Laser wind lidar and data processing and assimilation method thereof

Abstract

The application discloses a laser wind measurement radar and a data processing and assimilation method thereof, and relates to the field of meteorological radar data processing and assimilation.The method comprises the following steps: acquiring laser wind measurement radar observation data, atmospheric stability parameters and terrain undulation characteristic data; constructing a three-dimensional space weight distribution function based on the above parameters, which represents the initial influence weight of an observation point on surrounding grids; dynamically adjusting the horizontal influence range and vertical attenuation characteristics of the function according to the local terrain complexity and atmospheric vertical stratification change information, and generating a dynamic weight distribution matched with local atmospheric physics and terrain characteristics; introducing the distribution as a spatial adjustment factor of an observation error covariance matrix into a cost function, and simultaneously optimizing a wind field analysis field and the dynamic weight distribution by minimizing the cost function; and outputting the optimized wind field analysis field for numerical weather prediction.The application realizes accurate and efficient fusion of observation information in three-dimensional variational assimilation, and significantly improves the authenticity of wind field analysis and the reliability of prediction under complex conditions.

Description

A laser wind-measuring radar and its data processing and assimilation method Technical Field

[0001] This invention relates to the field of laser wind measurement radar application technology, and in particular to a laser wind measurement radar and its data processing and assimilation method. Background Technology

[0002] As a key piece of equipment in modern atmospheric sounding, laser wind radar can acquire the fine three-dimensional structure of wind fields with high spatiotemporal resolution. Its data is invaluable for improving the accuracy of numerical weather prediction, especially for forecasting severe weather events such as strong convection and low-level wind shear. However, how to scientifically and effectively integrate these valuable point and linear observation data into three-dimensional numerical models covering a wide area—that is, the data assimilation process—has always been a core technical challenge in the meteorological field.

[0003] Current mainstream assimilation methods, especially three-dimensional variational assimilation systems, typically employ fixed weighting schemes based on distance or empirical functions to allocate the influence of observational information on the surrounding grid when processing such observations. While this method is simple and easy to implement, it has a fundamental limitation: it struggles to accurately reflect the true, physically driven spatial propagation patterns of observational information in the atmosphere. The atmosphere is not a homogeneous medium; its stability profoundly affects the efficiency of momentum and energy exchange in the vertical direction. Complex terrain forces airflow to change direction, resulting in bypassing or climbing, thus significantly distorting the horizontal structure of the near-surface wind field. Existing static, uniform weighting distribution models cannot adaptively couple these spatially varying dynamic and thermodynamic processes. This leads to systematic deviations in the influence range and intensity distribution of observational information from the actual situation in complex terrain areas or regions with unstable atmospheric stratification. This not only fails to fully leverage the high precision advantages of lidar wind measurement data but can also sometimes introduce inaccurate errors into the analysis field, hindering further improvements in forecast accuracy. Summary of the Invention

[0004] Therefore, the technical problem to be solved by the present invention is to overcome the shortcomings of the prior art in accurately reflecting the real physical propagation law of laser wind measurement radar observation information in complex terrain and changing atmospheric stratification, and to provide a laser wind measurement radar and its data processing and assimilation method, which can generate a physically adaptive observation weight distribution by dynamically coupling terrain undulation and atmospheric stability parameters, thereby realizing the accurate and efficient fusion of observation information in three-dimensional variational assimilation, and significantly improving the authenticity of wind field analysis and forecast reliability under complex conditions.

[0005] To address the aforementioned technical problems, this invention provides a data processing and assimilation method for laser wind radar, comprising the following steps:

[0006] Acquire laser wind radar observation data, corresponding atmospheric stability parameters, and terrain undulation characteristics of the assimilation target area;

[0007] Based on atmospheric stability parameters and topographic relief characteristics, a three-dimensional spatial weight distribution function is constructed to characterize the initial influence weight of a single laser wind radar observation point on its surrounding three-dimensional grid points.

[0008] Based on topographic relief data, regions with complex topography are identified. Combined with information on vertical stratification changes reflected by atmospheric stability parameters, the horizontal influence range and vertical attenuation characteristics of the three-dimensional spatial weight distribution function are dynamically adjusted to generate a dynamic weight distribution that matches local atmospheric physical processes and topographic features.

[0009] The dynamic weight distribution is introduced into the cost function of the three-dimensional variational assimilation system as a spatial adjustment factor of the observation error covariance matrix. By minimizing the cost function, the wind field analysis field and the dynamic weight distribution are optimized simultaneously.

[0010] The optimized wind field analysis is output for numerical weather prediction.

[0011] In one embodiment of the present invention, a three-dimensional spatial weight distribution function is constructed based on atmospheric stability parameters and terrain undulation feature data, including:

[0012] The atmospheric stability parameters and topographic relief characteristics data are standardized by converting parameters of different dimensions and ranges into a preset dimensionless range to obtain standardized stability parameters and standardized topographic parameters.

[0013] Based on the principles of meteorological fluid dynamics, a basic influence template is defined: a three-dimensional spatial grid, in which each grid point stores an initial scalar value. The initial scalar value is determined by the preset horizontal distance function and vertical distance function from the grid point to the central observation point, which characterizes the theoretical attenuation in an ideal homogeneous medium.

[0014] The standardized stability parameter is mapped to a vertical adjustment coefficient to modulate the decay rate of the vertical distance function in the basic influence template; at the same time, the standardized terrain parameter is mapped to an anisotropic adjustment field in the horizontal direction to modulate the decay rate of the horizontal distance function in the basic influence template in each direction.

[0015] The modulated vertical distance function and the horizontal distance function are combined in three-dimensional space to form a three-dimensional spatial weight distribution function, the output of which is the initial influence weight of the corresponding three-dimensional grid point.

[0016] In one embodiment of the present invention, identifying areas with complex terrain based on terrain undulation feature data includes:

[0017] Based on terrain undulation feature data, calculate the slope value and terrain undulation degree of all points in the target area;

[0018] Sort the slope values ​​and terrain relief values ​​from largest to smallest respectively;

[0019] Based on the preset complex area ratio coefficient, the high-slope subset in the slope value ranking and the high-undulation subset in the terrain undulation ranking are determined respectively.

[0020] Points that are simultaneously located in both the high-slope subset and the high-undulation subset are identified as points with complex terrain.

[0021] Spatial connectivity analysis is performed on points with complex terrain, and the area formed by interconnected points with complex terrain is identified as a complex terrain area.

[0022] In one embodiment of the present invention, dynamically adjusting the horizontal influence range of the three-dimensional spatial weight distribution function based on complex terrain areas includes:

[0023] In areas with complex terrain, extract the main terrain orientation and main slope direction at the current location of the laser wind radar observation point;

[0024] Establish a local horizontal influence ellipse model centered on the observation point, with the major axis of the ellipse aligned with the main terrain direction and the minor axis aligned with the main slope direction;

[0025] The eccentricity of the ellipse is dynamically adjusted according to the slope: the greater the slope, the greater the eccentricity of the ellipse, so that the horizontal influence range of the weight value represented by the three-dimensional spatial weight distribution function on the main slope is more contracted than the horizontal influence range on the main terrain direction.

