Meteorological element three-dimensional analysis method combining remote sensing and ground observation

Abstract

The invention provides a meteorological element three-dimensional analysis method combining remote sensing and ground observation, and relates to the technical field of meteorological monitoring, and the method comprises the steps: obtaining remote sensing and ground observation data, and obtaining meteorological element data of a unified space-time reference; dividing the data into space-time grid units and performing quality evaluation to obtain a quality evaluation index; calculating a fusion weight coefficient according to the quality evaluation index based on an adaptive fusion algorithm of a dynamic weight, and performing weighted fusion on the meteorological element data in the grid units to generate an initial three-dimensional meteorological field; performing scale decomposition and reconstruction by adopting a spectrum analysis method to obtain an optimal three-dimensional meteorological field, constructing a self-adaptive tree-shaped composite analysis grid, analyzing evolution characteristics of a weather system in a multi-layer progressive mode, determining topological evolution parameters, and outputting a three-dimensional analysis result of meteorological elements.

Description

Technical Field

[0001] The present invention relates to the technical field of meteorological monitoring, and particularly to a three-dimensional analysis method for meteorological elements combining remote sensing and ground observations. Background Art

[0002] The three-dimensional analysis of meteorological elements is of great significance for weather forecasting and meteorological monitoring. Traditional analysis of meteorological elements mainly relies on data from ground observation stations. Limited by the spatial distribution of observation stations, it is difficult to obtain fine-structure information in the vertical direction of the atmosphere. With the development of remote sensing technology, remote sensing devices such as satellites and radars can obtain atmospheric information at different altitudes, providing new data sources for the three-dimensional analysis of meteorological elements. At present, the analysis of meteorological elements by combining remote sensing and ground observation data has become a research hotspot, but there are still many challenges in effectively integrating multi-source heterogeneous data and achieving refined analysis of meteorological elements.

[0003] However, there are still some deficiencies in the existing technologies. Remote sensing and ground observation data have different spatio-temporal resolutions and observation accuracies, and lack a unified data quality assessment standard, making it difficult to ensure the accuracy of the fusion results. Traditional data fusion methods often use fixed weights or simple linear superposition, without fully considering the physical consistency of the data, resulting in deviations between the fusion results and the actual atmospheric state. Existing analysis methods mostly adopt a fixed grid structure, which is difficult to adapt to the multi-scale characteristics of weather systems and cannot effectively depict the local details and evolution characteristics of meteorological elements.

[0004] In summary, there is an urgent need for a three-dimensional analysis method for meteorological elements combining remote sensing and ground observations. By establishing a unified spatio-temporal reference and data quality assessment system, effective fusion of multi-source data can be achieved; an adaptive fusion algorithm with dynamic weights is adopted to ensure the physical consistency of the fusion results; based on an adaptive tree-like composite analysis grid and a multi-layer progressive analysis method, refined analysis of meteorological elements can be realized, providing technical support for accurately depicting the evolution characteristics of weather systems. Summary of the Invention

[0005] An embodiment of the present invention provides a three-dimensional analysis method for meteorological elements combining remote sensing and ground observations, which can solve the problems in the existing technologies.

[0006] In the first aspect of the embodiment of the present invention, A three-dimensional analysis method for meteorological elements combining remote sensing and ground observations is provided, including: Obtaining remote sensing and ground observation data of a specified area, and performing spatio-temporal registration and data standardization processing to obtain meteorological element data with a unified spatio-temporal reference; Dividing the meteorological element data into different spatio-temporal grid units, and performing data quality assessment on the observation data within each spatio-temporal grid unit to obtain data quality assessment indicators for the spatio-temporal grid units; An adaptive fusion algorithm based on dynamic weights calculates the fusion weight coefficients of meteorological element data within a spatio-temporal grid cell according to the data quality evaluation index of the spatio-temporal grid cell, and performs weighted fusion on the meteorological element data within the spatio-temporal grid cell to generate an initial three-dimensional meteorological field with physical consistency; Based on the initial three-dimensional meteorological field, a spectral analysis method is used to perform scale decomposition and reconstruction on the meteorological element field to obtain an optimal three-dimensional meteorological field; An adaptive tree-shaped composite analysis grid is constructed based on the optimal three-dimensional meteorological field, and the evolution characteristics of the weather system are analyzed through a multi-layer progressive method to determine the topological evolution parameters and output the three-dimensional stereo analysis results of meteorological elements.

[0007] In an alternative embodiment, The meteorological element data is divided into different spatio-temporal grid cells, and data quality evaluation is performed on the observation data within each spatio-temporal grid cell. The data quality evaluation indexes of the spatio-temporal grid cell include: The meteorological element data is divided into multiple spatio-temporal grid cells according to a preset time interval and spatial resolution; Calculate the difference in observed values between any two observation points within the spatio-temporal grid cell to obtain the numerical consistency deviation; calculate the change amplitude of the observed values of each observation point at two adjacent moments within the spatio-temporal grid cell to obtain the time series volatility deviation; calculate the difference degree between the spatial distribution density of the observation points within the spatio-temporal grid cell and a preset density threshold to obtain the spatial density distribution deviation; The numerical consistency deviation, the time series volatility deviation, and the spatial density distribution deviation are linearly combined after normalization processing to obtain the data quality evaluation index of each spatio-temporal grid cell.

[0008] In an alternative embodiment, An adaptive fusion algorithm based on dynamic weights calculates the fusion weight coefficients of meteorological element data within a spatio-temporal grid cell according to the data quality evaluation index of the spatio-temporal grid cell, and performs weighted fusion on the meteorological element data within the spatio-temporal grid cell to generate an initial three-dimensional meteorological field with physical consistency, including: According to the data quality evaluation index of the spatio-temporal grid cell, calculate the weight influence factor of the spatio-temporal grid cell. The weight influence factor adopts the form of a Gaussian function and is calculated with the data quality evaluation index as the independent variable; Normalize the weight influence factor of the spatio-temporal grid cell to obtain the fusion weight coefficient of the meteorological element data within the spatio-temporal grid cell; perform weighted calculation on the meteorological element data within the spatio-temporal grid cell according to the fusion weight coefficient to obtain the observed value fitting term; construct a mass conservation constraint term based on the air density continuity equation, and simultaneously construct a momentum conservation constraint term based on the atmospheric dynamics equation; Combine the observed value fitting term, the mass conservation constraint term, and the momentum conservation constraint term to form a fusion objective function; Use an optimization algorithm based on adaptive step size and orthonormalized quasi-Newton iteration to solve the fusion objective function to obtain a fused meteorological field; Calculate the mass conservation residual and the momentum conservation residual of the fused meteorological field. When the mass conservation residual is less than a first preset threshold and the momentum conservation residual is less than a second preset threshold, determine that the fused meteorological field is an initial three-dimensional meteorological field with physical consistency.

[0009] In an alternative embodiment, Using an optimization algorithm based on adaptive step size and orthonormalized quasi-Newton iteration to solve the fusion objective function to obtain a fused meteorological field includes: Take the meteorological element data within the spatio-temporal grid cell as the initial solution, and based on the fusion objective function, calculate the gradient value of the fusion objective function corresponding to the initial solution; According to the initial solution and the gradient value of the fusion objective function, construct an iterative format using the adaptive step size factor and the approximate inverse matrix of the quasi-Newton matrix. When the change rate of the fusion objective function values between two adjacent iterations is greater than zero, decrease the adaptive step size factor according to a preset proportional coefficient, otherwise increase it according to the preset proportional coefficient. The value of the adaptive step size factor is restricted within a preset upper and lower limit range; In each iteration process, calculate the difference between the solution vectors of two adjacent iterations and the corresponding gradient difference vector of the fusion objective function; perform orthonormalization processing on the difference between the solution vectors and the gradient difference vector of the fusion objective function to obtain an orthonormalized difference vector; According to a preset iteration evaluation round, update the approximate inverse matrix of the quasi-Newton matrix in a two-way recursive manner based on the corresponding orthonormalized difference vector; use the updated approximate inverse matrix of the quasi-Newton matrix to calculate a new solution vector; Calculate the fusion objective function value corresponding to the new solution vector, and calculate the distance value between the solution vectors of two adjacent iterations; When the relative change amount of the fusion objective function value is less than a preset change threshold and the distance value between the solution vectors is less than a preset distance threshold, determine the new solution vector as the fused meteorological field, otherwise use the new solution vector as the initial solution for the next iteration and continue the iteration.

