Meteorological data assimilation method fusing topographic features

Abstract

The invention discloses a meteorological data assimilation method fusing topographic features, and relates to the technical field of meteorological data assimilation, and the method comprises the steps: combining a meteorological variable-topographic factor correlation map with an atmospheric state numerical field, and obtaining a multivariable initial meteorological field model fusing topographic factors; identifying space-time dynamic deviation distribution characteristics of the multivariable initial meteorological field model by using real-time multi-source environment observation data, and generating a terrain disturbance response matrix; carrying out dynamic correction and space-time error reconstruction based on a terrain disturbance sensitive weight on the terrain disturbance response matrix and the multivariable initial meteorological field model through an assimilation optimization algorithm to obtain a meteorological assimilation result; according to the method, dynamic correction and space-time error reconstruction are carried out on the initial meteorological field through the terrain disturbance sensitive weight, directional perception and intelligent correction of observation deviation are realized, and the response capability and the error control level of the model in a dynamic change scene are effectively enhanced.

Description

TECHNICAL FIELD

[0001] The present application relates to the field of meteorological data assimilation, in particular to a meteorological data assimilation method fusing terrain features. BACKGROUND

[0002] In recent years, meteorological data assimilation technology has become the core means to improve the accuracy of numerical weather prediction. Traditional assimilation methods mainly include three-dimensional variation (3D-Var), four-dimensional variation (4D-Var) and integrated Kalman filter (EnKF), etc. Such methods can effectively fuse multi-source observation data and numerical prediction model output to construct a physically consistent atmospheric initial field. At the same time, with the development of high-resolution satellite remote sensing, ground automatic weather station network and high-altitude sounding technology, the temporal and spatial coverage and quality of observation data have been significantly improved, providing more abundant input for meteorological data assimilation.

[0003] The existing technology still has deficiencies in the coupling of terrain features and meteorological variables: most methods only fix the weighting of terrain factors in the preprocessing stage or before assimilation, and the weighting factor often remains unchanged in space and time, lacking dynamic response to seasonal, daily changes and sudden weather processes; simple spatial filtering usually based on a preset filtering radius or neighborhood weight, ignoring the sensitive effect of terrain undulation in the vertical direction, and it is difficult to take into account the differences between different scales and geomorphic units; empirical parameters are generally used in the assimilation process rather than quantitative models driven by data, which cannot accurately represent the nonlinear coupling relationship between terrain and meteorological variables, so there is still a large deviation in the prediction of complex terrain areas such as mountain valleys, hilly tablelands, etc. SUMMARY

[0004] In view of the above existing problems, the present application is proposed.

[0005] Therefore, the present application provides a meteorological data assimilation method fusing terrain features to solve the problem of low prediction accuracy in complex terrain areas caused by the inability to dynamically quantify the coupling effect of terrain and meteorological variables.

[0006] To solve the above technical problems, the present application provides the following technical solutions: In a first aspect, the present application provides a meteorological data assimilation method fusing terrain features, which comprises collecting multi-source environmental observation data for preprocessing, obtaining terrain feature model data, and performing multivariate regression analysis and mutual information calculation with the preprocessed multi-source environmental observation data to obtain a meteorological variable-terrain factor correlation atlas; combining the meteorological variable-terrain factor correlation atlas with the atmospheric state numerical field to obtain a multi-variable initial meteorological field model fusing terrain factors; using real-time multi-source environmental observation data to identify the spatiotemporal dynamic bias distribution characteristics of the multi-variable initial meteorological field model, and generating a terrain disturbance response matrix; The terrain disturbance response matrix and the multivariate initial meteorological field model are dynamically corrected and the space-time error is reconstructed based on the terrain disturbance sensitive weight through the assimilation optimization algorithm, and a meteorological assimilation result is obtained. The meteorological assimilation result is subjected to multi-scale adaptive error decomposition, and fine error correction is performed in combination with the pretreated multi-source environmental observation data, and a high-precision assimilated meteorological field data set is obtained.

[0007] As a preferred scheme of the meteorological data assimilation method fusing terrain features, wherein: the multi-source environmental observation data includes ground meteorological observation elements, high-altitude meteorological detection elements and landform features; The preprocessing includes data cleaning, time series alignment and data standardization conversion.

[0008] As a preferred scheme of the meteorological data assimilation method fusing terrain features, wherein: the terrain feature model data is obtained, and the specific steps are as follows, The pretreated landform features are subjected to multi-scale terrain factor decomposition to obtain a multi-scale terrain factor set; The multi-scale terrain factor set is subjected to high-frequency detail amplification and low-frequency smooth fusion by applying a feature enhancement method based on wavelet transform to generate an enhanced terrain factor set; The enhanced terrain factor set is mapped to the geographical coordinate grid of the meteorological monitoring area for spatial interpolation to obtain terrain feature model data.

[0009] As a preferred scheme of the meteorological data assimilation method fusing terrain features, wherein: the meteorological variable-terrain factor correlation graph is obtained, and the specific steps are as follows, The pretreated ground meteorological observation elements, the pretreated high-altitude meteorological detection elements and the enhanced terrain factor set are subjected to multiple regression analysis to obtain a regression coefficient matrix; The regression coefficient matrix, the terrain feature model data, the pretreated ground meteorological observation elements and the pretreated high-altitude meteorological detection elements are subjected to mutual information calculation to obtain a mutual information matrix; The regression coefficient matrix and the mutual information matrix are subjected to joint analysis and weighted fusion to obtain a meteorological variable-terrain factor correlation graph.

[0010] As a preferred scheme of the meteorological data assimilation method fusing terrain features, wherein: the meteorological variable-terrain factor correlation graph is combined with the atmospheric state numerical field to obtain a multivariate initial meteorological field model fusing terrain factors, and the specific steps are as follows, The pretreated ground meteorological observation elements and high-altitude meteorological detection elements are subjected to joint modeling to obtain an atmospheric state numerical field; The terrain factor weight in the meteorological variable-terrain factor correlation atlas is mapped to the geographical coordinate grid of the meteorological monitoring area of the atmospheric state numerical field to obtain a terrain influence weight grid; The weighted influence coefficient of the terrain factor on the meteorological variable in the terrain influence weight grid is calculated, and the terrain factor is weighted and adjusted to the atmospheric state numerical field to form a terrain influence response field; The terrain influence response field and the atmospheric state numerical field are weighted and fused to construct a multivariate initial meteorological field model fused with terrain factors.