[0026] The boundary defined by the adjusted local horizontal influence ellipse model is used as the new effective boundary of the three-dimensional spatial weight distribution function in the horizontal direction.

[0027] In one embodiment of the present invention, the vertical stratification variation information reflected by the atmospheric stability parameters is extracted in the following manner:

[0028] Obtain atmospheric stability parameters, including at least the vertical gradient of potential temperature or the Richardson number;

[0029] Based on atmospheric stability parameters, an index is calculated to characterize the vertical mixing capacity of the atmosphere;

[0030] Based on the numerical range of the index, the vertical stratification of the atmosphere is classified into stable, neutral, and unstable states.

[0031] In one embodiment of the present invention, the vertical attenuation characteristics of the three-dimensional spatial weight distribution function are dynamically adjusted through a vertical attenuation scale modulation step, including:

[0032] Based on the extracted atmospheric vertical stratification state, a vertical attenuation scale modulation factor is determined.

[0033] If the atmospheric vertical stratification is in a stable state, a vertical attenuation scale modulation factor less than 1 is applied to shorten the vertical attenuation scale of the three-dimensional spatial weight distribution function.

[0034] If the atmospheric vertical stratification is unstable, a vertical attenuation scale modulation factor greater than 1 is applied to extend the vertical attenuation scale of the three-dimensional spatial weight distribution function.

[0035] If the atmospheric vertical stratification state is neutral, a vertical attenuation scale modulation factor equal to 1 is applied to keep the vertical attenuation scale of the three-dimensional spatial weight distribution function unchanged.

[0036] In one embodiment of the present invention, a dynamic weight distribution is introduced into the cost function of the three-dimensional variational assimilation system as a spatial adjustment factor for the observation error covariance matrix, including:

[0037] The values ​​of the dynamic weight distribution are normalized to form a spatial adjustment factor field.

[0038] The spatial adjustment factor field and the observation error covariance matrix of the three-dimensional variational assimilation system are correlated according to spatial location to generate a composite observation error covariance matrix.

[0039] The original observation error covariance matrix is ​​replaced by a composite observation error covariance matrix and substituted into the cost function of the three-dimensional variational assimilation system.

[0040] In one embodiment of the present invention, the wind field analysis field and dynamic weight distribution are simultaneously optimized by minimizing the cost function, including:

[0041] Set the dynamic weight distribution as the first optimizable variable and the wind field analysis field as the second optimizable variable.

[0042] Construct an extended cost function that includes a composite observation error covariance matrix and a dynamic weight distribution regularization constraint term;

[0043] In the extended cost function, the first variable and the second variable are coupled through the composite observation error covariance matrix;

[0044] An iterative optimization algorithm is used to solve the extended cost function. In each iteration, the first variable and the second variable are updated simultaneously, and the composite observation error covariance matrix is ​​recalculated using the updated first variable.

[0045] The iteration stops when the update amounts of the first and second variables tend to stabilize between successive iterations.

[0046] The second variable obtained from the final iteration is used as the output of the optimized wind field analysis field.

[0047] In one embodiment of the present invention, after outputting the optimized wind field analysis field, the method further includes:

[0048] The dynamic weight distribution generated during this assimilation process is then subjected to feature extraction and standardization.

[0049] The extracted features are associated with and stored with the corresponding weather pattern type, seasonal information and terrain classification labels to form a weighted pattern knowledge base;

[0050] In subsequent assimilation tasks, historical patterns similar to the current conditions are retrieved from the weight pattern knowledge base first, serving as a reference for constructing the initial three-dimensional spatial weight distribution function.

[0051] To address the aforementioned technical problems, this invention also provides a laser wind-measuring radar, integrating a three-dimensional variational assimilation system for performing the above-mentioned method, including:

[0052] The data acquisition module is configured to acquire laser wind radar observation data, corresponding atmospheric stability parameters, and terrain undulation feature data of the target area for assimilation.

[0053] The initial weight function construction module is configured to construct a three-dimensional spatial weight distribution function based on atmospheric stability parameters and terrain undulation feature data to characterize the initial influence weight of a single laser wind radar observation point on its surrounding three-dimensional grid points.

[0054] The dynamic weight generation module is configured to identify complex terrain areas based on terrain undulation feature data and, in conjunction with the vertical stratification change information reflected by atmospheric stability parameters, dynamically adjust the horizontal influence range and vertical attenuation characteristics of the three-dimensional spatial weight distribution function to generate a dynamic weight distribution that matches local atmospheric physical processes and terrain features.

[0055] The variational assimilation optimization module is configured to introduce the dynamic weight distribution into the cost function of the three-dimensional variational assimilation system as a spatial adjustment factor for the observation error covariance matrix, and simultaneously optimize the wind field analysis field and dynamic weight distribution by minimizing the cost function.

[0056] The results output module is configured to output an optimized wind field analysis field for use in numerical weather prediction.

[0057] The technical solution of the present invention has the following advantages compared with the prior art:

[0058] The data processing and assimilation method for laser wind-measuring radar described in this invention is applied to a three-dimensional variational assimilation system. It acquires laser wind-measuring radar observation data, atmospheric stability parameters, and topographic relief characteristics. Based on this information, it dynamically adjusts the horizontal influence range and vertical attenuation characteristics of the three-dimensional spatial weight distribution function, thereby generating a dynamic weight distribution that matches local atmospheric physical processes and topographic features. This dynamic weight distribution is then introduced into the cost function of the three-dimensional variational assimilation system as a spatial adjustment factor for the observation error covariance matrix. By minimizing the cost function, the wind field analysis field and the dynamic weight distribution are simultaneously optimized. This method overcomes the limitations of traditional fixed-weight schemes in complex terrain and unstable atmospheric stratification, enabling laser wind-measuring radar observation data to be more scientifically and effectively integrated into the three-dimensional variational assimilation system. This significantly improves the accuracy of the wind field analysis field, particularly in the forecasting of severe weather such as strong convection and low-level wind shear. Attached Figure Description

[0059] To make the content of this invention easier to understand, the invention will be further described in detail below with reference to specific embodiments and accompanying drawings, wherein:

[0060] Figure 1 is a flowchart of the data processing and assimilation method of the laser wind radar of the present invention.

[0061] Figure 2 is a flowchart of the steps in constructing a three-dimensional spatial weight distribution function based on atmospheric stability parameters and terrain undulation feature data according to the present invention;

[0062] Figure 3 is a flowchart of the steps of dynamically adjusting the horizontal influence range of the three-dimensional spatial weight distribution function according to the present invention;

[0063] Figure 4 is a flowchart of the steps for dynamically adjusting the vertical attenuation characteristics of the three-dimensional spatial weight distribution function according to the present invention;

[0064] Figure 5 is a flowchart of the steps of introducing dynamic weight distribution into the cost function of the three-dimensional variational assimilation system in this invention.

