Integrated ensemble updating assimilation method based on deterministic variational perturbation

Abstract

The present invention discloses an integrated ensemble update assimilation method based on deterministic variational perturbation, including: taking the ensemble forecast field of the initial ensemble members as the background field set to perform ensemble-variation hybrid assimilation to update the ensemble average; calculating the new information vector based on the forward observation of the ensemble initial guess field and the analysis field, and the forward observation of the ensemble members, taking the ensemble average analysis field as the initial guess field, performing minimization analysis using the ensemble perturbation update function based on deterministic variational perturbation, obtaining the updated ensemble perturbation and superimposing it with the updated ensemble average to obtain the ensemble analysis field; taking the updated ensemble average and the ensemble analysis field as the initial field, respectively performing deterministic forecast and ensemble forecast. The present invention improves the continuity and balance characteristics of the ensemble perturbation and ensemble average variables in hybrid assimilation, provides high-quality flow-dependent background error covariance for the transmission of observation information in the assimilation system, and thus improves the level of data assimilation and numerical weather forecasting.

Description

Technical Field

[0001] The invention belongs to the field of atmospheric science technology, and specifically relates to an integrated ensemble update assimilation method based on deterministic variational perturbations. Background Art

[0002] Hybrid data assimilation in numerical weather forecasting is a new data assimilation scheme developed by combining the advantages of ensemble forecasting and variational assimilation methods using an ensemble-variational hybrid assimilation system. Based on the variational data assimilation framework, it takes advantage of the ensemble forecast disturbance field information to construct a background error covariance structure with flow-dependent properties, overcomes the defects of fixed, uniform and isotropic background error covariance in variational assimilation, improves the ability to express the forecast error covariance of different weather systems, and thus improves the quality of analysis and forecasting. In recent years, hybrid data assimilation based on ensemble and variational methods has quickly become a new trend in the development of numerical forecast data assimilation, with good business application potential.

[0003] However, the current ensemble-variation hybrid assimilation system is implemented by combining the Ensemble Kalman Filter (EnKF) system and the variational system. The EnKF system exists independently of the variational system. The two have their own observation operators and different technical implementation schemes, which increases the complexity of system maintenance and is obviously very time-consuming and labor-intensive. Coupling two different systems together will also cause some system errors.

[0004] Moreover, the EnKF method also has a series of problems, such as systematic underestimation of covariance, sampling error, uncoordinated dynamic balance, and localization of observation space that is not suitable for satellite data. In addition, the ensemble mean analysis field and ensemble perturbation analysis field generated by the traditional hybrid assimilation framework of two different mechanisms and their superposition may have discontinuity and imbalance problems.

[0005] To address these problems, there is currently a variational ensemble error covariance update method that uses a cost function to directly perturb observations. This method is equivalent to the original EnKF method of perturbed observations, but it will introduce additional new problems such as sampling error and underestimated error covariance. There is also an ensemble root mean square filter, or EnSRF, which is a method that can improve problems such as sampling error and underestimated ensemble error covariance introduced by perturbed observations. However, this method is difficult to derive the form of the variational cost function, that is, it is difficult to implement EnSRF within the variational framework. Summary of the invention

[0006] The present invention aims to solve one of the technical problems existing in the related art at least to a certain extent.

[0007] The present invention aims to provide an integrated ensemble update assimilation method based on deterministic variational perturbations, which updates the ensemble perturbations through a deterministic variational perturbation scheme, updates the ensemble perturbations and the ensemble average under a unified variational framework, and further improves the assimilation and forecasting level of the numerical forecasting system on the basis of the ensemble-variational assimilation method.

