Low-altitude weather guarantee method and system

Abstract

The application is suitable for the technical field of low-altitude meteorological guarantee, and provides a low-altitude meteorological guarantee method and system.The method comprises the following steps: fusing ground-based, air-based and space-based multi-source observation data to generate a high-temporal and high-spatial resolution low-altitude three-dimensional meteorological real-time field; inputting the low-altitude three-dimensional meteorological real-time field into a boundary layer numerical model to perform a rapid update cycle assimilation, and obtaining a short-term and nowcast prediction field; calculating threat factors such as wind shear and turbulence based on a physical engine; generating a grid risk probability map through a space-time graph convolution network; fusing and deciding the wind shear and the grid risk probability map through a dynamic alarm rule base, and outputting a hierarchical early warning instruction and a risk geographic fence; and converting the early warning information into airspace switch instructions and three-dimensional avoidance routes in real time through airspace grid modeling and meteorological corridor intelligent generation, and publishing the airspace switch instructions and the three-dimensional avoidance routes through a multi-terminal service architecture.The application realizes a full closed-loop guarantee from fine perception, intelligent early warning to dynamic decision of low-altitude meteorology, and significantly improves the safety and airspace operation efficiency of low-altitude flight.

Description

Technical Field

[0001] This invention belongs to the field of low-altitude meteorological support technology, and particularly relates to a low-altitude meteorological support method and system. Background Technology

[0002] With the rapid development of low-altitude economic applications such as drone logistics, urban air traffic, and low-altitude inspection, the frequency and complexity of low-altitude flight activities have increased significantly, placing higher demands on low-altitude meteorological support capabilities. Low-altitude airspace typically refers to the airspace from the ground to a certain altitude. The atmospheric environment within this altitude range is significantly affected by topography, underlying surface, and human activities, and is characterized by rapid changes and complex vertical structures. Various local convective weather phenomena pose a serious threat to low-altitude flight safety.

[0003] Currently, low-altitude meteorological monitoring mainly relies on traditional ground-based meteorological station networks, radiosondes, lidar remote sensing, and meteorological satellite remote sensing. Ground-based meteorological stations can only provide single-point surface meteorological elements; radiosondes have low spatiotemporal resolution, making it difficult to capture dramatic weather changes on a minute to hourly scale; lidar remote sensing has a limited detection range; and while meteorological satellites have wide coverage, their vertical resolution and near-surface observation accuracy are limited. At the data application level, existing technologies mostly use single numerical weather prediction models to output guidance products. Their boundary layer parameterization schemes are insufficient in characterizing complex underlying surfaces and microscale processes, leading to significant errors in short-term forecasts, particularly limited forecasting capabilities for dynamic processes such as wind shear and turbulence. Furthermore, current meteorological warning products are mostly issued in a regionalized and qualitative manner, lacking quantitative, three-dimensional visualization of risk indicators directly coupled with airspace structure and flight paths, and cannot support minute-level dynamic airspace reconstruction and real-time flight path planning.

[0004] Therefore, existing technologies suffer from the following major shortcomings: insufficient sensing capabilities, limited observation methods with constraints on resolution, coverage, and detection elements, lack of a high-resolution, three-dimensional low-altitude monitoring network with multiple devices working together, making it difficult to accurately acquire key micro-scale meteorological information such as turbulence and wind shear; insufficient forecast and early warning accuracy and timeliness, limited assimilation capabilities of traditional numerical models, low accuracy of short-term forecasts, and coarse early warning information, failing to meet the needs of low-altitude flight for precise obstacle avoidance at the meter and minute levels; and disconnected decision support, with meteorological information isolated from airspace management and flight control systems, making it impossible to achieve dynamic airspace optimization and intelligent route planning based on real-time meteorological threats. Summary of the Invention

[0005] The purpose of this application is to provide a low-altitude meteorological support method, which aims to solve the problems of large systematic bias in existing forecasts, unstable meteorological element inversion results, which affect the reliability of subsequent forecasts and risk identification, and the need to improve forecast accuracy.

[0006] This application provides a low-altitude meteorological protection method, the method comprising:

[0007] Acquire multi-source observation data of the target area from ground-based, air-based, and space-based sources; perform spatiotemporal benchmark unification and multi-level quality control on the multi-source observation data to generate a standardized observation dataset;

[0008] Standardized data is fused to generate a three-dimensional dynamic field with a resolution of 100 meters;

[0009] The three-dimensional dynamic field is input into the boundary layer numerical model, and a fast update cyclic assimilation technique is used to generate a short-term low-altitude meteorological forecast field.