[0065] Figure 6 is a flowchart of the steps of the present invention to simultaneously optimize the wind field analysis field and dynamic weight distribution by minimizing the cost function. Detailed Implementation

[0066] The present invention will be further described below with reference to the accompanying drawings and specific embodiments, so that those skilled in the art can better understand and implement the present invention. However, the embodiments described are not intended to limit the present invention.

[0067] Referring to Figure 1, this invention discloses a data processing and assimilation method for laser wind radar, which is applied in a three-dimensional variational assimilation system, and includes the following steps:

[0068] First, the method acquires laser wind-measuring radar observation data, corresponding atmospheric stability parameters, and topographic relief feature data of the target area for assimilation. Specifically, the laser wind-measuring radar observation data can be received in real time from laser wind-measuring radar equipment deployed within the target area. This data will serve as the core target observation values ​​for the subsequent assimilation optimization process, used to construct the initial weight function and constrain the wind field analysis field to approximate its measured state. The atmospheric stability parameters can be extracted from upper-air sounding data (e.g., radiosonde data) or the output results of numerical weather prediction models. The topographic relief feature data can be obtained from a high-resolution digital elevation model and preprocessed to extract the required features. This method systematically integrates laser wind-measuring radar observation data, atmospheric stability parameters, and high-precision topographic data, which forms the basis for dynamic weight calculation. The atmospheric stability parameters quantitatively describe the turbulent mixing capacity in the vertical direction, while the topographic data characterizes the dynamic forcing effect of the underlying surface.

[0069] Secondly, based on atmospheric stability parameters and topographic relief characteristics, a three-dimensional spatial weight distribution function is constructed to characterize the initial influence weight of a single lidar wind measurement radar observation point on its surrounding three-dimensional grid points. Based on these key physical parameters, the method constructs a three-dimensional spatial weight distribution function that characterizes the a priori law of the diffusion and attenuation of observation point information into the surrounding space in an idealized homogeneous medium under a given atmospheric stability background and macroscopic topographic features. The output value of the three-dimensional spatial weight distribution function is the initial influence weight of the corresponding three-dimensional grid point, used to quantify the initial correlation strength between the observation point and the grid points. This initial influence weight constitutes the physical benchmark and input basis for the subsequent dynamic adjustment process.

[0070] Furthermore, based on topographic relief data, complex terrain areas are identified. Combined with vertical stratification changes reflected by atmospheric stability parameters, the horizontal influence range and vertical attenuation characteristics of the three-dimensional spatial weight distribution function are dynamically adjusted to generate a dynamic weight distribution that matches local atmospheric physical processes and terrain features. The method introduces a dynamic adjustment step: it automatically identifies complex terrain areas such as ridges and valleys based on terrain data and simultaneously combines this with vertical stratification changes revealed by atmospheric stability to correct the horizontal influence range and vertical attenuation characteristics of the initial weight function in real time and online. For example, under stable atmospheric stratification, the vertical transmission of observational information is suppressed, and the system automatically shrinks the attenuation scale of the weights in the vertical direction. Near steep terrain, the horizontal influence range is asymmetrically adjusted based on the obstruction or channeling effect of the terrain. This allows the final generated dynamic weight distribution to realistically simulate the "diffusion" and "attenuation" processes of observational information under specific weather conditions and terrain, achieving intelligent matching of weights with "local atmospheric physical processes and terrain features."

[0071] Subsequently, the dynamic weight distribution is introduced into the cost function of the three-dimensional variational assimilation system as a spatial adjustment factor for the observation error covariance matrix. By minimizing the cost function, the wind field analysis field and the dynamic weight distribution are simultaneously optimized. During the optimization process of minimizing the cost function, the wind field analysis field and these dynamic weight distributions are optimized simultaneously. This means that the assimilation system not only finds the wind field that best matches the observation and background fields, but also continuously fine-tunes the way observation information should be trusted and propagated based on the current atmospheric conditions. This fundamentally ensures that observation information can be absorbed by numerical weather prediction models in a way that best conforms to the actual atmospheric dynamics.

[0072] Finally, the optimized wind field analysis field is output for numerical weather prediction. This optimized wind field analysis field can be saved in a standard meteorological data format and transmitted to downstream numerical weather prediction models as their initial field or boundary conditions.

[0073] Through the above technical solution, this invention brings several significant benefits. First, it fundamentally overcomes the physical inaccuracies of traditional fixed-weight schemes, significantly improving the efficiency and accuracy of laser wind radar observation data in the assimilation process. Observational information is distributed in three-dimensional space in a manner more consistent with fluid dynamics, reducing spurious analysis field structures introduced by improper weight allocation. Second, this method enhances the ability to analyze wind fields under complex underlying surfaces and special weather conditions. Whether it is local circulation caused by topography or vertical mixing differences due to stability variations, the dynamic weighting mechanism can adaptively respond, thereby more accurately capturing the fine wind field structure of these key areas, which is crucial for applications such as mountain meteorology, aviation safety, and wind energy assessment.

[0074] To implement the above method, referring to Figure 2, this application further proposes a method for constructing a three-dimensional spatial weight distribution function based on atmospheric stability parameters and topographic relief feature data. Specifically, this includes: standardizing the atmospheric stability parameters and topographic relief feature data, uniformly transforming parameters with different dimensions and ranges to a preset dimensionless interval, resulting in standardized stability parameters and standardized topographic parameters. The standardization process aims to eliminate the influence of different physical dimensions and numerical ranges on subsequent calculations, ensuring the comparability and consistency of all parameters during the construction of the weight function. For example, atmospheric stability parameters may be represented by temperature gradients or Richardson numbers, while topographic relief feature data may be represented by altitude or slope; these parameters have vastly different dimensions and numerical ranges. Through standardization, such as using min-max normalization or Z-score standardization, these parameters can be uniformly scaled to a preset dimensionless interval such as [0, 1] or [-1, 1]. The standardized parameters, namely the standardized stability parameters and standardized topographic parameters, can participate more fairly and effectively in subsequent weight modulation calculations, avoiding unreasonable dominance of the results by a single parameter due to dimensional differences.

[0075] Based on the principles of meteorological fluid dynamics, a fundamental influence template is defined: a three-dimensional spatial grid, where each grid point stores an initial scalar value. This initial scalar value is determined by a pre-defined horizontal distance function and a vertical distance function from the grid point to the central observation point, characterizing the theoretical attenuation in an ideal homogeneous medium. The fundamental influence template provides a theoretical attenuation model under ideal homogeneous atmospheric and topographic conditions. This template is typically a three-dimensional grid centered on the observation point, with each grid point pre-calculated and storing an initial scalar value. This initial scalar value reflects the theoretical influence strength of the observation point on that grid point, and its attenuation characteristics are jointly determined by a pre-defined horizontal distance function (e.g., Gaussian function, exponential function, etc.) and a vertical distance function (e.g., an attenuation function considering atmospheric boundary layer height). These functions, based on the principles of meteorological fluid dynamics, simulate the diffusion and attenuation of observed information with distance in the absence of complex terrain and atmospheric stratification.