[0008] In order to achieve the above-mentioned object, the present invention provides, on one hand, an integrated ensemble updating assimilation method based on deterministic variational perturbation, comprising the following steps:

[0009] S1. Obtain the global ensemble forecast analysis field as the initial ensemble member, and perform ensemble forecast to obtain the initial background field set for assimilation analysis cycle;

[0010] S2, using the background field ensemble obtained in step S1 to calculate the ensemble mean and ensemble disturbance, perform ensemble-variational hybrid assimilation to update the ensemble mean, and save the ensemble mean analysis field in the control variable space, the forward observation of the ensemble mean initial guess field, and the forward observation of the ensemble mean analysis field;

[0011] S3, based on the variational assimilation outer loop, the average analysis field in the control variable space is used as the initial guess field; at the same time, the forward observation of the ensemble average initial guess field saved in step S2, the forward observation of the ensemble average analysis field and the forward observation of the ensemble member to be updated are used to recalculate the new information vector;

[0012] S4. Perform minimization analysis through the deterministic variational set perturbation update function, superimpose the analysis increments on the set members and then subtract the set average to obtain the updated set perturbation;

[0013] S5. Based on the ensemble perturbations before and after the update, the flow-dependent background error covariance relaxation expansion method is used to obtain the final ensemble perturbation field, and the updated ensemble mean and ensemble perturbation are superimposed to obtain the final ensemble analysis field;

[0014] S6. Use the updated ensemble mean obtained in step S2 as the initial field for deterministic forecasting, use the final ensemble analysis field obtained in step S5 as the initial ensemble member for ensemble forecasting, and the obtained ensemble forecast field is used for the next assimilation analysis cycle according to steps S2 to S6.

[0015] A further preferred technical solution of the present invention is that, in step S1, the global ensemble forecast analysis field is obtained as an initial ensemble member, and an ensemble forecast is performed to obtain an initial background field set for the assimilation analysis cycle; specifically,

[0016] Get Global Ensemble Forecast Analysis Fields As the initial ensemble members, regional dynamical downscaling ensemble forecast is carried out to obtain the initial background field ensemble members for assimilation analysis cycle .

[0017] Preferably, step S2 uses the background field set obtained in step S1 to calculate the ensemble mean and ensemble disturbance, performs ensemble-variational hybrid assimilation to update the ensemble mean, and saves the ensemble mean analysis field in the control variable space, the forward observation of the ensemble mean initial guess field, and the forward observation of the ensemble mean analysis field; specifically,

[0018] S21. Calculate the ensemble average based on the background field set , and then subtract the ensemble mean from each ensemble member to get the ensemble perturbation ;

[0019] S22. Take the ensemble mean as the initial guess field, and the ensemble perturbation provides the flow-dependent background error covariance. Combined with the balance constraint of the static background error covariance, read the multi-source observation data, perform ensemble-variational hybrid assimilation, and obtain the optimal analysis field of the ensemble mean. ;

[0020] S23, save the ensemble average analysis field of the control variable space at the same time , forward observation of the ensemble mean initial guess field Forward observation of ensemble mean analysis field .

[0021] Preferably, the variational assimilation outer loop described in step S3 uses the average analysis field in the control variable space as the initial guess field; and simultaneously uses the forward observation of the ensemble average initial guess field saved in step S2, the forward observation of the ensemble average analysis field, and the forward observation of the ensemble member to be updated to recalculate the innovation vector; specifically:

[0022] S31, based on the outer loop of variational assimilation, with each set member As the background field, read the ensemble average analysis field saved in step S23 , forward observation of the ensemble mean initial guess field Forward observation of ensemble mean analysis field ;

[0023] S32, Analyze the field with ensemble average As the initial guess of the variational iteration during the disturbance update; based on the ensemble mean analysis field , forward observation of the ensemble mean initial guess field , forward observation of ensemble mean analysis field , and the forward observations of the set members to be updated , recalculate the innovation vector .

[0024] Preferably, in step S4, the deterministic variational set perturbation update function is used to perform minimization analysis, the analysis increment is superimposed on the set members and then the set average is subtracted to obtain the updated set perturbation; specifically:

[0025] S41, for each set member, use the set perturbation except the members of this set Provides flow-dependent background error covariance and localizes the pattern space, combined with the static background error covariance , through the deterministic variational set perturbation update function, expressed as:

[0026] ;

[0027] in, Extend control variables for collections, is the covariance weight coefficient, is the analytical field in the control variable space, is a localization matrix, which constrains the spatial variation of the set expansion control variable 𝛂 and localizes the set error covariance. is the observation operator that projects the background field into the observation space, is the control variable conversion matrix, is the collective perturbation, is the observation error covariance matrix, is the new information vector, and the calculation formula is:

[0028] ;

[0029] in, is the forward observation of the ensemble mean analysis field, is the forward observation of the ensemble mean initial guess field, is the forward observation of the set member to be updated;

[0030] S42, perform minimization analysis to obtain the analytical increment of the disturbance field , superimpose the analysis increment on the members of this set, traverse each member, and obtain the updated set members , minus the ensemble mean, leaving the updated ensemble perturbation .

