Three-dimensional variational assimilation method and system for atmospheric chemical aerosol data assimilation

Abstract

The present invention discloses a three-dimensional variational assimilation method and system for atmospheric chemical aerosol data assimilation, comprising designing a multi-component aerosol assimilation framework, obtaining a first guess field, obtaining a new background error based on the PMAAP background error by performing equal proportion distribution and incremental compensation, spatially transforming a normalized control variable based on the new background error covariance to obtain an analysis increment, obtaining an observation residual and a total objective function based on the analysis increment and observation information, calculating the total objective function gradient based on the observation residual, performing a minimization search based on the total objective function and the total objective function gradient to obtain an optimal normalized control variable, and obtaining an optimal analysis field based on the optimal normalized control variable and the first guess field. This method not only improves the efficiency and accuracy of atmospheric chemical aerosol data assimilation, but also has good interpretability and can be directly applied to a three-dimensional variational assimilation system for atmospheric chemical aerosol data assimilation.

Description

Technical Field

[0001] The present invention relates to the technical field of atmospheric chemical data assimilation, and in particular to a three-dimensional variational assimilation method and system for atmospheric chemical aerosol data assimilation. Background Art

[0002] Atmospheric aerosols have a significant impact on climate and the environment. Accurate monitoring and forecasting of their concentration and distribution are crucial to environmental science and meteorology. Therefore, assimilation of aerosol observation data is a key technology to improve the forecast quality of atmospheric chemistry models. By integrating atmospheric chemistry observation data with the atmospheric chemistry forecast information of numerical models, a high-precision initial field of atmospheric chemistry state is provided to the model, thereby significantly improving the accuracy of atmospheric pollutant concentration forecasts.

[0003] Traditional three-dimensional variational systems for coupled atmospheric chemistry and weather are widely used, but they still have many shortcomings. First, the PM Aerosol Assimilation Platform (PMAAP) uses a single element for atmospheric chemistry variable analysis, while the CMA-CW v1.0 / HAZE-FOG numerical prediction model system requires multi-component aerosol variables as initial fields. Therefore, the existing PMAAP system for coupled atmospheric chemistry and weather is not compatible with the CMA-CW v1.0 / HAZE-FOG numerical prediction model. Second, it is difficult to establish a relationship between aerosol mass concentration variables in the existing three-dimensional variational assimilation system for coupled atmospheric chemistry and weather and satellite and radar observation variables such as atmospheric chemical optical depth and extinction coefficient, making it difficult to directly assimilate these observational data. To address the above problems, the present invention proposes a three-dimensional variational assimilation system for assimilating atmospheric chemical aerosol data, namely the Multi-Component Aerosol Assimilation Platform (MCAAP). By setting the multi-component aerosol variables as control variables, the background errors of the MCAAP multi-component aerosol variables are obtained by adopting the proportional distribution model and the background error incremental compensation method for the background errors of the PMAAP assimilation system. The total objective function and the corresponding objective function gradient are calculated, and a minimization search is performed to obtain the optimal normalized control variables. The optimal analysis field is then output. This effectively overcomes the shortcomings of traditional technologies, can provide matching atmospheric chemical and weather initial fields for the regional chemical weather model CMA-CW v1.0 / HAZE-FOG, and can more easily achieve the assimilation of observational data such as optical depth and extinction coefficient. Summary of the Invention

[0004] The purpose of the present invention is to provide a three-dimensional variational assimilation method and system for atmospheric chemical aerosol data assimilation.

[0005] To achieve the above object, the present invention is implemented according to the following technical solutions:

[0006] The present invention comprises the following steps:

[0007] Design a multi-component aerosol assimilation framework (MCAAP) to obtain atmospheric chemical background fields and weather element background fields, and input them into a three-dimensional variational assimilation system as the first guess field; the first guess field includes the atmospheric chemical background field and the weather element background field;

[0008] According to the atmospheric chemical background field and the background error of the PM aerosol assimilation framework PMAAP, the background error is proportionally distributed and incrementally compensated to obtain a new background error of the multi-component aerosol assimilation framework, and a related model is constructed to describe the horizontal correlation coefficient and vertical correlation coefficient of the background error applicable to the multi-component aerosol assimilation system MCAAP, thereby obtaining a horizontal transformation operator and a vertical transformation operator;

[0009] Constructing a background error covariance and performing variable transformation on the normalized control variable to obtain an analysis increment, obtaining an observation residual and a total objective function based on the analysis increment and observation information, and calculating the gradient of the total objective function based on the observation residual; the variable transformation includes vertical transformation, horizontal transformation, and physical transformation; and the analysis increment includes an atmospheric chemical analysis increment and a weather element analysis increment;

[0010] An optimal normalized control variable is obtained by performing a minimization search based on the total objective function and the gradient of the total objective function, and an optimal analysis field is obtained based on the optimal normalized control variable and the first initial guess field; the optimal analysis field includes an optimal atmospheric chemistry analysis field and an optimal weather element analysis field.

