Variation deviation correction method and device

Abstract

The invention discloses a variational deviation correction method and device, and relates to the field of weather forecast, and the method comprises the steps: obtaining the observation data of a current time window and an updated deviation correction coefficient of a previous time window; calculating a satellite observation deviation correction amount based on the deviation correction coefficient, the observation information and the mode background field, calculating a deviation between simulation observation and actual observation based on an observation operator, and carrying out correction and quality control on the deviation; according to the satellite observation data after quality control, on the basis of the improved target functional and the gradient thereof, in the minimization iteration process, under the joint constraint of the observation data and a numerical weather forecast mode, a deviation correction coefficient is updated online; wherein the improvement process of the target functional comprises the steps that a deviation correction coefficient is used as a control variable to be combined with a mode state variable to construct a new target functional, and variable transformation is carried out in an increment form. The assimilation analysis quality of observation data is improved, and the numerical weather forecast level is improved.

Description

Technical Field

[0001] The present application relates to the field of weather forecasting, and in particular to a variational bias correction method and device. Background Art

[0002] Numerical weather forecasting is first of all an initial value problem. The quality of initial conditions is becoming the key to the effect of numerical forecasting. Data assimilation is the process of using all existing information to provide the numerical forecast model with the most accurate initial state possible. With the development of atmospheric detection technology, the era of detection systems based on non-conventional observations such as satellites has arrived. This change has greatly improved the coverage and spatial and temporal resolution of data. These data make up for the shortage of data in the ocean and sparsely populated areas, and play a vital role in improving the accuracy of global forecasts.

[0003] However, due to the calibration, positioning, sensitivity and aging of satellite detection instruments, there are systematic errors in satellite observation data. The existence of such systematic deviations restricts the full utilization of satellite radiation data and affects the accuracy of numerical forecasts. Therefore, it is very important to correct these systematic deviations. Summary of the invention

[0004] The purpose of this application is to provide a variational bias correction method and device, which can improve the quality of assimilation analysis of observation data and enhance the level of numerical weather forecasting.

[0005] To achieve the above objectives, this application provides the following solutions:

[0006] In a first aspect, the present application provides a variational bias correction method, comprising:

[0007] Obtain the observation data of the current time window and the deviation correction coefficient updated in the previous time window;

[0008] Calculate the satellite observation bias correction based on the observation information, model background field and the bias correction coefficient updated in the previous time window;

[0009] According to the satellite observation deviation correction amount and the observation data, the deviation between the simulated observation and the actual observation is calculated based on the observation operator, and the deviation is corrected and quality controlled to obtain the satellite observation data after quality control;

[0010] According to the satellite observation data after quality control, based on the improved target functional and its gradient, in the minimization iteration process, under the joint constraints of the observation data and the numerical weather forecast model, the bias correction coefficient is updated online to obtain the updated bias correction coefficient of the current time window;

[0011] The improvement process of the target functional includes: taking the deviation correction coefficient as the control variable and combining the mode state variable to construct a new target functional; transforming the new target functional into an incremental form and performing variable transformation to obtain the improved target functional and its gradient.

[0012] Optionally, the deviation correction coefficient is used as a control variable and the mode state variable is combined to construct a new target functional, specifically including:

[0013] Define the satellite observation bias correction equation based on the bias correction coefficient;

[0014] Based on the satellite observation bias correction equation, the observation term of the target functional of the differential assimilation is modified to obtain a modified observation term;

[0015] Define parameter terms based on deviation correction coefficients;

[0016] A new target functional is constructed according to the background term of the assimilated target functional, the modified observation term and the parameter term.

[0017] Optionally, the satellite observation bias correction equation is:

[0018]

[0019] Among them, bias(β) is the satellite observation bias correction, b scan is the scanning angle correction, b air Corrected for air mass, N p is the number of air mass correction factors, β i is the correction factor coefficient of the ith air mass, p i (x b ) is the correction factor of the ith air mass, x b For background vector.

[0020] Optionally, the new target functional is:

[0021]

[0022] Among them, J(x,β) is the value of the new target functional, is the background term of the variational assimilation objective functional, is a parameter item, is the modified observation item, x is the original analysis variable vector, B x is the background error covariance matrix, β is the deviation correction coefficient, β b is the initial estimate of β, B β is the parameter background error covariance matrix, H x (x) is the observation operator, y o is the observation vector, and R is the observation error covariance matrix.