[0031] Preferably, in step S5, based on the ensemble disturbances before and after the update, a flow-dependent background error covariance relaxation expansion method is used to obtain the final ensemble disturbance field, and the updated ensemble mean and ensemble disturbance are superimposed to obtain the final ensemble analysis field; specifically,

[0032] Based on the set perturbation before update and the updated set perturbation , using the flow-dependent background error covariance relaxation expansion method, the final updated collective disturbance field is obtained , superimpose the updated ensemble mean and the updated ensemble disturbance field to obtain the final ensemble analysis field .

[0033] Preferably, in step S6, the updated ensemble mean obtained in step S2 is used as the initial field for deterministic forecasting, and the final ensemble analysis field obtained in step S5 is used as the initial ensemble member for ensemble forecasting, and the obtained ensemble forecast field is subjected to the next assimilation analysis cycle through steps S2 to S6; specifically, as follows:

[0034] Updated ensemble average As the initial field, a deterministic forecast is made for n days, and the forecast results are output every t hours; and the final ensemble analysis field is used as the The ensemble forecast with a forecast validity of t hours is made as the initial field, and the obtained ensemble forecast field is passed to the next assimilation analysis cycle.

[0035] The present invention also provides a non-transitory computer-readable storage medium on which computer instructions are stored, and the computer instructions enable the computer to execute the above-mentioned integrated set update assimilation method based on deterministic variational perturbations.

[0036] Yet another aspect of the present invention provides an electronic device, comprising: a processor, a communication interface, a memory and a communication bus, wherein the processor, the communication interface and the memory communicate with each other via the communication bus, and the processor calls the logic instructions in the memory to execute the above-mentioned integrated set update assimilation method based on deterministic variational perturbations.

[0037] On the other hand, the present invention provides a computer program product, which includes a computer program stored on a non-transitory computer-readable storage medium. When the computer program is executed by a processor, the computer executes the above-mentioned integrated ensemble update assimilation method based on deterministic variational perturbations.

[0038] Beneficial effect: The present invention integrates the ensemble perturbation update process in ensemble-variational hybrid assimilation into the variational framework, and realizes the update of the background field and its flow-dependent error covariance in a unified variational framework. This method is the implementation of Monte Carlo theory in the variational framework, and its theoretical basis is equivalent to the ensemble perturbation update of ensemble Kalman filtering.

[0039] The variational system's inherent ability to coordinate and assimilate multi-source observations under equilibrium constraints gives the variational perturbation method a potential advantage. Updating the ensemble perturbation and ensemble average under a unified variational framework is conducive to enhancing the representativeness of the ensemble perturbation for the error characteristics of the variational framework and improving the problem of analytical variable discontinuity that occurs when the ensemble perturbation field and the average field updated by different frameworks are superimposed in the traditional scheme. BRIEF DESCRIPTION OF THE DRAWINGS

[0040] Figure 1 It is a flow chart of the integrated ensemble update assimilation method based on deterministic variational perturbation of the present invention;

[0041] Figure 2 The parameter calculation flow chart of the integrated ensemble update assimilation method based on deterministic variational perturbation of the present invention. DETAILED DESCRIPTION

[0042] In order to make the purpose, technical solution and advantages of the present invention clearer, the technical solution of the present invention will be clearly and completely described below in conjunction with the drawings in the present invention. Obviously, the described embodiments are part of the embodiments of the present invention, not all of the embodiments, and they should not be understood as limitations on the present invention. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention. In the description of the present invention, it should be understood that the terms used are only for descriptive purposes and cannot be understood as indicating or implying relative importance.

[0043] Combine the following Figure 1-Figure 2 The present invention describes an integrated ensemble update assimilation method based on deterministic variational perturbations, a non-transitory computer-readable storage medium, an electronic device, and a computer program product.

[0044] Prior to this, technical terms that may be used in the present invention are first explained.