[0011] As a preferred scheme of the meteorological data assimilation method fused with terrain characteristics, wherein: the real-time multi-source environmental observation data is used to identify the spatio-temporal dynamic deviation distribution characteristics of the multivariate initial meteorological field model, and a terrain disturbance response matrix is generated, and the specific steps are as follows, The preprocessed real-time multi-source environmental observation data is subjected to spatio-temporal analysis to identify the dynamic change characteristics of the meteorological variable in different time and space dimensions, and a meteorological variable spatio-temporal deviation data set is obtained; The meteorological variable spatio-temporal deviation data set is compared with the multivariate initial meteorological field model to identify the spatio-temporal dynamic deviation distribution characteristics; The spatio-temporal dynamic deviation distribution characteristics are subjected to feature transformation and visualization processing to obtain a meteorological variable spatio-temporal dynamic deviation distribution map; The meteorological variable spatio-temporal dynamic deviation distribution map and the terrain influence response field are subjected to quantitative calculation of terrain disturbance to generate a terrain disturbance response matrix.

[0012] As a preferred scheme of the meteorological data assimilation method fused with terrain characteristics, wherein: the terrain disturbance response matrix and the multivariate initial meteorological field model are subjected to dynamic correction and spatio-temporal error reconstruction based on terrain disturbance sensitive weights by a assimilation optimization algorithm to obtain a meteorological assimilation result, and the specific steps are as follows, The terrain disturbance response matrix and the multivariate initial meteorological field model are subjected to dynamic correction by the assimilation optimization algorithm to obtain an optimized terrain influence response field; The optimized terrain influence response field and the multivariate initial meteorological field model are subjected to spatio-temporal error reconstruction to obtain a meteorological assimilation result.

[0013] As a preferred scheme of the meteorological data assimilation method fused with terrain characteristics, wherein: the meteorological assimilation result is subjected to multi-scale adaptive error decomposition, and the preprocessed multi-source environmental observation data is combined for fine error correction to obtain a high-precision assimilated meteorological field data set, and the specific steps are as follows, The meteorological assimilation result is subjected to multi-scale adaptive error decomposition to identify the error distribution characteristics at different scales to obtain an error decomposition result; The error distribution is finely corrected and compensated by using the preprocessed multi-source environmental observation data, to obtain a fine correction error distribution. The fine correction error distribution is used to perform weighted correction on the meteorological assimilation result, to obtain a high-precision assimilated meteorological field data set.

[0014] In a second aspect, the present application provides a computer device comprising a memory and a processor, wherein the memory stores a computer program, and wherein the computer program, when executed by the processor, implements any step of the method for meteorological data assimilation fusing terrain features according to the first aspect of the present application.

[0015] In a third aspect, the present application provides a computer-readable storage medium having a computer program stored thereon, wherein the computer program, when executed by a processor, implements any step of the method for meteorological data assimilation fusing terrain features according to the first aspect of the present application.

[0016] The present application has the following beneficial effects: by constructing a meteorological variable-terrain factor correlation graph, the multi-scale response relationship between terrain features and meteorological variables is quantitatively modeled, so that the initial meteorological field can fully reflect the influence of terrain on meteorological evolution, thereby improving the spatial adaptability and initial value accuracy of the model in complex terrain regions; further, the initial meteorological field is dynamically corrected and spatio-temporal error is reconstructed based on the terrain disturbance sensitive weight, directional perception and intelligent correction of observation bias are realized, the response capability and error control level of the model in dynamic change scenarios are effectively enhanced, and finally a meteorological assimilation result with higher precision and stronger terrain consistency is formed. BRIEF DESCRIPTION OF DRAWINGS

[0017] In order to more clearly illustrate the technical solutions of the embodiments of the present application, the drawings needed in the embodiment description will be briefly introduced. Obviously, the drawings in the following description are only some embodiments of the present application, and other drawings can be obtained by those skilled in the art without creative labor.

[0018] Fig. 1 The flowchart of the method for meteorological data assimilation fusing terrain features.

[0019] Fig. 2 The flowchart for obtaining terrain feature model data.

[0020] Fig. 3 The flowchart for outputting the meteorological variable-terrain factor correlation graph.

[0021] Fig. 4 The flowchart for outputting the high-precision assimilated meteorological field data set. DETAILED DESCRIPTION

[0022] In order to make the above objectives, features and advantages of the present application more obvious and easy to understand, the specific embodiments of the present application are described in detail below with reference to the accompanying drawings.

[0023] In the following description, a lot of specific details are set forth in order to give a thorough understanding of the present application, but the present application can also be implemented in other different ways from those described herein, and those skilled in the art can make similar generalizations without departing from the concept of the present application, so the present application is not limited to the specific embodiments disclosed below.

[0024] Secondly, "one embodiment" or "embodiment" referred to herein means that a specific feature, structure or characteristic can be included in at least one implementation of the present application. "In one embodiment" appearing in different places in the specification does not mean the same embodiment, nor is it an embodiment that is independent of or mutually exclusive of other embodiments.

[0025] Reference Figs. 1-4 For one embodiment of the present application, the embodiment provides a weather data assimilation method fusing terrain features, comprising the following steps: S1, collecting multi-source environmental observation data for preprocessing.

[0026] S1.1, the multi-source environmental observation data includes ground meteorological observation elements, upper air meteorological detection elements and landform features.

[0027] Specifically, ground meteorological observation elements are collected, temperature, humidity, wind speed, wind direction, air pressure and precipitation are obtained using ground automatic weather stations, monitoring data is recorded by continuous sampling, sampling interval is set to 5 minutes as an example, sampling period covers 24 hours as an example, to ensure that the diurnal meteorological variation characteristics are covered; High-altitude meteorological detection elements are obtained, temperature, humidity, wind speed and air pressure at 700 hPa to 100 hPa height layer are obtained using high-altitude balloon sounding device, sampling data is transmitted in real time through radio equipment, the ascending rate of the sounding balloon is set to 5 meters per second as an example, and the transmission sampling interval is set to 2 seconds as an example; Select the public DEM data set and high-resolution remote sensing image corresponding to the meteorological monitoring area, respectively obtain the elevation, slope, slope direction and curvature to obtain the landform attributes, all landform attributes are divided into geographical coordinate grids according to 30 meters resolution as an example, to obtain the landform features.

[0028] S1.2, preprocessing includes data cleaning, time series alignment and data standardization conversion.