[0010] In an alternative embodiment, Based on the initial three-dimensional meteorological field, the spectral analysis method is used to decompose and reconstruct the meteorological element field, and the optimal three-dimensional meteorological field obtained includes: Perform Fourier transform on the initial three-dimensional meteorological element field to obtain spectral coefficients, perform wavelet transform on the spectral coefficients and a local window function to obtain decomposition coefficients of the meteorological element field at different scales and spatial positions, construct a multi-dimensional tensor based on the decomposition coefficients and perform nuclear norm calculation and tensor decomposition, and construct a wave number-energy spectrum matrix by combining the method of feature truncation extraction; Obtain the dominant wave number mode through singular value decomposition of the wave number-energy spectrum matrix, and determine the characteristic wave number sequence according to the energy contribution rate of the dominant wave number mode; Construct an orthogonal filter bank based on the characteristic wave number sequence, the orthogonal filter bank includes a plurality of Gaussian filters, the center frequency of each Gaussian filter corresponds to a characteristic wave number in the characteristic wave number sequence, and the bandwidth parameter of the Gaussian filter is determined according to the interval between adjacent characteristic wave numbers; Apply the orthogonal filter bank to the spectral coefficients to obtain a plurality of scale components, and construct a spectral reconstruction tensor based on the local energy distribution of the plurality of scale components; Calculate the local reconstruction weight according to the spectral reconstruction tensor, and perform weighted superposition on the plurality of scale components to obtain the optimal meteorological element field.

[0011] In an alternative embodiment, Constructing a multi-dimensional tensor based on the decomposition coefficients and performing nuclear norm calculation and tensor decomposition, and constructing a wave number-energy spectrum matrix by combining the method of feature truncation extraction includes: Construct a multi-dimensional tensor according to different scale levels of the decomposition coefficients, and calculate the nuclear norm of the multi-dimensional tensor to obtain a scale correlation matrix; Perform tensor decomposition on the scale correlation matrix to obtain a scale core tensor and a set of projection matrices, the set of projection matrices includes a spatial position projection matrix, a wave number projection matrix, and an energy projection matrix; Calculate the singular value sequence of the scale core tensor, use 80% of the maximum value of the singular value sequence as the feature truncation threshold, and extract the projection matrix components corresponding to the singular values in the singular value sequence that are greater than the feature truncation threshold; Reconstruct the projection matrix components to obtain the wave number-energy spectrum matrix.

[0012] In an alternative embodiment, Construct an adaptive tree-like composite analysis grid based on the optimal three-dimensional meteorological field, analyze the evolution characteristics of the weather system through a multi-layer progressive method, determine the topological evolution parameters, and output the three-dimensional stereo analysis result of the meteorological elements including: Obtain the meteorological field data in the optimal three-dimensional meteorological field, determine the reference resolution and layer coefficient according to the spatial distribution characteristics and gradient distribution of the meteorological field data, calculate the first resolution, and determine the backbone grid; calculate the second resolution according to the local change characteristics of the meteorological field data, determine the branch grid, and combine the backbone grid and the branch grid to generate an adaptive tree-shaped composite analysis grid; Calculate the local gradient field on the adaptive tree-shaped composite analysis grid, determine the grid branch growth direction, construct a multi-layer progressive interpolation function, and establish a numerical association between the backbone grid and the branch grid; Construct a characteristic metric tensor based on the local gradient field and the numerical association, calculate the evolution parameters of the characteristic metric tensor, and construct a characteristic transfer function between grid nodes according to the evolution parameters; Calculate the grid topological characteristics using the characteristic transfer function, obtain the topological stability index, structure evolution rate, and system deformation parameters, and construct a topological evolution characteristic operator; Construct a recursive operator according to the historical characteristics, current characteristics, and change characteristics, and use the recursive operator to perform multi-layer progressive fusion on the topological evolution characteristic operator, and output the three-dimensional analysis result of meteorological elements.

[0013] In the embodiments of the present invention, by performing spatio-temporal registration and standardization processing on remote sensing and ground observation data, the problem of inconsistent spatio-temporal benchmarks of data from different sources is solved, ensuring the basic accuracy of data fusion, and significantly reducing the systematic error in the data fusion process; based on the dynamic weight adaptive fusion algorithm for data quality assessment, the weights of different observation data are reasonably allocated, avoiding the negative impact of poor-quality data on the fusion result, and making the fusion result more reliable and stable; using the spectral analysis method for scale decomposition and reconstruction can retain the characteristic information of weather systems at different scales, improve the spatial resolution of the meteorological element field, and make the identification and analysis of small-scale weather phenomena more accurate; the construction method of the adaptive tree-shaped composite analysis grid dynamically adjusts the grid density according to the characteristics of the weather system, reducing the consumption of computing resources while ensuring the analysis accuracy, and improving the system operation efficiency; the multi-layer progressive topological feature analysis method can comprehensively capture the evolution characteristics of weather systems, accurately identify key weather processes, and provide a more reliable decision-making basis for meteorological forecasting and early warning. Brief Description of the Drawings

[0014] Figure 1 It is a schematic flow chart of the three-dimensional analysis method of meteorological elements combining remote sensing and ground observation in the embodiments of the present invention; Figure 2 It is a comparison curve of the convergence of the objective function; Figure 3 It is a bar chart of the residual value analysis; Figure 4It is a scatter plot effect diagram of the wavelet transform method and the tensor decomposition - feature truncation method in the three - dimensional feature space; Specific embodiments

[0015] To make the objectives, technical solutions, and advantages of the embodiments of the present invention clearer, the technical solutions in the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings in the embodiments of the present invention. Obviously, the described embodiments are only a part of the embodiments of the present invention, rather than all the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those of ordinary skill in the art without creative efforts shall fall within the protection scope of the present invention.

[0016] The technical solutions of the present invention will be described in detail below with specific embodiments. These specific embodiments can be combined with each other, and the same or similar concepts or processes may not be repeated in some embodiments.

[0017] Figure 1 It is a flow chart of a three - dimensional analysis method for meteorological elements combining remote sensing and ground observations in an embodiment of the present invention. As Figure 1 shown, the method includes: Obtain remote sensing and ground observation data of a specified area, and perform spatio - temporal registration and data standardization processing to obtain meteorological element data with a unified spatio - temporal benchmark; Divide the meteorological element data into different spatio - temporal grid cells, and perform data quality assessment on the observation data within each spatio - temporal grid cell to obtain the data quality assessment index of the spatio - temporal grid cell; Based on the adaptive fusion algorithm with dynamic weights, calculate the fusion weight coefficients of the meteorological element data within the spatio - temporal grid cell according to the data quality assessment index of the spatio - temporal grid cell, and perform weighted fusion on the meteorological element data within the spatio - temporal grid cell to generate an initial three - dimensional meteorological field with physical consistency; Based on the initial three - dimensional meteorological field, use the spectral analysis method to perform scale decomposition and reconstruction on the meteorological element field to obtain the optimal three - dimensional meteorological field; Based on the optimal three - dimensional meteorological field, construct an adaptive tree - shaped composite analysis grid, analyze the evolution characteristics of the weather system through a multi - layer progressive method, determine the topological evolution parameters, and output the three - dimensional stereoscopic analysis results of meteorological elements.

[0018] In a specific implementation, during the data acquisition and spatio-temporal registration stage, various types of remote sensing observation data within the study area are acquired, including meteorological elements such as temperature, humidity, and wind field observed by satellites. At the same time, conventional observation data from ground meteorological stations are acquired. During the spatio-temporal registration process, in view of the time resolution differences of different data sources, a time series interpolation method is used to unify the data onto the same time series. For spatial registration, a bilinear interpolation algorithm based on the geographic coordinate system is used to map data with different spatial resolutions into a unified grid system. Data standardization processing includes unit unification and numerical normalization. An improved min-max normalization method is used for numerical conversion, and at the same time, the moving window method is combined for outlier identification and processing.

[0019] In the spatio-temporal grid division and quality assessment link, based on the spatial scope and data characteristics of the study area, a three-dimensional grid system is established. Grid division considers three dimensions: longitude, latitude, and altitude, ensuring that the grid size is suitable for the spatial scale of weather systems. The data quality assessment adopts a multi-index comprehensive assessment method, including calculating the data integrity index, evaluating the spatial representativeness of data based on the spatio-temporal coverage rate of observation data; calculating the data consistency index, and using an improved correlation analysis method to evaluate the degree of consistency between data from different sources; calculating the data reliability index, and estimating the credibility of data according to the error characteristics of observation equipment and the observation environment. Finally, these indexes are synthesized into a unified quality assessment index through an adaptive weight method.

[0020] In the adaptive data fusion stage, based on the aforementioned quality assessment index, a dynamic weight function is constructed. The weight calculation adopts an adaptive weight adjustment algorithm based on data quality, which can automatically adjust the fusion weight according to the dynamic changes in data quality. For the observation data within each spatio-temporal grid cell, a weighted average method is applied for fusion, and at the same time, physical constraint conditions are introduced to ensure that the fusion result satisfies the basic laws of atmospheric dynamics. For the area where observation data is missing, an interpolation algorithm is used for spatial interpolation to generate a complete initial three-dimensional meteorological field.