[0045] Background Field: Background field refers to the estimate of the system state obtained based on the numerical model prediction before considering the latest observation data. It is the prior information in the assimilation system and represents the best guess of the system state before adding the observation data. The background field usually comes from the previous analysis field or a forecast result integrated over a longer time.

[0046] Observation Field: An observation field is a field consisting of actual observation data, which represents the actual value of the system state observed at a specific time and place. The data of the observation field may include meteorological elements such as temperature, pressure, humidity, wind speed, or ocean currents, salinity, temperature, etc. in oceanography. Observation data are usually used to verify and improve the estimation of the background field.

[0047] Analysis Field: The analysis field is the result of combining the background field and the observation field through the data assimilation process. It is an optimal estimate that takes into account both the prediction of the model and the information of the observation data. The analysis field aims to provide the most accurate description of the system state and is usually more reliable than the background field or the observation field alone.

[0048] Forecast Field: Forecast field refers to the result of further predicting the future system state through numerical models based on the analysis field. The accuracy of the forecast field depends on the quality of the analysis field and the forecasting ability of the model.

[0049] Error Covariance Matrix: In the assimilation process, the error covariance matrix describes the statistical characteristics of the system state estimation error, including the variance of the error and the covariance between different variables. The background error covariance matrix and the observation error covariance matrix are two important components of the assimilation system, which affect the estimation of the analysis field.

[0050] Innovation Vector: The innovation vector is the difference between the observed field and the background field, usually expressed as the observed value minus the value obtained by the observation operator from the background field. The innovation vector reflects the inconsistency between the observed data and the background field prediction.

[0051] Observation Operator: An observation operator is a mathematical model that defines how to derive observation data from system state variables. During the assimilation process, the observation operator is used to calculate the expected value of the background field at the observation point in order to compare it with the actual observation data.

[0052] Forward Modeling: Forward modeling is a term used in geophysical exploration research. It refers to the process of calculating the theoretical value (mathematical simulation) of a geological body by constructing a mathematical model, or observing the numerical value (physical simulation) of the geophysical effect produced by the model by constructing a physical model based on the shape, occurrence and physical property data of the geological body. In this invention, it refers to the process of using ensemble members to simulate in an ensemble-variation hybrid assimilation system to generate simulated observation data, which can then be used in the assimilation process to improve the estimation of the real system state.

[0053] Embodiment 1: This embodiment provides an integrated set update assimilation method for deterministic variational perturbations.

[0054] In the traditional ensemble-variation hybrid assimilation system, the EnKF system exists independently of the variational system, and the two have their own observation operators and different technical implementation schemes, which increases the complexity of system maintenance. The integrated ensemble update assimilation method of deterministic variational perturbations provided in this embodiment can significantly reduce the complexity of hybrid assimilation system development and maintenance, and the background error covariance provided provides high-quality flow-dependent information for the transmission of observation information in the assimilation system, thereby providing a more reasonable and convenient new hybrid assimilation scheme for numerical weather forecasting.

[0055] The integrated ensemble update assimilation method of deterministic variational perturbations provided in this embodiment can directly utilize the dynamic or physical balance constraints in the variational framework, which is beneficial to improving the balance characteristics of the disturbance variables.

[0056] The integrated ensemble updating assimilation method based on deterministic variational perturbation in this embodiment is as follows: Figure 1-Figure 2 As shown, the following steps are included:

[0057] S1. Acquisition Global Ensemble Forecast Analysis Fields As the initial ensemble members, regional dynamical downscaling ensemble forecast is carried out to obtain the initial background field ensemble members for assimilation analysis cycle ;

[0058] In this embodiment, the number of members of the initial forecast field set is 40; the forecast field assimilation interval is 6 hours.

