO model of the prior art, due to the central difference in the vertical direction, the closer the thickness of two adjacent layers is, the closer the difference calculation accuracy is to the second-order accuracy. Therefore, the layered method provided in this embodiment can be close to the second-order accuracy in the low and high layers.

[0077] Depend on Figure 2 As shown, from different of It can be seen that The smaller, The overall span is larger; The bigger The smaller the overall span, the The closer the layer thickness is to uniform hour, The closer it is to the first derivative of the uniform stratification function. and total number of layers When it remains unchanged, The smaller the value, the smaller the height of the first layer of the model, the denser the layers near the lower layers, and the sparser the layers at the upper layers. The larger it is, the stratification near the bottom layer of the model becomes sparser, while the higher layers become relatively denser, but the thickness of the middle layers is not affected.

[0078] Example 2

[0079] Please refer to Figure 3-Figure 7 Based on Example 1, this example takes 71 vertical layers as an example, and the details are as follows:

[0080] The existing technology CMA–MESOV6.0 adopts a vertical layering scheme of 71 layers (hereinafter referred to as: meso_71) obtained by a piecewise vertical layering function. The vertical layering method under this function is uniformly layered in the layers close to the bottom of the model and the layers close to the bottom of the model, and the rest are non-uniformly layered.

[0081] In the vertical layering method provided in this embodiment, layer 71 is taken as an example, hereinafter referred to as xu_71.

[0082] It is worth noting that, in combination with the hierarchical function provided in this embodiment , so that it The 71-layer delamination under the condition of xu_71 (hereinafter referred to as xu_71) is compared with the delamination scheme meso_71 in the prior art. Figure 3 It can be seen that both xu_71 and meso_71 satisfy the characteristics of sparse upper and dense lower layers, and from Figure 3As can be seen from (a) of , the number of layers obtained by the two layering methods is the same, but each layer is obviously different. Figure 3 In (b), the thickness of the first and second layers of meso_71 are the same, and the thickness gradually increases after the third layer. Compared with meso_71, in the stratification method provided in this embodiment, the stratification function used by xu_71 is a continuous function, which has higher flexibility, and its calculation process is more stable and the calculation accuracy is higher. In detail, the vertical stratification implemented by this method has different thicknesses of several layers close to the bottom of the model. It should be noted that the closer the thickness of adjacent layers of vertical stratification is, the higher the calculation accuracy is. In the stratification function of this implementation, under the condition of b=3, the thickness of several layers close to the bottom of the model and the thickness of several layers close to the top of the model of xu_71 are both smaller than meso_71, while the thickness of the middle layer is greater than meso_71.

[0083] It should be noted that the thickness of several layers close to the bottom of the model and the thickness of several layers close to the top of the model of xu_71 are both smaller than those of meso_71. Smaller thickness of the lower layers is beneficial to improving the forecast effect of lower-level elements, and smaller thickness of the upper layers is beneficial to enhancing the stability of model calculation. Since the middle atmosphere is relatively stable, the thickness of the middle layer of xu_71 close to the model is greater than that of meso_71. Appropriately reducing the layers can effectively reduce the requirements for computing power, thereby improving the efficiency of the entire model calculation.

[0084] It should be noted that Figure 3 The medium grey shaded area is the bell-shaped terrain.

[0085] It should be noted that Figure 4 The diagram shows the number of layers-height diagram (a), thickness-height diagram (b) and thickness-height diagram below 200 meters (c) for the distribution of different vertical stratification characteristics, specifically including 71 layers of the prior art, and 51 layers, 61 layers, 71 layers, 81 layers and 91 layers of the stratification method of this embodiment.

[0086] Furthermore, this embodiment uses a set of 7-day exemplary experiments to explain the stratification method of this embodiment in detail. In this experiment, the same number of layers and model top height are set to evaluate the impact of different Z(2) stratification on the model forecast results. The specific setting is from 00:00 (UTC) on August 20 to August 26, 2023, for a total of 7 days, with a forecast time of 36 hours. The experimental information parameters are as follows: Figure 5 As shown in the table, the experimental results of this exemplary experiment are described in detail as follows:

[0087] Using the layering function provided in this embodiment, the obtained layering (xu_71) is compared with the 71-layer layering (meso_71) of the prior art, and the isobaric surface element field forecast results are generally similar. The RMSE of wind speed and temperature forecasts is not much different from each other, please refer to Figure 6 (a) and (b). However, there are differences in the RMSE of the potential height field. For details, compared with the meso_71 in the prior art, the xu_71 series in this embodiment has better prediction results in each experiment; the RMSE of the potential height field tends to increase as the height of the first layer of the model increases. Please refer to Figure 6 (c); please refer to Figure 6 (d) In terms of precipitation, in the xu_71 series, 0.1mm, 25mm, and 50mm levels of xu_71_b40 and xu_71_b50 are better than meso_71.