[0076] The standardized stability parameter is mapped to a vertical adjustment coefficient to modulate the decay rate of the vertical distance function in the fundamental influence template. The standardized stability parameter reflects the ease of atmospheric vertical mixing; for example, in a stable atmosphere, vertical mixing is suppressed, and information decays rapidly vertically; in an unstable atmosphere, vertical mixing is vigorous, and information decays slowly. Therefore, this adjustment coefficient can dynamically adjust the decay rate of the vertical distance function to reflect the influence of the actual atmospheric stratification on vertical information propagation. For example, when the stability parameter indicates atmospheric stability, the adjustment coefficient will make vertical decay faster; when it indicates instability, it will make vertical decay slower.

[0077] Simultaneously, standardized terrain parameters are mapped to an anisotropic modulating field in the horizontal direction to modulate the attenuation rate of the horizontal distance function in the basic influence template in various directions. The standardized terrain parameters reflect the complexity and undulation characteristics of the terrain, exhibiting a significant anisotropic impact on wind field propagation in the horizontal direction. For example, in a valley, wind information may propagate further along the valley direction but attenuate faster in the direction perpendicular to the valley. Therefore, this modulating field can dynamically modulate the attenuation rate of the horizontal distance function in different directions based on terrain features. For instance, in the direction of terrain orientation, the modulating field may slow down horizontal attenuation; while in the direction perpendicular to the terrain orientation, it may speed up horizontal attenuation. This mapping relationship can be implemented using a predefined lookup table, empirical formula, or machine learning model.

[0078] Finally, the modulated vertical distance function and the horizontal distance function are combined in three-dimensional space to form a three-dimensional weight distribution function, the output of which is the initial influence weight of the corresponding three-dimensional grid point. This combination operation usually uses multiplication or weighted averaging. For example, the modulated horizontal distance function can be multiplied by the modulated vertical distance function to obtain the initial influence weight of each three-dimensional grid point. This multiplicative combination ensures that the attenuation characteristics in both the horizontal and vertical directions are reflected in the final weight value.

[0079] In some of the embodiments described above in this application, it is proposed to identify complex terrain areas based on terrain undulation feature data. However, in practical applications, if the identification method for complex terrain areas is not accurate enough or fails to fully consider the subtle features of the terrain, it may lead to the inability to accurately reflect local atmospheric physical processes when dynamically adjusting the horizontal influence range of the three-dimensional spatial weight distribution function, thereby affecting the optimization accuracy of the wind field analysis field.

[0080] To address this, this application further proposes steps for identifying terrain-complex areas based on terrain undulation feature data, including: firstly, calculating the slope value and terrain undulation degree of all points within the target area based on the terrain undulation feature data. The slope value characterizes the degree of inclination of a surface unit in the horizontal direction and can be calculated using digital elevation model (DEM) data, utilizing the elevation difference and horizontal distance between adjacent grid points, for example, using a gradient algorithm. The terrain undulation degree reflects the degree of elevation change within a local area. Typically, within a preset sliding window (e.g., a 3x3 or 5x5 grid), the difference between the maximum and minimum elevation values ​​of all points within the window is calculated, and this difference is assigned to the center point of the window.

[0081] Subsequently, to identify the points with the most significant terrain features within the region, the slope values ​​and terrain undulations of all grid points within the entire target region are sorted from largest to smallest. Based on this, according to a preset complex region proportion coefficient, a high-slope subset in the slope value sorting and a high-undulation subset in the terrain undulation sorting are determined respectively. This proportion coefficient is a preset percentage or threshold used to define which points belong to the "high-position" feature; for example, if set to 10%, the top 10% of grid points after sorting are selected as the high-position subset.

[0082] Furthermore, points simultaneously located in both the high-slope subset and the high-relief subset are identified as terrain-complex points. This dual-screening mechanism, by combining two independent indicators—slope value and terrain relief—more comprehensively and accurately identifies points with truly significant complexity characteristics, avoiding misjudgments that might occur based on a single indicator. Finally, spatial connectivity analysis is performed on these identified terrain-complex points, classifying the regions formed by interconnected terrain-complex points as terrain-complex areas. Spatial connectivity analysis aims to aggregate discrete terrain-complex points into meaningful regions, avoiding the misclassification of isolated, potentially noise-prone, complex points as important areas. For example, graph-theoretic breadth-first search (BFS) or depth-first search (DFS) algorithms can be used to check whether adjacent grid points are also terrain-complex points, thus grouping them into the same connected region.

[0083] After identifying areas with complex terrain, as shown in Figure 3, this application further proposes a specific method for dynamically adjusting the horizontal influence range of the three-dimensional spatial weight distribution function based on areas with complex terrain. This includes: Within areas with complex terrain, it is necessary to first extract the main topographic orientation and main slope aspect of the current laser wind radar observation point. This step aims to obtain the key geometric features of the local terrain at the observation point. The main topographic orientation refers to the main direction of terrain extension in the area, such as the orientation of mountains or valleys. The main slope aspect refers to the direction of the most significant slope change in the area, usually perpendicular to contour lines. This information can be extracted by performing local analysis of the terrain undulation feature data within a certain range around the observation point. For example, principal component analysis (PCA) or Fourier transform can be used to process the local elevation data to identify the dominant direction.

[0084] Based on this, a local horizontal influence ellipse model centered on the observation point is established, with the major axis aligned with the main terrain orientation and the minor axis aligned with the main slope aspect. This step constructs a geometric model capable of characterizing the anisotropic horizontal influence range. Traditional circular influence range models assume that wind field influence is isotropic, but in complex terrain, the distance wind fields propagate along valleys or ridges is often much greater than their propagation distance perpendicular to the valley or ridge direction. By establishing an ellipse model and aligning the major axis with the main terrain orientation and the minor axis with the main slope aspect, this anisotropic propagation characteristic can be simulated. The major axis of the ellipse represents the main influence range of the wind field in the terrain guidance direction, while the minor axis represents the influence range perpendicular to the terrain guidance direction.