[0011] Furthermore, the method for designing a multi-component aerosol assimilation framework includes:

[0012] Black carbon (BC), organic carbon (OC), road dust (SD), sea salt (SS), sulfate (SF), and nitrate (NT) are divided into 24 control variables according to four particle size segments. The 24 control variables and the unsegmented ammonium salt (AM) form a total of 25 atmospheric chemical control variables. The assimilation system with the 25 multi-component aerosol variables as control variables is defined as the multi-component aerosol assimilation framework (MCAAP). The diameters of the four particle size segments include 0.01-1.28 μm, 1.28-2.56 μm, 2.56-10.24 μm, and 10.24-40.98 μm, respectively. The atmospheric chemical control variables are uncorrelated with each other and with weather factor control variables. The weather factor control variables include wind field, temperature field, surface air pressure, and specific humidity.

[0013] The atmospheric chemical control variables, and Composition of atmospheric chemical analysis variables; and In the MCAAP assimilation system, it is the cumulative amount of the atmospheric chemistry control variables.

[0014] Furthermore, the method of performing equal proportion distribution includes:

[0015] The background error of the PMAAP assimilation system is assimilated using the proportional distribution model. 、 The proportion of the forecast field of different particle size segments in the total forecast field is converted into the background error of different particle size segments. The background error of the MCAAP assimilation system is calculated based on the atmospheric chemical background field and the known background error of the PMAAP assimilation system. The expression is:

[0016]

[0017]

[0018]

[0019]

[0020] in Variables in the PMAAP assimilation system The background error, Variables in the PMAAP assimilation system In the multi-component aerosol assimilation framework MCAAP assimilation system, and Respectively The background error and background field of the aerosol variable X for each particle size segment, Take 1 to 4, is the atmospheric chemical aerosol variable set, and are the background error and background field of ammonium salt respectively;

[0021] The background error of the aerosol variable in the fourth particle size segment is set to 1 / 5 of the background error of the aerosol in the third particle size segment, and the expression is:

[0022]

[0023] Using the equal-proportional distribution model, the system aerosol variables were assimilated from PMAAP. and The background errors of seven types of aerosol variables in four particle size segments required by the multi-component aerosol assimilation system MCAAP are calculated.

[0024] Furthermore, the method for performing the incremental compensation includes:

[0025] Based on the assimilation of the same single-point ground aerosol observation PM 2.5 and PM 10 , the analysis increment principle of the PMAAP assimilation system and the MCAAP assimilation system is adopted to calculate the background error increment of the control variables in the MCAAP assimilation system. Based on the background error increment, the atmospheric chemistry control variables in the MCAAP assimilation system are incrementally compensated to obtain the new background error. The expression of the new background error is:

[0026]

[0027]

[0028]

[0029]

[0030]

[0031] in and Respectively The background error of the aerosol variable X in each particle size segment and the new background error after background error increment compensation, Take 1 to 4, is the atmospheric chemical aerosol variable set, and are the background error of ammonium salt and the new background error after background error incremental compensation;

[0032] 、 is the correction parameter for the background error of the control variable in different particle size segments, where the parameter and The expression is:

[0033]

[0034]

[0035] in Variables in the PMAAP assimilation system The background error, Variables in the PMAAP assimilation system background error.

[0036] Furthermore, the method for obtaining the horizontal correlation coefficient and the vertical correlation coefficient includes:

[0037] The Gaussian correlation model is used to describe the background error level correlation coefficient, and the expression is:

[0038]

[0039] in is the background error level correlation coefficient, is the horizontal distance, is the horizontal correlation scale;

[0040] The vertical correlation coefficient of the background error is described using a correlation model that varies with height. The expression is:

[0041]

[0042] For the Layer and The vertical correlation coefficient of the background error of the layer, 、 Respectively Layer and The height of the layer, is the height sensitivity coefficient, is the acceleration due to gravity, is the dry air gas constant, is the average temperature on the pattern surface, is the particle size parameter.