[0023] Optionally, the new target functional is transformed into an incremental form, specifically including:

[0024] Taylor expansion is performed on the observation operator and the satellite observation bias correction, and the first-order term is retained to obtain the incremental observation operator and the incremental satellite observation bias correction;

[0025] According to the incremental observation operator and the incremental satellite observation bias correction amount, the new target functional is deformed by using a series of transformations, and the parameter background error covariance matrix is ​​adjusted to a diagonal matrix to obtain the target functional in incremental form.

[0026] Optionally, the incremental observation operator is:

[0027] H x (x) = H x (x b +δx)≈H x (x b )+H x ′δx;

[0028] Among them, δx is the assimilation increment, δx=xx b , H x ′ is the gradient of the observation operator with respect to the analysis variable,

[0029] The incremental satellite observation bias correction is:

[0030] bias(β)=b scan +b air (β) = b scan +b air (β b +δβ)≈b scan +b air (β b )-H′ β δβ;

[0031] Among them, δβ is the deviation correction coefficient increment, δβ=β-β b , H′ β is the negative gradient of the observation operator with respect to the parameter,

[0032] Optionally, the target functional in incremental form is:

[0033]

[0034] Among them, J(δx,δβ) is the value of the target functional in incremental form, is the corrected innovation vector,

[0035] Optionally, a variable transformation is performed on the incremental target functional to obtain an improved target functional and its gradient, specifically including:

[0036] Decomposing the background error covariance matrix and the parameter background error covariance matrix to obtain a background error covariance decomposition matrix and a parameter background error covariance decomposition matrix;

[0037] Defining new control variables according to the background error covariance decomposition matrix and the parameter background error covariance decomposition matrix;

[0038] According to the background error covariance decomposition matrix, the parameter background error covariance decomposition matrix and the new control variable, the target functional in incremental form is transformed to obtain an improved target functional and its gradient.

[0039] Optionally, the improved objective functional is:

[0040]

[0041] Among them, J inc is the value of the improved objective functional, w x is the new control variable corresponding to the original analysis variable, w x =U x δx,U x is the background error covariance decomposition matrix, w β is the new control variable corresponding to the deviation correction parameter, w β =U β δβ,U β is the parameter background error covariance decomposition matrix,

[0042] The gradient of the control variable of the improved objective functional is:

[0043]

[0044] in, w x The gradient of w β gradient.

[0045] In a second aspect, the present application provides a variational deviation correction device, comprising:

[0046] The data acquisition module is used to obtain the observation data of the current time window and the deviation correction coefficient updated in the previous time window;

[0047] The deviation correction calculation module calculates the satellite observation deviation correction based on the observation information, the model background field and the deviation correction coefficient updated in the previous time window;

[0048] A quality control module is used to calculate the deviation between the simulated observation and the actual observation based on the observation operator according to the satellite observation deviation correction amount and the observation data, and to correct and quality control the deviation to obtain the satellite observation data after quality control;

[0049] The bias correction coefficient update module is used to update the bias correction coefficient online based on the satellite observation data after quality control, based on the improved target functional and its gradient, in the minimization iteration process, under the joint constraints of the observation data and the numerical weather forecast model, to obtain the updated bias correction coefficient of the current time window;

[0050] The improvement process of the target functional includes: taking the deviation correction coefficient as the control variable and combining the mode state variable to construct a new target functional; transforming the new target functional into an incremental form and performing variable transformation to obtain the improved target functional and its gradient.

[0051] According to the specific embodiments provided by the present application, the present application has the following technical effects: the present application provides a variational deviation correction method and device, which constructs a new target functional by taking the deviation correction coefficient as the control variable and the model state variable, and adopts an incremental form to solve the parameter background error estimation problem, and further performs variable transformation on the incremental form of the target functional to obtain an improved target functional and its gradient. In the minimization iterative process, based on the improved target functional and its gradient, under the joint constraints of the observation data and the numerical weather forecast model, the deviation correction coefficient is updated online, which can effectively improve the assimilation analysis quality of the observation data, thereby improving the level of numerical weather forecasting. BRIEF DESCRIPTION OF THE DRAWINGS

[0052] In order to more clearly illustrate the embodiments of the present application or the technical solutions in the prior art, the drawings required for use in the embodiments will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present application. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying creative work.

