A method and system for optimizing an empirical model of atmospheric pressure in a large-elevation-difference area

Abstract

The application discloses a method for optimizing a barometric pressure empirical model in a large-elevation-difference area, comprising the following steps: obtaining the space-time coordinates of a to-be-detected station and the average residual error of the estimated barometric pressure and the actually measured barometric pressure of all stations in a preset historical period; inputting the obtained space-time coordinates and the average residual error into a trained estimated barometric pressure residual error optimization model, and outputting the residual error prediction value of the estimated barometric pressure of the to-be-detected station, wherein the estimated barometric pressure residual error optimization model adopts a random forest regression algorithm, takes the space-time coordinates and the average residual error of the estimated barometric pressure and the actually measured barometric pressure of all ephemeris barometric pressure empirical models of the selected stations in the preset historical period as inputs, and outputs the residual error prediction value of the estimated barometric pressure of the to-be-detected station; and obtaining the estimated barometric pressure of the to-be-detected station, combining the residual error prediction value of the estimated barometric pressure, and obtaining the optimized value of the estimated barometric pressure of the to-be-detected station.

Description

TECHNICAL FIELD

[0001] The present application relates to the application of global navigation satellite system technology in the field of meteorology, in particular to a pressure empirical value optimization method for retrieving atmospheric water vapor using Beidou or GNSS, and specifically to a pressure empirical model optimization method and system for large elevation difference areas. BACKGROUND

[0002] Beidou / GNSS atmospheric water vapor content (PWV) retrieval technology is a new atmospheric water vapor content monitoring method, which has been widely used in meteorology and has the advantages of high precision, high resolution and all-weather. When retrieving PWV using Beidou / GNSS, the zenith tropospheric wet delay (ZWD) needs to be separated from the total delay (ZTD) through the zenith tropospheric dry delay (ZHD), and then converted to PWV through a conversion coefficient. ZHD can be accurately calculated by measuring the pressure and the Saastamoinen model, with an accuracy of mm level. Therefore, pressure is a necessary parameter for Beidou / GNSS to retrieve PWV, and obtaining accurate pressure priori value is very important, which plays a decisive role in the accuracy of Beidou / GNSS PWV retrieval. At the same time, ZHD plays a crucial role in high-precision positioning and application data processing of Beidou / GNSS and very long baseline interferometry (VLBI). However, not all Beidou / GNSS stations are equipped with meteorological equipment, making it very difficult to obtain accurate pressure priori value, which greatly reduces the accuracy of PWV retrieval by dense Beidou / GNSS station network, and restricts the popularization and deepening of Beidou / GNSS in the field of meteorology. Therefore, it is necessary to obtain the pressure of Beidou / GNSS station by non-measured means.

[0003] Most existing studies use numerical weather prediction products, nearby meteorological station data or grid empirical values to generate the pressure of Beidou / GNSS station through interpolation models. However, the models that interpolate through numerical weather prediction products and nearby meteorological station data need additional measured meteorological data assistance, among which the commonly used one is T0 model. The most commonly used model for interpolation based on grid empirical values is GPT3. GPT3 model does not need additional meteorological data assistance, but only needs to input the three-dimensional coordinate information and time of the station. However, GPT3 model assumes that the atmospheric temperature is at an isothermal condition, which is inconsistent with the actual temperature vertical decrement rate of the troposphere, so when the reference height and station height differ greatly, the error of the pressure output by the model will increase significantly, which will affect the calculation accuracy of ZHD.

[0004] In summary, the current two shortcomings of the pressure estimation model can be summarized as follows: first, it depends on additional measured meteorological data assistance, and second, the interpolation accuracy is poor under large elevation difference conditions. SUMMARY

[0005] In view of the current situation that the estimated air pressure model relies on additional measured meteorological data for assistance or has poor interpolation accuracy under large elevation difference conditions, the application provides a large elevation difference area air pressure empirical model optimization method, which is based on the strong nonlinear approximation capability of random forest to more accurately depict the relationship between the GPT3 air pressure prior value and the measured air pressure of different height layers, so that the established optimization model has better adaptability and can also achieve high precision under large elevation difference conditions, thereby significantly improving the precision and stability of air pressure prediction, ultimately improving the calculation deviation of ZHD caused by air pressure estimation error, and further improving the reliability of Beidou / GNSS inversion PWV.