[0021] In the spectral analysis and reconstruction link, spectral decomposition is performed on the initial three-dimensional meteorological field. The fast Fourier transform algorithm is used to decompose the meteorological field into different frequency domains to analyze the characteristics of weather systems at different scales. An optimized band-pass filter is designed in the frequency domain to effectively remove noise signals while retaining the characteristics of the main weather systems. The time-domain signal is reconstructed through inverse transformation, and an iterative optimization method is used to adjust the reconstruction parameters during the reconstruction process to ensure that the reconstructed three-dimensional meteorological field retains the physical characteristics of the weather system to the greatest extent.

[0022] In the final stage of adaptive grid construction and topological analysis, based on the reconstructed optimal three-dimensional meteorological field, a tree-like grid construction algorithm is adopted to adaptively adjust the grid density according to the gradient characteristics of the meteorological element field, and the grid is encrypted in the active area of the weather system. The topological feature analysis adopts a multi-layer progressive analysis method to extract the characteristic parameters such as the position, intensity, and movement of the weather system layer by layer from large scales to small scales. By calculating the temporal evolution of the characteristic parameters, the development trend of the system is determined, and finally a complete three-dimensional analysis result is generated, providing a quantitative description of the evolution characteristics of the weather system.

[0023] In this embodiment, by combining dynamic weight adaptive fusion and physical constraints, the problem of weight allocation in the fusion process of remote sensing and ground observation data is solved, ensuring the physical consistency of the fusion result, effectively improving the analysis accuracy of the three-dimensional meteorological field; realizing the effective separation and retention of the characteristics of weather systems at different scales, overcoming the limitations of traditional methods in dealing with multi-scale weather systems, and improving the spatial resolution of the meteorological element field; enabling the grid density to be automatically adjusted according to the activity level of the weather system, improving the analysis accuracy of key areas while ensuring the calculation efficiency, and achieving the optimal allocation of computing resources; being able to accurately capture the evolution characteristics of the weather system, providing a quantitative description basis for the development trend of the weather system, and enhancing the understanding level of the evolution law of the weather system.

[0024] In an alternative embodiment, the meteorological element data is divided into different spatio-temporal grid cells, and the data quality of the observation data within each spatio-temporal grid cell is evaluated. The data quality evaluation indicators for the spatio-temporal grid cells include: The meteorological element data is divided into multiple spatio-temporal grid cells according to a preset time interval and spatial resolution; The difference in the observed values between any two observation points within the spatio-temporal grid cell is calculated to obtain the numerical consistency deviation; the change amplitude of the observed values at each observation point between two adjacent moments within the spatio-temporal grid cell is calculated to obtain the time series volatility deviation; the difference degree between the spatial distribution density of the observation points within the spatio-temporal grid cell and the preset density threshold is calculated to obtain the spatial density distribution deviation; The numerical consistency deviation, the time series volatility deviation, and the spatial density distribution deviation are linearly combined after being normalized to obtain the data quality evaluation indicator for each spatio-temporal grid cell.

[0025] In a specific embodiment, the data is divided according to time and space dimensions. A suitable time interval is selected, for example, 1 hour, and the observed data is grouped according to this time interval. At the same time, the grid size is set in space, for example, 10 km × 10 km, and the study area is divided into regular grid cells. In this way, a grid cell system in two dimensions of time and space is formed, and each grid cell contains all the observed data within a specific time period and a specific spatial range.

[0026] For each spatio-temporal grid cell, three types of deviation indicators are calculated respectively. The first type is the numerical consistency deviation: within the same grid cell, any two observation points are selected, and the degree of difference between their observed values is calculated. This process is repeated for all possible pairs of observation points within the grid, and finally the average of all the difference values is taken to obtain the numerical consistency deviation of this grid cell. The second type is the time series volatility deviation: for each observation point within the grid cell, the change amount between the observed values at two adjacent moments is calculated, and it is evaluated whether this change exceeds a reasonable range. The time change characteristics of all the observation points within the grid are statistically analyzed to obtain the volatility deviation reflecting the time continuity of the data. The third type is the spatial density distribution deviation: an ideal observation point density threshold is set, the distribution density of the observation points within the actual grid cell is calculated, and the two are compared to evaluate the representativeness of the observed data in space.

[0027] Finally, the calculation of the comprehensive evaluation index is carried out. First, the above three types of deviations are normalized to make their numerical ranges unified between 0 and 1. Then, weight coefficients are set according to the influence degree of each type of deviation on the data quality, and the three normalized deviations are weighted and combined to obtain the final data quality evaluation index. The smaller this index value is, the better the data quality within this grid cell.

[0028] Exemplarily: Suppose in a certain urban meteorological observation network, a 10 km × 10 km area is selected, and the time period is from 10:00 to 11:00 on March 10, 2024. There are 5 surface automatic weather stations and 2 radiosonde stations in this spatio-temporal grid cell, and temperature data is observed.

[0029] Calculation of numerical consistency deviation: For example, at 10:30, the temperatures measured by the 5 automatic stations are 25.3 °C, 25.5 °C, 25.2 °C, 25.4 °C, 25.6 °C respectively, and the temperatures measured by the 2 radiosonde stations are 25.4 °C and 25.3 °C. Calculate the temperature difference between any two stations, and take the average of all the differences. Suppose an average deviation of 0.2 °C is obtained.

[0030] Calculation of time - series volatility deviation: Taking one of the automatic weather stations as an example, the temperature series every 10 minutes during the period from 10:00 to 11:00 is: 25.1°C, 25.2°C, 25.3°C, 25.5°C, 25.4°C, 25.3°C. Calculate the temperature change between adjacent moments, evaluate whether this change conforms to the normal temperature change law, and obtain an index reflecting the temporal continuity of the data.

[0031] Calculation of spatial density distribution deviation: The ideal number of observation stations in this 10 - kilometer × 10 - kilometer area should be 10 (preset density threshold), but there are actually only 7 observation stations. Calculate the difference between the actual density of observation points and the ideal density to obtain the deviation value of spatial representativeness.

[0032] After normalizing these three deviation values, assume that we get 0.2, 0.15, and 0.3 respectively. Set the weights to be 0.4, 0.3, and 0.3 respectively. Finally, the data quality assessment index of this spatio - temporal grid cell is calculated to be 0.217, indicating that the data quality within this grid cell is good.

[0033] In this embodiment, through the division of spatio - temporal grid cells and the design of multi - dimensional evaluation indicators, a refined evaluation of the quality of observation data is achieved. By combining the numerical comparison between adjacent observation points and the analysis of time - series volatility, outliers and mutation points in the observation data can be effectively identified. This method not only considers the data consistency in space but also pays attention to the data continuity in the time dimension, improving the recognition accuracy of abnormal data; introducing the evaluation of spatial density distribution deviation solves the problem of the impact of uneven spatial distribution of observation data on data quality assessment. By comparing with the preset density threshold, the spatial representativeness of observation data can be accurately evaluated, providing a reliable quality reference basis for subsequent data processing; based on the comprehensive evaluation method of normalization processing and linear combination, the effective fusion of evaluation indicators in different dimensions is realized. This method not only maintains the independence of each dimension index but also reflects the influence degree of different indexes on data quality through weight configuration, making the final evaluation result more scientific and practical.

[0034] In an alternative embodiment, based on the adaptive fusion algorithm with dynamic weights, according to the data quality assessment index of the spatio - temporal grid cell, calculate the fusion weight coefficient of the meteorological element data within the spatio - temporal grid cell, and perform weighted fusion on the meteorological element data within the spatio - temporal grid cell to generate an initial three - dimensional meteorological field with physical consistency, including: According to the data quality assessment index of the spatio - temporal grid cell, calculate the weight influence factor of the spatio - temporal grid cell. The weight influence factor adopts the form of a Gaussian function and is calculated with the data quality assessment index as the independent variable; Normalize the weight influence factor of the spatio-temporal grid cell to obtain the fusion weight coefficient of the meteorological element data within the spatio-temporal grid cell; perform weighted calculation on the meteorological element data within the spatio-temporal grid cell according to the fusion weight coefficient to obtain the observed value fitting term; construct a mass conservation constraint term based on the air density continuity equation, and at the same time construct a momentum conservation constraint term based on the atmospheric dynamics equation; Combine the observed value fitting term, the mass conservation constraint term, and the momentum conservation constraint term to form a fusion objective function; Use an optimization algorithm based on adaptive step size and orthogonalized quasi-Newton iteration to solve the fusion objective function to obtain a fused meteorological field; Calculate the mass conservation residual and the momentum conservation residual of the fused meteorological field. When the mass conservation residual is less than the first preset threshold and the momentum conservation residual is less than the second preset threshold, determine that the fused meteorological field is an initial three-dimensional meteorological field with physical consistency.

[0035] In a specific embodiment, calculate the weight influence factor according to the data quality evaluation index of the spatio-temporal grid cell. The specific process is as follows: for each grid cell, substitute its data quality evaluation index into the Gaussian function to calculate the weight influence factor. The better the data quality of the grid cell, the larger the obtained weight influence factor. Then normalize the weight influence factors of all grid cells so that their sum is 1 to obtain the fusion weight coefficient of the meteorological element data within each grid cell.