[0059] S2, using the background field ensemble obtained in step S1 to calculate the ensemble mean and ensemble disturbance, perform ensemble-variational hybrid assimilation to update the ensemble mean, and save the ensemble mean analysis field in the control variable space, the forward observation of the ensemble mean initial guess field, and the forward observation of the ensemble mean analysis field; specifically:

[0060] S21. Calculate the ensemble average based on the background field set , and then subtract the ensemble mean from each ensemble member to get the ensemble perturbation ;

[0061] S22, using the ensemble mean as the initial guess, the ensemble perturbation provides the flow-dependent background error covariance, combined with the static background error covariance The balance constraint is , in which the weights of static background error covariance and ensemble background error covariance are half each, and the local scale is 200km; read multi-source observation data (mainly including land reports and ship reports, sounding data, aviation routine weather reports, aircraft reports, satellite thickness data, ship data, wind measurement data, buoy data, wind profile data, etc.), perform ensemble-variation mixed assimilation, and obtain the optimal analysis field of the ensemble average , which is the updated ensemble average;

[0062] At the same time, save the ensemble mean analysis field of the control variable space , forward observation of the ensemble mean initial guess field and the forward observations of the ensemble mean analysis field and .

[0063] S3, based on the outer loop of variational assimilation, the average analysis field in the control variable space is used as the initial guess field; at the same time, the forward observation of the ensemble average initial guess field saved in step S2, the forward observation of the ensemble average analysis field and the forward observation of the ensemble member to be updated are used to recalculate the innovation vector; specifically:

[0064] S31, based on the outer loop of variational assimilation, with each set member As the background field, read the ensemble average analysis field saved in step S23 , forward observation of the ensemble mean initial guess field and the forward observations of the ensemble mean analysis field and ;

[0065] S32, Analyze the field with ensemble average As the initial guess of the variational iteration during the disturbance update; based on the ensemble mean analysis field , forward observation of the ensemble mean initial guess field , forward observation of ensemble mean analysis field , and the forward observations of the set members to be updated , recalculate the innovation vector .

[0066] S4. Perform minimization analysis through the deterministic variational set perturbation update function, superimpose the analysis increment on the set members and then subtract the set average to obtain the updated set perturbation; specifically:

[0067] S41, for each set member, use the set perturbation except the members of this set Provides flow-dependent background error covariance and localizes the pattern space, combined with the static background error covariance , through the deterministic variational set perturbation update function, expressed as:

[0068] ;

[0069] in, Extend control variables for collections, is the covariance weight coefficient, is the analytical field in the control variable space, is a localization matrix, which constrains the spatial variation of the set expansion control variable 𝛂 and localizes the set error covariance. is the observation operator that projects the background field into the observation space, is the control variable conversion matrix, is the collective perturbation, is the observation error covariance matrix, is the new information vector, and the calculation formula is:

[0070] ;

[0071] in, is the forward observation of the ensemble mean analysis field, is the forward observation of the ensemble mean initial guess field, is the forward observation of the set member to be updated;

[0072] S42, perform minimization analysis to obtain the analytical increment of the disturbance field , superimpose the analysis increment on the members of this set, traverse each member, and obtain the updated set members , minus the ensemble mean, leaving the updated ensemble perturbation .

[0073] S5. Based on the set perturbation before updating and the updated set perturbation , using the flow-dependent background error covariance relaxation expansion method, the final updated collective disturbance field is obtained , superimpose the updated ensemble mean and the updated ensemble disturbance field to obtain the final ensemble analysis field .

[0074] S6, using the updated ensemble average The initial field is used for a 2-day deterministic forecast, with the forecast results output every 6 hours; and the final ensemble analysis field is used as the An ensemble forecast with a forecast validity of 6 hours is made as the initial field.

[0075] S7. The obtained ensemble forecast field is subjected to the next assimilation analysis cycle according to steps S2 to S6.

[0076] Embodiment 2: This embodiment provides a non-transitory computer-readable storage medium having computer instructions stored thereon, the computer instructions causing the computer to execute an integrated ensemble update assimilation method based on deterministic variational perturbations, the method comprising the following steps:

[0077] S1. Obtain the global ensemble forecast analysis field as the initial ensemble member, and perform ensemble forecast to obtain the initial background field set for assimilation analysis cycle;

[0078] S2, using the background field ensemble obtained in step S1 to calculate the ensemble mean and ensemble disturbance, perform ensemble-variational hybrid assimilation to update the ensemble mean, and save the ensemble mean analysis field in the control variable space, the forward observation of the ensemble mean initial guess field, and the forward observation of the ensemble mean analysis field;

[0079] S3, based on the variational assimilation outer loop, the average analysis field in the control variable space is used as the initial guess field; at the same time, the forward observation of the ensemble average initial guess field saved in step S2, the forward observation of the ensemble average analysis field and the forward observation of the ensemble member to be updated are used to recalculate the new information vector;