[0043] Furthermore, the method for calculating the gradient of the total objective function includes:

[0044] According to the normalized control variables Calculate the analysis increment and use EOF decomposition, recursive filtering method and physical transformation operator to normalize the control variable Perform vertical, horizontal, and physical transformations to obtain analysis increments:

[0045]

[0046] in To analyze the increment, is the space transformation operator, is the horizontal transformation, is the vertical transformation, is the physical transformation operator, is a normalized control variable; the spatial transformation operator consists of a horizontal transformation and a vertical transformation; the horizontal transformation realizes the horizontal propagation of observation information through recursive filtering; the vertical transformation realizes the propagation of vertical direction information through EOF decomposition;

[0047] According to the analysis increment Calculate observation equivalent increment , according to the background field and observation fields Determine the observation increment , according to the observed increment Equivalent to the observed increment Determine the observation residual , according to the observed residual Calculate the overall objective function , the expression is:

[0048]

[0049]

[0050] in is the normalized control variable The corresponding overall objective function is, , is the linear observation operator that projects the atmospheric state into the observation space, , is the observation error covariance matrix The inverse matrix of is the normalized control variable The adjoint operator of is the observed residual The adjoint operator of ;

[0051] Define the adjoint operator of the linear observation operator, the adjoint operator of the physical transformation and the adjoint operator of the spatial transformation, combined with the observation residual Calculate the total objective function gradient, the expression is:

[0052]

[0053] in is the total objective function gradient, is the normalized control variable, is the adjoint operator of the linear observation operator, is the adjoint operator of the physical transformation, is the adjoint operator of the space transformation, is the observation error covariance matrix The inverse matrix of .

[0054] Furthermore, the method for obtaining the optimal normalized control variable includes:

[0055] According to the total objective function and the total objective function gradient, the minimization search direction and step size information are obtained through the finite memory variable scale minimization algorithm LBFGS, and the new normalized increment is calculated;

[0056] Calculate the new total objective function and the total objective function gradient according to the new normalized increment, update the normalized increment again, repeat the iteration until the total objective function is minimized, and output the optimal normalized control variable .

[0057] The second aspect is a three-dimensional variational assimilation system for atmospheric chemical aerosol data assimilation, including:

[0058] Control variable module: used to design atmospheric chemical control variables and atmospheric chemical analysis variables, obtain atmospheric chemical background fields and weather element background fields, and input them into the three-dimensional variational assimilation system to generate the first guess field;

[0059] Background error module: used to determine the background error of the MCAAP assimilation system using the proportional distribution model according to the background error of the PMAAP assimilation system, and to obtain a new background error suitable for the MCAAP assimilation system by performing an incremental compensation method on the background error;

[0060] Variable transformation module: used to perform spatial transformation and physical transformation on the normalized control variable to obtain analysis increments, and to perform variable transformation on the optimal normalized control variable to obtain the optimal analysis increment; the variable transformation includes vertical transformation, horizontal transformation and physical transformation; the analysis increment includes atmospheric chemical analysis increment and weather element analysis increment;

[0061] Total objective function module: used to obtain the observation residual and the total objective function according to the analysis increment, and calculate the total objective function gradient according to the observation residual;

[0062] Search module: performing a minimization search according to the total objective function and the gradient of the total objective function to obtain the optimal normalized control variable;

[0063] Output module: used to generate an optimal analysis field according to the optimal analysis increment and the first initial guess field.

[0064] The beneficial effects of the present invention are:

[0065] The present invention is a three-dimensional variational assimilation method and system for assimilating atmospheric chemical aerosol data. Compared with the prior art, the present invention has the following technical effects:

[0066] The present invention improves the optimization preprocessing capability in atmospheric chemical aerosol data assimilation by constructing a multi-component aerosol assimilation system (MCAAP), an equal-proportion distribution model, background error increment compensation, spatial transformation, physical transformation, and minimization search steps. The system more easily assimilates observational data such as optical depth and extinction coefficient, thereby enhancing the adaptability of atmospheric chemical aerosol data in the model and further improving the model's simulation capability of atmospheric chemical processes and the accuracy of simulation of aerosol-related elements. The optimization of meteorological data assimilation technology can significantly save resources and improve work efficiency. It can effectively assimilate atmospheric chemical observations and weather element observations simultaneously, providing matching atmospheric chemical and weather element initial fields for the CMA-CW v1.0 / HAZE-FOG chemical-weather coupling model, offering more reliable technical support for meteorological forecasting, and achieving high-precision analysis of both atmospheric chemical and weather states. The system can adapt to different three-dimensional variational assimilation systems for atmospheric chemical aerosol data assimilation and meet the atmospheric chemical aerosol data assimilation needs of different users, thus possessing a certain degree of universality. BRIEF DESCRIPTION OF THE DRAWINGS

[0067] Figure 1 The figure is a flowchart of the steps of a three-dimensional variational assimilation method for assimilating atmospheric chemical aerosol data according to the present invention. DETAILED DESCRIPTION

[0068] The present invention will be further described below through specific examples. The illustrative examples and descriptions of the present invention are used to explain the present invention but are not intended to limit the present invention.

