Method for extracting characteristics of ocean low-level waveguide based on meteorological gradient

Abstract

This invention discloses a method for extracting oceanic low-altitude waveguide features based on meteorological gradients, relating to the fields of marine atmospheric environment detection and radio wave propagation technology. This method constructs a standardized multi-scale meteorological gradient dataset through data preprocessing, and then uses a gradient fusion algorithm to achieve global optimization of the physical constraint fusion factor of the mesoscale meteorological gradient and local modeling of small-scale non-uniform features, forming a high-resolution feature field. Based on this, combined with radiosonde data, a waveguide comprehensive inversion algorithm is used to invert and obtain a full-space waveguide feature parameter set. Finally, the coefficients are adaptively optimized through closed-loop verification using independent measured data. This invention achieves high-resolution integration of mesoscale global constraints and small-scale local features, improving the accuracy of oceanic low-altitude atmospheric environment characterization. Relying on the waveguide comprehensive inversion and closed-loop optimization mechanism, it integrates measured data and multi-dimensional verification, adaptively optimizing core coefficients, enhancing the reliability of waveguide feature extraction and its adaptability to complex ocean-atmosphere environments.

Description

Technical Field

[0001] This invention relates to the field of marine atmospheric environment detection and radio wave propagation technology, specifically a method for extracting marine low-altitude waveguide features based on meteorological gradients. Background Technology

[0002] Ocean low-altitude waveguides are a typical and unique refraction phenomenon in the marine atmospheric environment. Their formation and development are closely related to the meteorological gradient characteristics in the marine atmosphere. The spatiotemporal variations of the meteorological gradient directly determine the structure, strength, and distribution patterns of the waveguide. This phenomenon has a crucial impact on the propagation path and efficiency of low-altitude electromagnetic waves in the ocean. Electromagnetic wave propagation is the core foundation for technologies such as marine communication, radar detection, and maritime early warning. Therefore, accurately extracting the characteristics of ocean low-altitude waveguides has become an important topic in marine atmospheric environment research and marine engineering technology applications. With the development of marine observation technology, a multi-source data acquisition system has been formed for monitoring marine meteorological gradients. Numerical model gridded data, station radiosonde data, and near-sea surface in-situ monitoring data can reflect the characteristics of marine meteorological gradients from different dimensions. At the same time, the marine region is affected by multiple factors such as air-sea interaction, ocean currents, and monsoon changes, and the meteorological gradient exhibits significant multi-scale characteristics. The global distribution patterns at the mesoscale and the local non-uniform fluctuations at the small scale are coupled with each other, making it necessary to consider the analysis and fusion of multi-scale characteristics when extracting waveguide features based on meteorological gradients.

[0003] Traditional methods for extracting features from low-altitude ocean waveguides have many limitations in processing meteorological gradient data and analyzing features. They are ill-suited to the multi-scale characteristics of ocean meteorological gradients and the need for refined extraction of waveguide features. These methods often rely on a single type of meteorological gradient data for analysis, failing to effectively integrate complementary information from multiple data sources. This results in an incomplete representation of meteorological gradient features. Furthermore, traditional methods lack targeted analysis of the multi-scale features of meteorological gradients, either focusing only on the global meteorological distribution at the mesoscale and ignoring the impact of small-scale local meteorological gradient fluctuations on the local waveguide structure, or being limited to the small scale. The single-point measured data cannot reflect the spatial global distribution law of waveguide characteristics. In addition, traditional methods do not fully integrate the key physical constraints of waveguide formation when calculating meteorological gradient correlation factors, nor do they scientifically unify the dimensional differences of different meteorological gradient characteristics. This results in insufficient fit between the factor calculation and the physical laws of the actual sea-atmosphere environment. The subsequent waveguide inversion algorithm also lacks the coordinated adaptation of mesoscale and small-scale features. The accuracy and spatial resolution of the inversion results are difficult to meet the needs of practical applications. Furthermore, the lack of a closed-loop verification and optimization mechanism further reduces the adaptability of the inversion results to different sea-atmosphere environments. Summary of the Invention

[0004] The purpose of this invention is to overcome the shortcomings of existing technologies and provide a method for extracting features of marine low-altitude waveguides based on meteorological gradients. This method constructs a standardized multi-scale meteorological gradient dataset through multi-source data preprocessing; it then uses a gradient fusion algorithm to successively achieve global optimization of the mesoscale physical constraint fusion factor and local modeling of small-scale non-uniform features, generating a high-resolution feature field; combined with waveguide comprehensive inversion algorithm and radiosonde measurement data, it inverts to obtain a full-space feature parameter set including waveguide type, intensity, height, and other parameters; finally, it performs closed-loop verification and adaptive coefficient optimization using independent measurement data.

[0005] To solve the above-mentioned technical problems, this invention provides the following technical solution: a method for extracting features of ocean low-altitude waveguides based on meteorological gradients, the specific steps of which are as follows:

[0006] S100. Data Preprocessing: Define the spatial range and study period of the target sea area, collect multi-source marine meteorological gradient related data, complete the spatiotemporal registration, cleaning, feature extraction and standardization of the data, divide the dataset and output a standardized multi-scale meteorological gradient dataset.

[0007] S200, Mesoscale Optimization: Using a standardized multi-scale meteorological gradient dataset as input, the mesoscale meteorological gradient physical constraint fusion factor is calculated based on the gradient fusion algorithm to form an initial feature field and complete grid-by-grid calibration, and outputs the corrected mesoscale meteorological gradient physical constraint fusion factor feature field.

[0008] S300, Small-scale modeling: Using the modified mesoscale meteorological gradient physical constraint fusion factor feature field as the boundary constraint, the meteorological gradient physical constraint fusion factor adapted to small-scale non-uniform features is calculated based on the gradient fusion algorithm, spatial interpolation is performed, and a high-resolution small-scale meteorological gradient physical constraint fusion factor feature field is output.

[0009] S400 Waveguide Inversion: Using the physical constraint fusion factor feature field of mesoscale and small-scale meteorological gradient and high-precision radiosonde data as input, the waveguide comprehensive inversion algorithm is used to calculate the comprehensive inversion value of marine low-altitude waveguide features, form the inversion value feature field and complete grid-by-grid analysis, and output the full-space feature parameter set of marine low-altitude waveguides.

[0010] S500, Closed-loop optimization: Collect independent measured verification data, conduct multi-dimensional accuracy verification of the full-space characteristic parameter set of ocean low-altitude waveguides, optimize the adaptive coefficients of gradient fusion algorithm and waveguide integrated inversion algorithm based on the verification deviation results, repeat the mesoscale optimization, small-scale modeling, and waveguide inversion steps and verify again until the coefficient optimization is completed.