[0006] According to an aspect of the application, a large elevation difference area air pressure empirical model optimization method is provided, comprising:

[0007] Obtain the space-time coordinates of the to-be-measured station and the residual average value of the estimated air pressure and the measured air pressure of all stations in a preset historical period;

[0008] Input the obtained space-time coordinates and residual average value into the trained estimated air pressure residual optimization model to output the residual prediction value of the estimated air pressure of the to-be-measured station, wherein the estimated air pressure residual optimization model uses a random forest regression algorithm, takes the space-time coordinates and the residual average value of the estimated air pressure and the measured air pressure of all epochs of the selected station in the preset historical period as input, and outputs the residual prediction value of the estimated air pressure of the to-be-measured station;

[0009] Obtain the estimated air pressure of the to-be-measured station, and combine the residual prediction value of the estimated air pressure to obtain the optimized value of the estimated air pressure of the to-be-measured station.

[0010] As a further technical solution, the space-time coordinates include year, year cumulative day, day cumulative hour, latitude, longitude and elevation.

[0011] As a further technical solution, the expression of the estimated air pressure residual optimization model is:

[0012] ,

[0013] Wherein, represents the residual measured value and the output prediction value of the estimated air pressure relative to the measured air pressure of different height layers during training and use respectively; represent year, year cumulative day, day cumulative hour, latitude, longitude and elevation respectively; represents the residual average value of the estimated air pressure and the measured air pressure of all epochs of the selected station in the preset historical period.

[0014] As a further technical solution, the training of the estimated air pressure residual optimization model comprises:

[0015] Randomly select observation data of several stations from sounding stations in a set region to construct a training set;

[0016] Determine the optimal number of decision trees by performing ten-fold cross-validation to minimize the root mean square error of the verification set;

[0017] According to the determined optimal number of decision trees, a random forest regression algorithm is used, the average residual error of the estimated pressure and the measured pressure of the selected stations in the preset historical period is used as input, and the difference between the estimated pressure and the measured pressure at each epoch is used as the training target to obtain the trained estimated pressure residual error optimization model.

[0018] As a further technical solution, the average residual error of the estimated pressure and the measured pressure of all epochs of the selected stations in the preset historical period is the same value during model training and use, which is used to constrain the input.

[0019] According to an aspect of the specification, a large-elevation-area pressure empirical model optimization system is provided, comprising:

[0020] The first main module is used to obtain the space-time coordinates of the to-be-measured station and the average residual error of the estimated pressure and the measured pressure of all stations in the preset historical period;

[0021] The second main module is used to input the obtained space-time coordinates and average residual error into the trained estimated pressure residual error optimization model, and output the residual prediction value of the estimated pressure of the to-be-measured station, wherein the estimated pressure residual error optimization model uses a random forest regression algorithm, and the average residual error of the estimated pressure and the measured pressure of all epochs of the selected stations in the preset historical period is used as input, and the residual prediction value of the estimated pressure of the to-be-measured station is output.

[0022] The third main module is used to obtain the estimated pressure of the to-be-measured station, and combine the residual prediction value of the estimated pressure to obtain the optimized value of the estimated pressure of the to-be-measured station.

[0023] As a further technical solution, the second main module is also used to perform training of the estimated pressure residual error optimization model, comprising:

[0024] Randomly select observation data of several stations from sounding stations in a set region to construct a training set;

[0025] Determine the optimal number of decision trees by performing ten-fold cross-validation to minimize the root mean square error of the verification set;

[0026] According to the determined optimal decision tree number, a random forest regression algorithm is adopted, the average value of the residual error of the estimated pressure and the measured pressure of the selected station in the preset historical period is taken as the input, and each epoch difference value between the estimated pressure and the measured pressure of the pressure empirical model is taken as the training target, so as to obtain the trained estimated pressure residual error optimization model.

[0027] According to an aspect of the specification of the present application, an optimization device is provided, comprising a processor, a memory, and a computer program stored in the memory and configured to be executed by the processor, wherein the processor implements the method for optimizing the pressure empirical model in a large-elevation-area when executing the computer program.