[0080] S4. Perform minimization analysis through the deterministic variational set perturbation update function, superimpose the analysis increments on the set members and then subtract the set average to obtain the updated set perturbation;

[0081] S5. Based on the ensemble perturbations before and after the update, the flow-dependent background error covariance relaxation expansion method is used to obtain the final ensemble perturbation field, and the updated ensemble mean and ensemble perturbation are superimposed to obtain the final ensemble analysis field;

[0082] S6. Use the updated ensemble mean obtained in step S2 as the initial field for deterministic forecasting, use the final ensemble analysis field obtained in step S5 as the initial ensemble member for ensemble forecasting, and the obtained ensemble forecast field is used for the next assimilation analysis cycle according to steps S2 to S6.

[0083] Embodiment 3: This embodiment provides an electronic device, which may include: a processor, a communication interface, a memory, and a communication bus, wherein the processor, the communication interface, and the memory communicate with each other through the communication bus. The processor may call the logic instructions in the memory to execute an integrated set update assimilation method based on deterministic variational perturbations, and the method includes the following steps:

[0084] S1. Obtain the global ensemble forecast analysis field as the initial ensemble member, and perform ensemble forecast to obtain the initial background field set for assimilation analysis cycle;

[0085] S2, using the background field ensemble obtained in step S1 to calculate the ensemble mean and ensemble disturbance, perform ensemble-variational hybrid assimilation to update the ensemble mean, and save the ensemble mean analysis field in the control variable space, the forward observation of the ensemble mean initial guess field, and the forward observation of the ensemble mean analysis field;

[0086] S3, based on the variational assimilation outer loop, the average analysis field in the control variable space is used as the initial guess field; at the same time, the forward observation of the ensemble average initial guess field saved in step S2, the forward observation of the ensemble average analysis field and the forward observation of the ensemble member to be updated are used to recalculate the new information vector;

[0087] S4. Perform minimization analysis through the deterministic variational set perturbation update function, superimpose the analysis increments on the set members and then subtract the set average to obtain the updated set perturbation;

[0088] S5. Based on the ensemble perturbations before and after the update, the flow-dependent background error covariance relaxation expansion method is used to obtain the final ensemble perturbation field, and the updated ensemble mean and ensemble perturbation are superimposed to obtain the final ensemble analysis field;

[0089] S6. Use the updated ensemble mean obtained in step S2 as the initial field for deterministic forecasting, use the final ensemble analysis field obtained in step S5 as the initial ensemble member for ensemble forecasting, and the obtained ensemble forecast field is used for the next assimilation analysis cycle according to steps S2 to S6.

[0090] In addition, the logic instructions in the above-mentioned memory can be implemented in the form of software functional units and can be stored in a computer-readable storage medium when sold or used as an independent product. Based on this understanding, the technical solution of the present invention, in essence, or the part that contributes to the prior art or the part of the technical solution, can be embodied in the form of a software product, which is stored in a storage medium and includes several instructions for a computer device (which can be a personal computer, a server, or a network device, etc.) to perform all or part of the steps of the method described in each embodiment of the present invention. The aforementioned storage medium includes: U disk, mobile hard disk, read-only memory (ROM, Read-Only Memory), random access memory (RAM, Random Access Memory), disk or optical disk, etc., which can store program code.

[0091] Embodiment 4: This embodiment provides a computer program product, which includes a computer program. The computer program can be stored on a non-transitory computer-readable storage medium. When the computer program is executed by a processor, the computer can execute an integrated ensemble update assimilation method based on deterministic variational perturbations. The method includes the following steps:

[0092] S1. Obtain the global ensemble forecast analysis field as the initial ensemble member, and perform ensemble forecast to obtain the initial background field set for assimilation analysis cycle;

[0093] S2, using the background field ensemble obtained in step S1 to calculate the ensemble mean and ensemble disturbance, perform ensemble-variational hybrid assimilation to update the ensemble mean, and save the ensemble mean analysis field in the control variable space, the forward observation of the ensemble mean initial guess field, and the forward observation of the ensemble mean analysis field;