[0011] Furthermore, the target sea area is defined as a marine region within a specified latitude and longitude range based on actual research needs. The research period is a complete time series of at least one year, covering different meteorological characteristic cycles of the four seasons to ensure the spatiotemporal representativeness of the data. The multi-source marine meteorological gradient related data specifically includes: mesoscale numerical model gridded data acquired through a meteorological data platform, covering 37 vertically layered data points from sea surface to 30km altitude, including temperature, air pressure, relative humidity, wind speed, sea surface temperature, salinity, ocean current velocity, waveguide interface tilt angle, and Richardson number of the ocean-atmospheric boundary layer, with a spatial resolution of 25km × 100km. 25km, with a time resolution of 1h; Measured data from fixed radiosonde stations are collected daily at 08:00 and 20:00 using radiosonde equipment at different levels from the sea surface to 30km altitude, including temperature, pressure, humidity, wind speed, and wind direction profiles, as well as the tilt angle of the waveguide interface and Richardson number around the station; In-situ data from near-sea surface meteorological gradient instruments are collected at a frequency of 10min / time by sensors deployed in at least 10 layers within a height of 40m above the sea surface, including temperature, humidity, and air pressure data for each layer. All data undergoes spatiotemporal registration, cleaning, feature extraction, and standardization, and the dataset is divided and a standardized multi-scale meteorological gradient dataset is output.

[0012] Furthermore, the feature extraction process involves extracting four core meteorological gradient features: comprehensive temperature gradient, comprehensive humidity gradient, vertical pressure gradient, and horizontal wind speed gradient. Simultaneously, it extracts the waveguide interface tilt angle and the Richardson number of the ocean-atmospheric boundary layer. The comprehensive temperature gradient integrates the vertical temperature gradient from the sea surface to the waveguide critical height and the horizontal gradient of the sea-air temperature difference; the comprehensive humidity gradient integrates the vertical relative humidity gradient and the horizontal water vapor advection gradient; the vertical pressure gradient is taken as the rate of vertical pressure change from the sea surface to a height of 30 km; and the horizontal wind speed gradient represents the horizontal wind speed changes caused by ocean currents and monsoons. Two types of physical constraint features, the waveguide interface tilt angle and the Richardson number of the ocean-atmospheric boundary layer, are also extracted simultaneously. Standardization processing is then performed. The temperature gradient, humidity gradient, vertical pressure gradient, and horizontal wind speed gradient were mapped to the [0,1] interval using the min-max normalization method. The upper and lower limits of the normalization were taken as the extreme values ​​of the same type of meteorological gradient data in the target sea area in the past 5 years. The waveguide interface tilt angle was linearly scaled to [0,1] in the [-90°, 90°] interval. The Richardson number of the ocean-atmospheric boundary layer was standardized and scaled in the [0,10] interval to eliminate the dimensional differences of different features. Dataset partitioning: The standardized data were divided into training set, validation set and test set in a 7:2:1 ratio according to the time series dimension. The spatiotemporal distribution of each subset was ensured to be uniform during partitioning. The radiosonde measurement data were archived separately.

[0013] Furthermore, the mathematical expression for the gradient fusion algorithm is:

[0014] ;

[0015] in, ∇T is the meteorological gradient physical constraint fusion factor, with a value range of [−10,5]. Negative values ​​represent the effective range for waveguide formation. The larger the absolute value of the negative value, the easier it is for the meteorological gradient conditions to form a waveguide. ∇T is the temperature comprehensive gradient, in ℃ / km. It is the normalized value of the fusion of the vertical temperature gradient and the horizontal gradient of the sea-air temperature difference, representing the core influence of the temperature field on the atmospheric refractive index. ∇H is the humidity comprehensive gradient, in % / km. It is the normalized value of the fusion of the vertical relative humidity gradient and the horizontal gradient of water vapor advection. ∇V is the vertical pressure gradient, in hPa / km, which is the normalized value of the rate of change of vertical pressure from sea surface to 30km altitude; ∇V is the horizontal wind speed gradient, in m / (s·km), which characterizes the influence of horizontal wind speed changes caused by ocean currents and monsoons on waveguide inhomogeneity. , , , The contribution weighting coefficients for each meteorological gradient are adaptively adjusted according to the marine climate characteristics of the target sea area, and the sum of the weighting coefficients is 1. The waveguide interface tilt angle correction term is expressed as follows: =1−0.01∣ |, The angle of inclination between the waveguide interface and the horizontal plane, with a value range of [−90, 90]; The physical constraint term for atmospheric stability is the Ricci number. The constraint function, when <5 o'clock =Ri / 5, when ≥5 =1, Richardson number for the ocean-atmosphere boundary layer, characterizing the stability of ocean-atmosphere turbulence.

[0016] Furthermore, the small-scale non-uniformity features refer to the local variations in meteorological gradients and waveguide structures within a spatial scale of less than ten kilometers that cannot be covered by the mesoscale grid in the target sea area. These include subtle fluctuations in the vertical gradients of temperature, humidity, and pressure captured by near-surface meteorological gradiometers, local deflections of the tilt angles of waveguide interfaces in different sub-regions, small-scale differences in the Richardson number of the ocean-atmospheric boundary layer, and local abrupt changes in the horizontal wind speed gradient caused by ocean currents and monsoons. When calculating the meteorological gradient physical constraint fusion factor adapted to the small-scale non-uniformity features, the characteristic field of the mesoscale meteorological gradient physical constraint fusion factor is used as the boundary constraint. Combined with high-resolution measured data from the near-surface, the mesoscale grid is divided into sub-regions. The actual tilt angles and Richardson numbers of the waveguide interfaces in each sub-region are obtained. These are substituted into the gradient fusion algorithm for iterative calculation. The calculation is iterated at least three times for each sub-region, and the average value is taken to obtain the meteorological gradient physical constraint fusion factor adapted to the small-scale non-uniformity features. This makes the fusion factor fit the local meteorological gradient changes in the sub-regions, corrects the lack of description of small-scale non-uniformity by the mesoscale model, and provides an accurate small-scale gradient feature basis for subsequent spatial interpolation.

[0017] Furthermore, the spatial interpolation operation specifically involves: using the modified mesoscale meteorological gradient physical constraint fusion factor feature field as the global boundary constraint, and using the meteorological gradient physical constraint fusion factor adapted to the small-scale non-uniform characteristics of each sub-region as the local constraint, dividing the 25km×25km coarse-resolution mesoscale grid into 9km×9km fine-grained sub-grids to determine the spatial topology of the sub-grids; employing a spatial interpolation method based on Monte Carlo simulation to perform random interpolation, for each fine-grained sub-grid, combining the mesoscale fusion factor of its parent grid with the small-scale fusion factor of the surrounding sub-regions, calculating the fusion factor value of the sub-grid through multiple random samplings, preserving the global distribution law of the mesoscale feature field during the interpolation process, while restoring the subtle changes and non-uniform characteristics of the meteorological gradient in the sub-regions; after the interpolation is completed, boundary verification is performed on the fusion factor values ​​of all fine-grained sub-grids, and the neighborhood mean smoothing method is used to eliminate numerical abrupt changes between adjacent sub-grids to ensure the spatial continuity of the feature field, finally outputting a high-resolution small-scale meteorological gradient physical constraint fusion factor feature field corresponding one-to-one with the fine-grained sub-grids.