[0085] Furthermore, the eccentricity of the ellipse is dynamically adjusted according to the slope: the steeper the slope, the greater the eccentricity of the ellipse, causing the horizontal influence range of the weight value represented by the three-dimensional spatial weight distribution function to shrink more in the main slope direction compared to the horizontal influence range in the main terrain direction. This step further refines the parameters of the ellipse model, enabling it to dynamically adapt to different terrain complexities. Eccentricity is a key parameter of the ellipse shape, determining its flatness. In areas with steep terrain slopes, the wind field is more significantly hindered and guided by the terrain, resulting in stronger restrictions on its propagation perpendicular to the slope direction, while its propagation along the slope direction or terrain direction is relatively longer. Therefore, by increasing the eccentricity, the influence range of the ellipse in the minor axis direction (main slope direction) can shrink more significantly compared to the major axis direction (main terrain direction), thus more accurately characterizing the anisotropic propagation characteristics of the wind field in steep terrain.

[0086] Finally, the boundary defined by the adjusted local horizontal influence ellipse model is used as the new effective boundary for the 3D spatial weight distribution function in the horizontal direction. This step clarifies how the dynamically adjusted ellipse model is applied to the 3D spatial weight distribution function. The boundary of the ellipse model is no longer a simple circle, but an ellipse that dynamically changes according to local terrain features. This means that when calculating the horizontal influence weight of an observation point on surrounding grid points, only grid points located within the ellipse boundary will be assigned non-zero weights, or the rate of weight decay will be adjusted according to their position relative to the ellipse boundary.

[0087] In some of the embodiments described above in this application, a method for dynamically adjusting the vertical attenuation characteristics of a three-dimensional spatial weight distribution function based on vertical stratification change information reflected by atmospheric stability parameters is proposed. However, in its implementation, the key to ensuring the physical rationality and accuracy of the dynamic adjustment lies in how to accurately and effectively extract the vertical stratification change information that can directly guide the adjustment of the vertical attenuation characteristics from the original atmospheric stability parameters and transform it into an operable stratification state.

[0088] In response, this application further proposes the vertical stratification change information reflected by atmospheric stability parameters, which is extracted in the following way: obtaining atmospheric stability parameters, including at least the vertical gradient of potential temperature or Richardson number; calculating an index to characterize the vertical mixing capacity of the atmosphere based on the atmospheric stability parameters; and classifying the vertical stratification state of the atmosphere into stable, neutral and unstable states according to the numerical range of the index.

[0089] Specifically, when obtaining atmospheric stability parameters, at least the vertical gradient of potential temperature or the Richardson number should be included. Atmospheric stability parameters are key physical quantities describing the resistance or tendency of vertical atmospheric motion. The vertical gradient of potential temperature (dθ / dz) directly reflects the static stability of atmospheric stratification: when dθ / dz is greater than zero, the atmosphere is stable; when dθ / dz is less than zero, the atmosphere is unstable; and when dθ / dz is equal to zero, the atmosphere is neutral. The Richardson number (Ri) comprehensively considers buoyancy and shear effects and is an important indicator for judging the intensity of atmospheric turbulence and mixing; its value is closely related to the vertical mixing capacity of the atmosphere. These parameters can be obtained through various means, such as calculation using temperature, humidity, air pressure, and wind speed profile data measured by instruments such as radiosondes and tethered balloons; or directly extracted from the output results of numerical weather prediction models. These parameters are chosen as the basis because they can directly and quantitatively reflect the vertical stratification characteristics and mixing capacity of the atmosphere, providing a solid physical basis for subsequent dynamic adjustments.

[0090] Based on this, an index characterizing atmospheric vertical mixing capacity is calculated using the acquired atmospheric stability parameters. This index aims to transform the original atmospheric stability parameters into a more intuitive and easily quantifiable indicator of atmospheric vertical mixing intensity. For example, the Richardson number can be used directly as this index, as it inherently reflects the atmosphere's resistance to vertical mixing. Alternatively, it can be mapped to a mixing capacity index between 0 and 1 based on the vertical gradient of potential temperature through specific functional relationships (such as reciprocals, exponential functions, etc.), where a higher index value indicates stronger vertical mixing capacity. The calculation of this index aims to provide a unified quantitative standard for subsequent clear distinction of atmospheric vertical stratification states.

[0091] Subsequently, based on the numerical range of the index, the vertical stratification state of the atmosphere is classified into stable, neutral, and unstable states. This step transforms continuously changing physical quantities into discrete stratification states with clear physical meaning. For example, if the Richardson number is used as the index, different thresholds can be set: when the Richardson number is greater than a preset stability threshold (e.g., 0.25), it is considered a stable state, where atmospheric vertical mixing is suppressed; when the Richardson number is less than a preset instability threshold (e.g., 0), it is considered an unstable state, where atmospheric vertical mixing is active; when the Richardson number is between the instability and stability thresholds, it is considered a neutral state, where atmospheric vertical mixing is in a transitional or equilibrium state. In this way, the complex physical process is simplified into a clear stratification classification, providing a direct basis for the subsequent dynamic adjustment of the vertical attenuation characteristics of the three-dimensional spatial weight distribution function.

[0092] In practical applications, how to effectively transform this vertical stratification information into a dynamic adjustment of the vertical attenuation characteristics of the three-dimensional spatial weight distribution function to more accurately reflect the vertical correlation under different atmospheric stratifications remains a problem to be solved. If the vertical attenuation characteristics fail to match the actual atmospheric stratification state, it may lead to inaccurate propagation of observational information in the vertical direction, thereby affecting the accuracy of wind field analysis.

[0093] Referring to Figure 4, this application further proposes a method for dynamically adjusting the vertical attenuation characteristics of the three-dimensional spatial weight distribution function. This is achieved through a vertical attenuation scale modulation step, specifically including: determining a vertical attenuation scale modulation factor based on the extracted atmospheric vertical stratification state; if the atmospheric vertical stratification state is stable, applying a vertical attenuation scale modulation factor less than 1 to shorten the vertical attenuation scale of the three-dimensional spatial weight distribution function; if the atmospheric vertical stratification state is unstable, applying a vertical attenuation scale modulation factor greater than 1 to lengthen the vertical attenuation scale of the three-dimensional spatial weight distribution function; and if the atmospheric vertical stratification state is neutral, applying a vertical attenuation scale modulation factor equal to 1 to maintain the vertical attenuation scale of the three-dimensional spatial weight distribution function unchanged.