[0028] According to an aspect of the specification of the present application, a computer-readable storage medium is provided, comprising a stored computer program; wherein the computer program controls the device where the computer-readable storage medium is located to execute the method for optimizing the pressure empirical model in a large-elevation-area when running.

[0029] According to an aspect of the specification of the present application, a computer program product is provided, comprising computer programs / instructions, which, when executed by a processor, implement the method for optimizing the pressure empirical model in a large-elevation-area.

[0030] The present application provides a method for optimizing the pressure empirical model in a large-elevation-area (RF-GPT3), which has the following advantages compared with the current commonly used estimated pressure method (T0 and GPT3):

[0031] (1) No additional measured meteorological data is needed;

[0032] (2) Under the condition of large elevation difference between the reference station and the station to be measured, the accuracy is still significantly higher than that of other methods. BRIEF DESCRIPTION OF DRAWINGS

[0033] In order to more clearly illustrate the technical solutions in the embodiments of the present application or the prior art, the drawings used in the embodiment or prior art description will be briefly introduced. Obviously, the drawings in the following description are some embodiments of the present application, and other drawings can be obtained by those skilled in the art without creative labor.

[0034] Figure 1 A flowchart of a method for optimizing the pressure empirical model in a large-elevation-area provided by an embodiment of the present application.

[0035] Figure 2 A 2020 different height layer RF-GPT3 pressure and RS pressure density scatter plot (the larger the number of layers, the larger the elevation difference) provided by an embodiment of the present application. Detailed Implementation

[0036] To make the objectives, technical solutions, and advantages of the embodiments of the present invention clearer, the technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention. In addition, the technical features of the various embodiments or individual embodiments provided by the present invention can be arbitrarily combined to form new technical solutions. Such combinations are not bound by the order of steps and / or structural composition patterns, but must be based on the ability of those skilled in the art to implement them. When the combination of technical solutions is contradictory or cannot be implemented, it should be considered that such a combination of technical solutions does not exist and is not within the scope of protection claimed by the present invention.

[0037] In view of the shortcomings of existing methods for estimating air pressure, namely (1) the reliance on additional measured meteorological data and (2) the poor interpolation accuracy under large elevation differences, this invention proposes the following solutions.

[0038] (1) First, using the GPT3 model as the empirical model for air pressure, a method is proposed to model the GPT3 model using random forest (RF) to estimate air pressure. Relative to measured air pressure (air pressure provided by sounding data), The residuals of the model are expressed as follows:

[0039] (1)

[0040] In the formula, During training and during use, respectively represent Relative to different height levels The measured residuals and the predicted output values; These represent year, yearly days, dayly hours, latitude, longitude, and elevation, respectively. This represents all epochs of several selected sounding stations within a preset historical period (e.g., 5 years). and residuals ( The average of the residuals is used. This residual average is the same value during training and use, and its purpose is to constrain the input value, thereby avoiding excessive bias in the output prediction of outliers.

[0041] (2) In the residual modelization using RF, selecting a reasonable number of decision trees is the key to obtaining ideal results. Therefore, from 145 sounding stations in a certain area, 116 sounding stations were randomly selected, and 80% of the observation data from 2015 to 2019 was randomly selected as the training set. According to the long-term experimental experience of scholars, the number of decision trees was set between 5 and 95, and ten-fold cross-validation was performed. When the root mean square error (RMS) of ten-fold cross-validation is the smallest, the corresponding number of decision trees is the final number of decision trees of the model, and the GPT3 model is used to estimate the pressure residual optimization model.

[0042] (3) Based on the model constructed in (2), only the year, day of year (doy), hour of day (hod), latitude (lat), longitude (lon), height and residual average value are input into the model, and the prediction value of the GPT3 model estimated pressure residual at any time and any location in a certain area can be obtained .