[0018] Furthermore, the mathematical expression for the waveguide synthesis inversion algorithm is:

[0019] ;

[0020] in, The value is the comprehensive inversion value of the characteristics of low-altitude ocean waveguides, with a range of (−8, 4), which is a comprehensive quantity characterizing the waveguide type, intensity, and height. The meteorological gradient physical constraint fusion factor calculated by the gradient fusion algorithm; The value range is [0.8, 1.2]. It is adaptively adjusted according to the spatial resolution of the mesoscale numerical model and represents the contribution of the mesoscale global gradient features to waveguide inversion. The small-scale feature adaptation coefficient has a value range of [0.9, 1.1] and is adjusted according to the resolution of the small-scale fine-grained grid. It characterizes the correction effect of small-scale local non-uniform gradient features on waveguide inversion. The calibration coefficient for radiosonde measurements ranges from [0.05, 0.2] and is adjusted based on the spatial density of radiosonde stations in the target sea area. The baseline value of the meteorological gradient measured by radiosonde is the average value of the physical constraint fusion factor of the meteorological gradient measured by radiosonde stations in the target sea area.

[0021] Furthermore, the full-space feature parameter set of the ocean low-altitude waveguide specifically includes the waveguide existence identifier, waveguide type identifier, waveguide intensity level, waveguide top height, waveguide bottom height, waveguide thickness, and waveguide spatial distribution range of each fine-grained grid node, and is also equipped with time-series labels for corresponding time nodes, forming a spatiotemporally integrated feature dataset; wherein the waveguide existence identifier is based on the comprehensive inversion value M of the ocean low-altitude waveguide features. ω Division, M ω When M < 0, it is marked as the presence of a waveguide. ω A value ≥0 is marked as no waveguide; waveguide type is identified by M. ω Interval differentiation, M ω When M >-2 and <0, it is marked as a surface waveguide. ω A waveguide with a strength ≤-2 is designated as a suspended waveguide; the waveguide strength rating is based on |M ω | is the numerical range marker, |M ω |∈(0,3) are marked as weak waveguides, |M ω |∈[3,5) are marked as the middle waveguide, |M ω |≥5 is marked as a strong waveguide; waveguide apex height is defined as H=100−10|M ω |Calculation, unit is m; the bottom height of the surface waveguide is taken as the sea surface height of 0m, and the lower limit of the corresponding height range is taken when the waveguide is suspended; the waveguide thickness is the difference between the top height and the bottom height; the spatial distribution range of the waveguide is marked by the latitude and longitude range of the fine-grained grid, covering the entire space of the target sea area.

[0022] Furthermore, the independent measured verification data includes measured data from a microwave refractometer and measured data from a shore-based radar across three frequency bands. The microwave refractometer collects vertical profile data of atmospheric refractive index from the sea surface to a height of 30 km in the target sea area, with a vertical sampling interval of no more than 100 meters. The temporal resolution is consistent with the inversion results, and it can be directly used to determine the actual waveguide structure parameters. The shore-based radar uses three frequency bands—ultra-shortwave, microwave, and millimeter wave—to conduct simultaneous measurements, collecting data on the over-the-horizon propagation distance, signal strength, and attenuation characteristics of electromagnetic waves in the marine atmospheric environment. The continuous collection time is no less than one month, covering propagation scenarios in different seasons and under different marine atmospheric environments. All measured verification data have undergone spatiotemporal registration and data cleaning. After removing abnormal abrupt changes, the data is matched with the spatiotemporal nodes of the full-space characteristic parameter set of marine low-altitude waveguides, serving as unified benchmark data for waveguide characteristic accuracy verification.

[0023] Compared with existing technologies, this method for extracting ocean low-altitude waveguide features based on meteorological gradients has the following advantages:

[0024] I. This invention constructs a standardized processing system for multi-source marine meteorological gradient data, combines gradient fusion algorithms to construct and calibrate mesoscale meteorological gradient feature fields grid-by-grid, and uses the mesoscale feature fields as boundary constraints to achieve refined modeling and high-resolution feature field output of small-scale non-uniform meteorological gradient features. It opens up a linkage and adaptation path between mesoscale global constraints and small-scale local features, effectively making up for the shortcomings of traditional methods in capturing subtle changes in marine low-altitude meteorological gradients. At the same time, it introduces physical constraint terms to achieve the fusion and optimization of multiple types of meteorological gradient features, ensuring the spatial continuity and physical rationality of the feature fields, significantly improving the characterization accuracy of marine low-altitude atmospheric environment features, providing a stable and high-resolution input foundation for waveguide feature inversion, and solving the industry pain points of poor multi-source data adaptability and scale connection discontinuity in traditional methods.

[0025] II. This invention utilizes a complete technical system of waveguide feature comprehensive inversion and full-process closed-loop optimization. Relying on the waveguide comprehensive inversion algorithm, it achieves the fusion inversion of multi-scale meteorological gradient feature fields and measured high-precision data, and can output a waveguide feature parameter set covering the entire space. This enables the systematic analysis and precise quantification of multi-dimensional waveguide features. At the same time, it introduces independent measured data to carry out multi-dimensional accuracy verification. Based on the verification deviation results, it realizes adaptive optimization and iterative closed-loop of the algorithm's core coefficients, effectively improving the reliability of waveguide feature extraction results and adaptability to complex ocean-atmosphere environments. It breaks through the limitations of traditional inversion methods, such as weak generalization ability and lack of accuracy verification system, and can provide precise technical support for related application scenarios such as ocean over-the-horizon electromagnetic wave propagation and ocean-atmosphere environment monitoring and early warning.

[0026] Other advantages, objectives and features of the invention will be set forth in part in the description which follows, and in part will be apparent to those skilled in the art from the following examination or study, or may be learned from the practice of the invention. Attached Figure Description

[0027] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the accompanying drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are merely some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without any creative effort.

[0028] Figure 1 This is a flowchart of a method for extracting features of marine low-altitude waveguides based on meteorological gradients.

[0029] Figure 2 This is the mathematical expression for a method of extracting features from ocean low-altitude waveguides based on meteorological gradients.

