A variational multiscale assimilation method and system based on multiscale control variables

Abstract

The invention belongs to the field of atmospheric science and technology, and discloses a variational multiscale assimilation method and system based on multiscale control variables, comprising: S1, generating background error forecast samples; S2, performing scale separation on background error samples to obtain large-scale and small-scale background error statistical samples; S3, constructing a multi-scale background error covariance matrix, and statistically analyzing large-scale background error and small-scale background error characteristics based on a control variable conversion algorithm; S4, establishing cross-scale correlation between large-scale and small-scale background errors in the background error covariance matrix; S5, based on a three-dimensional variational assimilation framework, introducing large-scale control variables and small-scale control variables to construct a multi-scale three-dimensional variational assimilation method; S6, introducing a numerical model background field and an observation field, solving the minimization of a multi-scale three-dimensional variational assimilation cost function, and realizing simultaneous and coordinated updating of weather information of different scales in a single assimilation analysis.

Description

Technical Field

[0001] The present invention belongs to the field of atmospheric science technology, and in particular relates to a variational multi-scale assimilation method and system based on multi-scale control variables. Background Art

[0002] Conducting research on multiscale assimilation methods to improve the multiscale analysis in the initial field of convective resolvable models is the key to improving the effectiveness of convective weather forecasting.

[0003] Traditional multiscale assimilation methods focus on how to extract weather information of different scales from multi-source observations. For example, conventional observations with a wide distribution but coarse temporal and spatial resolution and uneven distribution often contain weather information of larger scales, while observations with high temporal and spatial resolution, such as radar observations, contain more convective scale information. Multiscale analysis in variational methods usually adopts a step-by-step assimilation strategy, that is, first assimilate conventional observations with a wider distribution, and then assimilate observations with a denser distribution, such as radar observations, based on the previous analysis; in the process of step-by-step assimilation, the characteristic length scale in the background error covariance is gradually reduced, so as to achieve the effective assimilation of observation information of different scales by the assimilation system.

[0004] However, the step-by-step assimilation strategy ignores the multi-scale nature of the observation network. Conventional observations with sparse and uneven distribution also contain weather information at a smaller scale, while observation networks with high spatiotemporal resolution, such as Doppler radar observations, can also provide weather information at a larger scale. At the same time, when assimilating observations from different batches, the adjustment of the characteristic length scale of the background error covariance is subjective and empirical, and it is impossible to achieve objective and adaptive analysis of weather information at different scales. Therefore, in the variational assimilation framework, Li et al. (2015) decomposed the static background error covariance matrix into large-scale and small-scale parts, thereby decomposing the three-dimensional variational cost function into large-scale assimilation and small-scale assimilation. By minimizing the large-scale and small-scale cost functions, the large-scale and small-scale analysis increments are obtained, and the multi-scale state quantity is updated.

[0005] However, the control variables in the current variational framework are of a single scale, and the background error characteristics of weather at different scales have not been considered at the same time, so the multi-scale weather state cannot be updated at the same time. On the other hand, weather systems at different scales influence each other; this inter-scale correlation feature has not been fully considered in the variational framework, resulting in the inability to reasonably transfer large (small) scale analysis to small (large) scale analysis, which limits the application of multi-source observation data in variational assimilation and ensemble-variational mixed assimilation.

[0006] Through the above analysis, the problems and defects of the existing technology are as follows: in the current variational framework, (1) the control variable is of a single scale, and the background error characteristics of weather at different scales are not considered at the same time, and the multi-scale weather status cannot be updated simultaneously; (2) the cross-scale correlation is not reasonably considered and established, which limits the reasonable transmission of analysis increments at different scales. Summary of the invention

[0007] In view of the problems existing in the prior art, the present invention provides a variational multiscale assimilation method and system based on multi-scale control variables. By introducing multi-scale control variables and constructing a multi-scale background error covariance matrix, the simultaneous updating of state variables of different scales in three-dimensional variational assimilation is realized; at the same time, by establishing the cross-scale correlation between multi-scale weather background errors, the coordinated analysis of weather information of different scales is realized.

[0008] The present invention is implemented as follows: a variational multiscale assimilation method based on multiscale control variables, comprising:

[0009] S1, background error prediction sample generation;

[0010] S2, scale separation of background error samples to obtain large-scale and small-scale background error statistical samples;

[0011] S3, construct a multi-scale background error covariance matrix, based on the control variable conversion algorithm, statistically analyze the characteristics of large-scale background errors and small-scale background errors;

[0012] S4, establish the cross-scale correlation between large-scale and small-scale background errors in the background error covariance matrix;

[0013] S5, based on the three-dimensional variational assimilation framework, introduces large-scale control variables and small-scale control variables to construct a multi-scale three-dimensional variational assimilation method;

[0014] S6, introduces the numerical model background field and observation field, solves the minimization of the multi-scale three-dimensional variational assimilation cost function, and realizes the simultaneous and coordinated updating of weather information at different scales in a single assimilation analysis.