[0069] The present invention provides a three-dimensional variational assimilation method and system for atmospheric chemical aerosol data assimilation, comprising the following steps:

[0070] like Figure 1 As shown, in this embodiment, the following steps are included:

[0071] Design a multi-component aerosol assimilation framework (MCAAP) to obtain atmospheric chemical background fields and weather element background fields, and input them into a three-dimensional variational assimilation system as the first guess field; the first guess field includes the atmospheric chemical background field and the weather element background field;

[0072] According to the atmospheric chemical background field and the background error of the PM aerosol assimilation framework PMAAP, the background error is proportionally distributed and incrementally compensated to obtain a new background error of the multi-component aerosol assimilation framework, and a related model is constructed to describe the horizontal correlation coefficient and vertical correlation coefficient of the background error applicable to the multi-component aerosol assimilation system MCAAP, thereby obtaining a horizontal transformation operator and a vertical transformation operator;

[0073] Constructing a background error covariance and performing variable transformation on the normalized control variable to obtain an analysis increment, obtaining an observation residual and a total objective function based on the analysis increment and observation information, and calculating the gradient of the total objective function based on the observation residual; the variable transformation includes vertical transformation, horizontal transformation, and physical transformation; and the analysis increment includes an atmospheric chemical analysis increment and a weather element analysis increment;

[0074] An optimal normalized control variable is obtained by performing a minimization search based on the total objective function and the gradient of the total objective function, and an optimal analysis field is obtained based on the optimal normalized control variable and the first initial guess field; the optimal analysis field includes an optimal atmospheric chemistry analysis field and an optimal weather element analysis field.

[0075] In this embodiment, the method for designing a multi-component aerosol assimilation framework includes:

[0076] Black carbon (BC), organic carbon (OC), road dust (SD), sea salt (SS), sulfate (SF), and nitrate (NT) are divided into 24 control variables according to four particle size segments. The 24 control variables and the unsegmented ammonium salt (AM) form a total of 25 atmospheric chemical control variables. The assimilation system with the 25 multi-component aerosol variables as control variables is defined as the multi-component aerosol assimilation framework (MCAAP). The diameters of the four particle size segments include 0.01-1.28 μm, 1.28-2.56 μm, 2.56-10.24 μm, and 10.24-40.98 μm, respectively. The atmospheric chemical control variables are uncorrelated with each other and with weather factor control variables. The weather factor control variables include wind field, temperature field, surface air pressure, and specific humidity.

[0077] The atmospheric chemical control variables, and Composition of atmospheric chemical analysis variables; and In the MCAAP assimilation system, it is the cumulative amount of the atmospheric chemistry control variables.

[0078] In this embodiment, the method for performing equal proportion distribution includes:

[0079] The background error of the PMAAP assimilation system is assimilated using the proportional distribution model. 、 The proportion of the forecast field of different particle size segments in the total forecast field is converted into the background error of different particle size segments. The background error of the MCAAP assimilation system is calculated based on the atmospheric chemical background field and the known background error of the PMAAP assimilation system. The expression is:

[0080]

[0081]

[0082]

[0083]

[0084] in Variables in the PMAAP assimilation system The background error, Variables in the PMAAP assimilation system In the multi-component aerosol assimilation framework MCAAP assimilation system, and Respectively The background error and background field of the aerosol variable X for each particle size segment, Take 1 to 4, is the atmospheric chemical aerosol variable set, and are the background error and background field of ammonium salt respectively;

[0085] The background error of the aerosol variable in the fourth particle size segment is set to 1 / 5 of the background error of the aerosol in the third particle size segment, and the expression is:

[0086]

[0087] Using the equal-proportional distribution model, the system aerosol variables were assimilated from PMAAP. and The background errors of the seven types of aerosol variables in the four particle size segments required by the multi-component aerosol assimilation system MCAAP are calculated.

[0088] In the actual evaluation, taking the three-dimensional variational assimilation of regional atmospheric chemical aerosol data as an example, the atmospheric chemical background field and weather element background field are obtained by using the numerical model forecast results of the previous moment, and compared with the single-point observation data (wind field (u, v) is (5,3) m / s, temperature field is 290K, surface pressure is 1010hPa, specific humidity is 0.01kg / kg, PM 2.5 The observed value is 30μg / m³, PM 10 The observed value is 50 μg / m³) and is input into the three-dimensional variational assimilation system 3D-Var to obtain the optimal analysis field;

[0089] According to the numerical model forecast results at the previous moment, the initial values ​​of the atmospheric chemical background field control variables black carbon BC, organic carbon OC, road dust SD, sea salt SS, sulfate SF, and nitrate NT in the four particle size segments are [1,2,3,4], [2,3,4,5], [3,4,5,6], [4,5,6,7], [5,6,7,8], [6,7,8,9] (μg / m³), and the concentration of ammonium salt (AM) is 10μg / m³. and is the cumulative amount of the background field of each component, then The background field is 58μg / m³, The background field is 33 μg / m³, and the background error is known in the PMAAP assimilation system: μg / m³, ;

[0090] Based on the proportional distribution model, the background error of each control variable in the MCAAP assimilation system at the observation point is calculated. Taking the conversion of black carbon bc as an example, the first two particle size segments of bc are The proportions of the background field are 1 / 58 and 2 / 58 respectively. The background errors in the MCAAP assimilation system are calculated to be 30 / 58=0.51μg / m³ and 30*2 / 58=1.02μg / m³ respectively. The proportion of the background field is 3 / 33, and the background error in the MCAAP assimilation system is calculated to be 3*10 / 33=0.9μg / m³. The background error of the atmospheric chemical control variable in the fourth particle size segment of the MCAAP assimilation system is set to 0.2 of the background error of the atmospheric chemical control variable in the third particle size segment of the MCAAP assimilation system, and the background error is 0.18μg / m³.