[0030] Figure 3 A schematic diagram of data transmission for modeling small-scale non-uniform features. Detailed Implementation

[0031] To further illustrate the technical means and effects of the present invention in achieving its intended purpose, the following detailed description of the specific implementation methods, structures, features, and effects of the present invention, in conjunction with the accompanying drawings and preferred embodiments, is provided below.

[0032] Example 1:

[0033] Feature extraction of low-altitude waveguides in open ocean areas

[0034] This embodiment is applied to a designated latitude and longitude range in the western Pacific Ocean, conducting high-precision extraction of low-altitude waveguide features across the entire space. This area exhibits significant air-sea interaction, with abrupt changes in wind speed gradients caused by ocean currents and monsoons being the main manifestation of small-scale non-uniformity. The method based on meteorological gradients for low-altitude waveguide feature extraction in this invention is used to accurately acquire waveguide feature parameters, such as... Figure 1 As shown, the specific implementation steps are as follows:

[0035] S100. Data Preprocessing: The western Pacific Ocean was designated as the target sea area within a specified latitude and longitude range. A continuous one-year study period was selected to achieve full coverage of different meteorological characteristic cycles throughout the four seasons, ensuring the spatiotemporal representativeness of the data. Multi-source marine meteorological gradient correlation data for this region were collected. Among them, mesoscale numerical model gridded data were obtained from the meteorological data platform, covering 37 vertical layers from sea surface to 30km altitude, including temperature, air pressure, relative humidity, wind speed, sea surface temperature, salinity, ocean current velocity, waveguide interface tilt angle, and Richardson number of the ocean-atmospheric boundary layer. This type of data is spatially divided. The resolution is 25km×25km, and the time resolution is 1h. The measured data of the fixed radiosonde station is collected by the equipment of the radiosonde station around the sea area. At 08:00 and 20:00 every day, the temperature, pressure, humidity, wind speed and wind direction profile data of different layers from the sea surface to 30km altitude are collected simultaneously, as well as the tilt angle of the waveguide interface and Richardson number around the station. The in-situ data of the near-sea surface meteorological gradient instrument is collected by the sensors deployed in at least 10 layers within a height of 40m above the sea surface. The temperature, humidity and air pressure data of each layer are acquired at a frequency of 10min / time, so as to realize the high frequency and fine collection of near-sea surface meteorological gradient data. Spatiotemporal registration and cleaning were performed on the above three types of multi-source data to eliminate spatiotemporal biases and abnormal interference data between different data sources. Subsequently, four core meteorological gradient features were extracted: comprehensive temperature gradient, comprehensive humidity gradient, vertical pressure gradient, and horizontal wind speed gradient. Simultaneously, two types of physical constraint features were extracted: waveguide interface tilt angle and Richardson number of the ocean-atmospheric boundary layer. Focusing on key meteorological influencing factors of waveguide formation, standardization processing was then carried out. The four core meteorological gradient features were mapped to the [0,1] interval using the min-max normalization method. The upper and lower limits of normalization were taken as the extreme values ​​of similar meteorological gradient data from the past 5 years for the target offshore sea area. The tilt angle of the guide interface is linearly scaled to [0,1] in the interval [-90°, 90°], and the Richardson number of the ocean-atmospheric boundary layer is standardized and scaled in the interval [0,10] to eliminate the dimensional differences of different features and ensure the uniformity of subsequent algorithm calculations. Finally, the standardized data is divided into training, validation and test sets in a 7:2:1 ratio according to the time series dimension, ensuring that the spatiotemporal distribution of each subset is uniform. At the same time, the radiosonde measurement data is archived separately to provide high-precision measurement reference for subsequent waveguide inversion. A standardized multi-scale meteorological gradient dataset is output to provide standardized and unified input data for subsequent mesoscale optimization.

[0036] S200, Mesoscale Optimization: Using the standardized multi-scale meteorological gradient dataset output above as input data, the physical constraint fusion factor of the mesoscale meteorological gradient for this offshore area is calculated using the gradient fusion algorithm. The mathematical expression of the gradient fusion algorithm is:

[0037] ;

[0038] in, The fusion factor for physical constraints of meteorological gradients; For the overall temperature gradient; The overall humidity gradient; This represents the vertical pressure gradient; This represents the horizontal gradient of wind speed. , , , The contribution weighting coefficients for each meteorological gradient; This is a correction term for the tilt angle of the waveguide interface; As a physical constraint term for atmospheric stability, this method integrates multi-dimensional meteorological gradient characteristics with physical constraints, conforming to the laws of mesoscale air-sea interaction in the open ocean. Based on the calculation results, a mesoscale initial feature field is formed. Subsequently, a grid-by-grid calibration operation is performed on the initial feature field to correct the systematic bias of the mesoscale numerical model data in the open ocean, making the mesoscale feature field more consistent with the actual meteorological gradient distribution in the open ocean. Finally, the corrected mesoscale meteorological gradient physical constraint fusion factor feature field is output, laying a solid global boundary foundation for subsequent small-scale modeling.

[0039] S300, Small-scale modeling: Using the modified mesoscale meteorological gradient physical constraint fusion factor feature field as the boundary constraint, and combining the local abrupt changes in wind speed horizontal gradients caused by ocean currents and monsoons captured in the high-resolution measured data of the nearshore surface of this offshore sea area, the gradient fusion algorithm is used to calculate the meteorological gradient physical constraint fusion factor adapted to this type of small-scale non-uniform characteristics, accurately capturing the non-uniform variation law of local meteorological gradients in the offshore sea. In the calculation process, the mesoscale 25km×25km grid is divided into several sub-regions. The actual tilt angle and Richardson number of the waveguide interface of each sub-region are obtained and then entered into the gradient fusion algorithm for iterative calculation. Each sub-region is iterated at least 3 times and the average value is taken to reduce the random error of a single calculation result and ensure that the fusion factor fits the local meteorological gradient changes of the sub-region. After calculating the small-scale fusion factor, spatial interpolation is performed. Using the corrected mesoscale feature field as the global boundary constraint and the small-scale fusion factor of each sub-region as the local constraint, the coarse-resolution 25km×25km mesoscale grid is divided into 9km×9km fine-grained sub-grids, and their spatial topology is determined to improve the spatial resolution of the feature field. Monte Carlo simulation is used for random interpolation. For each fine-grained sub-grid, the fusion factor value is calculated through multiple random samplings, combining the mesoscale fusion factor of its parent grid and the small-scale fusion factor of the surrounding sub-regions. This ensures that the interpolation results simultaneously preserve both global distribution patterns and local non-uniform characteristics. After interpolation, boundary verification is performed on the fusion factor values ​​of all fine-grained sub-grids. Neighborhood mean smoothing is used to eliminate numerical abrupt changes between adjacent sub-grids, ensuring the spatial continuity of the feature field and avoiding unreasonable numerical jumps at grid boundaries. Finally, a high-resolution small-scale meteorological gradient physical constraint fusion factor feature field is output, providing refined meteorological gradient feature support for waveguide inversion. Figure 3 As shown.