[0029] Specifically, the ground meteorological observation elements, upper-air meteorological detection elements and landform features are preprocessed to remove records with missing values, outliers or invalid labels. The outlier identification method is based on the degree of deviation between the mean and median within the sliding window. For example, the deviation range is set to ±3 times the standard deviation, and the deletion is determined by comparison. The time series of ground meteorological observation elements and upper-air meteorological detection elements are aligned according to an exemplary 5-minute time interval, and missing time periods are filled by interpolation under similar meteorological conditions or by interpolation of the mean of effective observations before and after. All observations are converted to standard units, such as temperature to °C, air pressure to hPa, wind speed to m / s, and humidity to percentage. For ground meteorological observation elements, upper-air meteorological detection elements, and landform features, the maximum and minimum value normalization method is used to standardize the data. For example, the upper and lower limits of standardization are set to 0 and 1, respectively, to complete the preprocessing of ground meteorological observation elements, upper-air meteorological detection elements, and landform features.

[0030] S2. Obtain terrain feature model data.

[0031] S2.1. Perform multi-scale topographic factor decomposition on the preprocessed geomorphic features to obtain a multi-scale topographic factor set.

[0032] Specifically, the preprocessed landform features are decomposed using discrete wavelet transform. An exemplary two-dimensional Daubechies wavelet basis function is selected to decompose the landform features into low-frequency and high-frequency components of different scales layer by layer according to the spatial grid with a resolution of 30 meters in the example. The number of decomposition layers is set to 3 for example. High-frequency terrain factors in the detail layer and low-frequency terrain factors in the approximate layer are extracted layer by layer. The frequency components obtained from each decomposition layer are saved as terrain factors of the corresponding scale to form a multi-scale terrain factor set. During the decomposition process, wavelet coefficients are subjected to wavelet coefficient noise suppression threshold processing. The wavelet coefficient noise suppression threshold is set to 0.1 times the mean absolute value of the wavelet coefficients. Wavelet coefficients below the wavelet coefficient noise suppression threshold are set to zero to suppress noise, and a multi-scale terrain factor set is output. It should be noted that the mean of all wavelet coefficients is obtained, and the mean is determined by an exemplary ratio of 0.1.

[0033] S2.2. For the multi-scale terrain factor set, apply a wavelet transform-based feature enhancement method to amplify high-frequency details and smoothly fuse low-frequency features to generate an enhanced terrain factor set.

[0034] Specifically, for a multi-scale set of terrain factors, the two-dimensional Daubechies wavelet basis function in discrete wavelet transform is used to process the high-frequency and low-frequency terrain factors at each scale. For the high-frequency terrain factors, an amplitude enhancement is performed using an exemplary amplification factor of 1.5. The enhancement process includes point-by-point multiplication with the amplification factor to amplify the detailed features of the terrain. For low-frequency topographic factors, a smoothing filtering algorithm is applied, using a weighted average filter. The filtering weights are, for example, the topographic factor values ​​within the geographic coordinate grid of the surrounding 3×3 meteorological monitoring area, to reduce noise and fluctuations in the low-frequency components. The processed high-frequency and low-frequency terrain factors are fused according to an exemplary ratio. The weight of the high-frequency terrain factor is set to 0.6 and the weight of the low-frequency terrain factor is set to 0.4 to ensure a balance between detail magnification and overall smoothness. The fusion result is used as an enhanced terrain factor set.

[0035] S2.3. Map the enhanced terrain factor set to the geographic coordinate grid of the meteorological monitoring area for spatial interpolation to obtain terrain feature model data.

[0036] Specifically, each topographic factor value in the enhanced topographic factor set is mapped to its corresponding spatial location within the geographic coordinate grid of the meteorological monitoring area. If the spatial distribution points in the enhanced topographic factor set do not completely overlap with the geographic coordinate grid nodes, a spatial interpolation method is used to fill the gaps. The spatial interpolation uses the exemplary inverse distance weighted interpolation method, with an exemplary search radius of 150 meters. Within each geographic coordinate grid node of the meteorological monitoring area to be interpolated, several known enhanced topographic factor value points closest to the node are collected. For example, the eight nearest enhanced topographic factor value points are selected. Based on the distance between the enhanced topographic factor value points and the geographic coordinate grid node of the meteorological monitoring area to be interpolated, the inverse distance weight is calculated. The closer the enhanced terrain factor value points are to the geographic coordinate grid nodes of the meteorological monitoring area to be interpolated, the greater the inverse distance weight. A weighted average is assigned to the geographic coordinate grid nodes of the meteorological monitoring area. During interpolation, enhanced terrain factor value points whose distance exceeds the search radius are not included in the inverse distance weight calculation to ensure local spatial influence. After interpolation, all geographic coordinate grid nodes of the meteorological monitoring area form a continuous spatial distribution of enhanced terrain factors covering the meteorological monitoring area, and the output is terrain feature model data. It should be noted that the expression for calculating the inverse distance weight, based on the distance between the enhanced terrain factor value points and the geographic coordinate grid nodes of the meteorological monitoring area to be interpolated, is as follows: ; in, It is the first Inverse distance weights for each enhanced terrain factor value point. It is an index variable that enhances the topographic factor value points. is the distance from the i-th enhanced terrain factor value point to the geographical coordinate grid node of the meteorological monitoring area to be interpolated is the power index parameter, and an exemplary value of 2 is taken is the total number of geographical coordinate grid nodes of the meteorological monitoring area to be interpolated is an index variable of the geographical coordinate grid nodes of the meteorological monitoring area to be interpolated.

[0037] S3, performing multiple regression analysis and mutual information calculation on the preprocessed multi-source environmental observation data to obtain a meteorological variable-terrain factor correlation atlas.

[0038] S3.1, performing multiple regression analysis on the preprocessed ground meteorological observation elements, preprocessed upper air meteorological detection elements and enhanced terrain factor set to obtain a regression coefficient matrix.

[0039] Specifically, the preprocessed ground meteorological observation elements, upper air meteorological detection elements and enhanced terrain factor set are correspondingly summarized into a statistical sample table according to a unified time stamp and a spatial grid, the independent variable column contains each factor value in the ground meteorological observation elements, upper air meteorological detection elements and enhanced terrain factor set, and the dependent variable column is the target meteorological variable observation value. The least squares method is used to solve the equation, and the ratio of the sample number to the number of independent variables is exemplarily set to be not less than 10 times, and the regression coefficient is calculated by matrix operation; multiple collinearity diagnosis is performed on the regression coefficient, and the variance inflation factor diagnosis method is exemplarily selected, and the independent variables exceeding the example of 10 are removed or combined according to the variance contribution rate; all the diagnosed and regression coefficients are arranged in order of independent variables, and the regression coefficient matrix is output. It should be noted that the expression for calculating the regression coefficient by matrix operation is: Wherein, is the regression coefficient, is a symmetric matrix, is the independent variable characteristic matrix, is a transpose operator, is the dependent variable column vector.