[0091] Similarly, the background errors of the four particle sizes of black carbon (bc), organic carbon (oc), road dust (sd), sea salt (ss), sulfate (sf), and nitrate (nt) at the observation point in the MCAAP assimilation system are calculated to be [0.51, 0.102, 0.9, 0.18], [1.02, 1.55, 1.21, 0.242], [1.55, 2.06, 1.51, 0.302], [2.06, 2.58, 1.81, 0.362], [2.58, 3.10, 2.12, 0.424], and [3.10, 3.62, 2.42, 0.484] (μg / m³). The background error of ammonium in the MCAAP assimilation system is 5.17 μg / m³.

[0092] In this embodiment, the method for performing the incremental compensation includes:

[0093] Based on the assimilation of the same single-point ground aerosol observation PM 2.5 and PM10 , the analysis increment principle of the PMAAP assimilation system and the MCAAP assimilation system is adopted to calculate the background error increment of the control variables in the MCAAP assimilation system. Based on the background error increment, the atmospheric chemistry control variables in the MCAAP assimilation system are incrementally compensated to obtain the new background error. The expression of the new background error is:

[0094]

[0095]

[0096]

[0097]

[0098]

[0099] in and Respectively The background error of the aerosol variable X in each particle size segment and the new background error after background error increment compensation, Take 1 to 4, is the atmospheric chemical aerosol variable set, and are the background error of ammonium salt and the new background error after background error incremental compensation;

[0100] 、 is the correction parameter for the background error of the control variable in different particle size segments, where the parameter and The expression is:

[0101]

[0102]

[0103] in Variables in the PMAAP assimilation system The background error, Variables in the PMAAP assimilation system Background error;

[0104] In the actual evaluation, in the three-dimensional variational assimilation of regional atmospheric chemical aerosol data, in the 1st to 2nd particle size segment control variables, the new background error calculation of black carbon in the 2nd particle size segment in the BC assimilation system is taken as an example. The correction parameters of the 1st to 2nd particle size segment control variables are calculated based on the background error of the atmospheric chemical control variables in the BC assimilation system. In the third particle size segment control variable, taking the background error increment calculation of black carbon in the third particle size segment in the BC assimilation system as an example, according to the correction parameters of the third particle size segment control variable of the atmospheric chemistry control variable in the BC assimilation system The new background errors of atmospheric chemical control variables in all particle size segments in the BC assimilation system are calculated as follows: the new background errors of black carbon (BC), organic carbon (OC), road dust (SD), sea salt (SS), sulfate (SF), and nitrate (NT) in the four particle size segments are [6.435, 6.945, 3.284, 2.564], [6.945, 7.475, 3.594, 2.626], [7.475, 7.985, 3.894, 2.686], [7.985, 8.505, 4.194, 2.746], [8.505, 9.025, 4.504, 2.808], and [9.025, 9.545, 4.804, 2.868] (μg / m³), and the new background error of ammonium salt (AM) is 11.095 μg / m³.

[0105] In this embodiment, the method for obtaining the horizontal correlation coefficient and the vertical correlation coefficient includes:

[0106] The Gaussian correlation model is used to describe the background error level correlation coefficient, and the expression is:

[0107]

[0108] in is the background error level correlation coefficient, is the horizontal distance, is the horizontal correlation scale;

[0109] The vertical correlation coefficient of the background error is described using a correlation model that varies with height. The expression is:

[0110]

[0111] For the Layer and The vertical correlation coefficient of the background error of the layer, 、 Respectively Layer and The height of the layer, is the height sensitivity coefficient, is the acceleration due to gravity, is the dry air gas constant, is the average temperature on the pattern surface, is the particle size segment parameter;

[0112] In actual evaluation, the horizontal distance Point A and point B, where the height of point A is 1000m and the height of point B is 3000m, and the acceleration due to gravity is 9.8m / s 2 , the dry air gas constant is 287 J / (kg·K), and the average temperature on the model surface is 290 K; when black carbon bc, organic carbon oc, road dust sd, sea salt ss, sulfate sf, and nitrate nt are controlled variables, the horizontal correlation scale of aerosol variables and ammonium salt in particle size segments 1 and 2 is 100 km, and the horizontal correlation scale of aerosol variables in particle size segments 3 and 4 is 60 km. For the parameters of aerosol variables and ammonium salt in particle size segments 1 and 2, Take the aerosol variables of the 1st, 3rd and 4th particle size segments Take 2;