[0010] Using threat factor values ​​calculated from short-term forecast fields as input, combined with risk probability maps generated by spatiotemporal graph convolutional networks, and fused and analyzed by a dynamic alarm rule base, hierarchical early warning instructions are output.

[0011] The risk probability map and graded early warning instructions are mapped to a digital airspace grid to obtain airspace status information; the passability of each grid cell is assessed in real time, and high-risk airspace is automatically closed; based on avoidance algorithms, three-dimensional routes are dynamically generated or adjusted for affected aircraft.

[0012] By using multiple platforms, the integrated risk probability map and airspace status information are released in real time, completing the closed loop of protection services.

[0013] Another objective of this application is to provide a low-altitude weather protection system, including a memory and a processor, wherein the memory stores a computer program, and when the computer program is executed by the processor, the processor performs the steps of the low-altitude weather protection method as described above.

[0014] The low-altitude meteorological support method provided in this application has the significant advantage of effectively integrating ground-based, airborne, and space-based observation data through feature-level fusion and assimilation of multi-source heterogeneous data. This solves the problems of insufficient resolution and limited coverage of single remote sensing methods, generating a three-dimensional meteorological field with spatiotemporal resolution at the hundred-meter and minute levels, providing a data foundation for accurate analysis. It employs a dual-engine collaborative early warning system driven by both physics and data, relying on numerical models to provide the physical background while using spatiotemporal graph convolutional networks to identify local risk probabilities, significantly improving the accuracy of short-term warnings. Through dynamic scoring of airspace grids and intelligent path planning algorithms, it automatically closes high-risk airspace and generates optimal three-dimensional obstacle avoidance routes, directly translating warnings into flight control instructions. The entire system provides a complete intelligent solution for low-altitude safety and efficient operation. Attached Figure Description

[0015] Figure 1An application environment diagram for a low-altitude meteorological protection method provided in this application embodiment;

[0016] Figure 2 A flowchart illustrating a low-altitude meteorological protection method provided in this application embodiment;

[0017] Figure 3 This is a block diagram of the internal structure of a computer device in one embodiment. Detailed Implementation

[0018] To make the objectives, technical solutions, and advantages of this invention clearer, the invention will be further described in detail below with reference to the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are merely illustrative and not intended to limit the invention.

[0019] It is understood that the terms "first," "second," etc., used in this application may be used herein to describe various elements, but unless otherwise stated, these elements are not limited by these terms. These terms are used only to distinguish the first unit or module from another unit or module. For example, without departing from the scope of this application, the first script may be referred to as the second script, and similarly, the second script may be referred to as the first script.

[0020] Figure 1 An application environment diagram for the low-altitude meteorological protection method provided in the embodiments of this application, such as... Figure 1 As shown, this application environment includes an environmental data acquisition input device 110 and a computer device 120.

[0021] Computer device 120 can be a standalone physical server or terminal, or a server cluster consisting of multiple physical servers, such as a desktop computer.

[0022] The environmental data acquisition input device 110 can be a boundary layer meteorological tower, wind lidar, millimeter-wave cloud radar, microwave radiometer, tethered balloon, etc., but is not limited to these.

[0023] The environmental data acquisition input device 110 and the computer device 120 can be connected via a network, which is not limited herein.

[0024] like Figure 2 As shown, in one embodiment, a low-altitude meteorological support method is proposed. This embodiment mainly applies this method to the above-mentioned... Figure 1 Let's take computer equipment 120 as an example. A low-altitude weather protection method may specifically include the following steps:

[0025] Step S10: Obtain multi-source observation data of the target area from ground-based, air-based, and space-based sources; perform spatiotemporal benchmark unification and multi-level quality control on the multi-source observation data to generate a standardized observation dataset.

[0026] In this application embodiment, the multi-source low-altitude meteorological observation data includes one or more of the following: ground-based, air-based, and space-based data.

[0027] In this embodiment, in order to construct a low-altitude meteorological integrated sensing network based on the fusion of multi-source heterogeneous observation equipment and achieve near real-time monitoring with high spatiotemporal resolution, the monitoring elements need to cover several elements that are crucial to low-altitude flight safety. For example, one or more of the following can be selected: wind direction, wind speed, temperature, humidity, air pressure, visibility, cloud base height, vertical visibility, atmospheric turbulence intensity, precipitation type and intensity, and key stratification information such as the location and intensity of the inversion layer and the height of the boundary layer.