[0053] Figure 1 This is an application environment diagram of a variational deviation correction method in one embodiment of the present application;

[0054] Figure 2 A flow chart of a variational deviation correction method provided in one embodiment of the present application;

[0055] Figure 3 A flow chart of a variational deviation correction method provided in another embodiment of the present application;

[0056] Figure 4A schematic diagram of the functional modules of a variational deviation correction device provided in one embodiment of the present application. DETAILED DESCRIPTION

[0057] The following will be combined with the drawings in the embodiments of the present application to clearly and completely describe the technical solutions in the embodiments of the present application. 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 this application.

[0058] The purpose of bias correction is to eliminate or reduce the system bias. So far, bias correction has mainly gone through the process of off-line bias correction to on-line bias correction. The traditional static method is to correct the radiation data before the minimization process. It is a statistical method. Its bias correction equation is established based on historical data as a sample. It contains the characteristics of the detection instrument and the historical evolution of atmospheric motion. The relationship or law, but does not necessarily include the latest atmospheric motion information, such as seasonal changes, solar activity, instrument aging, observation calibration, etc. The relationship or law existing in the historical data may not be maintained in the future, so the statistical method is often embarrassed by this.

[0059] The on-line variational bias correction is performed simultaneously with the four-dimensional variational analysis of data assimilation, that is, the bias correction is performed during the minimization iteration process and is used as part of the solution to the variational problem; during the assimilation process, all observational data affect the bias correction effect through observation operators, background field constraints, and model constraints. This application combines the characteristics of the China Meteorological Administration's Global Forecast System (CMA-Global Forecast System, CMA-GFS) to develop a variational bias correction scheme suitable for the CMA-GFS four-dimensional variational assimilation system, laying a good foundation for the operational application of satellite radiance data in the numerical weather forecast data assimilation system.

[0060] In order to make the above-mentioned objects, features and advantages of the present application more obvious and easy to understand, the present application is further described in detail below in conjunction with the accompanying drawings and specific implementation methods.

[0061] The variational deviation correction method provided in the embodiment of the present application can be applied to Figure 1In the application environment shown. Among them, the terminal 102 communicates with the server 104 through the network. The data storage system can store the data that the server 104 needs to process. The data storage system can be set up separately, or it can be integrated on the server 104, or it can be placed on the cloud or other servers. The terminal 102 can send a control instruction to the server 104. After receiving the instruction, the server 104 performs deviation correction and quality control on the observation data, and dynamically updates the deviation correction coefficient. The server 104 can feedback the calculation results to the terminal 102. Among them, the terminal 102 can be but not limited to various desktop computers, laptops, and tablet computers. The server 104 can be implemented with an independent server or a server cluster consisting of multiple servers, and can also be a cloud server.

[0062] In an exemplary embodiment, Figure 2 As shown, a variational deviation correction method is provided. The method is executed by a computer device, and can be executed by a computer device such as a terminal or a server alone, or by a terminal and a server together. In the embodiment of the present application, the method is applied to Figure 1 The server 104 in the example is used for explanation, and the steps include the following steps 201 to 204.

[0063] Step 201, obtain the observation data of the current time window and the deviation correction coefficient updated in the previous time window.

[0064] Step 202, calculate the satellite observation bias correction based on the observation information, the model background field and the bias correction coefficient updated in the previous time window.

[0065] Step 203, according to the satellite observation deviation correction amount and the observation data, the deviation between the simulated observation and the actual observation is calculated based on the observation operator, and the deviation is corrected and quality controlled to obtain the satellite observation data after quality control.

[0066] Step 204, based on the satellite observation data after quality control, based on the improved target functional and its gradient, in the minimization iteration process, under the joint constraints of the observation data and the numerical weather forecast model, the bias correction coefficient is updated online to obtain the updated bias correction coefficient for the current time window.

[0067] The target functional improvement process includes the following steps 301 to 303 .

[0068] Step 301, using the deviation correction coefficient as a control variable and the model state variable to construct a new target functional.