[0043] (4) The GPT3 model estimated pressure is subtracted from the residual prediction value calculated in step (3) , so as to realize the optimization of the GPT3 model estimated pressure and obtain a high-precision pressure estimation value, and the specific formula is as follows:

[0044] (2)

[0045] Among them , , respectively represent the optimized value of the GPT3 model estimated pressure, the GPT3 model estimated pressure and the prediction value of the GPT3 model estimated pressure residual optimization model. It should be noted that when using this method to estimate the pressure, only the , 3D coordinates and time of the station to be measured obtained in the modeling process are needed, which effectively eliminates the dependence on the acquisition of high-precision pressure values on the meteorological equipment.

[0046] As a preferred embodiment, the embodiment of the application uses the 2020 estimated pressure experiment at different height layers of the sounding station as an example to illustrate the specific embodiment of the application. Referring to the attached Figure 1 , the embodiment of the application provides a large-elevation-area pressure empirical model optimization method, which comprises the following steps:

[0047] Step one, using 80% of the data from 116 sounding stations from 2015 to 2019, setting the number of decision trees between 5 and 95, modeling, and carrying out ten-fold cross-validation, i.e. using 90% of the data to train and 10% of the data to verify, determining the optimal number of decision trees by minimizing the RMS of the verification set, and finally determining the number of decision trees as 55.

[0048] Step two, based on the optimized number of decision trees in step one (set to 55), using the random forest regression algorithm, taking the spatial and temporal coordinates (year, year-day, day-time, latitude, longitude, and geodetic height) and the 5-year average residual of the GPT3 model estimated pressure and the RS measured pressure as input values, and taking the difference between the GPT3 model estimated pressure and the RS measured pressure at each epoch as the training target, to build an RF-GPT3 model (i.e. an estimated pressure residual optimization model).

[0049] Step three, based on the model built in step two, inputting the year, year-day, day-time, latitude, longitude, elevation, and residual average value to obtain the predicted value of the GPT3 estimated pressure residual of the station to be measured. ) of the station to be measured.

[0050] Step four, using the predicted value of the GPT3 estimated pressure residual obtained in step three ; subtracting the GPT3 estimated pressure from , which can realize the optimization of the estimated pressure of the empirical model of the pressure in a certain area, and obtain a high-precision pressure estimation value. In this embodiment, the estimated pressure is compared with the actual pressure of the sounding station in 2020, and the average Bias and RMS are calculated to be -0.2 hPa and 3.2 hPa, respectively.

[0051] Table 1 shows the precision of the RF-GPT3, T0 and GPT3 estimated pressure using the stratified pressure data of 145 sounding stations in 2020 which did not participate in the modeling.

[0052] Table 1 2020 145 sounding stations stratified data to verify the precision of RF-GPT3, T0 and GPT3 estimated pressure

[0053] .

[0054] As can be seen from Table 1, the precision of the RF-GPT3 estimated pressure without additional measured data assistance is surprisingly higher than that of the T0 model which needs additional measured meteorological data assistance, and the bias range is smaller, the model is more stable, which proves the superiority of the solution of the present application. At the same time, the precision of RF-GPT3 relative to GPT3 is more significantly improved, and the root mean square error (RMS) is reduced by 87.0%.

[0055] Figure 2The figure shows the fitting effect of the atmospheric pressure at different altitudes calculated by the RF-GPT3 method and the GPT3 method with the atmospheric pressure at the sounding station. Figure 2 It can be seen that the method of the present application can more accurately estimate the atmospheric pressure than the GPT3, and the atmospheric pressure at the sounding station is well fitted (the slope of the fitting line is 1, and the intercept is constantly approaching 0), especially under the condition of large height difference, the advantage is more significant. The specific performance is: when the height difference is small, the prediction error is small, as the height difference increases, the RF-GPT3 can still maintain a high or even higher estimation accuracy than when the height difference is small, while the estimation accuracy of the GPT3 decreases significantly as the height difference increases, the systematic deviation increases (the slope of the fitting line is obviously more deviated from 1, and the intercept is getting larger and larger), and the data distribution is obviously divergent. This fully proves that the method of the present application can still achieve significantly higher accuracy than other methods under the condition of large height difference between the reference site and the site to be measured.

[0056] The implementation basis of each embodiment of the present application is realized by the programmed processing of the device with the processor function. Therefore, in engineering practice, the technical solutions and functions of each embodiment of the present application are packaged into various modules. Based on this actual situation, on the basis of the above-mentioned embodiments, the embodiments of the present application provide a large-height-difference-area atmospheric pressure empirical model optimization system, which is used to execute the large-height-difference-area atmospheric pressure empirical model optimization method in the above-mentioned method embodiments.