[0040] S3.2, performing mutual information calculation on the regression coefficient matrix, terrain feature model data, preprocessed ground meteorological observation elements and preprocessed upper air meteorological detection elements to obtain a mutual information matrix.

[0041] ​​​​Specifically, taking each meteorological monitoring area's geographic coordinate grid node in the terrain feature model data as a unit, the regression coefficient matrix corresponding to the independent variable factors, the preprocessed ground meteorological observation element values, the preprocessed upper-air meteorological detection element values, and the values ​​of the corresponding enhanced terrain factor set in the terrain feature model data at the same location are extracted to form a joint sample set; For each pair of variables in the joint sample set, the continuous variables are discretized using the equal-interval binning method. For example, the number of bins is set to 10. Based on the joint frequency and marginal frequency of each bin combination obtained after binning, the joint probability distribution and marginal probability distribution are estimated. Substitute the joint probability distribution and the marginal probability distribution into the definition formula of mutual information to calculate the mutual information value. A mutual information value greater than zero indicates the existence of an information correlation. The mutual information values ​​between the independent variable factors corresponding to the regression coefficient matrix, the preprocessed ground meteorological observation element values, the preprocessed upper-air meteorological detection element values, and the values ​​of the enhanced terrain factor set corresponding to the values ​​in the terrain feature model data are recorded in the form of a two-dimensional matrix. The row and column names of the matrix correspond to the names of the independent variable factors in the regression coefficient matrix, the preprocessed ground meteorological observation elements, the preprocessed upper-air meteorological detection elements, and the enhanced terrain factor set in the terrain feature model data, respectively. The mutual information matrix is ​​then output. It should be noted that, by substituting the joint probability distribution and marginal probability distribution into the definition formula of mutual information, the expression for calculating the mutual information value is as follows: ; in, It is a variable With variables Mutual information value between them It is a variable A specific value included in the range of values. , It is a variable A specific value included in the range of values. , It is a variable Values and variables Values The joint probability, It is a variable Values The marginal probability, It is a variable Values The marginal probability.

[0042] S3.3. Perform joint analysis and weighted fusion of the regression coefficient matrix and mutual information matrix to obtain the correlation map of meteorological variables and topographic factors.

[0043] Specifically, the regression coefficient matrix and mutual information matrix are arranged according to the order of independent variable factors. The absolute value elements of the regression coefficients of the corresponding independent variable factors in the regression coefficient matrix and the mutual information value elements of the corresponding independent variable factors in the mutual information matrix are extracted to form a regression coefficient weight vector and a mutual information correlation strength vector. The regression coefficient weight vector is, for example, normalized using the absolute value of the regression coefficients, and the mutual information correlation strength vector is, normalized using the mutual information value. For each pair of independent variable factors, the corresponding weights and correlation strengths are weighted and fused. The fusion weights are, for example, set to 0.5 for the regression coefficient weight vector and 0.5 for the mutual information correlation strength vector. The fusion result is calculated as the joint correlation strength. Based on the joint correlation strength, independent variable factor pairs with a value higher than 0.3 are selected to form preliminary correlation edges. The node names of the preliminary correlation edges are labeled with the corresponding independent variable factor names, and the correlation edge weights are labeled with the joint correlation strength values. A meteorological variable-topographic factor correlation map is constructed based on all selected correlation edges and nodes. It should be noted that the expression for calculating the fusion result as the joint correlation strength is as follows: ; in, It is the joint correlation strength. It represents the proportional weight of the regression coefficient weight vector. It is the normalized regression coefficient weight vector. It is the normalized mutual information correlation strength vector.

[0044] S4. Combine the meteorological variable-topographic factor correlation map with the atmospheric state numerical field to obtain a multivariate initial meteorological field model that incorporates topographic factors.

[0045] S4.1 Perform joint modeling on the preprocessed ground meteorological observation elements and upper-air meteorological detection elements to obtain the atmospheric state numerical field.

[0046] Specifically, the preprocessed ground meteorological observation elements and the preprocessed upper-air meteorological detection elements are matched according to a unified timestamp and the correspondence between the geographic coordinate grid of the meteorological monitoring area to form a joint observation sample set; the discrete meteorological observation values ​​in the joint observation sample set refer to the ground meteorological observation element values ​​and upper-air meteorological detection element values ​​retained after matching. Based on historical meteorological data and numerical model forecasts, an initial atmospheric state field that meets the resolution requirements of the meteorological monitoring area is generated through spatial interpolation and variable transformation. A three-dimensional interpolation method is used on the joint observation sample set to interpolate discrete meteorological observations to a uniform geographic coordinate grid of the meteorological monitoring area, thereby generating a continuous spatial distribution of meteorological physical quantities. Using the variational assimilation technique in numerical weather prediction, the continuous spatial distribution of meteorological physical quantities is data assimilated with the initial atmospheric state field to adjust the background field to approach the observation sample and iteratively optimize the numerical field state. The assimilation process is set to 50 iterations, and the atmospheric state numerical field after assimilation is output.

[0047] S4.2, the terrain factor weight in the meteorological variable-terrain factor correlation atlas is mapped to the meteorological monitoring area geographic coordinate grid of the atmospheric state numerical field to obtain a terrain influence weight grid.

[0048] Specifically, the terrain factor weight in the meteorological variable-terrain factor correlation atlas is point-by-point corresponding according to the spatial position of the meteorological monitoring area geographic coordinate grid, and each meteorological monitoring area geographic coordinate grid node obtains the terrain factor weight value at the corresponding position; For the case that the terrain factor weight and the meteorological monitoring area geographic coordinate grid node space of the atmospheric state numerical field do not completely coincide, a spatial interpolation method is used to fill the missing weight position, an inverse distance weighted interpolation method is used, an example search radius of 150 meters is set, a number of adjacent weight known nodes are collected for each to-be-mapped node, the distance weight is obtained, and a weighted average calculation is performed to obtain the mapping weight value; In the interpolation process, the weight points beyond the search radius range are excluded to ensure local spatial correlation. After completing the weight mapping of all meteorological monitoring area geographic coordinate grid nodes, a terrain influence weight grid covering the entire region is output.