[0015] Furthermore, S1 specifically includes: preparing global model analysis fields, conducting regional numerical forecasts for a period of two months with an interval of 12 hours and a forecast validity of 24 hours based on numerical weather forecast models, calculating the difference between the forecast fields at the same forecast time and with a forecast validity of 12 hours and 24 hours respectively, and using it as a background error sample.

[0016] Furthermore, S2 specifically includes: performing scale separation on the background error samples. For the weight function f(k,k c ), the mode state quantity x can be expressed as:

[0017] x=x L +x S =∑ k f(k,k c )x(k)+∑ k [1-f(k,k c )]x(k), (1)

[0018] Among them, k c is the filtered wave number, and the first term on the right side of the equation represents the large-scale spectrum (wave number less than k c ), the second term represents the small-scale spectrum (wave number greater than k c ). In this way, the background error e of the mode state quantity x b It can also be decomposed into large scale and small-scale parts

[0019]

[0020] Further, S3 specifically includes: constructing a multi-scale background error covariance matrix B. Assume first and Satisfies Gaussian distribution, and Not relevant, then The estimation of the background error covariance matrix B of the state quantity x can be expressed as:

[0021]

[0022] That is, the large-scale background error covariance matrix B L and the small-scale background error covariance matrix B S sum.

[0023] Preprocess the square root of B by controlling the variable conversion operator U: B = UU T , where U includes the physical transformation U p , vertical transformation U v and the horizontal transformation U h :

[0024] U=U p U v U h , (4)

[0025] Physical Transformation p Convert the relevant model variables into assimilation system control variables; vertical transformation U v The vertical component of the background error covariance is decomposed into eigenvalues ​​and eigenvectors in the EOF space by using the empirical orthogonal function (EOF); ​​the horizontal transformation U hThe horizontal correlation of the background error is estimated using a recursive filtering method.

[0026] The GEN_BEv2.0 software was used to count the eigenvalues ​​and eigenvectors of vertical transformation and the characteristic length scale of horizontal transformation in the background error covariance based on the decomposed large-scale and small-scale error samples.

[0027] Furthermore, S4 specifically includes: establishing cross-scale correlation in background error covariance; large-scale background error and small-scale background error are related, that is The physical transformation U of the background error covariance p In , the cross-scale correlation between large-scale control variables and small-scale control variables is established:

[0028]

[0029] Among them, U L ,V L ,T L , RH L is the large-scale analysis variable, U S ,V S ,T S , RH S is a small-scale analysis variable; U on the right side of the formula L , and are large-scale control variables and small-scale control variables respectively. In a 10×10 dimensional matrix, C A,B is the correlation between variables A and B, I represents the autocorrelation of the variables, and 0 represents no correlation between the variables. The subscript u indicates that the control variable is a non-equilibrium field, which is usually obtained by subtracting the equilibrium part of the complete model variable from the other variables. Take the control variable as an example:

[0030]

[0031] The relative humidity (RH S ) minus its large-scale control variable U L , The related equilibrium part can obtain the final non-equilibrium small-scale relative humidity control variable.

[0032] Furthermore, S5 specifically includes: based on the three-dimensional variational incremental cost function framework, introducing large-scale and small-scale control variables v L and v S, and the large-scale and small-scale control variable conversion operators U L and U S , establish a multi-scale three-dimensional variational cost function:

[0033]

[0034] Where δx=δx L +δx S =U L v L +U S v S is the analysis increment, which is the sum of the analysis field x and the background field x b The difference; d = y-Hx b , is the new information increment, y is the observation vector, H is the nonlinear observation operator; H is the linearized observation operator; R represents the observation error covariance matrix. Based on the multi-scale three-dimensional variational cost function, a three-dimensional variational multi-scale assimilation method based on multi-scale control variables is constructed.

[0035] Furthermore, S6 specifically includes: using the numerical model forecast field as the assimilated background field x b , taking the multi-source observation data after quality control as the observation field y, and conducting multi-scale assimilation analysis based on the constructed three-dimensional variational multi-scale assimilation system based on multi-scale control variables. By minimizing the multi-scale three-dimensional variational cost function, the large-scale analysis increment δx is solved. L and the small-scale analytical increment δx S , adding them together can get the multi-scale analysis increment δx, and the final multi-scale analysis x=δx+x b .