[0028] Monitoring methods may include: ground-based point observation, boundary layer meteorological towers, wind profiler radar, millimeter-wave cloud radar, wind-measuring lidar, microwave radiometers, balloon sounding, UAV-borne meteorological sounding systems, meteorological rocket sounding, geostationary meteorological satellites, polar-orbiting meteorological satellites, etc.

[0029] Through the collaborative observation of multiple platforms and sensors, a three-dimensional, multi-element collaborative low-altitude meteorological information perception capability covering the near-ground to the target low-altitude layer is formed, providing high-quality input for subsequent data processing and forecasting and early warning.

[0030] Understandably, in this embodiment, the ground meteorological station network forms the foundation of the observation network, providing dense point observations of ground meteorological elements such as temperature, humidity, pressure, wind, precipitation, and visibility. Boundary layer meteorological towers are used to provide high temporal resolution vertical gradient observations of near-surface meteorological elements at key locations, serving as important facilities for studying boundary layer processes and validating models. Wind profiler radar provides continuous vertical wind profiles with high temporal resolution, serving as an important supplement to wind field monitoring. Millimeter-wave cloud radar accurately detects the vertical structure of low clouds and fog, cloud base height, cloud particle phases, and weak precipitation, crucial for monitoring low-visibility weather. Wind-measuring lidar is the core three-dimensional wind field monitoring device, possessing scanning capabilities to acquire planar or three-dimensional wind field information, with a long detection range and high resolution, particularly adept at detecting clear-sky turbulence. Microwave radiometers are the core temperature and humidity profile monitoring device, capable of continuous all-weather operation, providing vertical distribution information of temperature, humidity, and liquid water content within the boundary layer, compensating for insufficient radiosonde frequency. Balloon sounding provides high-precision vertical profile baseline data for temperature, humidity, pressure, and wind, which is crucial for calibrating other remote sensing equipment. UAV meteorological payloads, in addition to carrying sounding instruments, also include miniature meteorological sensors fixedly mounted on the UAV for refined supplementary observations of specific routes or areas. Meteorological rocket sounding provides high-vertical-resolution, high-precision direct sounding data, serving as an important supplement and verification tool for other observation methods. Meteorological satellite data provides large-scale, macroscopic meteorological information and is an important input for background field analysis and short-term weather warnings.

[0031] Step S20: Perform data fusion on the standardized data to generate a three-dimensional dynamic field with a resolution of 100 meters;

[0032] In this embodiment, the system performs spatiotemporal standardization and quality control preprocessing on multi-source observation data. Then, a variational optimization algorithm based on physical constraints is used to perform feature-level fusion of the continuous temperature and humidity profiles retrieved from microwave radiometers, the three-dimensional wind field data retrieved from lidar, and the discrete high-precision observation points from radiosondes. This process simultaneously incorporates the dynamic boundary layer height, jointly determined by the lidar aerosol depolarization ratio and the liquid water path from the microwave radiometer, as a physical constraint. Simultaneously, machine learning-assisted vector field fusion technology, such as improved DDA, is employed to fuse the scanning data from multiple lidars, the vertical profiles from wind profiler radars, and the fixed-point wind observations from the UAV platform. This ultimately generates an integrated three-dimensional real-time field of temperature, humidity, and wind speed with a time resolution of minutes and a vertical resolution of 10 meters within a height range of 0-1500 meters, significantly improving the retrieval accuracy of key stratigraphic information such as low-level inversion structures.

[0033] Step S30: Input the three-dimensional dynamic field into the boundary layer numerical model and use the fast update cyclic assimilation technique to generate a short-term forecast field for low-altitude weather.

[0034] In this embodiment, the observation operator serves as the core interface connecting multi-source observation data and numerical models. Its core function is to convert state variables in the numerical model (such as three-dimensional wind speed, temperature, and humidity) into theoretically equivalent values ​​(such as radial wind speed from lidar and brightness temperature from microwave radiometers) that can be directly compared with actual remote sensing observations. By customizing high-precision observation operators for such non-model direct variables, the challenges of nonlinear transformation and direct application of remote sensing data in data assimilation are effectively solved.