[0036] Construct a fusion objective function, which includes three parts: the first part is the observed value fitting term, obtained by performing weighted calculation on the meteorological element data within the grid cell; the second part is the mass conservation constraint term, constructed based on the air density continuity equation to ensure that the fusion result satisfies the law of mass conservation; the third part is the momentum conservation constraint term, constructed based on the atmospheric dynamics equation to ensure that the fusion result conforms to the law of momentum conservation.

[0037] Use an optimization algorithm to solve the fusion objective function. Adopt the adaptive step size and orthogonalized quasi-Newton iteration method, and gradually optimize and solve through multiple iterations until a fused meteorological field that meets the convergence conditions is obtained.

[0038] Conduct a physical consistency test. Calculate the mass conservation residual and the momentum conservation residual of the fused meteorological field, and compare them with the preset thresholds. Only when both residuals are less than their respective thresholds can it be determined that the fused meteorological field has good physical consistency and can be used as an initial three-dimensional meteorological field.

[0039] Exemplarily, assume that meteorological data fusion is carried out in a certain area at 10:00 on March 10, 2024, with a spatial range of 100 km × 100 km and a vertical range from the ground to a height of 10 km.

[0040] The data quality evaluation index of a certain grid cell is 0.2 (the smaller the value, the better the quality). Substituting it into the Gaussian function, the weight influence factor 0.85 is calculated. Normalize the weight influence factors of all grid cells, and the final fusion weight coefficient obtained by this grid cell is 0.15.

[0041] The observed value fitting term includes the data of elements such as temperature, air pressure, and wind speed within this grid cell, and multiplies by the corresponding weight coefficient 0.15; the mass conservation constraint term considers that the change of air density over time must satisfy the continuity equation; the momentum conservation constraint term considers that the airflow movement must conform to the law of conservation of momentum.

[0042] For the optimization solution, set the initial step size to 0.1. Through iterative calculation, both the step size and the search direction are updated in each iteration. For example, if the objective function value decreases by 20% after the first iteration, then increase the step size to 0.12; if the objective function value increases after a certain iteration, then decrease the step size to 0.08. Finally, after 50 iterations, the fused meteorological field is obtained.

[0043] The calculated mass conservation residual is 0.001 (less than the preset first threshold 0.005), and the momentum conservation residual is 0.002 (less than the preset second threshold 0.008). Therefore, it is confirmed that this fused meteorological field has good physical consistency and can be used as the initial three-dimensional meteorological field.

[0044] In this embodiment, by introducing a Gaussian weight allocation mechanism based on data quality, the differential fusion of observation data is realized. It not only makes full use of the advantages of high-quality data, but also effectively reduces the negative impact of low-quality data on the fusion result, improving the reliability of the fusion result; adopting a fusion objective function design method that combines multiple physical constraints, while ensuring the data fitting accuracy, introducing mass conservation and momentum conservation constraints, not only ensures the consistency between the fusion result and the observation data, but also guarantees the physical rationality of the fusion field, avoiding the non-physical solutions that may be brought by simple mathematical interpolation; through a physical consistency test mechanism controlled by residual thresholds, the quality control of the fusion result is realized, ensuring that the finally obtained three-dimensional meteorological field not only satisfies the observation constraints, but also conforms to the basic atmospheric physical laws, providing a high-quality initial field for subsequent meteorological numerical forecasting.

[0045] In an alternative embodiment, an optimization algorithm based on adaptive step size and orthogonalized quasi-Newton iteration is used to solve the fusion objective function, and the obtained fused meteorological field includes: Take the meteorological element data within the spatio-temporal grid cell as the initial solution, and based on the fusion objective function, calculate the gradient value of the fusion objective function corresponding to the initial solution; According to the initial solution and the gradient value of the fusion objective function, an iterative format is constructed using an adaptive step size factor and an approximate inverse matrix of the quasi-Newton matrix. When the change rate of the fusion objective function values between two adjacent iterations is greater than zero, the adaptive step size factor is decreased according to a preset proportional coefficient; otherwise, it is increased according to the preset proportional coefficient. The value of the adaptive step size factor is restricted within a preset upper and lower limit range; In each iteration process, calculate the difference between the solution vectors of two adjacent iterations and the corresponding gradient difference vector of the fusion objective function; orthogonalize the difference between the solution vectors and the gradient difference vector of the fusion objective function to obtain an orthogonalized difference vector; According to the preset iteration evaluation rounds, update the approximate inverse matrix of the quasi-Newton matrix in a two-way recursive manner based on the corresponding orthogonalized difference vector; calculate a new solution vector using the updated approximate inverse matrix of the quasi-Newton matrix; Calculate the fusion objective function value corresponding to the new solution vector, and calculate the distance value between the solution vectors of two adjacent iterations; When the relative change amount of the fusion objective function value is less than a preset change threshold and the distance value between the solution vectors is less than a preset distance threshold, the new solution vector is determined as the fused meteorological field; otherwise, the new solution vector is used as the initial solution for the next iteration and the iteration continues.

[0046] In a specific embodiment, at the beginning, the existing meteorological element data in the spatio-temporal grid cell is used as the starting point for the solution, that is, the initial solution. Based on this initial solution, substitute it into the constructed fusion objective function to calculate the gradient value at the position of this initial solution, and this gradient value indicates the improvement direction of the optimization.

[0047] To efficiently search for the optimal solution, it is necessary to determine an appropriate search step size, and an adaptive step size factor is used to dynamically adjust the size of the search step. At the same time, to accelerate the convergence speed, an approximate inverse matrix of the quasi-Newton matrix is introduced to approximate the second-order derivative information of the objective function. Specifically, an initial step size factor is set. In each iteration, the step size factor is adjusted by comparing the fusion objective function values obtained from the previous and the current iterations. If the objective function value obtained from the current iteration is larger than the previous one, it indicates that the step size may be too large, and at this time, the step size factor is reduced according to a preset proportional coefficient; conversely, if the objective function value of the current iteration is smaller, it indicates that the current search direction is effective, and the step size factor is appropriately increased to accelerate the convergence. To avoid excessive step size adjustment, upper and lower limits are set to restrict the value range of the step size factor.

[0048] In each round of iterative calculation, it is necessary to update the approximate inverse matrix of the quasi-Newton matrix in order to more accurately approximate the second-order derivative information of the objective function. The update process specifically includes calculating the difference between the solution vectors obtained from two adjacent iterations, which reflects the direction of change of the solution; and at the same time calculating the difference vector of the corresponding fused objective function gradient, which reflects the change of the gradient. To improve the numerical stability of the calculation, it is necessary to orthogonalize these two difference vectors to make them orthogonal to each other, and obtain the orthogonalized difference vectors.

[0049] According to the preset number of iterative evaluation rounds, use the orthogonalized difference vectors to update the approximate inverse matrix of the quasi-Newton matrix. The update process is carried out in a two-way recursive manner, that is, considering both forward and backward information to construct a more accurate approximate inverse matrix. After the update is completed, use the new approximate inverse matrix and the current gradient information to calculate the next solution vector.

[0050] For each newly obtained solution vector, it is necessary to calculate the value of the fused objective function corresponding to this solution vector, and calculate the distance value between this new solution vector and the solution vector of the previous iteration. These two calculation results are used to judge whether the iteration can be terminated: when the relative change amount of the fused objective function value (that is, the difference between the objective function values of two adjacent iterations and the ratio of the current value) is less than the preset change threshold, and the distance value between the solution vectors obtained from two adjacent iterations is also less than the preset distance threshold, it means that the iteration has converged to a satisfactory result, and at this time, the last obtained solution vector can be determined as the final fused meteorological field. If the convergence condition is not met, the newly obtained solution vector is used as the initial solution for the next round of iteration, and the iterative calculation process continues.

[0051] Exemplarily, in a certain atmospheric observation area for meteorological data fusion, the initial data includes the temperature field, pressure field and wind field data at 10:00 on March 10, 2024. Select these data as the initial solution, and calculate that the initial fused objective function value is 100, and the corresponding gradient value is 0.5.

[0052] Set the initial step size factor to 0.1, the upper limit to 0.5, the lower limit to 0.01, and the adjustment ratio coefficient to 1.2. In the first iteration, the new solution vector reduces the objective function value to 80, indicating that the search direction is effective, so the step size factor is increased to 0.12; after the second iteration, the objective function value increases to 85, indicating that the step size is too large, so the step size factor is reduced to 0.1.

[0053] During the iteration process, the changes in the solution vector and the gradient vector are continuously calculated. For example, at the third iteration, the change value of the temperature field solution vector is 0.5 °C, and the corresponding gradient change value is 0.2. These change amounts are orthonormalized to obtain an orthonormalized difference vector, which is used to update the approximate inverse matrix of the quasi-Newton matrix.

[0054] Set the number of rounds of iterative evaluation to 10. At the fourth iteration, use the orthonormalized difference vector to update the approximate inverse matrix by a two-way recurrence method. Recalculate using the updated approximate inverse matrix to obtain a new solution vector value of the temperature field of 15.2 °C.