[0036] Another object of the present invention is to provide a variational multiscale assimilation system based on multiscale control variables for implementing the variational multiscale assimilation method based on multiscale control variables, comprising:

[0037] A prediction sample generation module is used to generate background error prediction samples;

[0038] A statistical sample acquisition module is used to perform scale separation on background error samples and obtain large-scale and small-scale background error statistical samples;

[0039] Background error feature statistics module, used to construct a multi-scale background error covariance matrix, based on the control variable conversion algorithm, to count the large-scale background error and small-scale background error features;

[0040] A cross-scale correlation building module is used to build cross-scale correlation between large-scale and small-scale background errors in the background error covariance matrix;

[0041] The three-dimensional variational assimilation method construction module is used to introduce large-scale control variables and small-scale control variables based on the three-dimensional variational assimilation framework to construct a multi-scale three-dimensional variational assimilation method;

[0042] The weather information update module is used to introduce the numerical model background field and observation field, solve the minimization of the multi-scale three-dimensional variational assimilation cost function, and realize the simultaneous and coordinated update of weather information at different scales in a single assimilation analysis.

[0043] Another object of the present invention is to provide a computer device, which includes a memory and a processor, wherein the memory stores a computer program, and when the computer program is executed by the processor, the processor executes the steps of the variational multi-scale assimilation method based on multi-scale control variables.

[0044] Another object of the present invention is to provide a computer-readable storage medium storing a computer program, which, when executed by a processor, enables the processor to execute the steps of the variational multi-scale assimilation method based on multi-scale control variables.

[0045] Another object of the present invention is to provide an information data processing terminal, which is used to implement the variational multi-scale assimilation system based on multi-scale control variables.

[0046] In combination with the above technical solutions and the technical problems solved, the advantages and positive effects of the technical solutions to be protected by the present invention are as follows:

[0047] First, the traditional variational assimilation method only updates the weather state of a single scale. The present invention proposes a variational assimilation method based on multi-scale control variables. By introducing multi-scale control variables and constructing a multi-scale background error covariance matrix, the simultaneous update of weather states of information at different weather scales in a single assimilation is achieved; by establishing cross-scale correlations between multi-scale background errors, the cross-scale transmission of observation information between weather at different scales is achieved, and coordinated analysis of weather information at different scales is achieved.

[0048] The present invention innovatively proposes a cross-scale data assimilation method, which considers the structural characteristics of control variables of different scales in the background error covariance matrix and establishes cross-scale correlation, realizes the synchronous and coordinated updating of weather conditions of different scales, and improves the assimilation and forecasting effect of multi-source observation data. The system is conducive to reducing the forecasting burden of meteorological business departments.

[0049] Second, the expected benefits and commercial value of the technical solution of the present invention after transformation are:

[0050] (1) Integrate multi-source observation data to improve the accuracy of weather forecasts and help all industries make better decisions;

[0051] (2) In weather-dependent industries such as renewable energy, such as wind power generation, improving forecast accuracy can optimize operational efficiency.

[0052] The technical solution of the present invention solves the technical problems that people have been eager to solve but have never been able to solve successfully:

[0053] The control variables in the current variational framework are of a single scale, and the background error characteristics of weather at different scales have not been considered at the same time, so the multi-scale weather status cannot be updated at the same time. In addition, the correlation characteristics between scales are not fully considered under the variational framework, which limits the application of multi-source observation data in variational assimilation and ensemble-variational mixed assimilation. In response to this technical problem, the present invention establishes a variational assimilation method for multi-source observation data that can distinguish scales and consider cross-scale correlation, so as to achieve coordinated and synchronous updating of multi-scale state quantities of the model.

[0054] Third, the present invention provides a summary of the technical features, problems solved, and technical progress brought about by the variational multiscale assimilation method based on multiscale control variables (hereinafter referred to as "the method"):

[0055] 1) Prepare global model analysis fields: Based on the numerical weather prediction model, long-term and high-frequency regional numerical forecasts are carried out to calculate the forecast field differences as background error samples.

[0056] 2) Scale separation: The background error samples are separated into large scale and small scale to facilitate the subsequent multi-scale assimilation analysis.

[0057] 3) Construct a multi-scale background error covariance matrix: Through the Gaussian distribution assumption and scale separation, the large-scale and small-scale background error covariance matrices are estimated respectively, and the control variable conversion operator is used for preprocessing.

[0058] 4) Establish cross-scale correlation: In physical transformation, establish cross-scale correlation between large-scale control variables and small-scale control variables to reflect the mutual influence between different scales.