[0049] S4.3, calculate the weighted influence coefficient of the terrain factor on the meteorological variable in the terrain influence weight grid, and perform terrain factor weighted adjustment on the atmospheric state numerical field to form a terrain influence response field.

[0050] Specifically, according to each terrain factor weight in the terrain influence weight grid, the corresponding meteorological monitoring area geographic coordinate grid node position is selected, and the corresponding meteorological variable value in the atmospheric state numerical field at the same position is selected; For all corresponding positions, the value of the meteorological variable in the atmospheric state numerical field is multiplied point-by-point with the terrain factor weight to obtain the weighted influence coefficient; For the case that there is a missing weight or value in the spatial position, the weighted average value of the adjacent effective nodes is used for filling to ensure that each meteorological monitoring area geographic coordinate grid node has complete weighted influence coefficient; The weighted influence coefficient is uniformly processed, the scale is adjusted to match the range and distribution characteristics of the original atmospheric state numerical field, and numerical abnormalities or distortion are avoided; after completing the weighted adjustment of all meteorological monitoring area geographic coordinate grid nodes, a terrain influence response field covering the entire meteorological monitoring area is formed.

[0051] S4.4, the terrain influence response field and the atmospheric state numerical field are weighted and fused to construct a multi-variable initial meteorological field model with terrain factors.

[0052] Specifically, the terrain influence response field and the atmospheric state numerical field are matched point by point according to the geographical coordinate grid node correspondence relationship of the meteorological monitoring area; for each geographical coordinate grid node of the meteorological monitoring area, the corresponding meteorological variable values in the terrain influence response field and the atmospheric state numerical field are extracted; the weight coefficients of the corresponding meteorological variable values in the terrain influence response field and the atmospheric state numerical field are exemplarily set; the weight coefficients of the terrain influence response field and the atmospheric state numerical field are exemplarily set as 0.6 and 0.4 respectively; the weight coefficients are used for the terrain influence response field and the atmospheric state numerical field respectively; the product of the meteorological variable values at the corresponding positions and the respective weight coefficients is calculated, and then the weighted sum is obtained to obtain the fused meteorological variable value; the weighted sum operation is repeated for all geographical coordinate grid nodes of the meteorological monitoring area to form a multi-variable initial meteorological field model with terrain factors.

[0053] S5, using real-time multi-source environmental observation data, identifying the spatio-temporal dynamic deviation distribution characteristics of the multi-variable initial meteorological field model, and generating a terrain disturbance response matrix.

[0054] S5.1, spatio-temporal analysis is performed on the preprocessed real-time multi-source environmental observation data to identify the dynamic change characteristics of the meteorological variables in different time and space dimensions, and a meteorological variable spatio-temporal deviation data set is obtained.

[0055] Specifically, the observation values of the same meteorological variable are summarized in time sequence order based on the preprocessed real-time multi-source environmental observation data, the difference change in the time dimension is calculated based on the observation values at each time point, and the time window length is exemplarily set as 24 hours; the time difference is calculated in a sliding window manner; in the spatial dimension, the meteorological variable values of adjacent regions are spatially interpolated according to the geographical coordinates of the observation points, and the Kriging interpolation method is exemplarily used to generate a continuous spatial distribution. The meteorological variable deviation at each time and space position is calculated to obtain a set of meteorological variable spatio-temporal deviation values, which are the differences between the observation values and the mean values in the adjacent time points or spatial neighborhoods, and the deviation change trend in the time and space dimensions is calculated; the spatio-temporal points in the meteorological variable spatio-temporal deviation value set that exceed the meteorological anomaly discrimination threshold are screened, and whether the current meteorological variable spatio-temporal deviation is abnormal is determined by comparing the current meteorological variable spatio-temporal deviation with the meteorological anomaly discrimination threshold; all the calculated meteorological variable spatio-temporal deviation values and the corresponding time and space coordinates are integrated to form a meteorological variable spatio-temporal deviation data set. It should be noted that the setting process of the meteorological anomaly discrimination threshold is as follows: the deviation distribution of the historical similar meteorological field data is calculated, and the standard deviation multiple is used to set the threshold; the mean value of the meteorological variable spatio-temporal deviation value is exemplarily set as ±3 times the standard deviation.

[0056] S5.2, compare the meteorological variable spatiotemporal bias dataset with the multivariate initial meteorological field model to identify the spatiotemporal dynamic bias distribution characteristics.

[0057] Specifically, the meteorological variable spatiotemporal bias value corresponding to each group of time and space in the meteorological variable spatiotemporal bias dataset is matched point by point, the meteorological variable values of the same time and space coordinates in the multivariate initial meteorological field model are selected, the numerical comparison of the corresponding variables is performed, the relative position change of the meteorological variable spatiotemporal bias value in the multivariate initial meteorological field model is calculated, the bias direction and size at all spatial positions are counted, the abnormal trend of the corresponding variable at the corresponding position is judged through the sign and absolute value of the meteorological variable spatiotemporal bias value, for example, a positive bias greater than an exemplary set meteorological anomaly discrimination 3.2 indicates that the variable is abnormally increased, and a negative bias with an absolute value greater than 3.2 indicates that the variable is abnormally decreased, all bias points and spatial positions that meet the conditions are associated, and distribution data with spatial coordinates and meteorological variable bias directions are formed by summarizing, thereby completing the matching of the meteorological variable spatiotemporal bias dataset and the dynamic bias distribution characteristic identification based on the multivariate initial meteorological field model, and obtaining the spatiotemporal dynamic bias distribution characteristics.

[0058] S5.3, feature transformation and visualization processing are performed on the spatiotemporal dynamic bias distribution characteristics to obtain a meteorological variable spatiotemporal dynamic bias distribution map.

[0059] Specifically, a multidimensional data matrix is constructed according to the meteorological variable bias direction and size corresponding to each meteorological monitoring area geographic coordinate grid node in the spatiotemporal dynamic bias distribution characteristics, the spatial coordinates are used as two-dimensional coordinate axes, the time is used as the third dimension, the meteorological variable bias size and direction are used as numerical attributes, and color gradients and vector icons are used to represent the bias size and direction, respectively. The multidimensional data matrix is filtered to smooth noise and outliers by using a Gaussian smoothing algorithm, thereby ensuring the continuity and readability of the bias distribution; based on the filtered multidimensional data matrix, the missing areas are completed by using a spatial interpolation method, and the spatial coverage is ensured to be complete by using an inverse distance weighting method. The numerical attributes are mapped to visual elements by using a graphical rendering tool, specifically, the meteorological variable spatiotemporal bias value, change trend, or stability level of the meteorological variable at each meteorological monitoring area geographic coordinate grid node are respectively mapped to color depth, arrow length, point size, or isogram density, thereby forming a visual layer that directly expresses the spatiotemporal dynamic bias distribution, generating an animation effect in combination with the time sequence, and displaying the dynamic change of the meteorological variable bias at different times and spaces; the color scale range is adjusted, and the bias absolute value is exemplarily set to be mapped to dark red or dark blue at the maximum, and zero bias corresponds to a neutral color, thereby facilitating the differentiation of abnormal intensity; and a meteorological variable spatiotemporal dynamic bias distribution map is output.