[0057] The system comprises: a first main module for acquiring the space-time coordinates of the site to be measured, and the residual error average value of the estimated atmospheric pressure and the measured atmospheric pressure of all sites in a preset historical period; a second main module for inputting the acquired space-time coordinates and residual error average value into the trained estimated atmospheric pressure residual error optimization model, and outputting the residual error prediction value of the estimated atmospheric pressure of the site to be measured, wherein the estimated atmospheric pressure residual error optimization model adopts a random forest regression algorithm, takes the space-time coordinates and the residual error average value of the estimated atmospheric pressure and the measured atmospheric pressure of all ephemeris atmospheric pressure empirical models of the selected sites in the preset historical period as input, and outputs the residual error prediction value of the estimated atmospheric pressure of the site to be measured; and a third main module for acquiring the estimated atmospheric pressure of the site to be measured, combining the residual error prediction value of the estimated atmospheric pressure, and obtaining the optimized value of the estimated atmospheric pressure of the site to be measured.

[0058] This invention provides an optimization system for an empirical air pressure model in areas with large elevation differences. Addressing the current situation where air pressure estimation models rely on additional measured meteorological data or suffer from poor interpolation accuracy under large elevation differences, this system employs several modules and leverages the powerful nonlinear approximation capabilities of random forests to more accurately characterize the relationship between GPT3 air pressure priors and measured air pressure values ​​at different altitudes. This results in an optimized model with better adaptability and high accuracy even under large elevation differences, significantly improving the accuracy and stability of air pressure prediction. Ultimately, this improves the ZHD calculation bias caused by air pressure estimation errors, thereby enhancing the reliability of PWV retrieval via BeiDou / GNSS.

[0059] It should be noted that the system embodiments provided by the present invention are used not only to implement the methods in the above method embodiments, but also to implement the methods in other method embodiments provided by the present invention. The only difference is that corresponding functional modules are set. The principle is basically the same as that of the above system embodiments provided by the present invention. As long as those skilled in the art can improve the modules in the above system embodiments by referring to the specific technical solutions in other method embodiments and combining technical features to obtain corresponding technical means and technical solutions composed of these technical means, on the basis of the above system embodiments, and on the premise of ensuring the practicality of the technical solutions, they can obtain corresponding system-like embodiments for implementing the methods in other method-like embodiments.

[0060] Based on the above system embodiments, as a preferred embodiment, this invention provides a pressure empirical model optimization system for areas with large elevation differences. The second main module is further used to train the pressure residual estimation optimization model, including:

[0061] A training set is constructed by randomly selecting observation data from several radiosonde stations in a designated area.

[0062] The optimal number of decision trees is determined by setting a range of possible numbers and performing tenfold cross-validation to minimize the root mean square error of the validation set.

[0063] Based on the determined optimal number of decision trees, the random forest regression algorithm is adopted. The average residual between the estimated air pressure and the measured air pressure of all epochs of the selected stations within the preset historical period is used as input, and the difference between the estimated air pressure and the measured air pressure at each epoch is used as the training objective to obtain the trained optimized model of estimated air pressure residual.

[0064] Based on the same inventive concept as the foregoing embodiments, this embodiment of the invention also provides a computer-readable storage medium, the computer-readable storage medium including a stored computer program; wherein, when the computer program is running, it controls the device where the computer-readable storage medium is located to execute the pressure empirical model optimization method for large elevation difference areas as described in any of the above embodiments.

[0065] Based on the same inventive concept as the foregoing embodiments, this embodiment of the invention also provides a computer program product, including a computer program / instruction, which, when executed by a processor, implements the pressure empirical model optimization method for large elevation differences in any of the above embodiments.

[0066] Based on the same inventive concept as the foregoing embodiments, this embodiment of the invention also provides an optimization device, including a processor, a memory, and a computer program stored in the memory and configured to be executed by the processor.