[0040] S400 Waveguide Inversion: The corrected mesoscale meteorological gradient physical constraint fusion factor feature field, the high-resolution small-scale meteorological gradient physical constraint fusion factor feature field, and high-precision radiosonde data are used as input data. Mesoscale global features, small-scale local features, and high-precision measured data are integrated to improve the accuracy of waveguide inversion. The comprehensive inversion value of the marine low-altitude waveguide features for the distant target sea area is calculated using the waveguide comprehensive inversion algorithm. The mathematical expression of the waveguide comprehensive inversion algorithm is as follows:

[0041] ;

[0042] in, This is a comprehensive inversion value of the characteristics of low-altitude waveguides in the ocean; The fusion factor for physical constraints of meteorological gradients; The fitting coefficients are for mesoscale features; For small-scale feature adaptation coefficients; The calibration coefficient is the result of actual sounding measurements. To establish a benchmark value for the meteorological gradient measured by sounding, and to achieve accurate conversion of multi-scale meteorological gradient features into waveguide features, an inversion value feature field is formed based on the inversion values, and grid-by-grid analysis is performed to achieve refined extraction of waveguide features across the entire open ocean. After analysis, a full-space feature parameter set of marine low-altitude waveguides for the region is output. This parameter set includes waveguide presence identifiers, waveguide type identifiers, waveguide intensity levels, waveguide top height, waveguide bottom height, waveguide thickness, and waveguide spatial distribution range for each 9km×9km fine-grained grid node. It also includes time-series labels for corresponding time nodes, forming a spatiotemporally integrated feature dataset that clearly reconstructs the spatiotemporal evolution of waveguide features. The waveguide presence identifier is determined based on the comprehensive inversion values. A value less than 0 indicates the presence of a waveguide, while an inversion value greater than or equal to 0 indicates the absence of a waveguide. Waveguide type is distinguished by the range of inversion values: values ​​greater than -2 and less than 0 are marked as surface waveguides, and values ​​less than or equal to -2 are marked as suspended waveguides. Waveguide strength is marked according to the numerical range of the absolute value of the inversion value: values ​​in (0,3) indicate weak waveguides, values ​​in [3,5) indicate medium waveguides, and values ​​greater than or equal to 5 indicate strong waveguides. The top height of the waveguide is obtained according to the corresponding calculation rules. The bottom height of a surface waveguide is taken as the sea surface height of 0m, and the bottom height of a suspended waveguide is taken as the lower limit of the corresponding height range. The waveguide thickness is the difference between the top height and the bottom height. The spatial distribution range of the waveguide is marked by the latitude and longitude range of a fine-grained grid, realizing the quantitative and spatial characterization of waveguide features.

[0043] S500, Closed-Loop Optimization: Independent field-measured verification data are collected for the target sea area in the open ocean, including data measured by a microwave refractometer and data measured by a shore-based radar in three frequency bands. The microwave refractometer collects vertical profile data of atmospheric refractive index from the sea surface to a height of 30 km, with a vertical sampling interval of no more than 100 meters. The time resolution is consistent with the waveguide inversion results, which can directly and accurately determine the actual waveguide structure parameters. The shore-based radar uses three frequency bands of ultra-short wave, microwave, and millimeter wave to simultaneously conduct field measurements, collecting data on the over-the-horizon propagation distance, signal strength, and attenuation characteristics of electromagnetic waves in each frequency band in the open ocean environment. The continuous collection time is no less than one month, covering propagation scenarios in different seasons and different ocean environments, so that the verification data has comprehensive scenario representativeness. All independent measured verification data underwent spatiotemporal registration and data cleaning. After removing anomalous abrupt changes, the data was precisely matched with the spatiotemporal nodes of the full-space feature parameter set of marine low-altitude waveguides. This served as the unified benchmark data for waveguide feature accuracy verification. Based on this benchmark data, multi-dimensional accuracy verification of the full-space feature parameter set of waveguides was carried out. The accuracy of extracting waveguide existence, type, intensity, and geometric parameters was comprehensively tested. Based on the deviation results obtained from the verification, the adaptive coefficients of the gradient fusion algorithm and the waveguide comprehensive inversion algorithm were optimized to make the algorithm more suitable for the air-sea environment characteristics of open ocean areas. Subsequently, the steps of S200 mesoscale optimization, S300 small-scale modeling, and S400 waveguide inversion were repeated, and the accuracy was verified again using the benchmark data. The algorithm was continuously iterated and optimized until the optimal adjustment of the adaptive coefficients was achieved, resulting in high-precision marine low-altitude waveguide feature parameters for this open ocean area, meeting the accuracy requirements of open ocean waveguide feature applications.

[0044] This embodiment addresses the need for low-altitude waveguide feature extraction in the open waters of the western Pacific Ocean. It strictly follows a complete process: data preprocessing, mesoscale optimization, small-scale modeling, waveguide inversion, and closed-loop optimization. Each step is specifically addressed to address the significant air-sea interaction and the tendency for local abrupt changes in wind speed gradients in the open ocean. A gradient fusion algorithm is used to accurately calculate multi-scale meteorological gradient fusion factors. A waveguide comprehensive inversion algorithm is employed to convert multi-source features into waveguide features. Through iterative verification and algorithm coefficient adjustment in closed-loop optimization, systematic biases in the numerical model are effectively corrected, improving the spatial resolution and continuity of the feature field. The final output is a high-resolution spatiotemporal integrated waveguide feature parameter set that accurately captures waveguide variations caused by small-scale non-uniform features in the open ocean. The extraction results closely match the actual air-sea environment in the open ocean, providing high-precision marine atmospheric environmental data support for fields such as offshore radar detection and maritime communication.