[0060] S5.4 Quantify the topographic disturbance calculation on the spatiotemporal dynamic deviation distribution map of meteorological variables and the topographic impact response field, and generate the topographic disturbance response matrix.

[0061] Specifically, using the geographic coordinate grid nodes of the meteorological monitoring area as spatial analysis units, the spatiotemporal deviation values ​​of meteorological variables at different times are extracted from the spatiotemporal dynamic deviation distribution map of meteorological variables at each spatial location. Simultaneously, the topographic factors such as topographic relief, slope, and relative elevation difference at the corresponding coordinates in the topographic impact response field are acquired. A joint variable array containing the spatiotemporal deviation values ​​of meteorological variables and topographic factors is constructed. Corresponding values ​​are aligned according to the same coordinates. The response magnitude of the spatiotemporal deviation value of meteorological variables at each spatial location as a function of topographic factors is calculated. The change in the spatiotemporal deviation of meteorological variables at each spatial location is selected as the dependent variable, and the values ​​of topographic factors such as topographic relief, slope, and relative elevation difference are used as the dependent variable. As independent variables, a multiple linear regression equation is established, and the regression coefficients are estimated using the least squares method. The goodness of fit of the multiple linear regression equation is evaluated, and regression coefficients with significant statistical correlation are selected as meteorological response coefficients. The meteorological response coefficients corresponding to each topographic factor are obtained. For example, spatial units with response coefficients greater than 0.65 are marked as positive response areas, spatial units with response coefficients less than -0.65 are marked as negative response areas, and spatial units in the range of -0.65 to 0.65 are marked as weak response areas. The meteorological variable response coefficients are combined with spatial locations through spatial superposition, and a topographic disturbance response matrix containing the response intensity levels of all spatial units is generated. It should be noted that the calculation of the spatiotemporal deviation of meteorological variables at each spatial location as a function of topographic factors is performed is as follows: ; in ， It is the first The spatiotemporal deviation of meteorological variables at a spatial location affects the first The response magnitude of each terrain factor It is the first One terrain factor, It is in time The The disturbance or change of a meteorological state variable It is the first Small changes in the disturbance of each meteorological state variable It is the partial derivative sign. It refers to the change or disturbance of meteorological variables. It is a meteorological state variable. It is an index variable of meteorological state variables. It is a time variable. It is a terrain factor variable. It is the index variable of the terrain factor variable.

[0062] S6.1, dynamically correcting the terrain disturbance response matrix and the multi-variable initial meteorological field model by the assimilation optimization algorithm to obtain an optimized terrain influence response field.

[0063] S6.1, dynamically correcting the terrain disturbance response matrix and the multi-variable initial meteorological field model by the assimilation optimization algorithm to obtain an optimized terrain influence response field.

[0064] Specifically, the meteorological variable values of the corresponding geographical coordinate grid nodes in the terrain disturbance response matrix and the multi-variable initial meteorological field model are matched point by point, and the response intensity level of each spatial unit in the terrain disturbance response matrix is associated with the meteorological variable sensitive weight of the corresponding node in the multi-variable initial meteorological field model. The assimilation optimization algorithm is used to dynamically adjust the meteorological variable sensitive weight according to the response intensity of the terrain disturbance response matrix. For example, when the response intensity is positive, the meteorological variable sensitive weight is increased by 10% to 20%, when the response intensity is negative, the meteorological variable sensitive weight is decreased by 10% to 20%, and when the response intensity is weak, the meteorological variable sensitive weight remains unchanged. The adjusted meteorological variable sensitive weight is used to calculate the meteorological variable values in the multi-variable initial meteorological field model, and the weighted results are iteratively updated until the change amount of the sensitive weight and the meteorological variable value is less than 0.01, the dynamic correction of the meteorological variable sensitive weight based on the terrain factor is completed, and the optimized terrain influence response field is output.

[0065] S6.2, reconstructing the space-time error of the optimized terrain influence response field and the multi-variable initial meteorological field model to obtain the meteorological assimilation result.

[0066] Specifically, the meteorological variable values of the corresponding geographical coordinate grid nodes in the terrain disturbance response matrix and the multi-variable initial meteorological field model are matched point by point, and the response intensity level of each spatial unit in the terrain disturbance response matrix is associated with the meteorological variable sensitive weight of the corresponding node in the multi-variable initial meteorological field model. It should be noted that the process of completing the error space distribution by using the Kriging interpolation method includes: obtaining the correlation between the spatial positions of the meteorological monitoring region geographic coordinate grid nodes and the corresponding error values, determining the spatial correlation parameters by fitting the empirical semi-variation function, estimating the error values of the unobserved points by weighting the known error values around the unobserved points using the determined spatial correlation parameters and the observed error values, and realizing the continuous completion of the error in the entire spatial range; The process of processing the error time variation by using the time series smoothing algorithm includes: applying the moving average method or the exponential smoothing method to the error time series of each meteorological monitoring region geographic coordinate grid node, setting an exemplary smoothing window length of 12 hours, calculating the smoothed error values by weighting, removing short-term fluctuations and noise in the error time series, and ensuring the continuity and stability of the error variation trend.

[0067] S7, multi-scale adaptive error decomposition is performed on the meteorological assimilation result, fine error correction is performed in combination with the preprocessed multi-source environmental observation data, and a high-precision assimilated meteorological field data set is obtained.

[0068] S7.1, multi-scale adaptive error decomposition is performed on the meteorological assimilation result, and error distribution characteristics on different scales are identified to obtain an error decomposition result.