[0069] The target functional of the basic formula of traditional variational assimilation can be reduced to a quadratic functional minimization problem that characterizes the deviation between the analysis field and the observation field, and between the analysis field and the background field. The target functional of variational assimilation is generally defined as:

[0070]

[0071] Where J is the value of the variational assimilation objective functional, J b is the background term, J o is the observation term; to facilitate the distinction of the new target functional to be constructed, x is called the original analysis variable vector, whose length is N, x b is the background vector (usually the short-term forecast of the model), whose length is N, y o is the observation vector, whose length is M, B x is the background error covariance matrix, whose order is N×N, is the inverse of the background error covariance matrix, R is the observation error covariance matrix, whose order is M×M, R -1 is the inverse of the observation error covariance matrix, H x (x) is the observation operator, which consists of three parts: H x =H m H RT H In , operator H m is the numerical weather forecast model, operator H x is the radiation transmission mode, the operator H In For spatial interpolation, the observation operator represents a mapping from the pattern space to the observation space.

[0072] In this application, the basic idea of ​​the variational bias correction method is to embed the bias correction equation into the basic formula of four-dimensional variational assimilation - the target functional, and to achieve fine-tuning of the bias correction coefficient under the joint constraints of all observation data (conventional and unconventional) and numerical weather forecast models in the minimization iteration process to adapt to seasonal changes, etc.; that is, the target functional is changed from the original background term and observation term to background term, parameter term and modified observation term.

[0073] In an exemplary embodiment, step 301 includes:

[0074] (1) Define the satellite observation bias correction equation.

[0075] The satellite observation bias correction equation is:

[0076]

[0077] Where bias(β) is the satellite observation bias correction, that is, the satellite radiance bias correction, b scan is the scanning angle correction, which is the pre-statistical data, bair Corrected for air mass, N p is the number of air mass correction factors, β i is the correction factor coefficient of the ith air mass, p i (x b ) is the correction factor of the ith air mass.

[0078] (2) Based on the satellite observation bias correction equation, the observation term of the variational assimilation target functional is modified to obtain a modified observation term. The variational assimilation target functional includes a background term and an observation term.

[0079] The modified observations are:

[0080]

[0081] Among them, H x (x) is the observation operator, and the order of the observation error covariance matrix R is N p ×N p , the superscript T represents the transpose operation.

[0082] (3) Define the parameter term based on the deviation correction coefficient. That is, the distance between the weighted deviation correction coefficient and its initial estimate: Among them, β is the prediction factor coefficient of the deviation correction equation, referred to as the deviation correction coefficient, β b is the initial estimate of β, and its length is N p , B β is the parameter background error covariance matrix, B β -1 is the inverse of the parameter background error covariance matrix, length N p ×N p .

[0083] (4) constructing a new target functional according to the background term of the assimilated target functional, the modified observation term and the parameter term:

[0084]

[0085] Among them, J(x,β) is the value of the new target functional, is the background term of the variational assimilation objective functional.

[0086] In the dynamic updating process of the deviation correction coefficient, the deviation correction coefficient updated in the previous analysis is used as the initial estimate, and then the updated deviation correction coefficient is applied to the next time window.

[0087] Step 302: transform the new target functional into an incremental form to obtain the target functional in the incremental form.

[0088] Considering the incompleteness of the quality-controlled observation data set and reducing the amount of calculation, the incremental analysis method is used to transform the target functional formula into an incremental form.

[0089] In an exemplary embodiment, step 302 includes:

[0090] (1) Taylor expansion is performed on the observation operator and the satellite observation bias correction, and the first-order terms are retained to obtain the incremental observation operator and the incremental satellite observation bias correction.

[0091] The incremental observation operator is:

[0092] H x (x) = H x (x b +δx)≈H x (x b )+H x ′δx;

[0093] Among them, δx is the assimilation increment, δx=xx b , H x ′ is the gradient of the observation operator with respect to the analysis variable,

[0094] The incremental satellite observation bias correction is:

[0095] bias(β)=b scan +b air (β) = b scan +b air (β b +δβ)≈b scan +b air (β b )-H′ β δβ;

[0096] Among them, δβ is the deviation correction coefficient increment, δβ=β-β b , H′ β is the negative gradient of the observation operator with respect to the parameter,

[0097] (2) According to the incremental observation operator and the incremental satellite observation bias correction, the new target functional is deformed by a series of transformations, and the parameter background error covariance matrix is ​​adjusted to a diagonal matrix to obtain the target functional in incremental form.