[0094] Specifically, the vertical attenuation characteristic refers to the decreasing influence of a laser wind radar observation point on other grid points in the vertical direction with increasing distance. Dynamic adjustment means that this attenuation law is not fixed but adaptively changed according to real-time atmospheric conditions. This is achieved through a "vertical attenuation scale modulation step," aiming to provide a systematic and quantifiable method to control the rate of attenuation, ensuring that the propagation of observed information in the vertical direction conforms to actual atmospheric physical processes. Based on the extracted atmospheric vertical stratification state, a vertical attenuation scale modulation factor is determined, where the atmospheric vertical stratification state (stable, unstable, neutral) directly reflects the intensity of atmospheric vertical mixing and turbulent activity. For example, stable atmospheres inhibit vertical mixing, while unstable atmospheres promote it. The vertical attenuation scale modulation factor is a dimensionless numerical value that alters the attenuation rate of the vertical distance term in the three-dimensional spatial weight distribution function. This factor can be derived in advance through numerical simulation, empirical statistics, or physical models and stored in a lookup table or function for rapid retrieval based on the currently identified stratification state. If the atmospheric vertical stratification is stable, a vertical attenuation scale modulation factor less than 1 is applied to shorten the vertical attenuation scale of the three-dimensional spatial weight distribution function. Under stable atmospheric conditions, vertical turbulent mixing is suppressed, the atmospheric stratification is stable, and the vertical information propagation and influence range are small. Therefore, applying a modulation factor less than 1 will cause the weight attenuation in the vertical direction to be faster, meaning the influence of the observation point on distant grid points in the vertical direction will weaken rapidly, thus "shortening" the vertical attenuation scale and more realistically reflecting the weak vertical correlation characteristic of a stable atmosphere. If the atmospheric vertical stratification is unstable, a vertical attenuation scale modulation factor greater than 1 is applied to extend the vertical attenuation scale of the three-dimensional spatial weight distribution function. Under unstable atmospheric conditions, vertical convection and turbulence are vigorous, atmospheric mixing is intense, and observational information propagates more easily to greater distances in the vertical direction. Therefore, applying a modulation factor greater than 1 will cause the weight attenuation in the vertical direction to be slower, meaning the influence range of the observation point on distant grid points in the vertical direction will expand, thus "extending" the vertical attenuation scale and more accurately capturing the strong vertical correlation characteristic of an unstable atmosphere. If the atmospheric vertical stratification is neutral, a vertical attenuation scale modulation factor of 1 is applied to maintain the vertical attenuation scale of the three-dimensional spatial weight distribution function unchanged. Neutral atmospheric stratification lies between stable and unstable conditions, with moderate vertical mixing. In this case, the basic, unadjusted vertical attenuation characteristics are typically used as a reference. Therefore, applying a modulation factor of 1 means not further shortening or lengthening the vertical attenuation scale, maintaining its original setting to reflect the average vertical correlation under neutral atmospheric conditions.

[0095] In its implementation, to effectively integrate these dynamic weight distributions into the core computational mechanism of the three-dimensional variational assimilation system and enable them to play a precise spatial adjustment role in the cost function to accurately reflect the spatial heterogeneity of observation errors, a mechanism is needed to ensure that the dynamic weights are properly coupled with the core parameters of the assimilation system (such as the observation error covariance matrix). Without such a coupling mechanism, the dynamic weights may not be able to fully exert their spatial adjustment role, resulting in insufficiently refined utilization of observation information in the assimilation results, thus affecting the accuracy of the wind field analysis.

[0096] Referring to Figure 5, this application further proposes incorporating the dynamic weight distribution into the cost function of a three-dimensional variational assimilation system as a spatial adjustment factor for the observation error covariance matrix. This includes the following steps: First, the values ​​of the dynamic weight distribution are normalized to form a spatial adjustment factor field. The normalization process aims to transform the original values ​​of the dynamic weight distribution (whose range may vary depending on the calculation method) into a preset, dimensionless numerical range, such as [0, 1] or [0.5, 2.0]. This ensures that when the dynamic weights are subsequently used as adjustment factors, they can modulate the observation error covariance matrix in a stable and controllable manner, avoiding unnecessary bias or numerical instability caused by excessively large or small original numerical ranges.

[0097] Secondly, the spatial adjustment factor field is correlated with the observation error covariance matrix of the three-dimensional variational assimilation system according to spatial location to generate a composite observation error covariance matrix. The observation error covariance matrix is ​​a core parameter in the three-dimensional variational assimilation system, used to describe the magnitude of the observation error and its spatial correlation. Through correlation operations, the spatial adjustment factor field can dynamically modulate the elements of the observation error covariance matrix, especially its diagonal elements (i.e., the observation error variance). For example, at a certain spatial location, if the value of the spatial adjustment factor field indicates that the observations in that area are greatly affected by the environment (such as complex terrain leading to reduced observation representativeness), the observation error variance at that location can be increased accordingly; conversely, if it indicates that the environmental influence is small, the observation error variance can be decreased.

[0098] Finally, the original observation error covariance matrix is ​​replaced with the composite observation error covariance matrix and substituted into the cost function of the three-dimensional variational assimilation system. In three-dimensional variational assimilation, the cost function is a mathematical expression used to measure the differences between the analysis field and the background field, and between the analysis field and the observation field; its minimization process determines the final wind field analysis field. The observation term in the cost function typically contains the inverse of the observation error covariance matrix. By replacing the original static matrix with the composite observation error covariance matrix, the modulation effect of the dynamic weight distribution on the observation error is directly reflected in the calculation and minimization process of the cost function.

[0099] If the dynamic weight distribution is treated as a fixed parameter during the optimization process, its interaction with the wind field analysis field may not be fully considered, which may result in the final wind field analysis field failing to reach the optimal state. This is especially true in regions with complex or sparse observation data, where such fixed weights may not accurately reflect the actual contribution of observations to the analysis field.

[0100] Referring to Figure 6, this application further proposes a method for simultaneously optimizing the wind field analysis field and dynamic weight distribution by minimizing the cost function, specifically including the following steps:

[0101] First, the dynamic weight distribution is set as the first optimizable variable, and the wind field analysis field is set as the second optimizable variable. In traditional variational assimilation frameworks, the dynamic weight distribution is usually treated as a pre-calculated or fixed parameter, while the wind field analysis field is the primary optimization objective. By elevating the dynamic weight distribution to an optimizable variable, this method allows the system to adjust not only the wind field itself during optimization but also the contribution and spatial distribution of observed data to the wind field, thereby achieving a more refined assimilation effect that better reflects the actual physical process. The first variable represents the spatial distribution and intensity of the influence of observed data on the analysis field, while the second variable is the physical quantity ultimately output by the assimilation system. This setup enables the system to respond more flexibly to complex observation environments and atmospheric conditions.

[0102] Secondly, an extended cost function is constructed, incorporating a composite observation error covariance matrix and a dynamic weight distribution regularization constraint term. This extended cost function is the core of the variational assimilation system, aiming to minimize the difference between the analysis field and the observed values, while simultaneously considering background field information and various errors. Based on this, this application introduces two key components: the composite observation error covariance matrix and the dynamic weight distribution regularization constraint term. The composite observation error covariance matrix, resulting from the inclusion of the dynamic weight distribution in the cost function, reflects the spatial non-uniformity of the observation errors. The dynamic weight distribution regularization constraint term is used to impose restrictions on the optimization process of the first variable (dynamic weight distribution), preventing unreasonable fluctuations or overfitting. For example, smoothing constraints, sparsity constraints, or constraints based on physical prior knowledge can be used to ensure that the optimized dynamic weight distribution has physical meaning and stability, avoiding unreasonable weight distributions during the optimization process, thereby guaranteeing the physical rationality of the assimilation results.