[0055] Set the relative change threshold of the fusion objective function value to 0.001 and the solution vector distance threshold to 0.01. After 50 iterations, the ratio of the difference between the objective function values of two adjacent iterations to the current value is 0.0008, which is less than the change threshold of 0.001; at the same time, the distance value between two adjacent solution vectors is 0.008, which is less than the distance threshold of 0.01. This indicates that the iteration has converged, so the solution vector obtained at the 50th iteration is determined as the final fused meteorological field.

[0056] In the prior art, variational assimilation methods or optimal interpolation methods are usually adopted for meteorological data fusion. Variational assimilation methods mainly achieve data fusion by constructing a cost function and solving its minimum value. However, the optimization and solution process usually adopts a gradient descent method or a conjugate gradient method with a fixed step size. When dealing with meteorological element data with strong nonlinearity, these methods are prone to problems such as slow convergence speed and easy to fall into local optimal solutions. The optimal interpolation method mainly relies on statistical correlation information for interpolation calculation. Although the calculation efficiency is relatively high, it is difficult to ensure that the fusion result satisfies the atmospheric dynamics constraints, resulting in poor physical consistency.

[0057] In view of the problems existing in the prior art, improvements have been made from the following aspects: In terms of optimizing algorithm design: Introduce an adaptive step size mechanism, and dynamically adjust the search step size by monitoring the change of the objective function value in real time. Increase the step size to accelerate convergence when the optimization effect improves, and decrease the step size to ensure stability when the optimization effect deteriorates. At the same time, set the upper and lower limit constraints of the step size to avoid numerical instability caused by excessive step size adjustment. This adaptive mechanism can better balance the optimization efficiency and stability compared with the fixed step size method.

[0058] In terms of matrix approximation: Use a two-way recurrence method to update the approximate inverse matrix of the quasi-Newton matrix, and introduce orthonormalization to improve numerical stability. This method improves the accuracy of matrix approximation while maintaining the calculation efficiency compared with the traditional BFGS (Broyden-Fletcher-Goldfarb-Shanno) algorithm, enabling the optimization process to better capture the second-order information of the objective function.

[0059] In terms of convergence judgment: A dual convergence judgment mechanism based on the relative change of the integrated objective function value and the distance value of the solution vector is designed, which is more reliable than the single objective function value judgment and can more accurately determine whether the iterative process is truly convergent.

[0060] In Figure 2 the comparison of the convergence performance between the proposed technical solution and traditional methods is shown. The objective function value of the proposed technical solution has dropped to 45.2 at the 10th iteration, while the traditional variational assimilation method and the optimal interpolation method require 25 and 35 iterations respectively to reach a similar level. By the 50th iteration, the objective function value of the proposed technical solution has dropped to 8.3, which is 47.1% and 62.9% lower than that of the traditional variational assimilation method (15.7) and the optimal interpolation method (22.4) respectively. Especially in the first 20 iterations, the proposed technical solution shows a faster convergence speed, which fully proves the effectiveness of the adaptive step size and the two-way recursive update strategy.

[0061] In Figure 3 the performance of the three methods in maintaining physical consistency is shown. In terms of mass conservation, the residual value of the proposed technical solution is 0.015 kg / m³, which is 34.8% and 46.4% lower than that of the traditional variational assimilation method (0.023 kg / m³) and the optimal interpolation method (0.028 kg / m³) respectively. In terms of momentum conservation, the residual value of the proposed technical solution is 0.008 N·s / m³, which is also significantly lower than the other two methods.

[0062] In this embodiment, the starting point of the improvement is to solve three main problems existing in the existing meteorological data fusion methods: low convergence efficiency of the optimization process; insufficient numerical calculation stability; and poor physical consistency of the fusion results. Through the above improvements, significant effects have been achieved. Compared with the fixed-step method, the convergence speed has increased by about 40% and the number of iterations has decreased by about 30%; when dealing with strongly nonlinear meteorological element data, the probability of numerical divergence has decreased by about 50%; the root mean square error of the fusion result has decreased by about 25% compared with the traditional method, and the residual values of mass conservation and momentum conservation have decreased by about 35% and 30% respectively. The improvement effect enables the method of this application to achieve meteorological data fusion more efficiently and stably, while ensuring that the fusion result has better physical consistency, providing a more reliable initial field for meteorological numerical prediction.

[0063] In an alternative embodiment, based on the initial three-dimensional meteorological field, a spectral analysis method is used to perform scale decomposition and reconstruction on the meteorological element field, and the optimal three-dimensional meteorological field obtained includes: The Fourier transform is performed on the initial three-dimensional meteorological element field to obtain spectral coefficients. The wavelet transform is performed on the spectral coefficients and a local window function to obtain the decomposition coefficients of the meteorological element field at different scales and spatial positions. A multi-dimensional tensor is constructed based on the decomposition coefficients, and the nuclear norm calculation and tensor decomposition are performed. A wave number-energy spectrum matrix is constructed by combining the method of feature truncation extraction. The dominant wave number modes are obtained through the singular value decomposition of the wave number-energy spectrum matrix, and the characteristic wave number sequence is determined according to the energy contribution rate of the dominant wave number modes. An orthogonal filter bank is constructed based on the characteristic wave number sequence. The orthogonal filter bank includes a plurality of Gaussian filters. The center frequency of each Gaussian filter corresponds to a characteristic wave number in the characteristic wave number sequence, and the bandwidth parameter of the Gaussian filter is determined according to the interval between adjacent characteristic wave numbers. The orthogonal filter bank is applied to the spectral coefficients to obtain a plurality of scale components, and a spectral reconstruction tensor is constructed based on the local energy distribution of the plurality of scale components. The local reconstruction weight is calculated according to the spectral reconstruction tensor, and the plurality of scale components are weighted and superimposed to obtain the optimal meteorological element field.

[0064] In a specific embodiment, a frequency domain transformation process is performed on the initial three-dimensional meteorological element field. The data of the meteorological element field is transformed to the frequency domain space through the Fourier transform to obtain the corresponding spectral coefficients. Then, a suitable local window function is selected, and the wavelet transform is performed on these spectral coefficients to obtain the decomposition coefficients of the meteorological element field at different spatial scales and different positions. These decomposition coefficients reflect the energy distribution characteristics of the meteorological field at different scales.

[0065] Next, these decomposition coefficients are organized into the form of a multi-dimensional tensor according to different scale levels. The nuclear norm of the multi-dimensional tensor is calculated to obtain a matrix describing the correlation between different scales. Then, tensor decomposition is performed on this correlation matrix to extract its core features. In the obtained singular value sequence, 80% of the maximum singular value is selected as the threshold for feature truncation, and the eigenvectors corresponding to the singular values greater than this threshold are extracted, and the wave number-energy spectrum matrix is reconstructed using these eigenvectors.

[0066] The singular value decomposition is performed on the obtained wave number-energy spectrum matrix, and the dominant wave number modes are extracted from the decomposition results. By calculating the energy contribution rates of these dominant wave number modes, the wave numbers with a cumulative contribution rate reaching a preset threshold are selected as the characteristic wave number sequence in the order of decreasing energy contribution. These characteristic wave number sequences reflect the most important spatial scale features in the meteorological field.

[0067] Based on the selected characteristic wave number sequence, a set of orthogonal Gaussian filters is constructed. Each Gaussian filter takes a wave number in the characteristic wave number sequence as its center frequency, and the bandwidth of the filter is determined according to the interval between adjacent characteristic wave numbers. The filter bank constructed in this way can effectively separate meteorological features of different scales.

[0068] Apply the constructed orthogonal filter bank to the previously obtained spectral coefficients to obtain multiple components of different scales. Calculate the local energy distribution characteristics of these scale components and construct a spectral reconstruction tensor accordingly. Finally, calculate the weight coefficients of each scale component in the reconstruction process according to the spectral reconstruction tensor, and obtain the final optimal meteorological element field by weighted superposition of these scale components.

[0069] Exemplarily, assume that the temperature field data within a range of 1000 km × 1000 km × 10 km in a certain area at 10:00 on March 10, 2024 is processed.

[0070] First, perform Fourier transform on the temperature field data to obtain spectral coefficients. Select the Gaussian window function as the local window function, set the window size to 100 km, and perform wavelet transform on the spectral coefficients to obtain decomposition coefficients of different scales (such as 2 km, 10 km, 50 km, etc.) and different positions.

[0071] Organize these decomposition coefficients into a three-dimensional tensor (scale dimension × longitude dimension × latitude dimension), calculate the nuclear norm to obtain the scale correlation matrix. Perform tensor decomposition on the correlation matrix. Assume that the maximum singular value is 10, then take the threshold 8 (80% of the maximum singular value) for feature truncation, and extract the corresponding eigenvectors to reconstruct the wave number-energy spectrum matrix.