[0098] The target functional in the incremental form is:

[0099]

[0100] Among them, J(δx,δβ) is the value of the target functional in incremental form, is the corrected innovation vector, d=H x (x b )-y o .

[0101] Set the parameter background error covariance matrix to a diagonal matrix: in, is the control parameter β n The variance is an empirical value. After many tests, it is determined to be the deviation correction coefficient of 1%. The appropriateness of its value is very critical and determines the update amplitude and effect of the deviation correction coefficient.

[0102] Step 303: Perform variable transformation on the incremental target functional to obtain an improved target functional and its gradient. Pre-adjustment is achieved through a series of variable transformations from the physical pattern space to the abstract control space.

[0103] In an exemplary embodiment, step 303 includes:

[0104] (1) Decompose the background error covariance matrix and the parameter background error covariance matrix to obtain the background error covariance decomposition matrix and the parameter background error covariance decomposition matrix.

[0105] Specifically, for the background error covariance matrix B x and parameter background error covariance matrix B β This breaks down as follows:

[0106]

[0107] Among them, U x is the background error covariance decomposition matrix, U β is the parameter background error covariance decomposition matrix.

[0108] (2) According to the background error covariance decomposition matrix and the parameter background error covariance decomposition matrix, define new control variables:

[0109] w x =U x δx;

[0110] w β =U β δβ;

[0111] Among them, w x is the new control variable corresponding to the original analysis variable, w β is a new control variable corresponding to the deviation correction parameter.

[0112] (3) According to the background error covariance decomposition matrix, the parameter background error covariance decomposition matrix and the new control variable, the incremental target functional is transformed to obtain an improved target functional and its gradient.

[0113] The improved objective functional is:

[0114]

[0115] Among them, J inc is the value of the improved objective functional;

[0116] The gradient of the control variable of the improved objective functional is:

[0117]

[0118] in, w x The gradient of w β gradient.

[0119] And its second-order derivative is:

[0120]

[0121] Where I is the identity matrix.

[0122] and The condition number has been significantly improved, which means that the shape of the target functional is isotropic through variable transformation, and the minimization algorithm can quickly converge to the optimal solution.

[0123] In an exemplary embodiment, after determining the target functional and its gradient, a descent algorithm is used based on optimization theory to perform minimization iterative convergence; in this process, all observation data (including conventional observations such as sounding, ground, and aircraft reports, and unconventional observations such as satellites, radars, and occultations) are adjusted with deviation correction coefficients through observation operators and model constraints.

[0124] In order to better understand the technical solution of the present application, the variational bias correction method provided by the present application is applied to CMA-GFS to introduce the specific process, such as Figure 3 As shown. Among them, 2) to 5) are the outer loop, 6) to 8) are the inner loop, the outer loop uses a high-resolution nonlinear model to integrate forward to obtain the observation increment, the inner loop uses a low-resolution tangent linear model to integrate forward to obtain the observation forcing term, and uses a low-resolution adjoint model to integrate backward.

[0125] 1) CMA-GFS Global 4D-Var uses one executable program to perform a complete outer and inner loop calculation with a time window of 6 hours.

[0126] 2) In each time window, the satellite and other observation data of the corresponding time and the deviation correction coefficient updated in the previous time window are read in.

[0127] 3) Use the formula Calculate the observation bias correction of each satellite and each detector respectively.

[0128] 4) Based on the observation operator H x (x b ) Calculate the simulated observations and actual observations y of each detector o The deviation of x (x b )-y o -bias(β) corrects the bias.

[0129] 5) Perform quality control such as extreme value inspection, background field inspection, and cloud detection on the observation data of each detector that has been corrected for deviations.

[0130] 6) The satellite detector data after quality screening is used to calculate the target functional and its gradient and enter the inner loop.

[0131] 7) Perform minimization iterative convergence; in this process, all observation data (including conventional observations such as sounding, ground, and aircraft reports, and unconventional observations such as satellites, radars, and occultations) are adjusted with bias correction coefficients through observation operators and model constraints.