[0113] For the aerosol control variables and ammonium salts in size segments 1 and 2, the horizontal correlation scale is 100 km, and the size segment parameters Taking 1, the background error horizontal correlation coefficient and background error vertical correlation coefficient are calculated to be 0.88 and 0.95 respectively;

[0114] For the control variables of particle size segments 3 and 4, the horizontal correlation scale is 60 km, and the particle size segment parameters Taking 2, the background error horizontal correlation coefficient and background error vertical correlation coefficient are calculated to be 0.71 and 0.90 respectively;

[0115] The spatial transformation operator is determined by the background error horizontal correlation coefficient and the background error vertical correlation coefficient, and the background error covariance matrix is ​​described according to the new background error and the spatial transformation operator. , based on the background error and the spatial transformation operator by the normalized control variable Calculate the analysis increment.

[0116] In this embodiment, the method for calculating the gradient of the total objective function includes:

[0117] According to the normalized control variables Calculate the analysis increment and use EOF decomposition, recursive filtering method and physical transformation operator to normalize the control variable Perform vertical, horizontal, and physical transformations to obtain analysis increments:

[0118]

[0119] in To analyze the increment, is the space transformation operator, is the horizontal transformation, is the vertical transformation, is the physical transformation operator, is a normalized control variable; the spatial transformation operator consists of a horizontal transformation and a vertical transformation; the horizontal transformation realizes the horizontal propagation of observation information through recursive filtering; the vertical transformation realizes the propagation of vertical direction information through EOF decomposition;

[0120] According to the analysis increment Calculate observation equivalent increment , according to the background field and observation fields Determine the observation increment , according to the observed increment Equivalent to the observed increment Determine the observation residual , according to the observed residual Calculate the overall objective function , the expression is:

[0121]

[0122]

[0123] in is the normalized control variable The corresponding overall objective function is, , is the linear observation operator that projects the atmospheric state into the observation space, , is the observation error covariance matrix The inverse matrix of is the normalized control variable The adjoint operator of is the observed residual The adjoint operator of ;

[0124] Define the adjoint operator of the linear observation operator, the adjoint operator of the physical transformation and the adjoint operator of the spatial transformation, combined with the observation residual Calculate the total objective function gradient, the expression is:

[0125]

[0126] in is the total objective function gradient, is the normalized control variable, is the adjoint operator of the linear observation operator, is the adjoint operator of the physical transformation, is the adjoint operator of the space transformation, is the observation error covariance matrix The inverse matrix of .

[0127] In this embodiment, the method for obtaining the optimal normalized control variable includes:

[0128] According to the total objective function and the total objective function gradient, the minimization search direction and step size information are obtained through the finite memory variable scale minimization algorithm LBFGS, and the new normalized increment is calculated;

[0129] Calculate the new total objective function and the total objective function gradient according to the new normalized increment, update the normalized increment again, repeat the iteration until the total objective function is minimized, and output the optimal normalized control variable ;

[0130] In the actual evaluation, the dimension of the actual observation field is M, and the dimension of the aerosol control variable field is N. The dimension N includes the product of the three-dimensional spatial field and the number of control variables, that is, nx*ny*nz*25. The linear observation operator is taken It is an M*N matrix, the spatial operator and the physical transformation operator are both N*N matrices, and the N*N background error covariance matrix is ​​constructed based on the new background error. For an M*1 matrix, calculate the total objective function ;

[0131] Normalized control variables As the initial point, combined with the total objective function gradient , use the LBFGS algorithm to calculate the search direction and step length , calculate the new normalized control variable according to the search direction and step size , calculate the new total objective function and gradient, repeat the above operation until the total objective function obtained at the 15th iteration is minimum, and take the normalized control variable at the 15th iteration The optimal normalized control variable is transformed to obtain the analysis increment of atmospheric chemistry, and the optimal analysis field is determined based on the analysis increment of atmospheric chemistry and the first guess field.

[0132] The second aspect is a three-dimensional variational assimilation system for atmospheric chemical aerosol data assimilation, including:

[0133] Control variable module: used to design atmospheric chemical control variables and atmospheric chemical analysis variables, obtain atmospheric chemical background fields and weather element background fields, and input them into the three-dimensional variational assimilation system to generate the first guess field;

[0134] Background error module: used to determine the background error of the MCAAP assimilation system using the proportional distribution model according to the background error of the PMAAP assimilation system, and to obtain a new background error suitable for the MCAAP assimilation system by performing an incremental compensation method on the background error;

[0135] Variable transformation module: used to perform spatial transformation and physical transformation on the normalized control variable to obtain analysis increments, and to perform variable transformation on the optimal normalized control variable to obtain the optimal analysis increment; the variable transformation includes vertical transformation, horizontal transformation and physical transformation; the analysis increment includes atmospheric chemical analysis increment and weather element analysis increment;

[0136] Total objective function module: used to obtain the observation residual and the total objective function according to the analysis increment, and calculate the total objective function gradient according to the observation residual;

[0137] Search module: performing a minimization search according to the total objective function and the gradient of the total objective function to obtain the optimal normalized control variable;

[0138] Output module: used to generate an optimal analysis field according to the optimal analysis increment and the first initial guess field.