[0035] In boundary layer numerical models such as WRF-LES, the system constructs a rapid update cyclic assimilation process based on the aforementioned observation operators. Specifically, with an assimilation time window of no more than 10 minutes, an ensemble variational hybrid assimilation framework is adopted, injecting the 3D real-world field generated in the above steps (including simulated observation information transformed by operators) as the initial field of the model as an observation increment. Simultaneously, by implementing spatiotemporal localization of the background error covariance, key boundary layer parameterization schemes (such as the turbulence diffusion coefficient in the YSU scheme) are optimized. The entire system continuously completes a closed loop from observation, assimilation to forecasting with a 30-minute cycle, outputting real-time short-term low-air meteorological forecasts for the next 0-2 hours with a horizontal resolution of 100 meters. This high-precision forecast field provides core input for risk identification in subsequent S40 operations and forms a closed-loop dynamic optimization of the assimilation system itself based on real-time observation feedback, thereby continuously improving forecast accuracy.

[0036] Step S40: Using the threat factor values ​​calculated from the short-term forecast field as input, combined with the risk probability map generated by the spatiotemporal graph convolutional network, and after fusion analysis by the dynamic alarm rule base, a graded early warning instruction is output.

[0037] This embodiment incorporates a collaborative workflow driven by both physical and data engines. The physical engine calculates the gridded low-level wind shear index, eddy current dissipation rate, icing index, and low visibility index in real time based on short-term forecast fields. The data-driven engine takes the time series of the aforementioned threat factors as input and outputs a gridded risk probability map for the next 10 minutes and 100 meters using a pre-trained spatiotemporal graph convolutional network model. A dynamic warning rule base fuses and analyzes these two types of results. For example, when the physical calculation EDR value of a grid cell exceeds a threshold or the severe turbulence probability given by ST-GCN exceeds 90%, the corresponding rule is triggered, automatically generating red, yellow, and blue graded warning commands and their corresponding spatial geofences.

[0038] Step S50 involves mapping the risk probability map and graded early warning instructions to a digital airspace grid to obtain airspace status information; assessing the passability of each grid cell in real time and automatically closing high-risk airspace; and dynamically generating or adjusting three-dimensional routes for affected aircraft based on avoidance algorithms.

[0039] In this embodiment, the three-dimensional airway can refer to a meteorological corridor. The risk probability map and warning instructions output by S40 are mapped to a digital micro-airspace management grid. The system calculates the comprehensive meteorological safety score for each airspace grid (e.g., 1km×1km×100m) in real time and automatically closes or restricts passage for units with scores below the safety threshold. For flight plans affected by this, the system calls an intelligent path planning module based on the Dijkstra-A* hybrid algorithm. This module uses the real-time three-dimensional wind field as the environmental dynamic cost and the closed high-risk airspace as the obstacle avoidance constraint. By minimizing flight drag and risk crossing penalties, it dynamically generates or adjusts a continuous sequence of three-dimensional waypoints, i.e., meteorological corridors, and generates structured control instructions.

[0040] Step S60: Through multi-terminal platforms, the integrated risk probability map and airspace status information are released in real time to complete the closed loop of protection services.

[0041] In this embodiment, a unified service publishing platform for multiple user terminals is constructed. The system simultaneously publishes the integrated risk warning products, dynamic airspace status, and intelligently planned airway control instructions to the command center's 3D situation display screen, airborne lightweight terminals, and the mobile app of mission planners via low-latency communication links. The command screen enables full-domain 3D visualization monitoring and decision support; airborne terminals receive concise warning codes and airway change instructions; and the mobile app supports mission reporting and personalized warning push notifications. All instructions interact with the air traffic management system through standardized interfaces, ultimately forming a minute-level service closed loop from meteorological perception, intelligent warning, dynamic decision-making to instruction execution.

[0042] Through the above steps, the advantages of this method are as follows: At the data layer, based on multi-source fusion and rapid cyclic assimilation technology, it integrates ground-based, airborne, and space-based observation information, avoiding the limitations of single remote sensing methods in terms of resolution and feature coverage, and constructing high-precision three-dimensional real-time field and short-term forecast field at the hundred-meter and minute levels; At the early warning layer, it innovatively adopts a dual-engine collaborative identification of physics and data, relying on numerical models to ensure the reliability of mechanisms, and using spatiotemporal map convolutional networks to improve the ability to capture sudden and local risks, and outputting graded early warnings and risk geofencing through dynamic alarm rules; At the execution layer, it realizes the real-time conversion of meteorological information into control instructions, and relies on airspace grid assessment and intelligent route planning to automatically execute dynamic airspace switching and generate three-dimensional obstacle avoidance corridors. Finally, through a multi-terminal service system, it forms a minute-level closed-loop control of perception, early warning, decision-making, and control, improving the safety of low-altitude flight and the efficiency of airspace operation.