[0059] 5) Establish a multi-scale three-dimensional variational cost function: Introduce large-scale and small-scale control variables and their conversion operators to construct a multi-scale three-dimensional variational cost function, providing a mathematical basis for assimilation analysis.

[0060] 6) Multiscale assimilation analysis: Based on the constructed three-dimensional variational multiscale assimilation system, multiscale assimilation analysis is performed in combination with numerical model forecast fields and multi-source observation data after quality control to solve the multiscale analysis increment.

[0061] Technical issues solved:

[0062] 1) Improve assimilation accuracy: Through multi-scale analysis, the dynamic changes of the atmospheric system at different scales can be more accurately described and simulated, thereby improving the precision and accuracy of assimilation analysis.

[0063] 2) Considering the interaction between scales: Traditional assimilation methods often ignore the interaction between different scales. This method can better reflect such interaction and improve the physical consistency of assimilation analysis by establishing cross-scale correlations.

[0064] 3) Processing multi-source observation data: This method can integrate observation data from different sources and scales, improving the comprehensiveness and reliability of assimilation analysis.

[0065] Technological progress achieved:

[0066] 1) Innovation in assimilation technology: This method introduces the concept of multi-scale control variables in the field of numerical weather forecast assimilation, realizes the detailed description and simulation of the multi-scale characteristics of the atmospheric system, and promotes the innovative development of assimilation technology.

[0067] 2) Improved performance of the assimilation system: Through multi-scale assimilation analysis, this method can provide more accurate and reliable assimilation results, provide more precise initial conditions for numerical weather forecasting, and thus improve the accuracy and reliability of weather forecasts.

[0068] 3) Broad application prospects: This method is not only applicable to the field of numerical weather forecasting, but can also be extended to other fields that need to deal with multi-scale problems, such as climate simulation, ocean forecasting, etc., and has broad application prospects.

[0069] Fourth, the present invention provides a variational multi-scale assimilation method based on multi-scale control variables, which is used to solve the technical problem that a single-scale control variable in the existing weather forecast model is difficult to accurately express meteorological processes of different scales. Specific technical advances include:

[0070] 1. Scale separation and comprehensive processing: By separating the scales of background error samples and constructing an independent background error covariance matrix for each scale, the present invention allows more accurate processing of meteorological data of different scales. This scale subdivision processing improves the model's ability to express large-scale and small-scale meteorological processes, and improves the accuracy and reliability of weather forecasts.

[0071] 2. Introduction and optimization of multi-scale control variables: This invention introduces large-scale and small-scale control variables and corresponding conversion operators to establish a three-dimensional variational cost function that comprehensively considers multi-scale information. This design enables the assimilation model to process multi-scale data simultaneously, improving the response speed and prediction accuracy of complex weather systems.

[0072] 3. Real-time and dynamic data assimilation processing: Combined with real-time observation data, this method can dynamically adjust and optimize the assimilation process, so that the forecast model can reflect the latest weather changes in real time, enhancing the timeliness of weather forecasts and the ability to respond to sudden weather events.

[0073] Through the above-mentioned technological advances, the present invention has significantly improved the accuracy and efficiency of numerical weather forecasting, especially in dealing with multi-scale and nonlinear complex meteorological processes, and has played an important role in promoting the development of modern meteorological forecasting technology. BRIEF DESCRIPTION OF THE DRAWINGS

[0074] Figure 1 is a flow chart of a variational multi-scale assimilation method based on multi-scale control variables provided by an embodiment of the present invention;

[0075] Figure 2 Schematic diagram of an implementation scheme of a variational multi-scale assimilation method based on multi-scale control variables provided in an embodiment of the present invention;

[0076] Figure 3 is a structural diagram of a variational multi-scale assimilation system based on multi-scale control variables provided by an embodiment of the present invention;

[0077] Figure 4 It is a schematic diagram of an application example provided by an embodiment of the present invention in a weather forecast of a landing typhoon precipitation. DETAILED DESCRIPTION

[0078] In order to make the purpose, technical solution and advantages of the present invention more clearly understood, the present invention is further described in detail below in conjunction with the embodiments. It should be understood that the specific embodiments described herein are only used to explain the present invention and are not used to limit the present invention.

[0079] In view of the problems existing in the prior art, the present invention provides a variational multi-scale assimilation method and system based on multi-scale control variables. The present invention is described in detail below in conjunction with the accompanying drawings.

[0080] The variational multiscale assimilation method based on multiscale control variables has important applications in meteorology and earth science. The following are two specific application examples to demonstrate the practical application of this assimilation method:

[0081] Example 1: Storm prediction and management

[0082] The prediction of storm systems is crucial for taking disaster prevention measures in advance, especially in areas that are frequently affected by severe storms. This assimilation technology can improve the prediction accuracy of storms, thereby reducing casualties and economic losses.