[0103] Furthermore, in the extended cost function, the first and second variables are coupled through a composite observation error covariance matrix. This coupling mechanism is crucial for achieving synchronous optimization. Specifically, the first variable (dynamic weight distribution) influences the optimization process of the second variable (wind field analysis field) by affecting the composite observation error covariance matrix. The elements of the composite observation error covariance matrix directly depend on the dynamic weight distribution, which in turn determines the degree of influence of the observation data on the wind field analysis field. When the dynamic weight distribution changes, the composite observation error covariance matrix changes accordingly, thereby altering the gradient contribution of the observation terms in the cost function to the wind field analysis field, thus adjusting the update direction and magnitude of the wind field analysis field. This interdependent and influential relationship ensures that the wind field analysis field and the observed dynamic weight distribution can coordinate and evolve together during the optimization process, thereby achieving global optimum.

[0104] Subsequently, an iterative optimization algorithm is used to solve the extended cost function. In each iteration, the first and second variables are updated simultaneously, and the composite observation error covariance matrix is ​​recalculated using the updated first variable. Iterative optimization algorithms, such as the conjugate gradient method or the quasi-Newton method, are commonly used to solve complex nonlinear optimization problems. In each iteration, the algorithm calculates the gradient of the cost function based on the current variable values ​​and updates the variables along the negative direction of the gradient. The key here is "simultaneously updating the first and second variables," which means that in each iteration, the dynamic weight distribution and the wind field analysis field are adjusted according to the current cost function gradient. Furthermore, "recalculating the composite observation error covariance matrix using the updated first variable" is a crucial step in achieving coupling, ensuring that in each iteration, the observation error covariance matrix accurately reflects the impact of the current dynamic weight distribution on the observation error, thus providing accurate weight information for the next update of the wind field analysis field.

[0105] The iteration stops when the updates of the first and second variables tend to stabilize between consecutive iterations. The iteration stopping condition is a crucial indicator of the convergence of the optimization algorithm. When, in two consecutive iterations, the updates of the first variable (dynamic weight distribution) and the second variable (wind field analysis field) (e.g., the change in the L2 norm of the variable values) are both less than a preset convergence threshold, it indicates that the algorithm is close to the optimal solution, and the iteration can be stopped. This ensures the efficiency of the optimization process and the stability of the results, avoiding unnecessary consumption of computational resources.

[0106] Finally, the second variable obtained from the final iteration is used as the output of the optimized wind field analysis field. After the above iterative optimization process, when the algorithm converges, the final second variable is the wind field analysis field after synchronous optimization.

[0107] Based on the above embodiments, this application further proposes an optimization scheme after outputting the optimized wind field analysis field, including: extracting and standardizing the features of the final dynamic weight distribution generated in this assimilation process; associating and storing the extracted features with the corresponding weather situation type, seasonal information and terrain classification labels to form a weight pattern knowledge base; in subsequent assimilation tasks, prioritizing the retrieval of historical patterns similar to the current conditions from the weight pattern knowledge base as a reference for constructing the initial three-dimensional spatial weight distribution function.

[0108] Through the above technical solution, this application constructs a self-learning and experience reuse mechanism. Each successful assimilation process not only outputs an optimized wind field analysis field but also stores the characteristics of the dynamic weight distribution and associates them with the corresponding environmental conditions. This enables the system to avoid performing complex weight distribution calculations and optimizations from scratch in subsequent assimilation tasks. Instead, it can quickly retrieve and utilize historical experience from the weight pattern knowledge base to provide optimized initial weight distribution estimates for the current assimilation task, thereby significantly improving the initialization efficiency and convergence speed of the assimilation process. Simultaneously, by continuously accumulating and learning weight patterns under different meteorological conditions and terrain features, the assimilation system can better adapt to complex and changing environments, further improving the accuracy and stability of the wind field analysis field and providing more reliable and efficient basic data for numerical weather prediction.

[0109] To implement the above method, the application further proposes a laser wind measurement radar that integrates a three-dimensional variational assimilation system for executing the above method. The laser wind measurement radar includes: a data acquisition module, an initial weight function construction module, a dynamic weight generation module, a variational assimilation optimization module, and a result output module.

[0110] The data acquisition module is configured to acquire laser wind radar observation data, corresponding atmospheric stability parameters, and terrain undulation feature data of the target area.

[0111] The initial weight function construction module is configured to construct a three-dimensional spatial weight distribution function based on atmospheric stability parameters and terrain undulation feature data, in order to characterize the initial influence weight of a single laser wind radar observation point on its surrounding three-dimensional grid points.

[0112] The dynamic weight generation module is configured to identify complex terrain areas based on terrain undulation feature data, and dynamically adjust the horizontal influence range and vertical attenuation characteristics of the three-dimensional spatial weight distribution function in combination with the vertical stratification change information reflected by atmospheric stability parameters, so as to generate a dynamic weight distribution that matches the local atmospheric physical processes and terrain features.

[0113] The variational assimilation optimization module is configured to introduce the dynamic weight distribution into the cost function of the three-dimensional variational assimilation system as a spatial adjustment factor for the observation error covariance matrix, and simultaneously optimize the wind field analysis field and the dynamic weight distribution by minimizing the cost function.

[0114] The results output module is configured to output an optimized wind field analysis field for use in numerical weather prediction.

[0115] Obviously, the above embodiments are merely illustrative examples for clear explanation and are not intended to limit the implementation. Those skilled in the art will recognize that other variations or modifications can be made based on the above description. It is neither necessary nor possible to exhaustively list all possible implementations here. However, obvious variations or modifications derived therefrom are still within the scope of protection of this invention.

Claims

1. A data processing and assimilation method for a laser wind-measuring radar, characterized in that, Includes the following steps: Acquire laser wind radar observation data, corresponding atmospheric stability parameters, and terrain undulation feature data of the target area for assimilation; based on the atmospheric stability parameters and terrain undulation feature data, construct a three-dimensional spatial weight distribution function to characterize the initial influence weight of a single laser wind radar observation point on its surrounding three-dimensional grid points; including: standardizing the atmospheric stability parameters and terrain undulation feature data, uniformly converting parameters of different dimensions and ranges to a preset dimensionless interval to obtain standardized stability parameters and standardized terrain parameters; based on the principle of meteorological fluid dynamics, define a basic influence template: a three-dimensional spatial grid, where each grid point stores an initial scalar value, which is jointly determined by the preset horizontal distance function and vertical distance function from the grid point to the central observation point, characterizing the theoretical attenuation in an ideal homogeneous medium; The standardized stability parameter is mapped as a vertical adjustment coefficient to modulate the decay rate of the vertical distance function in the basic influence template. Simultaneously, the standardized topographic parameter is mapped as an anisotropic adjustment field in the horizontal direction to modulate the decay rate of the horizontal distance function in each direction. The modulated vertical and horizontal distance functions are combined in three-dimensional space to form a three-dimensional spatial weight distribution function, the output of which is the initial influence weight for the corresponding three-dimensional grid point. Based on topographic relief data, complex terrain areas are identified, and combined with the vertical stratification change information reflected by atmospheric stability parameters, the horizontal influence range and vertical decay characteristics of the three-dimensional spatial weight distribution function are dynamically adjusted to generate a dynamic weight distribution that matches local atmospheric physical processes and topographic features. The dynamic weight distribution is introduced into the cost function of the three-dimensional variational assimilation system as a spatial adjustment factor for the observation error covariance matrix. By minimizing the cost function, the wind field analysis field and the dynamic weight distribution are simultaneously optimized. The optimized wind field analysis field is output for numerical weather prediction.