[0072] Perform singular value decomposition on the wave number-energy spectrum matrix. Assume that it is found that the modal energy contribution rates of wave numbers 2, 5, and 10 (unit: per 100 km) are 50%, 30%, and 15% respectively, and their cumulative contribution rate reaches 95%, exceeding the preset threshold of 90%. Therefore, select these three wave numbers as the characteristic wave number sequence.

[0073] Based on these three characteristic wave numbers, construct three Gaussian filters with center frequencies of 2, 5, and 10 respectively, and the bandwidth parameters are determined according to the interval between adjacent wave numbers. For example, the bandwidth of the first filter is taken as 3 (the difference from the adjacent wave number 5), the second is taken as 5, and the third is taken as 5.

[0074] Apply this set of filters to the original spectral coefficients to obtain three scale components. Calculate the local energy distribution of each scale component and construct a spectral reconstruction tensor. The weights of the three scale components calculated according to the reconstruction tensor are 0.5, 0.3, and 0.2 respectively. Finally, perform weighted superposition of the three scale components according to these weights to obtain the final temperature field.

[0075] The processing method in this embodiment can not only isolate meteorological features at different spatial scales, but also retain the spatial position information of these features. Compared with the traditional single-scale analysis method, it improves the ability to identify local features, can better retain the features of small-scale weather systems, and reduces the information loss in the feature extraction process; compared with the direct spectral analysis method, it can more accurately capture the dominant wavenumber modes in the meteorological field and improve the accuracy of feature extraction; it considers the correlation between different scales, effectively reduces the recognition rate of false features; the adaptive filter design method improves the accuracy of scale separation and reduces the crosstalk between scale components compared with the fixed-parameter filtering scheme; through the weighted reconstruction scheme guided by the spectral reconstruction tensor, it ensures the physical rationality of the final reconstruction result, reduces the root mean square error of the reconstruction result and improves the spatial correlation coefficient compared with the traditional equal-weight superposition method.

[0076] In an alternative embodiment, constructing a multi-dimensional tensor based on the decomposition coefficients and performing nuclear norm calculation and tensor decomposition, and constructing a wavenumber-energy spectrum matrix in combination with the feature truncation extraction method includes: Construct a multi-dimensional tensor from the decomposition coefficients according to different scale levels, and calculate the nuclear norm of the multi-dimensional tensor to obtain a scale correlation matrix; Perform tensor decomposition on the scale correlation matrix to obtain a scale core tensor and a set of projection matrices, and the set of projection matrices includes a spatial position projection matrix, a wavenumber projection matrix, and an energy projection matrix; Calculate the singular value sequence of the scale core tensor, use 80% of the maximum value of the singular value sequence as the feature truncation threshold, and extract the projection matrix components corresponding to the singular values greater than the feature truncation threshold in the singular value sequence; Reconstruct the projection matrix components to obtain the wavenumber-energy spectrum matrix.

[0077] In a specific implementation manner, the first important task is to systematically organize the decomposition coefficients obtained in the previous steps. A complete three-dimensional indexing system needs to be established, including longitude and latitude indexing of spatial positions, wavenumber indexing, and energy level indexing. At each determined grid point position, first collect all the decomposition coefficients corresponding to different wavenumbers at this point. Then, for the coefficient values at each wavenumber, calculate their corresponding energy values and divide them into different energy levels according to the energy magnitude. Finally, fill the organized data into the pre-established tensor structure in the order of position, wavenumber, and energy to form a complete multi-dimensional tensor.

[0078] When calculating the nuclear norm of the constructed multi-dimensional tensor, it is first necessary to unfold this three-dimensional tensor into a series of matrix slices. Perform singular value decomposition operations on each matrix slice separately to obtain the singular values corresponding to this slice. Sum up the singular values of all slices to obtain the overall nuclear norm value. Then reorganize the nuclear norm values obtained under different dimension combinations to form a correlation matrix describing scale correlation features.

[0079] When performing tensor decomposition, the higher-order singular value decomposition method is adopted. First, unfold the tensor along the spatial position dimension, calculate the covariance matrix in this dimension, and perform eigenvalue decomposition on the covariance matrix to obtain the projection matrix corresponding to the spatial position. Using the same processing method, perform tensor unfolding and decomposition operations on the wave number dimension and the energy dimension respectively to obtain the projection matrices corresponding to these two dimensions. In the calculation of the spatial position projection matrix, it is necessary to construct a spatial correlation matrix and extract its eigenvectors. The wave number projection matrix is obtained by calculating the correlation coefficients between different wave numbers based on wave number spectrum analysis. The energy projection matrix is obtained by analyzing the energy distribution characteristics and solving after constructing an energy conversion relationship matrix.

[0080] When calculating the singular values of the core tensor, it is necessary to first unfold the core tensor into a two-dimensional matrix form. Calculate the eigenvalues and eigenvectors of this unfolded matrix, sort the obtained eigenvalues in descending order to obtain the singular value sequence. At the same time, it is necessary to record the eigenvectors corresponding to each singular value. When performing eigenvalue truncation processing, first find the maximum value in the singular value sequence, multiply it by 0.8 to obtain the truncation threshold. Then traverse the entire singular value sequence, mark the positions of all singular values greater than this threshold, and extract the eigenvectors corresponding to these positions as the main eigencomponents.

[0081] During the process of reconstructing the spectral matrix, first extract the corresponding column vectors from each projection matrix according to the positions of the retained singular values. Orthogonalize these extracted column vectors to ensure the orthogonality between vectors, and at the same time check and adjust the direction consistency of each vector. Then reorganize these processed eigenvectors according to the original dimensions, calculate the inner products between the eigenvectors to obtain the correlation coefficients. Finally, construct a wave number-energy spectral matrix based on these correlation coefficients and normalize the reconstructed matrix.

[0082] Taking the temperature field data of a 100×100 grid in a certain area as an example: First, organize the original decomposition coefficients into a four-dimensional tensor according to 100 longitude points, 100 latitude points, 10 wave numbers, and 5 energy levels. At each grid point, there are decomposition coefficients of 10 different wave numbers, and these coefficients are classified into 5 different energy levels after energy value calculation. Expand this four-dimensional tensor into 50 100×100 matrices, calculate the singular values for each matrix, for example, obtain a sequence like [15, 12, 8, 5, 3], sum all the singular values to get the nuclear norm value, and thus construct a 10×5 scale correlation matrix. Through tensor decomposition, obtain a 100×8-dimensional spatial position projection matrix, a 10×4-dimensional wave number projection matrix, a 5×3-dimensional energy projection matrix, and an 8×4×3 core tensor. The singular value sequence of the core tensor is [10, 8, 6, 4, 2, 1], take the truncation threshold as 8, retain the eigenvectors corresponding to the singular value 10, and finally reconstruct a 10×5 wave number-energy spectrum matrix.

[0083] During the entire processing process, special attention needs to be paid to the stability of numerical calculations to ensure the continuity and conservation of physical quantities. When calculating energy values, the units should be unified, the condition number should be controlled during matrix decomposition, and the orthogonality of eigenvectors should be ensured during the reconstruction process. These are all key elements to ensure the accuracy of calculation results.

[0084] This embodiment is based on the multi-scale analysis method in meteorology. Traditional meteorological multi-scale analysis mainly uses methods such as Fourier transform and wavelet transform, which have been widely used in dealing with the multi-scale characteristics of atmospheric motion. At the same time, this technology also draws on tensor decomposition and feature extraction technologies in the field of data science, which show good effects in dealing with multi-dimensional data analysis.

[0085] The implementation means in the prior art include: The traditional Fourier spectrum analysis method obtains the wave number spectrum by performing Fourier transform on the meteorological field, but this method cannot obtain spatial local information at the same time; The wavelet transform method can obtain time-frequency localization information, but has a high computational complexity when dealing with multi-dimensional data and is difficult to maintain the correlation of features in each dimension; The empirical mode decomposition method can adaptively decompose signals, but is prone to mode mixing, affecting the physical meaning of the decomposition results; The traditional filtering method uses a filter with fixed parameters for scale separation and is difficult to adapt to the characteristic differences of different weather systems.

[0086] In this embodiment, in view of the problems existing in the prior art, the following improvements are made: introducing the tensor decomposition technology into meteorological multi-scale analysis, constructing a multi-dimensional tensor to characterize the multi-scale features of the meteorological field, and maintaining the correlation between features in different dimensions; adopting the nuclear norm calculation and feature truncation methods to extract the dominant features, reducing the computational complexity and improving the accuracy of feature extraction; designing an adaptive filter bank based on the characteristic wave number to enable the filter parameters to be automatically adjusted according to the actual data features; and proposing a weighted reconstruction scheme based on the spectral reconstruction tensor to optimize the combination method of different scale components.

[0087] The improved technology is mainly based on the following considerations: the meteorological element field has obvious multi-scale features and more effective methods are needed to separate and identify these features; there are complex interactions between meteorological features at different scales and the correlation between these features needs to be maintained; different types of weather systems have different characteristic scales and the method needs to have good adaptability; the meteorological data volume is large and it is necessary to improve the computational efficiency and reduce the storage requirements.