[0083] 1) Model initialization: The numerical weather forecast model is initialized using a variational multiscale assimilation method based on multiscale control variables to ensure that the model reflects the most accurate initial state of the atmosphere.

[0084] 2) Data assimilation: Input multi-source observation data into the assimilation system. By executing step S6, the assimilation includes satellite remote sensing, ground observation station and aircraft detection data.

[0085] 3) Forecast execution: Based on the optimized initial conditions, short- to medium-term storm path and intensity forecasts are performed.

[0086] 4) Application of results: The forecast results will be provided to relevant meteorological and emergency management departments for the formulation of early warning and evacuation plans.

[0087] By applying this assimilation technology, earlier storm warnings can be achieved and the accuracy of forecasts can be improved, thereby effectively guiding the formulation and implementation of disaster response measures.

[0088] Example 2: Urban flood prediction and response

[0089] In many cities, especially in coastal and river-dense areas, flood forecasting plays a vital role in protecting people's lives and property. Predicting precipitation and surface runoff through high-precision meteorological models can more effectively manage urban flood risks.

[0090] 1) Regional model setting: Setting up a city-specific numerical model covering urban topography and hydrological characteristics.

[0091] 2) Multi-scale data assimilation: Assimilate multi-source observation data such as sounding, aircraft reports, ground automatic stations, urban hydrological stations, radar precipitation data, etc. to accurately capture precipitation events and surface water flow.

[0092] 3) Flood prediction: Run the flood prediction model based on the assimilated data to predict the flood occurrence area and water level.

[0093] 4) Emergency Preparedness: Communicate forecast information to city management and emergency response agencies to optimize emergency preparedness and response strategies.

[0094] This method can enhance the timeliness and accuracy of flood forecasting, provide a scientific basis for urban flood management and emergency response, and help minimize the damage caused by floods.

[0095] These two examples demonstrate the application of the variational multiscale assimilation method based on multiscale control variables in actual meteorological forecasting and disaster management. By improving the accuracy and timeliness of predictions, this technology significantly improves the effectiveness of meteorological disaster warnings and the scientific nature of response measures.

[0096] like Figure 1In the three-dimensional variational data multiscale assimilation method based on multi-scale control variables in the present invention, the background error samples are separated into large-scale and small-scale samples through the scale separation technology, thereby realizing the construction and characteristic analysis of multi-scale and cross-scale background error covariance; based on the variational assimilation framework and the above-mentioned ensemble error covariance, multi-scale control variables are introduced to establish a multi-scale variational assimilation method; finally, the model background field and multi-source observation data are input, and the final output of the multi-scale assimilation analysis is obtained by minimizing the multi-scale three-dimensional variational cost function.

[0097] Combination Figure 2 The specific implementation scheme of this method is as follows:

[0098] Step 1) Prepare global model analysis fields, which can be the US FNL reanalysis dataset, China CRA40 reanalysis dataset, European Center ERA-interim, etc. Based on numerical weather forecast models, which can be the Weather Research and Forecast Model (WRF) model, China Meteorological Administration GRAPES model, etc., carry out regional numerical forecasts for a period of two months, with an interval of 12 hours and a forecast time of 24 hours, calculate the difference between the forecast fields at the same forecast time and with a forecast time of 12 hours and 24 hours respectively, and use it as the background error sample.

[0099] Step 2) scale separation of background error samples. For the weight function f(k,k c ), the mode state quantity x can be expressed as:

[0100] x=x L +x S =∑ k f(k,k c )x(k)+∑ k [1-f(k,k c )]x(k), (1)

[0101] Among them, k c is the filtered wave number, and the first term on the right side of the equation represents the large-scale spectrum (wave number less than k c ), the second term represents the small-scale spectrum (wave number greater than k c ). In this way, the background error e of the mode state quantity x b It can also be decomposed into large scale and small-scale parts

[0102]

[0103] Step 3) Construct the multi-scale background error covariance matrix B. Assume first and Satisfies Gaussian distribution, and Not relevant, then The estimation of the background error covariance matrix B of the state quantity x can be expressed as:

[0104]

[0105] That is, the large-scale background error covariance matrix B L and the small-scale background error covariance matrix B S sum.

[0106] Preprocess the square root of B by controlling the variable conversion operator U: B = UU T , where U includes the physical transformation U p , vertical transformation U v and the horizontal transformation U h :

[0107] U=U p U v U h , (4)

[0108] Physical Transformation p Convert the relevant model variables into assimilation system control variables; vertical transformation U v The vertical component of the background error covariance is decomposed into eigenvalues ​​and eigenvectors in the EOF space by using the empirical orthogonal function (EOF); ​​the horizontal transformation U h The horizontal correlation of the background error is estimated using a recursive filtering method.