[0139] The above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions, improvements, etc. made within the spirit and principles of the present invention should be included in the scope of protection of the present invention.

Claims

1. A three-dimensional variational assimilation method for atmospheric chemical aerosol data assimilation, characterized by: The following steps are involved: S1. Design a multi-component aerosol assimilation framework (MCAAP) to obtain the atmospheric chemical background field and the weather element background field, and input them into a three-dimensional variational assimilation system as the first guess field; the first guess field includes the atmospheric chemical background field and the weather element background field; S2. Based on the atmospheric chemical background field and the background error of the PM aerosol assimilation framework PMAAP, the background error is proportionally distributed and incrementally compensated to obtain a new background error of the multi-component aerosol assimilation framework, and a correlation model is constructed to describe the horizontal correlation coefficient and vertical correlation coefficient of the background error applicable to the multi-component aerosol assimilation system MCAAP, thereby obtaining a horizontal transformation operator and a vertical transformation operator; S3. Constructing a background error covariance and performing variable transformation on the normalized control variable to obtain an analysis increment, obtaining an observation residual and a total objective function based on the analysis increment and observation information, and calculating a total objective function gradient based on the observation residual; the variable transformation includes vertical transformation, horizontal transformation, and physical transformation; and the analysis increment includes an atmospheric chemical analysis increment and a weather element analysis increment; S4. Perform a minimization search based on the total objective function and the gradient of the total objective function to obtain the optimal normalized control variable, and obtain the optimal analysis field based on the optimal normalized control variable and the first initial guess field; the optimal analysis field includes the optimal atmospheric chemistry analysis field and the optimal weather element analysis field.

2. A three-dimensional variational assimilation method for atmospheric chemical aerosol data assimilation according to claim 1, characterized in that: The method for designing a multi-component aerosol assimilation framework comprises: Black carbon (BC), organic carbon (OC), road dust (SD), sea salt (SS), sulfate (SF), and nitrate (NT) are divided into 24 control variables according to four particle size segments. The 24 control variables and the unsegmented ammonium salt (AM) form a total of 25 atmospheric chemical control variables. The assimilation system with the 25 multi-component aerosol variables as control variables is defined as the multi-component aerosol assimilation framework (MCAAP). The diameters of the four particle size segments include 0.01-1.28 μm, 1.28-2.56 μm, 2.56-10.24 μm, and 10.24-40.98 μm, respectively. The atmospheric chemical control variables are uncorrelated with each other and with weather factor control variables. The weather factor control variables include wind field, temperature field, surface air pressure, and specific humidity. The atmospheric chemical control variables, and Composition of atmospheric chemical analysis variables; and In the MCAAP assimilation system, it is the cumulative amount of the atmospheric chemistry control variables.

3. The three-dimensional variational assimilation method for atmospheric chemical aerosol data assimilation according to claim 1, characterized in that: The method for performing equal proportion distribution includes: The background error of the PMAAP assimilation system is assimilated using the proportional distribution model. 、 The proportion of the forecast field of different particle size segments in the total forecast field is converted into the background error of different particle size segments. The background error of the MCAAP assimilation system is calculated based on the atmospheric chemical background field and the known background error of the PMAAP assimilation system. The expression is: in Variables in the PMAAP assimilation system The background error, Variables in the PMAAP assimilation system In the multi-component aerosol assimilation framework MCAAP assimilation system, and Respectively The background error and background field of the aerosol variable X for each particle size segment, Take 1 to 4, is the atmospheric chemical aerosol variable set, and are the background error and background field of ammonium salt respectively; The background error of the aerosol variable in the fourth particle size segment is set to 1 / 5 of the background error of the aerosol in the third particle size segment, and the expression is: ； Using the equal-proportional distribution model, the system aerosol variables were assimilated from PMAAP. and The background errors of seven types of aerosol variables in four particle size segments required by the multi-component aerosol assimilation system MCAAP are calculated.