[0132] 8) After the minimization iteration converges, the output deviation correction coefficient increment δβ.

[0133] 9) Update the deviation correction coefficient β = β b +δβ, for satellite data bias correction in the next time window.

[0134] In order to solve the problem of satellite observation data drift and improve the application ability of satellite observation data, this application has developed a variational bias correction scheme suitable for the global four-dimensional variational assimilation system of the China Meteorological Administration. The coefficient of the prediction factor of the satellite observation data bias correction equation (referred to as the bias correction coefficient) is used as a control parameter, and a new control variable is defined together with the original analysis variable, which solves the parameter background error estimation problem and adopts an incremental form; the variational bias correction is carried out simultaneously with the variational analysis, that is, the bias correction is carried out in the minimization iteration process and is used as part of the solution to the variational problem. In the assimilation process, all observation data (including conventional observations such as sounding, ground, and aircraft reports, and unconventional observations such as satellites, radars, and occultations) affect the bias correction effect through observation operators and model constraints, thereby improving the assimilation analysis quality of the observation data and improving the level of numerical weather forecasting.

[0135] The Numerical Forecast Center of the China Meteorological Administration currently adopts a dynamic bias correction scheme, which uses the difference between recent (10 days) satellite observations and simulated observations (background fields) as historical samples, and uses a one-dimensional variational method before the four-dimensional variational data assimilation system is started to dynamically update the bias correction coefficients; the variational bias correction method provided in this application belongs to an adaptive on-line technology. Compared with the dynamic bias correction scheme, the variational bias correction method can not only improve the observation data drift problem caused by seasonal changes, but also improve the short-term disturbance of observation data accuracy caused by solar activity, etc.

[0136] The results of a one-year cyclic assimilation analysis forecast impact test show that after starting the variational bias correction method provided by this application, compared with the existing dynamic bias correction scheme, the model high-level and mid-tropospheric mass and wind field 0-hour to 240-hour forecast biases are overall reduced, far exceeding the confidence interval. The forecast scoring results show that most layers of the mass field and wind field, the main body of the northern and southern hemispheres and tropical regions show improvements, among which the quality field effect in the tropical region is the most significant. After testing each link, the results obtained by applying the variational bias correction method to the CMA-GFS four-dimensional variational assimilation system are correct. This method can not only continuously update the deviation amount during the minimization process, but also automatically sense the changes in the deviation of a specific channel under a given sensor and give corresponding corrections. The assimilation experiment also shows that compared with the dynamic bias correction, the variational bias correction scheme can automatically and quickly optimize the bias correction equation, maintain the effect of the bias correction, and run stably.

[0137] Based on the same inventive concept, the embodiment of the present application also provides a variational deviation correction device for implementing the variational deviation correction method involved above. The implementation scheme for solving the problem provided by the device is similar to the implementation scheme recorded in the above method, so the specific limitations in one or more variational deviation correction device embodiments provided below can refer to the limitations of the variational deviation correction method above, and will not be repeated here.

[0138] In an exemplary embodiment, Figure 4 As shown, a variational deviation correction device is provided, including: a data acquisition module 401, a deviation correction amount calculation module 402, a quality control module 403 and a deviation correction coefficient update module 404.

[0139] The data acquisition module 401 is used to obtain the observation data of the current time window and the deviation correction coefficient updated in the previous time window.

[0140] The deviation correction calculation module 402 is used to calculate the satellite observation deviation correction based on the observation information, the model background field and the deviation correction coefficient updated in the previous time window.

[0141] The quality control module 403 is used to calculate the deviation between the simulated observation and the actual observation based on the observation operator according to the satellite observation deviation correction amount and the observation data, and to correct and quality control the deviation to obtain the satellite observation data after quality control.

[0142] The deviation correction coefficient updating module 404 is used to update the deviation correction coefficient online based on the satellite observation data after quality control, based on the improved target functional and its gradient, in the minimization iteration process, under the joint constraints of the observation data and the numerical weather forecast model, to obtain the deviation correction coefficient updated in the current time window.

[0143] The improvement process of the target functional includes: taking the deviation correction coefficient as the control variable and combining the mode state variable to construct a new target functional; transforming the new target functional into an incremental form and performing variable transformation to obtain the improved target functional and its gradient.