2. The data processing and assimilation method for laser wind radar according to claim 1, characterized in that: Identifying complex terrain areas based on terrain undulation feature data includes: calculating the slope value and terrain undulation degree of all points within the target area based on the terrain undulation feature data; sorting the slope value and terrain undulation degree from largest to smallest; determining the high-slope subset in the slope value sorting and the high-undulation degree subset in the terrain undulation degree sorting according to a preset complex area proportion coefficient; identifying points that are simultaneously located in both the high-slope subset and the high-undulation degree subset as complex terrain points; and performing spatial connectivity analysis on the complex terrain points to determine the area formed by interconnected complex terrain points as a complex terrain area.

3. The data processing and assimilation method for laser wind radar according to claim 2, characterized in that: Based on complex terrain areas, the horizontal influence range of the three-dimensional spatial weight distribution function is dynamically adjusted, including: extracting the main terrain orientation and main slope direction at the current laser wind radar observation point within the complex terrain area; establishing a local horizontal influence ellipse model centered on the observation point, with the major axis aligned with the main terrain orientation and the minor axis aligned with the main slope direction; dynamically adjusting the eccentricity of the ellipse according to the slope: the greater the slope, the greater the eccentricity of the ellipse, so that the horizontal influence range of the weight value represented by the three-dimensional spatial weight distribution function on the main slope direction shrinks more compared to the horizontal influence range on the main terrain orientation; and using the boundary defined by the adjusted local horizontal influence ellipse model as the new effective boundary of the three-dimensional spatial weight distribution function in the horizontal direction.

4. The data processing and assimilation method for laser wind radar according to claim 1, characterized in that: The vertical stratification change information reflected by atmospheric stability parameters is extracted in the following ways: obtain atmospheric stability parameters, including at least the vertical gradient of potential temperature or Richardson number; calculate an index to characterize the vertical mixing capacity of the atmosphere based on the atmospheric stability parameters; and classify the vertical stratification state of the atmosphere into stable, neutral, and unstable states according to the numerical range of the index.

5. The data processing and assimilation method for laser wind radar according to claim 4, characterized in that: The vertical attenuation characteristics of the three-dimensional spatial weight distribution function are dynamically adjusted through a vertical attenuation scale modulation step, including: determining a vertical attenuation scale modulation factor based on the extracted atmospheric vertical stratification state; if the atmospheric vertical stratification state is stable, applying a vertical attenuation scale modulation factor less than 1 to shorten the vertical attenuation scale of the three-dimensional spatial weight distribution function; if the atmospheric vertical stratification state is unstable, applying a vertical attenuation scale modulation factor greater than 1 to lengthen the vertical attenuation scale of the three-dimensional spatial weight distribution function; if the atmospheric vertical stratification state is neutral, applying a vertical attenuation scale modulation factor equal to 1 to keep the vertical attenuation scale of the three-dimensional spatial weight distribution function unchanged.

6. The data processing and assimilation method for laser wind radar according to claim 1, characterized in that: Introducing the dynamic weight distribution into the cost function of the three-dimensional variational assimilation system as a spatial adjustment factor for the observation error covariance matrix includes: normalizing the values ​​of the dynamic weight distribution to form a spatial adjustment factor field; performing a correlation operation between the spatial adjustment factor field and the observation error covariance matrix of the three-dimensional variational assimilation system according to spatial location to generate a composite observation error covariance matrix; and replacing the original observation error covariance matrix with the composite observation error covariance matrix and substituting it into the cost function of the three-dimensional variational assimilation system.

7. The data processing and assimilation method for laser wind radar according to claim 6, characterized in that: By minimizing the cost function, the wind field analysis field and the dynamic weight distribution are simultaneously optimized, including: setting the dynamic weight distribution as the first optimizable variable and the wind field analysis field as the second optimizable variable; constructing an extended cost function that includes a composite observation error covariance matrix and a dynamic weight distribution regularization constraint term; in the extended cost function, the first and second variables are coupled through the composite observation error covariance matrix; using an iterative optimization algorithm to solve the extended cost function, updating the first and second variables simultaneously in each iteration, and recalculating the composite observation error covariance matrix using the updated first variable; stopping the iteration when the update amounts of the first and second variables tend to stabilize between consecutive iterations; and using the second variable obtained in the final iteration as the output of the optimized wind field analysis field.

8. The data processing and assimilation method for laser wind radar according to claim 1, characterized in that: After outputting the optimized wind field analysis field, the process further includes: extracting and standardizing the dynamic weight distribution generated during this assimilation process; associating and storing the extracted features with the corresponding weather pattern type, seasonal information, and terrain classification labels to form a weight pattern knowledge base; and in subsequent assimilation tasks, prioritizing the retrieval of historical patterns similar to the current conditions from the weight pattern knowledge base as a reference for constructing the initial three-dimensional spatial weight distribution function.

9. A laser wind-measuring radar, integrating a three-dimensional variational assimilation system, for performing the method described in any one of claims 1 to 8, characterized in that: include: The data acquisition module is configured to acquire laser wind radar observation data, corresponding atmospheric stability parameters, and terrain undulation feature data of the target area for assimilation. The initial weight function construction module is configured to construct a three-dimensional spatial weight distribution function based on atmospheric stability parameters and topographic relief feature data to characterize the initial influence weight of a single laser wind radar observation point on its surrounding three-dimensional grid points. The dynamic weight generation module is configured to identify complex terrain areas based on topographic relief feature data and, in conjunction with the vertical stratification change information reflected by atmospheric stability parameters, dynamically adjust the horizontal influence range and vertical attenuation characteristics of the three-dimensional spatial weight distribution function to generate a dynamic weight distribution that matches local atmospheric physical processes and topographic features. The variational assimilation optimization module is configured to introduce the dynamic weight distribution into the cost function of the three-dimensional variational assimilation system as a spatial adjustment factor for the observation error covariance matrix, and simultaneously optimize the wind field analysis field and the dynamic weight distribution by minimizing the cost function. The results output module is configured to output an optimized wind field analysis field for use in numerical weather prediction.

Images