[0043] In a preferred embodiment, the detection equipment needs to be deployed in the field. The specific steps can be as follows: First, determine the deployment principles: Spatial representativeness principle: The coverage area should include typical underlying surface types, such as urban building clusters, industrial areas, suburbs, farmland, water bodies, and mountains. Densified deployment should be implemented in key protection areas such as UAV take-off and landing sites, airports, air logistics corridors, chemical industrial parks, ecological protection zones, and airspace for major events. Site selection must consider the impact of different terrains on local weather. Safety principle: All equipment installations must meet safety standards such as lightning protection, wind resistance, and electromagnetic compatibility. UAV operations must strictly comply with airspace management regulations, avoiding no-fly zones, and emergency plans must be developed. Sounding balloon releases must be approved by air traffic control to ensure aviation safety. Three-dimensional coverage principle: Vertical coverage should focus on the near-ground to low-altitude layer of the target, ensuring direct observation or reliable inversion / fusion data at key altitude layers. The observation layer settings need to be dynamically optimized based on the minimum safe altitude and typical operating altitude of the specific protection mission.

[0044] Then, a tiered and differentiated deployment approach is adopted for different key protection areas: one coherent Doppler wind lidar is deployed in the core area, using PPI, RHI, and volumetric scanning modes to achieve high-resolution three-dimensional real-time monitoring of the wind field within a radius of ≥3 kilometers and a height of 0-1000 meters; ≥4 boundary layer meteorological towers are deployed at key locations around the take-off and landing field, with multi-layer gradient observations on the towers to obtain the vertical distribution of temperature, humidity, wind, and pressure at minute-level high frequency; an ultrasonic anemometer can be optionally equipped to obtain turbulent flux; one microwave radiometer is deployed to continuously monitor the temperature and humidity profiles and liquid water paths; and an unmanned aerial vehicle (UAV)-borne sounding system is equipped for precise vertical detection under emergency or specific needs. Along the corridor, miniature weather stations are integrated and deployed on existing facilities such as streetlights and utility poles at intervals of no more than 1 kilometer. These stations utilize solar power and low-power wide-area networks or 5G for data transmission. Tethered balloon platforms equipped with microwave radiometers are deployed at key nodes or meteorologically complex sections of the corridor to achieve continuous temperature and humidity profile monitoring at altitudes of ≥300 meters, enhancing vertical coverage. Small millimeter-wave cloud radars can be deployed at corridor entrances, exits, or bends to monitor low clouds / fog affecting visibility. The mobile area is equipped with a vehicle-mounted coherent Doppler wind lidar system, capable of rapid deployment and providing three-dimensional wind field monitoring with a scanning radius of ≥2 kilometers and an altitude of 0-800 meters. A vehicle-mounted UAV meteorological sounding system is also included, allowing for rapid deployment to obtain accurate vertical profiles of temperature, humidity, pressure, and wind. The system also features satellite communication capabilities to ensure data transmission in areas without public network access. All fixed and mobile observation nodes of the data link and processing center transmit raw or pre-processed data to the low-altitude meteorological support center in real time through a high-speed and reliable wired / wireless communication network. The support center deploys a data fusion and processing platform, which is responsible for data reception, decoding, quality control, format standardization, multi-source data fusion / assimilation, product generation and distribution.

[0045] In a preferred embodiment, the method for generating a standardized observation dataset by performing spatiotemporal benchmark unification and multi-level quality control on the multi-source observation data is as follows:

[0046] The timestamps of the multi-source observation data are corrected based on a unified time reference; for observation data with different time resolutions, interpolation or resampling is performed according to a preset time step to unify the time reference of the multi-source observation data.

[0047] Based on a unified geographic coordinate system, the multi-source observation data is spatially mapped; the observation data is transformed into a unified vertical height layer structure, so that the spatial reference of the multi-source observation data is unified.