[0067] When the processor executes the computer program, it implements the steps in the above-described embodiments of the empirical model optimization method for air pressure in areas with large elevation differences. Alternatively, when the processor executes the computer program, it implements the functions of each module / unit in the above-described system embodiments. For example, the computer program can be divided into one or more modules / units, which are stored in the memory and executed by the processor to complete the present invention. The one or more modules / units can be a series of computer program instruction segments capable of performing specific functions, which describe the execution process of the computer program in the optimization device.

[0068] The optimized device may include, but is not limited to, processors and memory. Those skilled in the art will understand that the device described is merely an example of an optimized device and does not constitute a limitation on the optimized device. It may include more or fewer components than currently described, or combine certain components, or different components. For example, the optimized device may also include input / output devices, network access devices, buses, etc.

[0069] The processor can be a Central Processing Unit (CPU), or other general-purpose processors, digital signal processors (DSPs), application-specific integrated circuits (ASICs), field-programmable gate arrays (FPGAs), or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components, etc. The general-purpose processor can be a microprocessor or any conventional processor. The processor is the control center of the optimization device, connecting all parts of the optimization device via various interfaces and lines.

[0070] The memory can be used to store the computer programs and / or modules. The processor implements various functions of the optimized device by running or executing the computer programs and / or modules stored in the memory and by calling data stored in the memory. The memory may mainly include a program storage area and a data storage area. The program storage area may store the operating system, at least one application program required for a function (such as sound playback function, image playback function, etc.), etc.; the data storage area may store data created according to the use of the mobile phone (such as audio data, phonebook, etc.). In addition, the memory may include high-speed random access memory, and may also include non-volatile memory, such as hard disk, memory, plug-in hard disk, smart media card (SMC), secure digital (SD) card, flash card, at least one disk storage device, flash memory device, or other volatile solid-state storage device.

[0071] Wherein, if the modules / units integrated in the optimized device are implemented as software functional units and sold or used as independent products, they can be stored in a computer-readable storage medium. Based on this understanding, all or part of the processes in the methods of the above embodiments of the present invention can also be implemented by a computer program instructing related hardware. The computer program can be stored in a computer-readable storage medium, and when executed by a processor, it can implement the steps of the various method embodiments described above. The computer program includes computer program code, which can be in the form of source code, object code, executable files, or certain intermediate forms. The computer-readable medium can include: any entity or device capable of carrying the computer program code, recording media, USB flash drives, portable hard drives, magnetic disks, optical disks, computer memory, read-only memory (ROM), random access memory (RAM), electrical carrier signals, telecommunication signals, and software distribution media, etc.

[0072] In summary, this invention addresses the problem that current pressure estimation models rely on additional measured meteorological data or have poor interpolation accuracy under large elevation differences. Based on the powerful nonlinear approximation capability of random forests, it more accurately characterizes the relationship between GPT3 pressure priors and measured pressure values ​​at different altitudes, establishing an optimized model with better adaptability. Using the optimized model described in this invention, pressure at different altitudes can be estimated more accurately without the need for additional meteorological data, thereby improving the accuracy and stability of atmospheric water vapor retrieval by BeiDou / GNSS.

[0073] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention, and not to limit them; 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; and these modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the technical solutions of the embodiments of the present invention.

Claims

1. A method for optimizing an empirical air pressure model in areas with large elevation differences, characterized in that, include: Obtain the spatiotemporal coordinates of the station to be tested, as well as the average residual between the estimated and measured air pressures of all stations within a preset historical time period; The obtained spatiotemporal coordinates and residual average values ​​are input into the trained estimation pressure residual optimization model, and the output is the residual prediction value of the estimated pressure of the test site. The estimation pressure residual optimization model adopts the random forest regression algorithm, and takes the spatiotemporal coordinates and the residual average value of the estimated pressure of the selected site in all epochs of the empirical pressure model and the measured pressure as input, and outputs the residual prediction value of the estimated pressure of the test site. The air pressure at the test site is estimated using an empirical air pressure model. The estimated air pressure is then combined with the residual prediction value to obtain the optimized value of the estimated air pressure at the test site. The expression for the optimization model of the estimated pressure residual is as follows: ， in, During training and during use, these represent the measured residual values ​​and the output predicted values ​​of the estimated air pressure relative to the measured air pressure at different altitudes, respectively. These represent year, yearly days, dayly hours, latitude, longitude, and elevation, respectively. This represents the average residual between the estimated and measured air pressures at all epochs of the selected station within a preset historical period.