[0109] The GEN_BEv2.0 software was used to count the eigenvalues ​​and eigenvectors of vertical transformation and the characteristic length scale of horizontal transformation in the background error covariance based on the decomposed large-scale and small-scale error samples.

[0110] Step 4) Establish the cross-scale correlation in the background error covariance. In fact, the large-scale background error and small-scale background error are related, that is The physical transformation U of the background error covariance p In , the cross-scale correlation between large-scale control variables and small-scale control variables is established:

[0111]

[0112] Among them, U L ,V L ,T L , RH L is the large-scale analysis variable, US ,V S ,T S , RH S is a small-scale analysis variable; U on the right side of the formula L , and are large-scale control variables and small-scale control variables respectively. In a 10×10 dimensional matrix, C A,B is the correlation between variables A and B, I represents the autocorrelation of the variables, and 0 represents no correlation between the variables. The subscript u indicates that the control variable is a non-equilibrium field, which is usually obtained by subtracting the equilibrium part of the complete model variable from the other variables. Take the control variable as an example:

[0113]

[0114] The relative humidity (RH S ) minus its large-scale control variable U L , The related equilibrium part can obtain the final non-equilibrium small-scale relative humidity control variable.

[0115] Step 5) Based on the three-dimensional variational incremental cost function framework, introduce large-scale and small-scale control variables v L and v S , and the large-scale and small-scale control variable conversion operators U L and U S , establish a multi-scale three-dimensional variational cost function:

[0116]

[0117] Where δx=δx L +δx S =U L v L +U S v S is the analysis increment, which is the sum of the analysis field x and the background field x b The difference; d = y-Hx b , is the new information increment, y is the observation vector, H is the nonlinear observation operator; H is the linearized observation operator; R represents the observation error covariance matrix. Based on the multi-scale three-dimensional variational cost function, a three-dimensional variational multi-scale assimilation method based on multi-scale control variables is constructed.

[0118] Step 6) Use the numerical model forecast field as the assimilated background field x b, taking the multi-source observation data after quality control as the observation field y, and conducting multi-scale assimilation analysis based on the constructed three-dimensional variational multi-scale assimilation system based on multi-scale control variables. By minimizing the multi-scale three-dimensional variational cost function (Formula 7), the large-scale analysis increment δx is solved L and the small-scale analytical increment δx S , adding them together can get the multi-scale analysis increment δx, and the final multi-scale analysis x=δx+x b .

[0119] In this embodiment, the background field in step 1) is a model forecast field, and may also be a numerical model analysis / forecast field from other sources at the same time.

[0120] The background error samples described in step 1) can also be obtained through a set of ensemble forecast samples.

[0121] The scale separation algorithm described in step 2) may adopt a filtering algorithm, such as Raymond filtering, recursive filtering, etc., or a simple smoothing algorithm.

[0122] The control variables listed in step 4) are only conventional variables. Variables closely related to clouds and precipitation, such as hydrometeor variables and vertical wind fields, can be regarded as small-scale variables without scale separation.

[0123] The assimilation framework in step 5) takes the three-dimensional variational method as an example, but the present invention is also applicable to the static background error covariance in the four-dimensional variational and ensemble-variational hybrid assimilation, and therefore also falls within the protection scope of the present invention.

[0124] The model background field in step 6) is the forecast field integrated from the previous moment to the current moment, and can also be the numerical model analysis / forecast field from other sources at the same moment.

[0125] like Figure 3 As shown, the variational multi-scale assimilation system based on multi-scale control variables provided by the embodiment of the present invention is characterized by comprising:

[0126] A prediction sample generation module is used to generate background error prediction samples;

[0127] A statistical sample acquisition module is used to perform scale separation on background error samples and obtain large-scale and small-scale background error statistical samples;

[0128] Background error feature statistics module, used to construct a multi-scale background error covariance matrix, based on the control variable conversion algorithm, to count the large-scale background error and small-scale background error features;

[0129] A cross-scale correlation building module is used to build cross-scale correlation between large-scale and small-scale background errors in the background error covariance matrix;

[0130] The three-dimensional variational assimilation method construction module is used to introduce large-scale control variables and small-scale control variables based on the three-dimensional variational assimilation framework to construct a multi-scale three-dimensional variational assimilation method;

[0131] The weather information update module is used to introduce the numerical model background field and observation field, solve the minimization of the multi-scale three-dimensional variational assimilation cost function, and realize the simultaneous and coordinated update of weather information at different scales in a single assimilation analysis.