[0088] Figure 4 Shows the distribution characteristics of the tensor decomposition - feature truncation method and the traditional wavelet transform method in the feature space. The horizontal axis represents the wave number (1 - 10), the vertical axis represents the energy level (1 - 5), and the depth axis represents the feature intensity (0 - 100). The feature points (square markers) of this technical solution show a clear hierarchical distribution, with a high degree of aggregation and uniform spatial distribution of the feature points in the high-energy region (>80), indicating that the extracted features have good physical continuity. The feature points (circular markers) of the traditional method based on wavelet transform show greater discreteness, especially an obvious blank band appears in the energy level conversion region, indicating the discontinuity of feature extraction. Through quantitative analysis, the spatial correlation coefficient of the feature points of this technical solution reaches 0.92, which is significantly higher than 0.76 of the wavelet transform method.

[0089] In this embodiment, compared with the traditional method, it can better maintain the correlation between multi-scale features, improve the accuracy and integrity of feature extraction; through feature truncation and tensor operations, the processing efficiency is significantly improved and the storage requirements are reduced; the adaptive filter bank can better adapt to the features of different weather systems and improve the universality of the method; the weighted reconstruction scheme improves the physical rationality of the final result and improves the reconstruction quality.

[0090] In an optional embodiment, an adaptive tree-shaped composite analysis grid is constructed based on the optimal three-dimensional meteorological field, the evolution characteristics of the weather system are analyzed in a multi-layer progressive manner, the topological evolution parameters are determined, and the three-dimensional stereoscopic analysis results of meteorological elements include: Acquire meteorological field data in the optimal three-dimensional meteorological field, determine the reference resolution and the hierarchical coefficient according to the spatial distribution characteristics and gradient distribution of the meteorological field data, calculate the first resolution, and determine the main grid; calculate the second resolution according to the local change characteristics of the meteorological field data, determine the branch grid, and combine the main grid and the branch grid to generate an adaptive tree-like composite analysis grid; Calculating the local gradient field on the adaptive tree-like composite analysis grid, determining the growth direction of the grid branches, constructing a multi-layer progressive interpolation function, and establishing a numerical association between the trunk grid and the branch grid; constructing a feature metric tensor based on the local gradient field and the numerical association, calculating an evolution parameter of the feature metric tensor, and constructing a feature transfer function between grid nodes according to the evolution parameter; The characteristic transfer function is used to calculate the grid topological characteristics, obtain the topological stability index, the structural evolution rate and the system deformation parameter, and construct the topological evolution characteristic operator; A recursive operator is constructed according to historical features, current features and change features, and the recursive operator is used to perform multi-layer progressive fusion on the topological evolution feature operator to output a three-dimensional analysis result of the meteorological elements.

[0091] In a specific implementation, after obtaining the data of the optimal three-dimensional meteorological field, it is necessary to analyze the spatial distribution characteristics of the data. Check the changes in meteorological elements in three-dimensional space, focusing on the sudden change areas and gentle areas of the values. At the same time, calculate the gradient distribution of the entire field, and identify areas with drastic gradient changes and areas with relatively gentle gradients. Based on these characteristics, determine the baseline resolution, and this resolution serves as a reference standard for the entire grid system. Then set the hierarchy coefficient to control the density of grids at different levels. Calculate the first resolution based on the baseline resolution and the hierarchy coefficient, and use this resolution to establish a backbone grid. The backbone grid covers the entire analysis area and constitutes the basic framework of the grid system.

[0092] Analyze the local change characteristics of meteorological field data, focusing on the change amplitude and frequency of local values. In areas with significant changes, calculate the required second resolution, which is usually finer than the first resolution. Establish branch grids based on the second resolution, and the branch grids are mainly distributed in areas where meteorological elements change dramatically. Combine the trunk grid and the branch grid in a tree structure to form an adaptive tree-like composite analysis grid.

[0093] On the established composite analysis grid, the local gradient field of each grid cell is calculated. The growth direction of the grid branch is determined according to the direction and size of the gradient field, which determines the extension method of the branch grid. A multi-layer progressive interpolation function system is constructed. These functions are used to establish numerical connections between the trunk grid and the branch grid to ensure smooth transition of data between grids of different resolutions.

[0094] Based on the calculated local gradient field and the numerical association between the grids, a feature metric tensor is constructed. This tensor contains the relationship information between grid nodes and local feature information. The evolution parameters of this feature metric tensor are calculated, which reflect the characteristics of the grid structure changing over time. The feature transfer function between grid nodes is constructed based on the evolution parameters, which describes the law of feature transfer between grid nodes.

[0095] The topological characteristics of the grid are analyzed using the characteristic transfer function, and three key indicators are calculated: topological stability index, structural evolution rate, and system deformation parameter. The topological stability index reflects the stability of the grid structure, the structural evolution rate describes the speed of change of the grid morphology, and the system deformation parameter characterizes the deformation of the overall structure. Based on these three indicators, a topological evolution characteristic operator is constructed.

[0096] Collect historical feature data, combine current features and change features, and construct a recursive operator. This recursive operator can capture the evolution law of features. Use the recursive operator to perform multi-layer progressive fusion of topological evolution feature operators, and finally generate a complete three-dimensional analysis result.

[0097] For example, take the analysis of the temperature field of a certain area at 10:00 on March 10, 2024. The area range is 1000km×1000km×10km, and the initial data resolution is 10km.

[0098] By analyzing the spatial distribution of the temperature field, it was found that there was an obvious temperature gradient at 0-2 km near the ground, while the temperature change was relatively gentle at 2-10 km. The baseline resolution was determined to be 5 km, and the layer coefficient was set to 2. The first resolution was calculated to be 10 km, and a backbone grid covering the entire area was established.

[0099] The analysis found that in areas with drastic temperature changes near the ground, the required second resolution was calculated to be 2.5km. Branch grids were established in these areas, mainly distributed in the 0-2km altitude layer. The two grids were combined into a tree structure to form a composite analysis grid.

[0100] The gradient field is calculated on the composite grid, and it is found that the temperature gradient is mainly distributed in the vertical direction. Based on this, it is determined that the growth direction of the branch grid mainly extends in the vertical direction. A cubic spline interpolation function is established for data transfer between the trunk grid and the branch grid.

[0101] Construct a characteristic metric tensor, which contains the temperature difference and spatial distance information between grid nodes. Calculate the evolution parameters of the tensor and find that the temporal changes of the temperature field are mainly concentrated in the near-surface layer. Based on this, construct a characteristic transfer function to describe the transfer law of temperature characteristics between grids.

[0102] The calculated topological stability index indicates that the near-surface grid structure needs to be updated more frequently. The structure evolution rate shows that the changes in the temperature field mainly occur during the warming process after sunrise. The system deformation parameter reflects the vertical structure characteristics of the temperature field.

[0103] Finally, combining the temperature field evolution characteristics of the previous hour, the current temperature distribution characteristics, and the temperature change trend, a recursive operator is constructed. Through the recursive operator, the topological features are fused, and finally, the complete three-dimensional temperature field analysis result is obtained.

[0104] In this embodiment, through the design of the adaptive tree-like composite analysis grid, a finer resolution is adopted in the areas where meteorological elements change violently, while a relatively sparse grid distribution is maintained in the areas with gentle changes, which not only ensures the analysis accuracy of key areas but also avoids the waste of computing resources; the numerical association between the backbone grid and the branch grid is established to achieve a smooth transition between grids with different resolutions; the introduction of the feature metric tensor ensures the physical rationality of feature transfer between grid nodes and effectively avoids the numerical false oscillations that occur in traditional methods; the grid system is organized in a tree-like structure, and by dynamically adjusting the grid distribution, the number of computing nodes is significantly reduced. At the same time, the design of the feature transfer function optimizes the numerical calculation process and reduces the computational complexity of the algorithm; the method of determining the grid branch growth direction based on the local gradient field enables the grid structure to adaptively follow the change characteristics of meteorological elements, and the introduction of the topological evolution feature operator enhances the adaptability of the method to different weather systems; by constructing a recursive operator, the historical features, current features, and change features are organically combined, ensuring the continuity of the analysis results in time and space and improving the spatio-temporal consistency of the analysis results.

[0105] The present invention can be a method, apparatus, system, and / or computer program product. The computer program product may include a computer-readable storage medium having thereon computer-readable program instructions for performing various aspects of the present invention.

[0106] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and are not intended to limit them; although the present invention has been described in detail with reference to the foregoing embodiments, those of ordinary skill in the art should understand that they can still modify the technical solutions described in the foregoing embodiments, or perform equivalent replacements for some or all of the technical features; and these modifications or replacements do not cause the essence of the corresponding technical solutions to deviate from the scope of the technical solutions of the embodiments of the present invention.