[0094] S3, based on the variational assimilation outer loop, the average analysis field in the control variable space is used as the initial guess field; at the same time, the forward observation of the ensemble average initial guess field saved in step S2, the forward observation of the ensemble average analysis field and the forward observation of the ensemble member to be updated are used to recalculate the new information vector;

[0095] S4. Perform minimization analysis through the deterministic variational set perturbation update function, superimpose the analysis increments on the set members and then subtract the set average to obtain the updated set perturbation;

[0096] S5. Based on the ensemble perturbations before and after the update, the flow-dependent background error covariance relaxation expansion method is used to obtain the final ensemble perturbation field, and the updated ensemble mean and ensemble perturbation are superimposed to obtain the final ensemble analysis field;

[0097] S6. Use the updated ensemble mean obtained in step S2 as the initial field for deterministic forecasting, use the final ensemble analysis field obtained in step S5 as the initial ensemble member for ensemble forecasting, and the obtained ensemble forecast field is used for the next assimilation analysis cycle according to steps S2 to S6.

[0098] The device embodiments described above are merely illustrative, wherein the units described as separate components may or may not be physically separated, and the components displayed as units may or may not be physical units, that is, they may be located in one place, or they may be distributed on multiple network units. Some or all of the modules may be selected according to actual needs to achieve the purpose of the scheme of this embodiment. Ordinary technicians in this field can understand and implement it without paying creative labor.

[0099] Through the description of the above implementation methods, those skilled in the art can clearly understand that each implementation method can be implemented by means of software plus a necessary general hardware platform, and of course, can also be implemented by hardware. Based on this understanding, the above technical solution is essentially or the part that contributes to the prior art can be embodied in the form of a software product, and the computer software product can be stored in a computer-readable storage medium, such as ROM / RAM, a disk, an optical disk, etc., including a number of instructions for a computer device (which can be a personal computer, a server, or a network device, etc.) to execute the methods described in each embodiment or some parts of the embodiments.

[0100] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention, rather than to limit it. Although the present invention has been described in detail with reference to the aforementioned embodiments, those skilled in the art should understand that they can still modify the technical solutions described in the aforementioned embodiments, or make equivalent replacements for some of the technical features therein. However, these modifications or replacements do not deviate the essence of the corresponding technical solutions from the spirit and scope of the technical solutions of the embodiments of the present invention.

Claims

1. An integrated ensemble update assimilation method based on deterministic variational perturbations, characterized in that: The steps include: S1. Obtain the global ensemble forecast analysis field as the initial ensemble member, and perform ensemble forecast to obtain the initial background field set for assimilation analysis cycle; S2, using the background field ensemble obtained in step S1 to calculate the ensemble mean and ensemble disturbance, perform ensemble-variational hybrid assimilation to update the ensemble mean, and save the ensemble mean analysis field in the control variable space, the forward observation of the ensemble mean initial guess field, and the forward observation of the ensemble mean analysis field; specifically: S21. Calculate the ensemble average based on the background field set Then subtract the collective mean from each member of the set to get the collective disturbance S22. Take the ensemble mean as the initial guess field, and the ensemble perturbation provides the flow-dependent background error covariance. Combined with the balance constraint of the static background error covariance, read the multi-source observation data, perform ensemble-variational hybrid assimilation, and obtain the optimal analysis field of the ensemble mean. S23, save the ensemble average analysis field of the control variable space at the same time Forward Observation of Ensemble Mean Initial Guess Field Forward observation of ensemble mean analysis field S3, based on the outer loop of variational assimilation, the average analysis field in the control variable space is used as the initial guess field; at the same time, the forward observation of the ensemble average initial guess field saved in step S2, the forward observation of the ensemble average analysis field and the forward observation of the ensemble member to be updated are used to recalculate the innovation vector; specifically: S31, based on the outer loop of variational assimilation, with each set member As the background field, read the ensemble average analysis field saved in step S23 Forward Observation of Ensemble Mean Initial Guess Field Forward observation of ensemble mean analysis field S32, Analyze the field with ensemble average As the initial guess of the variational iteration during the disturbance update; based on the ensemble mean analysis field Forward Observation of Ensemble Mean Initial Guess Field Forward Observation of Ensemble Mean Analysis Field And the forward observations of the set members to be updated Recalculate the new information vector d; S4. Perform minimization analysis through the deterministic variational set perturbation update function, superimpose the analysis increment on the set members and then subtract the set average to obtain the updated set perturbation; specifically: S41, for each set member, use the set perturbation except the members of this set Provide flow-dependent background error covariance and localize the pattern space, combined with the static background error covariance B, and update the function through deterministic variational set perturbation, expressed as: Among them, α i is the set expansion control variable, β is the covariance weight coefficient, v i is the analysis field of the control variable space, A is the localization matrix, which constrains the spatial variation of the set expansion control variable α and realizes the localization of the set error covariance, H is the observation operator that projects the background field to the observation space, and U is the control variable conversion matrix. is the collective disturbance, R is the observation error covariance matrix, is the new information vector, and the calculation formula is: in, is the forward observation of the ensemble mean analysis field, is the forward observation of the ensemble mean initial guess field, is the forward observation of the set member to be updated; S42, perform minimization analysis to obtain the analytical increment δx of the disturbance field i , superimpose the analysis increment on the members of this set, traverse each member, and obtain the updated set members After subtracting the ensemble mean, the updated ensemble perturbation remains S5. Based on the ensemble perturbations before and after the update, the flow-dependent background error covariance relaxation expansion method is used to obtain the final ensemble perturbation field, and the updated ensemble mean and ensemble perturbation are superimposed to obtain the final ensemble analysis field; S6. Use the updated ensemble mean obtained in step S2 as the initial field for deterministic forecasting, use the final ensemble analysis field obtained in step S5 as the initial ensemble member for ensemble forecasting, and the obtained ensemble forecast field is used for the next assimilation analysis cycle according to steps S2 to S6.