[0144] In an exemplary embodiment, a computer device is provided, including a memory and a processor, wherein a computer program is stored in the memory, and the processor implements the steps in the above-mentioned method embodiments when executing the computer program.

[0145] In an exemplary embodiment, a computer-readable storage medium is provided, storing a computer program, and when the computer program is executed by a processor, the steps in the above method embodiments are implemented.

[0146] In an exemplary embodiment, a computer program product is provided, including a computer program, and when the computer program is executed by a processor, the steps in the above method embodiments are implemented.

[0147] It should be noted that the user information (including but not limited to user device information, user personal information, etc.) and data (including but not limited to data used for analysis, stored data, displayed data, etc.) involved in this application are all information and data authorized by the user or fully authorized by all parties, and the collection, use and processing of relevant data must comply with relevant regulations.

[0148] In this application, all actions to obtain signals, information or data are carried out in compliance with the relevant data protection laws and policies of the country where they are located and with the authorization given by the owner of the corresponding device.

[0149] Those of ordinary skill in the art can understand that all or part of the processes in the above-mentioned embodiment methods can be completed by instructing the relevant hardware through a computer program, and the computer program can be stored in a non-volatile computer-readable storage medium. When the computer program is executed, it can include the processes of the embodiments of the above-mentioned methods. Among them, any reference to the memory, database or other medium used in the embodiments provided in the present application can include at least one of non-volatile and volatile memory. Non-volatile memory can include read-only memory (ROM), magnetic tape, floppy disk, flash memory, optical memory, high-density embedded non-volatile memory, resistive random access memory (ReRAM), magnetic random access memory (MRAM), ferroelectric random access memory (FRAM), phase change memory (PCM), graphene memory, etc. Volatile memory can include random access memory (RAM) or external cache memory, etc. By way of illustration and not limitation, RAM may be in various forms, such as static random access memory (SRAM) or dynamic random access memory (DRAM).

[0150] The database involved in each embodiment provided in this application may include at least one of a relational database and a non-relational database. The non-relational database may include a distributed database based on blockchain, etc., but is not limited thereto. The processor involved in each embodiment provided in this application may be a general-purpose processor, a central processing unit, a graphics processor, a digital signal processor, a programmable logic device, a data processing logic device based on quantum computing, etc., but is not limited thereto.

[0151] The technical features of the above embodiments may be combined arbitrarily. To make the description concise, not all possible combinations of the technical features in the above embodiments are described. However, as long as there is no contradiction in the combination of these technical features, they should be considered to be within the scope of this specification.

[0152] This article uses specific examples to illustrate the principles and implementation methods of this application. The description of the above embodiments is only used to help understand the method and core ideas of this application. At the same time, for those skilled in the art, according to the ideas of this application, there will be changes in the specific implementation methods and application scope. In summary, the content of this specification should not be understood as limiting this application.

Claims

1. A variational bias correction method, characterized in that: The variational bias correction method comprises: Obtain the observation data of the current time window and the deviation correction coefficient updated in the previous time window; Calculate the satellite observation bias correction based on the observation information, model background field and the bias correction coefficient updated in the previous time window; According to the satellite observation deviation correction amount and the observation data, the deviation between the simulated observation and the actual observation is calculated based on the observation operator, and the deviation is corrected and quality controlled to obtain the satellite observation data after quality control; According to the satellite observation data after quality control, based on the improved target functional and its gradient, in the minimization iteration process, under the joint constraints of the observation data and the numerical weather forecast model, the bias correction coefficient is updated online to obtain the updated bias correction coefficient of the current time window; The improvement process of the target functional includes: using the deviation correction coefficient as the control variable and the mode state variable to construct a new target functional, transforming the new target functional into an incremental form and performing variable transformation to obtain the improved target functional and its gradient.

2. The variational deviation correction method according to claim 1, characterized in that: The deviation correction coefficient is used as the control variable and the mode state variable is combined to construct a new target functional, including: Define the satellite observation bias correction equation based on the bias correction coefficient; Based on the satellite observation bias correction equation, the observation term of the target functional of the differential assimilation is modified to obtain a modified observation term; Define parameter terms based on deviation correction coefficients; A new target functional is constructed according to the background term of the assimilated target functional, the modified observation term and the parameter term.