[0045] Example 2:

[0046] Feature extraction of low-altitude waveguides in complex nearshore waters

[0047] This embodiment is applied to a complex nearshore target sea area within a specified latitude and longitude range in the Yellow Sea. It performs high-precision full-space extraction of marine low-altitude waveguide features in this area. This sea area is significantly affected by nearshore topography and sea-land breeze interaction. Subtle fluctuations in the vertical gradient of nearshore surface temperature, humidity, and pressure, and local deflection of the waveguide interface tilt angle are the main manifestations of small-scale non-uniform features. Based on the meteorological gradient-based marine low-altitude waveguide feature extraction method of this invention, the precise acquisition of waveguide feature parameters is achieved, such as... Figure 2 As shown, the specific implementation steps are as follows:

[0048] S100. Data Preprocessing: The nearshore area of ​​the Yellow Sea was designated as the target sea area within a specified latitude and longitude range. A continuous study period of no less than one year was selected to achieve full coverage of different meteorological characteristic cycles in all four seasons, taking into account the seasonal variation patterns of nearshore sea and land breezes, ensuring the spatiotemporal representativeness of the data. Multi-source marine meteorological gradient correlation data for this region were collected. Among them, the mesoscale numerical model gridded data was obtained from the meteorological data platform, acquiring 37 vertically layered data on temperature, air pressure, relative humidity, wind speed, sea surface temperature, salinity, ocean current velocity, waveguide interface tilt angle, and Richardson number of the ocean-atmospheric boundary layer within the range from sea surface to 30km altitude. The spatial resolution of this type of data is [missing information]. The measurement range is 25km × 25km with a time resolution of 1 hour. The measured data from the fixed radiosonde station are collected by the equipment at the coastal radiosonde station. At 08:00 and 20:00 every day, the temperature, pressure, humidity, wind speed and wind direction profiles of different layers from the sea surface to a height of 30km are collected simultaneously, as well as the tilt angle of the waveguide interface and Richardson number around the station. The in-situ data of the near-shore meteorological gradient instrument are collected by sensors deployed in at least 10 layers within a height of 40m above the near-shore sea surface. The temperature, humidity and air pressure data of each layer are acquired at a frequency of 10 minutes / time. The focus is on capturing the subtle changes in temperature, pressure and humidity data under the influence of near-shore topography, so that the in-situ data are more consistent with the characteristics of the small-scale meteorological gradient in the near-shore area. Spatiotemporal registration and cleaning were performed on the above three types of multi-source data to eliminate abnormal data caused by spatiotemporal bias and equipment interference in the land-sea boundary area. Then, four core meteorological gradient features were extracted: temperature gradient, humidity gradient, vertical pressure gradient, and horizontal wind speed gradient. Simultaneously, two types of physical constraint features were extracted: waveguide interface tilt angle and Richardson number of the ocean-atmospheric boundary layer. Meteorological gradient factors that play a crucial role in nearshore waveguide formation were screened out. Standardization was then performed, and the four core meteorological gradient features were mapped to the [0,1] interval using the min-max normalization method. The upper and lower limits of normalization were taken as the extreme values ​​of similar meteorological gradient data from the past five years for the target nearshore sea area. The waveguide interface tilt angle was set according to [-90°]. The Richardson number of the ocean-atmospheric boundary layer is linearly scaled from [0, 90°] to [0, 1], and then standardized and scaled according to the [0, 10] interval to eliminate the dimensional differences of different features and ensure the uniformity of subsequent algorithms in the land-sea interface region. Finally, the standardized data is divided into training, validation and test sets in a 7:2:1 ratio according to the time series dimension. During the division, the spatiotemporal distribution of each subset is ensured to be uniform in different sub-regions of nearshore and offshore areas, so that the dataset is more in line with the complex spatial characteristics of nearshore areas. At the same time, the coastal radiosonde measurement data is archived separately to provide high-precision measurement reference for coastal waveguide inversion. The standardized multi-scale meteorological gradient dataset is output to provide standardized input data adapted to the nearshore environment for subsequent mesoscale optimization.

[0049] S200, Mesoscale Optimization: Using the standardized multi-scale meteorological gradient dataset output above as input data, the mesoscale meteorological gradient physical constraint fusion factor for the nearshore sea area is calculated based on the gradient fusion algorithm. Combined with the mesoscale meteorological patterns of nearshore land-sea breeze interaction, the fusion factor is made to better fit the mesoscale sea-atmosphere characteristics of the nearshore sea. Based on the calculation results, a mesoscale initial feature field is formed. Subsequently, a grid-by-grid calibration operation is performed on the initial feature field to correct the deviation of the mesoscale numerical model data in the land-sea interface region of the complex nearshore sea area, and to solve the problem of the numerical model not matching the actual meteorological gradient in the land-sea interface region. Finally, the corrected mesoscale meteorological gradient physical constraint fusion factor feature field is output, providing global boundary support that fits the actual nearshore conditions for subsequent small-scale modeling.

[0050] S300, Small-scale Modeling: Using the modified mesoscale meteorological gradient physical constraint fusion factor feature field as the boundary constraint, and combining the subtle fluctuations in the vertical gradients of temperature, humidity, and pressure, and the local deflection of the waveguide interface tilt angle captured in the high-resolution measured data of the nearshore sea surface, such as small-scale non-uniform features, the meteorological gradient physical constraint fusion factor adapted to these small-scale non-uniform features is calculated based on the gradient fusion algorithm. This accurately restores the local meteorological gradient changes under the influence of nearshore topography and sea breeze. During the calculation, the mesoscale 25km×25km grid is divided into several sub-regions according to the nearshore topography, so that the sub-region division fits the topographic distribution characteristics of the nearshore. The actual tilt angle and Richardson number of the waveguide interface of each sub-region are accurately obtained and then entered into the gradient fusion algorithm for iterative calculation. Each sub-region is iterated at least 3 times and the average value is taken to reduce the measurement and calculation errors of the nearshore local data and correct the lack of description of the small-scale non-uniformity of the nearshore by the mesoscale model. After calculating the small-scale fusion factor, spatial interpolation is performed. Using the corrected mesoscale feature field as the global boundary constraint and the small-scale fusion factor of each sub-region as the local constraint, the 25km×25km coarse-resolution mesoscale grid is divided into 9km×9km fine-grained sub-grids, and their spatial topology is determined. The sub-grid division accuracy is optimized, focusing on improving the feature field resolution of key nearshore areas with complex topography. Monte Carlo simulation is used for random interpolation. For each fine-grained sub-grid, the small-scale fusion factor of its parent grid is combined with that of the surrounding sub-regions. The fusion factor value of the subgrid is obtained by multiple random sampling calculations, so that the interpolation result takes into account both the global meteorological patterns in the nearshore area and the non-uniform characteristics under the influence of local topography. After interpolation, the fusion factor values ​​of all fine-grained subgrids are checked by boundary. The neighborhood mean smoothing method is used to eliminate numerical abrupt changes between adjacent subgrids, especially to ensure the spatial continuity of the feature field in the land-sea interface area and avoid unreasonable numerical jumps at the topographic boundary. Finally, a high-resolution small-scale meteorological gradient physical constraint fusion factor feature field is output, which provides refined meteorological gradient features that fit the complex environment of the nearshore area for waveguide inversion.