4. The three-dimensional variational assimilation method for atmospheric chemical aerosol data assimilation according to claim 1, characterized in that: The method for performing the incremental compensation comprises: Based on the assimilation of the same single-point ground aerosol observation PM 2.5 and PM 10 , the analysis increment principle of the PMAAP assimilation system and the MCAAP assimilation system is adopted to calculate the background error increment of the control variables in the MCAAP assimilation system. Based on the background error increment, the atmospheric chemistry control variables in the MCAAP assimilation system are incrementally compensated to obtain the new background error. The expression of the new background error is: in and Respectively The background error of the aerosol variable X in each particle size segment and the new background error after background error increment compensation, Take 1 to 4, is the atmospheric chemical aerosol variable set, and are the background error of ammonium salt and the new background error after background error incremental compensation; 、 is the correction parameter for the background error of the control variable in different particle size segments, where the parameter and The expression is: in Variables in the PMAAP assimilation system The background error, Variables in the PMAAP assimilation system background error.

5. The three-dimensional variational assimilation method for atmospheric chemical aerosol data assimilation according to claim 1, characterized in that: The method for obtaining the horizontal correlation coefficient and the vertical correlation coefficient includes: The Gaussian correlation model is used to describe the background error level correlation coefficient, and the expression is: in is the background error level correlation coefficient, is the horizontal distance, is the horizontal correlation scale; The vertical correlation coefficient of the background error is described using a correlation model that varies with height. The expression is: For the Layer and The vertical correlation coefficient of the background error of the layer, 、 Respectively Layer and The height of the layer, is the height sensitivity coefficient, is the acceleration due to gravity, is the dry air gas constant, is the average temperature on the pattern surface, is the particle size parameter.

6. The three-dimensional variational assimilation method for atmospheric chemical aerosol data assimilation according to claim 1, characterized in that: The method for calculating the gradient of the total objective function includes: According to the normalized control variables Calculate the analysis increment and use EOF decomposition, recursive filtering method and physical transformation operator to normalize the control variable Perform vertical, horizontal, and physical transformations to obtain analysis increments: in To analyze the increment, is the space transformation operator, is the horizontal transformation, is the vertical transformation, is the physical transformation operator, is a normalized control variable; the spatial transformation operator consists of a horizontal transformation and a vertical transformation; the horizontal transformation realizes the horizontal propagation of observation information through recursive filtering; the vertical transformation realizes the propagation of vertical direction information through EOF decomposition; According to the analysis increment Calculate observation equivalent increment , according to the background field and observation fields Determine the observation increment , according to the observed increment Equivalent to the observed increment Determine the observation residual , according to the observed residual Calculate the overall objective function , the expression is: in is the normalized control variable The corresponding overall objective function is, , is the linear observation operator that projects the atmospheric state into the observation space, , is the observation error covariance matrix The inverse matrix of is the normalized control variable The adjoint operator of is the observed residual The adjoint operator of ; Define the adjoint operator of the linear observation operator, the adjoint operator of the physical transformation and the adjoint operator of the spatial transformation, combined with the observation residual Calculate the total objective function gradient, the expression is: in is the total objective function gradient, is the normalized control variable, is the adjoint operator of the linear observation operator, is the adjoint operator of the physical transformation, is the adjoint operator of the space transformation, is the observation error covariance matrix The inverse matrix of .

7. The three-dimensional variational assimilation method for atmospheric chemical aerosol data assimilation according to claim 1, characterized in that: The method for obtaining the optimal normalized control variable comprises: According to the total objective function and the total objective function gradient, the minimization search direction and step size information are obtained through the finite memory variable scale minimization algorithm LBFGS, and the new normalized increment is calculated; Calculate the new total objective function and the total objective function gradient according to the new normalized increment, update the normalized increment again, repeat the iteration until the total objective function is minimized, and output the optimal normalized control variable .

8. A three-dimensional variational assimilation system for atmospheric chemical aerosol data assimilation, used to execute the method according to any one of claims 1 to 7, characterized in that: include: Control variable module: used to design atmospheric chemical control variables and atmospheric chemical analysis variables, obtain atmospheric chemical background fields and weather element background fields, and input them into the three-dimensional variational assimilation system to generate the first guess field; Background error module: used to determine the background error of the MCAAP assimilation system using the proportional distribution model according to the background error of the PMAAP assimilation system, and to obtain a new background error suitable for the MCAAP assimilation system by performing an incremental compensation method on the background error; Variable transformation module: used to perform spatial transformation and physical transformation on the normalized control variable to obtain analysis increments, and to perform variable transformation on the optimal normalized control variable to obtain the optimal analysis increment; the variable transformation includes vertical transformation, horizontal transformation and physical transformation; the analysis increment includes atmospheric chemical analysis increment and weather element analysis increment; Total objective function module: used to obtain the observation residual and the total objective function according to the analysis increment, and calculate the total objective function gradient according to the observation residual; Search module: performing a minimization search according to the total objective function and the gradient of the total objective function to obtain the optimal normalized control variable; Output module: used to generate an optimal analysis field according to the optimal analysis increment and the first initial guess field.

Images