3. The variational deviation correction method according to claim 2, characterized in that: The satellite observation bias correction equation is: Among them, bias(β) is the satellite observation bias correction, b scan is the scanning angle correction, b air Corrected for air mass, N p is the number of air mass correction factors, β i is the correction factor coefficient of the ith air mass, p i (x b ) is the correction factor of the ith air mass, x b For background vector.

4. The variational deviation correction method according to claim 3, characterized in that: The new target functional is: Among them, J(x,β) is the value of the new target functional, is the background term of the variational assimilation objective functional, is a parameter item, is the modified observation item, x is the original analysis variable vector, B x is the background error covariance matrix, β is the deviation correction coefficient, β b is the initial estimate of β, B β is the parameter background error covariance matrix, H x (x) is the observation operator, y o is the observation vector, and R is the observation error covariance matrix.

5. The variational deviation correction method according to claim 4, characterized in that: Transform the new target functional into incremental form, including: Taylor expansion is performed on the observation operator and the satellite observation bias correction, and the first-order term is retained to obtain the incremental observation operator and the incremental satellite observation bias correction; According to the incremental observation operator and the incremental satellite observation bias correction amount, the new target functional is deformed by using a series of transformations, and the parameter background error covariance matrix is ​​adjusted to a diagonal matrix to obtain the target functional in incremental form.

6. The variational deviation correction method according to claim 5, characterized in that: The incremental observation operator is: H x (x)=H x (x b +δx)≈H x (x b )+H x ′δx; Among them, δx is the assimilation increment, δx=xx b , H x ′ is the gradient of the observation operator with respect to the analysis variable, The incremental satellite observation bias correction is: bias(β)=b scan +b air (b)=b scan +b air (b b +δβ)≈b scan +b air (b b )-H′ β db; Among them, δβ is the deviation correction coefficient increment, δβ=β-β b , H′ β is the negative gradient of the observation operator with respect to the parameter, 7. The variational deviation correction method according to claim 6, characterized in that: The target functional in the incremental form is: Among them, J(δx,δβ) is the value of the target functional in incremental form, is the corrected innovation vector, 8. The variational deviation correction method according to claim 7, characterized in that: The variable transformation of the incremental target functional is performed to obtain the improved target functional and its gradient, including: Decomposing the background error covariance matrix and the parameter background error covariance matrix to obtain a background error covariance decomposition matrix and a parameter background error covariance decomposition matrix; Defining new control variables according to the background error covariance decomposition matrix and the parameter background error covariance decomposition matrix; According to the background error covariance decomposition matrix, the parameter background error covariance decomposition matrix and the new control variable, the target functional in incremental form is transformed to obtain an improved target functional and its gradient.

9. The variational deviation correction method according to claim 8, characterized in that: The improved objective functional is: Among them, J inc is the value of the improved objective functional, w x is the new control variable corresponding to the original analysis variable, w x =U x δx,U x is the background error covariance decomposition matrix, w β is the new control variable corresponding to the deviation correction parameter, w β =U β δβ,U β is the parameter background error covariance decomposition matrix, The gradient of the control variable of the improved objective functional is: in, w x The gradient of w β gradient.

10. A variational deviation correction device, applied to the variational deviation correction method according to any one of claims 1 to 9, characterized in that: The variational deviation correction device comprises: The data acquisition module is used to obtain the observation data of the current time window and the deviation correction coefficient updated in the previous time window; The deviation correction calculation module calculates the satellite observation deviation correction based on the observation information, the model background field and the deviation correction coefficient updated in the previous time window; A quality control module is used to calculate the deviation between the simulated observation and the actual observation based on the observation operator according to the satellite observation deviation correction amount and the observation data, and to correct and quality control the deviation to obtain the satellite observation data after quality control; The bias correction coefficient update module is used to update the bias correction coefficient online based on the satellite observation data after quality control, based on the improved target functional and its gradient, in the minimization iteration process, under the joint constraints of the observation data and the numerical weather forecast model, to obtain the updated bias correction coefficient of the current time window; The improvement process of the target functional includes: using the deviation correction coefficient as the control variable and the mode state variable to construct a new target functional, transforming the new target functional into an incremental form and performing variable transformation to obtain the improved target functional and its gradient.

Images