[0132] An application embodiment of the present invention provides a computer device, which includes a memory and a processor. The memory stores a computer program. When the computer program is executed by the processor, the processor executes the steps of a variational multi-scale assimilation method based on multi-scale control variables.

[0133] An application embodiment of the present invention provides a computer-readable storage medium storing a computer program. When the computer program is executed by a processor, the processor executes the steps of a variational multi-scale assimilation method based on multi-scale control variables.

[0134] An application embodiment of the present invention provides an information data processing terminal, which is used to implement a variational multi-scale assimilation system based on multi-scale control variables.

[0135] The present invention is particularly suitable for dense observation data such as weather radar and encrypted automatic stations. Figure 4 This is an application embodiment of the present invention for improving typhoon landing precipitation weather forecast and enhancing the understanding of typhoon processes based on intensive observation data.

[0136] Multi-source observation data not only contain convective scale information, but also contain larger scale information, reflecting the weather characteristics of the surrounding environment field. Based on the variational assimilation framework, the present invention introduces multi-scale control variables to establish a cross-scale related multi-source observation multi-scale assimilation method, which can provide a cross-scale related analysis field for convective scale weather that is forced by large-scale circulation situations and affected by mesoscale weather systems and their internal thermal dynamics and microphysical processes.

[0137] It should be noted that the embodiments of the present invention can be implemented by hardware, software, or a combination of software and hardware. The hardware part can be implemented using dedicated logic; the software part can be stored in a memory and executed by an appropriate instruction execution system, such as a microprocessor or dedicated design hardware. It can be understood by a person of ordinary skill in the art that the above-mentioned devices and methods can be implemented using computer executable instructions and / or contained in a processor control code, such as a carrier medium such as a disk, CD or DVD-ROM, a programmable memory such as a read-only memory (firmware), or a data carrier such as an optical or electronic signal carrier. Such code is provided on the carrier medium. The device and its modules of the present invention can be implemented by hardware circuits such as very large-scale integrated circuits or gate arrays, semiconductors such as logic chips, transistors, etc., or programmable hardware devices such as field programmable gate arrays, programmable logic devices, etc., can also be implemented by software executed by various types of processors, and can also be implemented by a combination of the above-mentioned hardware circuits and software, such as firmware.

[0138] The above description is only a specific implementation mode of the present invention, but the protection scope of the present invention is not limited thereto. Any modifications, equivalent substitutions and improvements made by any technician familiar with the technical field within the technical scope disclosed by the present invention and within the spirit and principle of the present invention should be covered by the protection scope of the present invention.

Claims

1. A variational multiscale assimilation method based on multiscale control variables, characterized in that: include: S1, background error prediction sample generation; S2, scale separation of background error samples to obtain large-scale and small-scale background error statistical samples; S3, construct a multi-scale background error covariance matrix, based on the control variable conversion algorithm, statistically analyze the characteristics of large-scale background errors and small-scale background errors; S4, establish the cross-scale correlation between large-scale and small-scale background errors in the background error covariance matrix; S5, based on the three-dimensional variational assimilation framework, introduces large-scale control variables and small-scale control variables to construct a multi-scale three-dimensional variational assimilation method; S6, introduces the numerical model background field and observation field, solves the minimization of the multi-scale three-dimensional variational assimilation cost function, and realizes the simultaneous and coordinated update of weather information at different scales in a single assimilation analysis; S2 specifically includes: performing scale separation on the background error samples; for the weight function f(k,k corresponding to the wave number k); c ), the mode state quantity x is expressed as: x=x L +x S =∑ k f(k,k c )x(k)+∑ k [1-f(k,k c )]x(k), (1) Among them, k c is the filtering wave number. The first term on the right side of the equation represents the large-scale spectrum, and the wave number is less than k c , the second term represents the small-scale spectrum, with wave numbers greater than k c ; Thus, the background error e of the mode state quantity x b Decomposition into large scale and small-scale parts S3 specifically includes: constructing a multi-scale background error covariance matrix B; assuming first and Satisfies Gaussian distribution, and Not relevant, then The estimation of the background error covariance matrix B of the state quantity x is expressed as: That is, the large-scale background error covariance matrix B L and the small-scale background error covariance matrix B S sum; Preprocess the square root of B by controlling the variable conversion operator U: B = UU T , where U includes the physical transformation U p , vertical transformation U v and the horizontal transformation U h : U=U p IN v IN h , (4) Physical Transformation p Convert the relevant model variables into assimilation system control variables; vertical transformation U v The vertical component of the background error covariance is decomposed into eigenvalues ​​and eigenvectors in the EOF space through the empirical orthogonal function EOF; the horizontal transformation U h The horizontal correlation of background errors is estimated using a recursive filtering method; The GEN_BE v2.0 software was used to count the eigenvalues ​​and eigenvectors of vertical transformation and the characteristic length scale of horizontal transformation in the background error covariance based on the decomposed large-scale and small-scale error samples.