[0051] S400 Waveguide Inversion: This method uses the corrected mesoscale meteorological gradient physical constraint fusion factor feature field, the high-resolution small-scale meteorological gradient physical constraint fusion factor feature field, and high-precision coastal radiosonde data as input data. It integrates mesoscale global features, small-scale local features, and high-precision coastal data to improve the waveguide inversion accuracy in nearshore areas, especially the nearshore region. Based on the waveguide comprehensive inversion algorithm, it calculates the comprehensive inversion value of marine low-altitude waveguide features for the target nearshore sea area, achieving accurate conversion of multi-scale meteorological gradient features to nearshore waveguide features. Combining the characteristics of the nearshore air-sea environment, it performs grid-by-grid analysis of the inverted value feature field, focusing on improving the analysis accuracy in the nearshore region, making the analysis results more consistent with the waveguide distribution characteristics under the influence of nearshore topography. After analysis, it outputs the full-space feature parameter set of marine low-altitude waveguides for the sea area. The dataset contains waveguide presence identifiers, waveguide type identifiers, waveguide intensity levels, waveguide top heights, waveguide bottom heights, waveguide thicknesses, and waveguide spatial distribution ranges for each 9km×9km fine-grained grid node. It also includes time-series labels for corresponding time nodes, forming a spatiotemporally integrated feature dataset that clearly reconstructs the spatiotemporal evolution of nearshore waveguide characteristics. The waveguide presence identifiers, type identifiers, and intensity levels are all divided into corresponding intervals based on the comprehensive inversion values. The waveguide top height is obtained according to the corresponding calculation rules, the bottom height of surface waveguides is taken as 0m above sea level, and the bottom height of suspended waveguides is taken as the lower limit of the corresponding height interval. The waveguide thickness is the difference between the top height and the bottom height. The waveguide spatial distribution range is identified by the latitude and longitude range of the fine-grained grid, accurately marking the waveguide distribution characteristics of different nearshore topographic regions, and realizing the quantitative and refined spatial characterization of nearshore waveguide characteristics.

[0052] S500, Closed-Loop Optimization: Independent field-measured verification data are collected for the target nearshore sea area, including microwave refractometer data and shore-based radar data in three frequency bands. The microwave refractometer is deployed in different sub-regions near the coast and offshore, collecting vertical profile data of atmospheric refractive index from the sea surface to a height of 30km, with a vertical sampling interval of no more than 100 meters. The temporal resolution is consistent with the waveguide inversion results, ensuring that the verification data covers the entire nearshore space. The shore-based radar is deployed at multiple stations along the coast, using three frequency bands of ultra-shortwave, microwave, and millimeter wave for simultaneous field measurements. It collects data on the over-the-horizon propagation distance, signal strength, and attenuation characteristics of electromagnetic waves in each frequency band under different sea-atmosphere environments and nearshore topography in the nearshore area. The continuous collection time is no less than one month, covering propagation scenarios under different meteorological conditions in all four seasons, ensuring that the verification data conforms to the seasonal and topographical characteristics of the nearshore area. All independent measured verification data underwent spatiotemporal registration and data cleaning. After removing anomalous abrupt changes caused by nearshore topographic interference, the data was precisely matched with the spatiotemporal nodes of the full-space feature parameter set of marine low-altitude waveguides. This data served as a unified benchmark for waveguide feature accuracy verification. Based on this benchmark data, multi-dimensional accuracy verification of the full-space feature parameter set of waveguides was carried out, with a focus on verifying the parameter accuracy in complex nearshore topographic areas. The extraction effect in key nearshore areas was specifically tested. Based on the deviation results obtained from the verification, the adaptive coefficients of the gradient fusion algorithm and the waveguide comprehensive inversion algorithm were optimized to make the algorithm more adaptable to the complex sea-atmosphere environment under the influence of nearshore sea-land breezes and topography. Subsequently, the steps of S200 mesoscale optimization, S300 small-scale modeling, and S400 waveguide inversion were repeated, and the accuracy was verified again using the benchmark data. The algorithm was continuously iterated and optimized until the optimal adjustment of the adaptive coefficients was achieved, resulting in high-precision marine low-altitude waveguide feature parameters for this complex nearshore sea area, meeting the refined requirements of various nearshore waveguide feature applications.

[0053] This embodiment addresses the waveguide feature extraction requirements of the complex coastal waters of the Yellow Sea. Considering the significant influence of nearshore topography and sea-land breeze interaction on the sea area, targeted and refined processing is implemented in each step. Data acquisition focuses on capturing small-scale meteorological gradient changes nearshore; grid division and calibration focus on correcting deviations in the land-sea boundary region; and the verification phase emphasizes testing the parameter accuracy of the complex nearshore topography. Relying on gradient fusion and waveguide comprehensive inversion algorithms, effective fusion of mesoscale global features and small-scale local features is achieved. Closed-loop optimization is used to specifically adjust the algorithm's adaptive coefficients, making the algorithm more adaptable to the complex sea-atmosphere and topographic environment of the nearshore area. The resulting waveguide feature parameter set accurately labels the waveguide distribution characteristics of different nearshore topographic features, achieving refined extraction of nearshore waveguide features across the entire space. This provides practical waveguide feature data support for applications such as nearshore shipping navigation and nearshore communication support.

[0054] The above description is merely a preferred embodiment of the present invention and is not intended to limit the present invention in any way. Although the present invention has been disclosed above with reference to preferred embodiments, it is not intended to limit the present invention. Any person skilled in the art can make some modifications or alterations to the above-disclosed technical content to create equivalent embodiments without departing from the scope of the present invention. Any simple modifications, equivalent changes and alterations made to the above embodiments based on the technical essence of the present invention without departing from the scope of the present invention shall still fall within the scope of the present invention.

Claims

1. A method for extracting features of oceanic low-altitude waveguides based on meteorological gradients, characterized in that, The specific steps of this method are as follows: S100. Data Preprocessing: Define the spatial range and study period of the target sea area, collect multi-source marine meteorological gradient related data, complete the spatiotemporal registration, cleaning, feature extraction and standardization of the data, divide the dataset and output a standardized multi-scale meteorological gradient dataset. S200, Mesoscale Optimization: Using a standardized multi-scale meteorological gradient dataset as input, the mesoscale meteorological gradient physical constraint fusion factor is calculated based on the gradient fusion algorithm to form an initial feature field and complete grid-by-grid calibration, and outputs the corrected mesoscale meteorological gradient physical constraint fusion factor feature field. S300, Small-scale modeling: Using the modified mesoscale meteorological gradient physical constraint fusion factor feature field as the boundary constraint, the meteorological gradient physical constraint fusion factor adapted to small-scale non-uniform features is calculated based on the gradient fusion algorithm, spatial interpolation is performed, and a high-resolution small-scale meteorological gradient physical constraint fusion factor feature field is output. S400 Waveguide Inversion: Using the physical constraint fusion factor feature field of mesoscale and small-scale meteorological gradient and high-precision radiosonde data as input, the waveguide comprehensive inversion algorithm is used to calculate the comprehensive inversion value of marine low-altitude waveguide features, form the inversion value feature field and complete grid-by-grid analysis, and output the full-space feature parameter set of marine low-altitude waveguides. S500, Closed-loop optimization: Collect independent measured verification data, conduct multi-dimensional accuracy verification of the full-space characteristic parameter set of ocean low-altitude waveguides, optimize the adaptive coefficients of gradient fusion algorithm and waveguide integrated inversion algorithm based on the verification deviation results, repeat the mesoscale optimization, small-scale modeling, and waveguide inversion steps and verify again until the coefficient optimization is completed.