2. The method for optimizing an empirical air pressure model in areas with large elevation differences according to claim 1, characterized in that, The spatiotemporal coordinates include year, year-to-day, day-to-hour, latitude, longitude, and elevation.

3. The method for optimizing an empirical air pressure model in areas with large elevation differences according to claim 1, characterized in that, The training of the pressure estimation residual optimization model includes: A training set is constructed by randomly selecting observation data from several radiosonde stations in a designated area. The optimal number of decision trees is determined by setting a range of possible numbers and performing tenfold cross-validation to minimize the root mean square error of the validation set. Based on the determined optimal number of decision trees, the random forest regression algorithm is adopted. The average residual between the estimated air pressure and the measured air pressure of all epochs of the selected stations within the preset historical period is used as input, and the difference between the estimated air pressure and the measured air pressure at each epoch is used as the training objective to obtain the trained optimized model of estimated air pressure residual.

4. The method for optimizing an empirical air pressure model in areas with large elevation differences according to claim 1, characterized in that, The average residual between the estimated and measured air pressures of all epochs at the selected stations within the preset historical time period is the same value during model training and use, and is used to constrain the input.

5. A system for optimizing an empirical air pressure model in areas with large elevation differences, characterized in that, include: The first main module is used to obtain the spatiotemporal coordinates of the stations to be tested, as well as the average residual of the estimated air pressure and the measured air pressure of all stations within a preset historical period. The second main module is used to input the acquired spatiotemporal coordinates and the average residual into the trained estimation pressure residual optimization model, and output the residual prediction value of the estimated air pressure at the test site. The estimation pressure residual optimization model uses a random forest regression algorithm, taking the spatiotemporal coordinates and the average residual between the estimated air pressure and the measured air pressure of all epochs of the selected site within a preset historical time period as input, and outputs the residual prediction value of the estimated air pressure of the test site. The expression of the estimation pressure residual optimization model is as follows: ， in, During training and during use, these represent the measured residual values ​​and the output predicted values ​​of the estimated air pressure relative to the measured air pressure at different altitudes, respectively. These represent year, yearly days, dayly hours, latitude, longitude, and elevation, respectively. This represents the average residual between the estimated and measured air pressures at all epochs of the selected station within a preset historical time period. The third main module is used to obtain the estimated air pressure of the test site using an empirical air pressure model, and combine the estimated air pressure residual prediction value to obtain the optimized value of the estimated air pressure of the test site.

6. The pressure empirical model optimization system for areas with large elevation differences according to claim 5, characterized in that, The second main module is also used to train the following optimization model for estimating pressure residuals, including: A training set is constructed by randomly selecting observation data from several radiosonde stations in a designated area. The optimal number of decision trees is determined by setting a range of possible numbers and performing tenfold cross-validation to minimize the root mean square error of the validation set. Based on the determined optimal number of decision trees, the random forest regression algorithm is adopted. The average residual between the estimated air pressure and the measured air pressure of all epochs of the selected stations within the preset historical period is used as input, and the difference between the estimated air pressure and the measured air pressure at each epoch is used as the training objective to obtain the trained optimized model of estimated air pressure residual.

7. An optimization device, characterized in that, The system includes a processor, a memory, and a computer program stored in the memory and configured to be executed by the processor, wherein the processor, when executing the computer program, implements a method for optimizing an empirical model of atmospheric pressure in areas with large elevation differences as described in any one of claims 1 to 4.

8. A computer-readable storage medium, characterized in that, The computer-readable storage medium includes a stored computer program; wherein, when the computer program is executed, it controls the device on which the computer-readable storage medium is located to perform an empirical model optimization method for air pressure in areas with large elevation differences as described in any one of claims 1 to 4.

9. A computer program product comprising a computer program / instructions, characterized in that, When the computer program / instruction is executed by the processor, it implements the method for optimizing an empirical model of air pressure in areas with large elevation differences as described in any one of claims 1 to 4.

Images