2. The variational multi-scale assimilation method based on multi-scale control variables according to claim 1, characterized in that: S1 specifically includes: preparing global model analysis fields, conducting two-month regional numerical forecasts with an interval of 12 hours and a forecast time of 24 hours based on numerical weather forecast models, calculating the differences in forecast fields at the same forecast time and with forecast times of 12 hours and 24 hours respectively, and using them as background error samples.

3. The variational multi-scale assimilation method based on multi-scale control variables according to claim 1, characterized in that: S4 specifically includes: establishing cross-scale correlation in background error covariance; large-scale background error and small-scale background error are related, that is The physical transformation U of the background error covariance p In , the cross-scale correlation between large-scale control variables and small-scale control variables is established: Among them, U L ,V L ,T L , RH L is the large-scale analysis variable, U S ,V S ,T S , RH S is a small-scale analysis variable; U on the right side of the formula L , and are large-scale control variables and small-scale control variables respectively; in a 10×10 dimensional matrix, C A,B is the correlation between variables A and B, I represents the autocorrelation of variables, and 0 represents no correlation between variables; the subscript u indicates that the control variable is a non-equilibrium field, which is usually obtained by subtracting the equilibrium part of the complete model variable from the other variables; the non-equilibrium relative humidity at a small scale Take the control variable as an example: From the relative humidity small scale analysis variable RH S minus its large-scale control variable U L , The related equilibrium part can obtain the final non-equilibrium small-scale relative humidity control variable.

4. The variational multi-scale assimilation method based on multi-scale control variables according to claim 1, characterized in that: S5 specifically includes: Based on the three-dimensional variational incremental cost function framework, large-scale and small-scale control variables v are introduced L and v S , and the large-scale and small-scale control variable conversion operators U L and U S , establish a multi-scale three-dimensional variational cost function: Where δx=δx L +δx S =U L v L +U S v S is the analysis increment, which is the sum of the analysis field x and the background field x b The difference; d = yH (x b ), is the new information increment, y is the observation vector, H is the nonlinear observation operator; H is the linearized observation operator; R represents the observation error covariance matrix; based on the multi-scale three-dimensional variational cost function, a three-dimensional variational multi-scale assimilation method based on multi-scale control variables is constructed.

5. The variational multi-scale assimilation method based on multi-scale control variables according to claim 1, characterized in that: S6 specifically includes: using the numerical model forecast field as the assimilated background field x b , taking the multi-source observation data after quality control as the observation field y, and conducting multi-scale assimilation analysis based on the constructed three-dimensional variational multi-scale assimilation system based on multi-scale control variables; solving the large-scale analysis increment δx by minimizing the multi-scale three-dimensional variational cost function L and the small-scale analytical increment δx S , adding them together can get the multi-scale analysis increment δx, and the final multi-scale analysis x=δx+x b .

6. A variational multiscale assimilation system based on multiscale control variables for implementing the variational multiscale assimilation method based on multiscale control variables as claimed in any one of claims 1 to 5, characterized in that: include: A prediction sample generation module is used to generate background error prediction samples; A statistical sample acquisition module is used to perform scale separation on background error samples and obtain large-scale and small-scale background error statistical samples; Background error feature statistics module, used to construct a multi-scale background error covariance matrix, based on the control variable conversion algorithm, to count the large-scale background error and small-scale background error features; A cross-scale correlation building module is used to build cross-scale correlation between large-scale and small-scale background errors in the background error covariance matrix; The three-dimensional variational assimilation method construction module is used to introduce large-scale control variables and small-scale control variables based on the three-dimensional variational assimilation framework to construct a multi-scale three-dimensional variational assimilation method; The weather information update module is used to introduce the numerical model background field and observation field, solve the minimization of the multi-scale three-dimensional variational assimilation cost function, and realize the simultaneous and coordinated update of weather information at different scales in a single assimilation analysis.

7. A computer device, comprising a memory and a processor, wherein the memory stores a computer program, and when the computer program is executed by the processor, the processor executes the steps of the variational multi-scale assimilation method based on multi-scale control variables as described in any one of claims 1 to 5.

8. An information data processing terminal, used to implement the variational multi-scale assimilation system based on multi-scale control variables as described in claim 6.

Images