[0069] Specifically, multi-scale adaptive error decomposition is performed on the meteorological assimilation result, a sliding window hierarchical filtering method is used, the meteorological monitoring region geographic coordinate grid node is taken as a basic unit, a plurality of exemplary scale windows are set, the spatial scale window can be set to an exemplary 3x3, 5x5, 7x7 grid, the time scale window can be set to an exemplary 3-time sequence, 6-time sequence, 12-time sequence unit, the local average error value under different scales is extracted layer by layer, and the difference between the meteorological variable value and the local average value in the corresponding scale is calculated as a local disturbance error. The local disturbance error under different scale windows is extracted by using a sliding window method to obtain an error sample set in each scale window, the statistical characteristics of the disturbance error with the change of the scale are analyzed by calculating the variance of the error value in each scale window, the variance of the error value in each scale window is obtained, and the change trend of the variance under different scales is compared, the change law of the variance with the spatial scale and the time scale is counted, the change of the error fluctuation intensity with the increase or decrease of the scale is judged, the concentration degree and the distribution mode of the disturbance error on different scales are revealed, thereby revealing the distribution law and the dominant scale of the disturbance error in the spatial or time dimension, and the concentration degree of the error energy distribution on each scale is identified. The disturbance error field at each scale is normalized to generate a multi-scale disturbance error layer; the disturbance error contributions at different scales are superimposed at each spatial position to calculate the error proportion of each scale; based on the exemplary setting that the error proportion of the scale layer greater than 0.6 is the dominant error scale, the disturbance error at the corresponding scale is marked as the main control error; the main control error layer identified at each spatial position is classified and counted to extract the error distribution pattern of different spatial regions at the dominant scale, and the error decomposition result containing the main control scale, error intensity and spatial distribution relationship is output; It should be noted that the expression for calculating the error proportion of each scale is: ; Wherein, is the error proportion of scale number , is an index variable of scale number, is the disturbance error value at a spatial position under scale number , is the total number of scale numbers.

[0070] S7.2, using the preprocessed multi-source environmental observation data, the error decomposition result is refined and error compensation is carried out, and the refined error distribution is obtained.

[0071] Specifically, the error values of each main control scale layer in the error decomposition result and the preprocessed multi-source environmental observation data are compared point by point at the corresponding spatial position, the environmental variable observation values in the multi-source environmental observation data at the same scale are extracted, the error comparison relationship table is established according to the same spatial coordinates, and the error value size and the observation difference direction at the corresponding scale are set. The exemplary adjustment coefficient range is-0.4 to 0.4, the main control scale error value in the error decomposition result is guided by difference and corrected, the corrected error value and the non-main control scale error value in the original error decomposition result are weighted and fused, the fusion coefficient is set according to the error proportion of each scale. The exemplary weight range is 0.2 to 0.8, the error compensation operation in space is completed, the error result after fusion of all scale layers is subjected to spatial filtering and smoothing processing, and after removing the local abnormal peak value, a continuous and smooth error field is formed, and the refined error distribution is output.

[0072] S7.3, according to the refined error distribution, the weighted correction of the meteorological assimilation result is carried out, and the high-precision assimilated meteorological field data set is obtained.

[0073] Specifically, the error values of the meteorological assimilation results and the corresponding meteorological monitoring area geographical coordinate grid node positions in the refined correction error distribution are matched point by point, the original values of each meteorological variable in the meteorological assimilation results are extracted, the error values in the refined correction error distribution are used as correction factors, and the original meteorological variable values are weighted and corrected. The exemplary range of the weighting coefficient is 0.3 to 0.7 according to the spatial gradient and local density of the error value. In the weighting correction process, positive and negative adjustments are made according to the error value direction to ensure that the correction direction is consistent with the error direction. After correction, the meteorological variable values of all grid nodes are subjected to global consistency test, local mutation nodes are identified using spatial variability constraint method, and the mutation nodes are subjected to local mean interpolation method for re-smoothing processing. The corrected meteorological variable values are compared with the original meteorological assimilation results, and the values exceeding the exemplary 0.01 are further corrected by iteration. The maximum number of iterations is set to be exemplary 5 times. When the difference values of all meteorological monitoring area geographical coordinate grid nodes are lower than the exemplary 0.01, the high-precision assimilated meteorological field data set is output.

[0074] The embodiment also provides a computer device suitable for the case of the meteorological data assimilation method fusing terrain features, which comprises a memory and a processor; the memory is used to store computer executable instructions, and the processor is used to execute the computer executable instructions to realize the meteorological data assimilation method fusing terrain features proposed in the above embodiment.

[0075] The computer device can be a terminal, which comprises a processor, a memory, a communication interface, a display screen and an input device connected through a system bus. The processor of the computer device is used to provide computing and control capabilities. The memory of the computer device comprises a non-volatile storage medium and an internal memory. The non-volatile storage medium stores an operating system and a computer program. The internal memory provides an environment for the operation of the operating system and the computer program in the non-volatile storage medium. The communication interface of the computer device is used to communicate with external terminals in a wired or wireless manner. The wireless manner can be realized through WIFI, an operator network, NFC (near field communication) or other technologies. The display screen of the computer device can be a liquid crystal display screen or an electronic ink display screen. The input device of the computer device can be a touch layer overlaid on the display screen, or a key, trackball or touchpad arranged on the shell of the computer device. The input device can also be an external keyboard, touchpad or mouse, etc.

[0076] The embodiment also provides a storage medium on which a computer program is stored, the program being executed by a processor to implement the meteorological data assimilation method for fusing terrain features as proposed in the above embodiment; the storage medium can be implemented by any type of volatile or non-volatile storage device or a combination thereof, such as a static random access memory (SRAM), an electrically erasable programmable read-only memory (EEPROM), an erasable programmable read-only memory (EPROM), a programmable read-only memory (PROM), a read-only memory (ROM), a magnetic memory, a flash memory, a magnetic disk, or an optical disk.

[0077] To sum up, the present application realizes the quantitative modeling of the multi-scale response relationship between terrain features and meteorological variables by constructing a meteorological variable-terrain factor correlation atlas, so that the initial meteorological field can fully reflect the influence of terrain on meteorological evolution, thereby improving the spatial adaptability and initial value accuracy of the model in complex topography regions; further, the initial meteorological field is dynamically corrected and spatio-temporal error is reconstructed based on the terrain disturbance sensitive weight, so as to realize the directional perception and intelligent correction of observation bias, effectively enhancing the response capability and error control level of the model in dynamic change scenarios, and finally forming a meteorological assimilation result with higher precision and stronger terrain consistency.

[0078] It should be noted that the above embodiments are only used to illustrate the technical solutions of the present application but not limit the present application. Although the present application has been described in detail with reference to the preferred embodiments, those skilled in the art should understand that the technical solutions of the present application can be modified or replaced equivalently without departing from the spirit and scope of the technical solutions of the present application, and all of them should be covered in the scope of the claims of the present application.