2. The integrated ensemble update assimilation method based on deterministic variational perturbation according to claim 1 is characterized in that: Step S1 is to obtain the global ensemble forecast analysis field as the initial ensemble member, and perform ensemble forecast to obtain the initial background field set for the assimilation analysis cycle; specifically: Get N global ensemble forecast analysis fields As the initial ensemble members, regional dynamical downscaling ensemble forecast is carried out to obtain the initial background field ensemble members for assimilation analysis cycle 3. The integrated ensemble update assimilation method based on deterministic variational perturbation according to claim 1 is characterized in that: In step S5, based on the ensemble disturbances before and after the update, the flow-dependent background error covariance relaxation expansion method is adopted to obtain the final ensemble disturbance field, and the updated ensemble mean and ensemble disturbance are superimposed to obtain the final ensemble analysis field; specifically: Based on the set perturbation before update and the updated set perturbation The final updated ensemble disturbance field is obtained by using the flow-dependent background error covariance relaxation expansion method. The updated ensemble mean and the updated ensemble disturbance field are superimposed to obtain the final ensemble analysis field 4. The integrated ensemble update assimilation method based on deterministic variational perturbation according to claim 1 is characterized in that: In step S6, the updated ensemble mean obtained in step S2 is used as the initial field for deterministic forecasting, and the final ensemble analysis field obtained in step S5 is used as the initial ensemble member for ensemble forecasting. The obtained ensemble forecast field is subjected to the next assimilation analysis cycle through steps S2 to S6; specifically, as follows: Updated ensemble average As the initial field, a deterministic forecast is made for n days, and the forecast results are output every t hours; and the final ensemble analysis field is used as the The ensemble forecast with a forecast validity of t hours is made as the initial field, and the obtained ensemble forecast field is passed to the next assimilation analysis cycle.

5. A non-transitory computer-readable storage medium having computer instructions stored thereon, the computer instructions enabling a computer to execute the integrated ensemble update assimilation method based on deterministic variational perturbations as described in claim 1.

6. An electronic device comprising: A processor, a communication interface, a memory and a communication bus, wherein the processor, the communication interface and the memory communicate with each other via the communication bus, and the processor calls the logic instructions in the memory to execute the integrated set update assimilation method based on deterministic variational perturbation as described in claim 1.

7. A computer program product, comprising a computer program stored on a non-transitory computer-readable storage medium, wherein when the computer program is executed by a processor, the computer executes the integrated ensemble update assimilation method based on deterministic variational perturbations as described in claim 1.

Images