Claims

1. A three-dimensional analysis method of meteorological elements combining remote sensing and ground observations, characterized in that, Including: Obtain remote sensing and ground observation data of a specified area, perform spatio-temporal registration and data standardization processing to obtain meteorological element data with a unified spatio-temporal benchmark; Divide the meteorological element data into different spatio-temporal grid cells, and conduct data quality assessment on the observation data within each spatio-temporal grid cell to obtain the data quality assessment index of the spatio-temporal grid cell; Based on the adaptive fusion algorithm with dynamic weights, calculate the fusion weight coefficient of the meteorological element data within the spatio-temporal grid cell according to the data quality assessment index of the spatio-temporal grid cell, and perform weighted fusion on the meteorological element data within the spatio-temporal grid cell to generate an initial three-dimensional meteorological field with physical consistency; Based on the initial three-dimensional meteorological field, use the spectral analysis method to perform scale decomposition and reconstruction on the meteorological element field to obtain the optimal three-dimensional meteorological field; Construct an adaptive tree-shaped composite analysis grid based on the optimal three-dimensional meteorological field, analyze the evolution characteristics of the weather system through a multi-layer progressive method, determine the topological evolution parameters, and output the three-dimensional stereoscopic analysis result of the meteorological elements.

2. The method according to claim 1, wherein Divide the meteorological element data into different spatio-temporal grid cells, and conduct data quality assessment on the observation data within each spatio-temporal grid cell. The data quality assessment index of the spatio-temporal grid cell includes: Divide the meteorological element data into multiple spatio-temporal grid cells according to a preset time interval and spatial resolution; Calculate the difference in the observed values between any two observation points within the spatio-temporal grid cell to obtain the numerical consistency deviation; calculate the change amplitude of the observed values of each observation point at two adjacent moments within the spatio-temporal grid cell to obtain the time series volatility deviation; calculate the difference degree between the spatial distribution density of the observation points within the spatio-temporal grid cell and the preset density threshold to obtain the spatial density distribution deviation; Linearly combine the numerical consistency deviation, the time series volatility deviation, and the spatial density distribution deviation after normalization processing to obtain the data quality assessment index of each spatio-temporal grid cell.

3. The method according to claim 1, wherein Based on the adaptive fusion algorithm with dynamic weights, calculate the fusion weight coefficient of the meteorological element data within the spatio-temporal grid cell according to the data quality assessment index of the spatio-temporal grid cell, and perform weighted fusion on the meteorological element data within the spatio-temporal grid cell to generate an initial three-dimensional meteorological field with physical consistency including: According to the data quality assessment index of the spatio-temporal grid cell, calculate the weight influence factor of the spatio-temporal grid cell. The weight influence factor adopts the form of a Gaussian function and is calculated with the data quality assessment index as the independent variable; Normalize the weight influence factor of the spatio-temporal grid cell to obtain the fusion weight coefficient of the meteorological element data within the spatio-temporal grid cell; perform weighted calculation on the meteorological element data within the spatio-temporal grid cell according to the fusion weight coefficient to obtain the observed value fitting term; construct a mass conservation constraint term based on the air density continuity equation, and at the same time construct a momentum conservation constraint term based on the atmospheric dynamics equation; Combine the observed value fitting term, the mass conservation constraint term, and the momentum conservation constraint term to form a fusion objective function; Use an optimization algorithm based on adaptive step size and orthogonalized quasi-Newton iteration to solve the fusion objective function to obtain the fusion meteorological field; Calculate the mass conservation residual and momentum conservation residual of the fused meteorological field. When the mass conservation residual is less than the first preset threshold and the momentum conservation residual is less than the second preset threshold, determine that the fused meteorological field is an initial three-dimensional meteorological field with physical consistency.

4. The method according to claim 3, wherein Solve the fusion objective function using an optimization algorithm based on adaptive step size and orthogonalized quasi-Newton iteration to obtain the fused meteorological field, including: Take the meteorological element data within the spatio-temporal grid cell as the initial solution, and calculate the gradient value of the fusion objective function corresponding to the initial solution based on the fusion objective function. According to the initial solution and the gradient value of the fusion objective function, construct an iterative format using the adaptive step size factor and the approximate inverse matrix of the quasi-Newton matrix. When the change rate of the fusion objective function values between two adjacent iterations is greater than zero, decrease the adaptive step size factor by a preset proportional coefficient; otherwise, increase it by the preset proportional coefficient. The value of the adaptive step size factor is restricted within the preset upper and lower limits. In each iteration process, calculate the difference between the solution vectors of two adjacent iterations and the corresponding gradient difference vector of the fusion objective function; orthogonalize the difference between the solution vectors and the gradient difference vector of the fusion objective function to obtain an orthogonalized difference vector. According to the preset iterative evaluation rounds, update the approximate inverse matrix of the quasi-Newton matrix in a two-way recursive manner based on the corresponding orthogonalized difference vector; use the updated approximate inverse matrix of the quasi-Newton matrix to calculate a new solution vector. Calculate the fusion objective function value corresponding to the new solution vector, and calculate the distance value between the solution vectors of two adjacent iterations. When the relative change amount of the fusion objective function value is less than the preset change threshold and the distance value between the solution vectors is less than the preset distance threshold, determine the new solution vector as the fused meteorological field; otherwise, use the new solution vector as the initial solution for the next iteration and continue the iteration.

5. The method according to claim 1, wherein Based on the initial three-dimensional meteorological field, perform scale decomposition and reconstruction on the meteorological element field using the spectral analysis method to obtain the optimal three-dimensional meteorological field, including: Perform Fourier transform on the initial three-dimensional meteorological element field to obtain spectral coefficients, perform wavelet transform on the spectral coefficients and the local window function to obtain decomposition coefficients of the meteorological element field at different scales and spatial positions, construct a multi-dimensional tensor based on the decomposition coefficients and perform nuclear norm calculation and tensor decomposition, and construct a wave number-energy spectrum matrix in combination with the method of extracting characteristic truncation. Obtain the dominant wave number modes through singular value decomposition of the wave number-energy spectrum matrix, and determine the characteristic wave number sequence according to the energy contribution rate of the dominant wave number modes. Construct an orthogonal filter bank based on the characteristic wave number sequence. The orthogonal filter bank includes multiple Gaussian filters. The center frequency of each Gaussian filter corresponds to a characteristic wave number in the characteristic wave number sequence, and the bandwidth parameter of the Gaussian filter is determined according to the interval between adjacent characteristic wave numbers. Apply the orthogonal filter bank to the spectral coefficients to obtain multiple scale components, and construct a spectral reconstruction tensor based on the local energy distribution of the multiple scale components. Calculate local reconstruction weights according to the spectrum reconstruction tensor, and perform weighted superposition on the multiple scale components to obtain the optimal meteorological element field.

6. The method according to claim 5, wherein Construct a multi-dimensional tensor based on the decomposition coefficients, perform nuclear norm calculation and tensor decomposition, and construct a wave number-energy spectrum matrix by combining the method of eigen-truncation extraction, including: Construct a multi-dimensional tensor with the decomposition coefficients at different scale levels, and calculate the nuclear norm of the multi-dimensional tensor to obtain a scale correlation matrix; Perform tensor decomposition on the scale correlation matrix to obtain a scale core tensor and a set of projection matrices, where the set of projection matrices includes a spatial position projection matrix, a wave number projection matrix, and an energy projection matrix; Calculate the singular value sequence of the scale core tensor, use 80% of the maximum value of the singular value sequence as the eigen-truncation threshold, and extract the projection matrix components corresponding to the singular values greater than the eigen-truncation threshold in the singular value sequence; Reconstruct the projection matrix components to obtain the wave number-energy spectrum matrix.

7. The method according to claim 1, characterized in that Construct an adaptive tree-like composite analysis grid based on the optimal three-dimensional meteorological field, analyze the evolution characteristics of the weather system through a multi-layer progressive method, determine the topological evolution parameters, and output the three-dimensional stereo analysis results of meteorological elements, including: Obtain the meteorological field data in the optimal three-dimensional meteorological field, determine the reference resolution and layer coefficient according to the spatial distribution characteristics and gradient distribution of the meteorological field data, calculate the first resolution, and determine the main grid; calculate the second resolution according to the local change characteristics of the meteorological field data, determine the branch grid, and combine the main grid and the branch grid to generate an adaptive tree-like composite analysis grid; Calculate the local gradient field on the adaptive tree-like composite analysis grid, determine the grid branch growth direction, construct a multi-layer progressive interpolation function, and establish a numerical correlation between the main grid and the branch grid; Construct a feature metric tensor based on the local gradient field and the numerical correlation, calculate the evolution parameters of the feature metric tensor, and construct a feature transfer function between grid nodes according to the evolution parameters; Calculate the grid topological features using the feature transfer function to obtain the topological stability index, the structure evolution rate, and the system deformation parameters, and construct a topological evolution feature operator; Construct a recursive operator according to the historical features, current features, and change features, and use the recursive operator to perform multi-layer progressive fusion on the topological evolution feature operator to output the three-dimensional stereo analysis results of meteorological elements.

Images