Claims

1. A meteorological data assimilation method incorporating topographic features, characterized in that: include, Multi-source environmental observation data were collected and preprocessed to obtain topographic feature model data. Then, multiple regression analysis and mutual information calculation were performed on the preprocessed multi-source environmental observation data to obtain the meteorological variable-topographic factor correlation map. By combining the correlation map of meteorological variables and topographic factors with the numerical field of atmospheric state, a multivariate initial meteorological field model incorporating topographic factors is obtained. By utilizing real-time multi-source environmental observation data, the spatiotemporal dynamic deviation distribution characteristics of the multivariate initial meteorological field model are identified, and a terrain disturbance response matrix is ​​generated. By using an assimilation optimization algorithm, dynamic correction and spatiotemporal error reconstruction based on terrain disturbance sensitive weights are performed on the terrain disturbance response matrix and the multivariate initial meteorological field model to obtain meteorological assimilation results. Multi-scale adaptive error decomposition is performed on the meteorological assimilation results, and refined error correction is carried out in combination with preprocessed multi-source environmental observation data to obtain a high-precision assimilated meteorological field dataset.

2. The meteorological data assimilation method incorporating terrain features as described in claim 1, characterized in that: The multi-source environmental observation data includes ground meteorological observation elements, upper-air meteorological detection elements, and geomorphological features; The preprocessing includes data cleaning, time series alignment, and data standardization transformation.

3. The meteorological data assimilation method incorporating terrain features as described in claim 1, characterized in that: The specific steps for obtaining terrain feature model data are as follows: The preprocessed geomorphic features are decomposed into multi-scale topographic factors to obtain a set of multi-scale topographic factors. For a multi-scale terrain factor set, a wavelet transform-based feature enhancement method is applied to amplify high-frequency details and smoothly fuse low-frequency data to generate an enhanced terrain factor set. The enhanced terrain factor set is mapped to the geographic coordinate grid of the meteorological monitoring area for spatial interpolation to obtain terrain feature model data.

4. The meteorological data assimilation method incorporating terrain features as described in claim 1, characterized in that: The specific steps for obtaining the meteorological variable-topographic factor correlation map are as follows. Multiple regression analysis was performed on the preprocessed ground meteorological observation elements, the preprocessed upper-air meteorological detection elements, and the enhanced topographic factors to obtain the regression coefficient matrix. The mutual information matrix is ​​obtained by performing mutual information calculation on the regression coefficient matrix, terrain feature model data, preprocessed ground meteorological observation elements, and preprocessed upper-air meteorological detection elements. By jointly analyzing and weighting the regression coefficient matrix and mutual information matrix, a correlation map of meteorological variables and topographic factors is obtained.

5. The meteorological data assimilation method incorporating terrain features as described in claim 4, characterized in that: The process of combining meteorological variable-topographic factor correlation maps with atmospheric state numerical fields to obtain a multivariate initial meteorological field model incorporating topographic factors is described in the following steps. Joint modeling was performed on the preprocessed ground meteorological observation elements and upper-air meteorological sounding elements to obtain the numerical field of atmospheric state. The topographic factor weights in the meteorological variable-topographic factor correlation map are mapped to the geographic coordinate grid of the meteorological monitoring area in the atmospheric state numerical field to obtain the topographic influence weight grid. The weighted influence coefficients of topographic factors on meteorological variables in the topographic influence weight grid are calculated, and the atmospheric state numerical field is adjusted by topographic factors to form a topographic influence response field. The topographic impact response field and the atmospheric state numerical field are weighted and fused to construct a multivariate initial meteorological field model that incorporates topographic factors.

6. The meteorological data assimilation method incorporating terrain features as described in claim 5, characterized in that: The method utilizes real-time multi-source environmental observation data to identify the spatiotemporal dynamic deviation distribution characteristics of a multivariate initial meteorological field model and generate a terrain disturbance response matrix. The specific steps are as follows: Spatiotemporal analysis was performed on the preprocessed real-time multi-source environmental observation data to identify the dynamic change characteristics of meteorological variables in different time and spatial dimensions, and to obtain a spatiotemporal deviation dataset of meteorological variables. By comparing the spatiotemporal deviation dataset of meteorological variables with a multivariate initial meteorological field model, the spatiotemporal dynamic deviation distribution characteristics are identified. The spatiotemporal dynamic deviation distribution characteristics are transformed and visualized to obtain the spatiotemporal dynamic deviation distribution map of meteorological variables. The spatiotemporal dynamic deviation distribution map of meteorological variables and the response field of topographic influence are used to perform quantitative calculations of topographic disturbances and generate a topographic disturbance response matrix.

7. The meteorological data assimilation method incorporating terrain features as described in claim 6, characterized in that: The assimilation optimization algorithm is used to dynamically correct and reconstruct the spatiotemporal error of the terrain disturbance response matrix and the multivariate initial meteorological field model based on terrain disturbance sensitive weights, thereby obtaining the meteorological assimilation result. The specific steps are as follows. The terrain disturbance response matrix and the multivariate initial meteorological field model are dynamically corrected by the assimilation optimization algorithm to obtain the optimized terrain impact response field. Spatiotemporal error reconstruction was performed on the optimized terrain impact response field and the multivariate initial meteorological field model to obtain the meteorological assimilation results.

8. The meteorological data assimilation method incorporating terrain features as described in claim 7, characterized in that: The process involves performing multi-scale adaptive error decomposition on the meteorological assimilation results, and then combining this with preprocessed multi-source environmental observation data for refined error correction, resulting in a high-precision assimilated meteorological field dataset. The specific steps are as follows: Multi-scale adaptive error decomposition was performed on the meteorological assimilation results to identify the error distribution characteristics at different scales and obtain the error decomposition results; Using preprocessed multi-source environmental observation data, the error decomposition results are refined and error compensated to obtain the refined corrected error distribution; Based on the refined correction error distribution, the meteorological assimilation results are weighted and corrected to obtain a high-precision assimilated meteorological field dataset.

9. A computer device comprising a memory and a processor, wherein the memory stores a computer program, characterized in that: When the processor executes the computer program, it implements the steps of the meteorological data assimilation method for fusing terrain features as described in any one of claims 1 to 8.

10. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the computer program is executed by the processor, it implements the steps of the meteorological data assimilation method for fusing terrain features as described in any one of claims 1 to 8.

Images