[0088] It is worth noting that the layering method provided in this embodiment can freely adjust the height Z(2) of the first layer. Specifically, the parameters in the layering function , can be The height of the first layer of the model is a relatively sensitive quantity in the stratification, and the flux calculation of the near-ground layer is directly related to it. At the same time, the 2m temperature, 2m humidity, and 10m wind are all based on similarity theory in the near-ground layer scheme, and are diagnosed by the state variables on the ground and the first layer of the model. Please refer to Figure 7 , Figure 7 The root mean square error of the experimental 2m temperature at 12:00, 24:00, and 36:00 is given in the figure. Figure 7 It can be seen that as the height of the first layer of the model decreases, the root mean square error of the 2m temperature at each forecast time decreases. This improvement is not only reflected in the numerical forecast for the night, but also in the numerical forecast for the afternoon. In particular, the improvement is most obvious at night. Therefore, the stratified method provided in this embodiment, in which the first layer height is placed at a lower level, has a stable improvement in the forecast of the 2m temperature.

[0089] To sum up, the stratification method provided in this embodiment is not only flexible in adjustment compared to the stratification method in the prior art, but also enables the forecast model to reduce the forecast error of certain diagnostic variables by flexibly adjusting the height of the first layer while maintaining the original forecast level of the isobaric surface elements. This makes the stratification method provided in this embodiment more stable and less error-prone in the forecast of each isobaric surface element field and precipitation forecast, thereby improving the accuracy of the forecast.

[0090] The above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions, improvements, etc. made within the spirit and principles of the present invention should be included in the protection scope of the present invention.

Claims

1. A method for vertical stratification of a numerical forecast model adapted to terrain height tracking coordinates, the method being executed by a processor, characterized in that: include: Obtain meteorological parameters to form an initial data set; determining a vertical stratification parameter set based on an initial data set; The vertical stratification parameter set is input into the original prediction model and the parameters of the stratification function are adjusted to obtain the primary prediction model, where the stratification function is: in, is the total number of layers, is the model top height, The thickness of each layer of the model when it is evenly layered, To control the underlying parameters of the model, is the height of the first layer, The calculation formula is: The initial data set is input into the original prediction model to obtain the original data set; The original data set is input into the primary prediction model to obtain the primary data set; The primary data set is input into the simulation model to obtain the primary prediction parameter set; The original data set is input into the simulation model to obtain the comparison parameters; Comparing the parameters in the primary prediction parameter set with the comparison parameters to obtain a preferred prediction parameter set; Obtain observation data; Comparing the parameters in the preferred prediction parameter set with the observed data to obtain at least one optimal prediction parameter; Determine the layer number parameter corresponding to the best prediction parameter; The best prediction model is determined based on the hierarchical function corresponding to the number of layers parameter.

2. The vertical layering method according to claim 1, characterized in that: The layered function The stratification is sparse at the top and dense at the bottom, and its first-order derivative for:

3. The vertical layering method according to claim 2, characterized in that: When b>1, Always greater than 0, Keeping the single increase, the height of the first layer is , which satisfies the following formula:

4. The vertical layering method according to claim 1, characterized in that: The primary data set is input into the simulation model, and the simulation time is at least 3 days; the original data set is input into the simulation model, and the simulation time is at least 3 days.

5. The vertical layering method according to claim 3, characterized in that: In the primary prediction model, the horizontal resolution is .

6. The vertical layering method according to claim 1, characterized in that: The observation data at least include precipitation products, temperature verification data, and pressure layer element forecast verification data; The precipitation product is a fused grid precipitation product with a horizontal resolution of The temperature inspection data is a 2m temperature inspection data set, which is integrated with ground element data and has a horizontal resolution of ; The horizontal resolution of the atmospheric pressure layer element forecast verification data is , the vertical level is 32 layers.

Images