2. The method for extracting marine low-altitude waveguide features based on meteorological gradients according to claim 1, characterized in that, In step S100, the multi-source marine meteorological gradient related data specifically includes: mesoscale numerical model gridded data, obtained through a meteorological data platform, covering 37 vertical layers from sea surface to 30km altitude, including temperature, air pressure, relative humidity, wind speed, sea surface temperature, salinity, ocean current velocity, waveguide interface tilt angle, and Richardson number of the ocean-atmospheric boundary layer; fixed radiosonde station measured data, collected daily at 08:00 and 20:00, including temperature, pressure, humidity, wind speed, and wind direction profiles for different layers from sea surface to 30km altitude, as well as waveguide interface tilt angle and Richardson number around the station; and near-sea surface meteorological gradient instrument in-situ data, collected at least 10 layers of sensors within 40m above sea surface, at a frequency of 10min / time, including temperature, humidity, and air pressure data for each layer.

3. The method for extracting marine low-altitude waveguide features based on meteorological gradients according to claim 1, characterized in that, In step S100, the feature extraction involves extracting four core meteorological gradient features: temperature gradient, humidity gradient, vertical pressure gradient, and horizontal wind speed gradient. Simultaneously, the waveguide interface tilt angle and the Richardson number of the ocean-atmosphere boundary layer are extracted, along with two types of physical constraint features. Standardization processing involves mapping the temperature gradient, humidity gradient, vertical pressure gradient, and horizontal wind speed gradient to the [0,1] interval using min-max normalization. The waveguide interface tilt angle is linearly scaled to [0,1] within the [-90°, 90°] interval, and the Richardson number of the ocean-atmosphere boundary layer is standardized and scaled within the [0,10] interval. Dataset partitioning involves dividing the standardized data into training, validation, and test sets in a 7:2:1 ratio according to the time series dimension.

4. The method for extracting marine low-altitude waveguide features based on meteorological gradients according to claim 1, characterized in that, In step S200, the mathematical expression of the gradient fusion algorithm is: ； in, The fusion factor for physical constraints of meteorological gradients; For the overall temperature gradient; The overall humidity gradient; This represents the vertical pressure gradient; This represents the horizontal gradient of wind speed. , , , The contribution weighting coefficients for each meteorological gradient; This is a correction term for the tilt angle of the waveguide interface; This is a physical constraint term for atmospheric stability.

5. The method for extracting ocean low-altitude waveguide features based on meteorological gradients according to claim 1, characterized in that, In step S300, the small-scale non-uniform features refer to the local variations in meteorological gradients and waveguide structures within a spatial scale of less than ten kilometers that cannot be covered by the mesoscale grid in the target sea area. These include subtle fluctuations in the vertical gradients of temperature, humidity, and pressure captured by the near-sea surface meteorological gradiometer, local deflections of the tilt angles of the waveguide interfaces in different sub-regions, small-scale differences in the Richardson number of the ocean-atmospheric boundary layer, and local abrupt changes in the horizontal wind speed gradient caused by ocean currents and monsoons. When calculating the meteorological gradient physical constraint fusion factor adapted to the small-scale non-uniform features, the characteristic field of the mesoscale meteorological gradient physical constraint fusion factor is used as the boundary constraint. Combined with high-resolution measured data from the near-sea surface, the mesoscale grid is divided into sub-regions to obtain the actual tilt angles and Richardson numbers of the waveguide interfaces in each sub-region. These are then substituted into the gradient fusion algorithm for iterative calculation.

6. The method for extracting marine low-altitude waveguide features based on meteorological gradients according to claim 1, characterized in that, In step S300, the spatial interpolation operation specifically involves: using the modified mesoscale meteorological gradient physical constraint fusion factor feature field as the global boundary constraint, and using the meteorological gradient physical constraint fusion factor of each sub-region adapted to small-scale non-uniform features as the local constraint, dividing the mesoscale coarse-resolution grid into fine-grained sub-grids to determine the spatial topology of the sub-grids; and using a spatial interpolation method based on Monte Carlo simulation to perform random interpolation. For each fine-grained sub-grid, the fusion factor value of the sub-grid is obtained by combining the mesoscale fusion factor of its parent grid with the small-scale fusion factor of the surrounding sub-regions through multiple random sampling calculations. During the interpolation process, the global distribution law of the mesoscale feature field is preserved, while restoring the subtle changes and non-uniform characteristics of the meteorological gradient in the sub-region.

7. The method for extracting marine low-altitude waveguide features based on meteorological gradients according to claim 1, characterized in that, In step S400, the mathematical expression of the waveguide synthesis inversion algorithm is: ； in, This is a comprehensive inversion value of the characteristics of low-altitude waveguides in the ocean; The fusion factor for physical constraints of meteorological gradients; The fitting coefficients are for mesoscale features; For small-scale feature adaptation coefficients; The calibration coefficient is the result of actual sounding measurements. This is the baseline value for the meteorological gradient measured by radiosonde.

8. The method for extracting oceanic low-altitude waveguide features based on meteorological gradients according to claim 1, characterized in that, In step S400, the full-space feature parameter set of the ocean low-altitude waveguide specifically includes the waveguide existence identifier, waveguide type identifier, waveguide intensity level, waveguide top height, waveguide bottom height, waveguide thickness, and waveguide spatial distribution range of each fine-grained grid node, and is also equipped with the time series label of the corresponding time node to form a spatiotemporal integrated feature dataset.

9. The method for extracting marine low-altitude waveguide features based on meteorological gradients according to claim 1, characterized in that, In step S500, the independent measured verification data includes measured data from a microwave refractometer and measured data from a shore-based radar in three frequency bands. The microwave refractometer collects vertical profile data of atmospheric refractive index from the sea surface to a height of 30 km in the target sea area, with a vertical sampling interval of no more than 100 meters. The time resolution is consistent with the inversion results and can be directly used to determine the parameters of the real waveguide structure. The shore-based radar uses three frequency bands—ultra-shortwave, microwave, and millimeter wave—to conduct simultaneous measurements, collecting data on the over-the-horizon propagation distance, signal strength, and attenuation characteristics of electromagnetic waves in the marine atmospheric environment. The continuous collection time is no less than one month, covering propagation scenarios in different seasons and different marine atmospheric environments.

Images