[0048] By removing obvious anomalies, distortions, and / or physical meanings from the data processed through time and space benchmark unification, a standardized observation dataset is obtained.

[0049] In this embodiment, the system first accesses ground weather stations, boundary layer weather towers, wind-measuring lidar, millimeter-wave cloud radar, and microwave radiometers in real time to acquire meteorological data such as wind, temperature, humidity, pressure, and precipitation. Data from each device is uniformly decoded through a dedicated protocol parsing module. Lidar data uses a custom binary format, with a header including a timestamp, device ID, and scanning mode identifier. Ground station data is transmitted via MQTT protocol in JSON format, with fields including physical quantities such as temperature, humidity, and wind speed. Then, the multi-source data is unified to the following spatiotemporal reference: the spatial grid can be 200m×200m×50m, using the WGS84 coordinate system, EPSG:4326; terrain data is corrected using SRTM 30m DEM; UTC timestamps are aligned to whole minutes; lagging data is compensated using linear interpolation; and the time window is set to ±30 seconds. Lidar spherical coordinate system data is interpolated to grid points using inverse distance weighted interpolation, with a weight function w. i Satisfy: w i =1 / d i 2 , where d i This represents the distance from the grid point to the scan point.

[0050] Then, dynamic quality control is performed to remove outliers. For example, lidar data points with a signal-to-noise ratio <5dB are removed; microwave radiometers mark data with liquid water path >5g / m³ as precipitation interference; and ground station wind speed changes >10m / s per minute trigger anomaly marking. Using UAV sounding data as a benchmark, Kalman filtering is used to correct microwave radiometer temperature and humidity profile deviations. Different terminals employ different outlier removal methods, which will not be elaborated here. Through these operations, standardized low-air meteorological observation data is obtained, thereby improving the overall data acquisition accuracy.

[0051] In a preferred embodiment, the method for fusing standardized data to generate a three-dimensional dynamic field with a resolution of 100 meters is as follows:

[0052] By integrating data from microwave radiometers, lidar, and radiosondes, and using a variational optimization algorithm based on physical constraints, an integrated vertical profile of temperature, humidity, and wind at the minute level is generated within an altitude range of 0-1500m.

[0053] The integrated vertical profile is unified into a 200m×200m×50m spatial grid, and Kriging interpolation is used to map discrete points to the spatial grid to form a three-dimensional dynamic field.

[0054] In this embodiment, generating a 100-meter resolution three-dimensional dynamic field can specifically include the following steps: First, by fusing data from microwave radiometers, lidar, and radiosondes using a variational optimization algorithm, a minute-level integrated temperature-humidity-wind vertical profile is generated. The boundary layer height is then dynamically determined using lidar and microwave radiometer observations, optimizing the inversion layer retrieval accuracy. Next, based on machine learning-assisted vector field fusion technology, lidar, wind profiler radar, and UAV wind observations are integrated to generate a 100-meter resolution three-dimensional dynamic field, improving the identification capabilities of wind shear and turbulent vortices. Finally, through vector field fusion technology, the multi-source meteorological field data is unified into a 200m×200m×50m grid, where radial wind speed is decomposed into u / v / w components, and Kriging interpolation is used to map discrete points to the grid.

[0055] In this embodiment, the multi-source observation constraint is transformed into an optimizable problem, and the most consistent and physically reasonable low-air meteorological state estimate is obtained through variational optimization. For example, a variational optimization algorithm based on physical constraints generates an integrated vertical profile of temperature, humidity, and wind at the 0–1500 m minute level, and introduces boundary layer height (PBLH) as a dynamic constraint. PBLH can be jointly determined by the depolarization ratio of lidar aerosols and the liquid water path gradient of microwave radiometers, with a depolarization ratio threshold of 0.15, and the difference between the two PBLH determination heights should be <100 m; otherwise, an emergency calibration of UAV sounding is triggered. The fusion objective function can adopt the following variational structure:

[0056] ;

[0057] in, It is the low-altitude meteorological state that needs to be determined. It's a background scene. It is the observation operator that maps the mode state to the first... Observational space, Reflects observation error; Reflects background error, Let represent the objective function, which measures the degree to which the low-altitude meteorological state deviates from both the background field and the combined observation data. The optimal low-altitude meteorological state estimate is obtained by minimizing this objective function. Let represent the observation vector formed by the nth type of observation data. The inverse matrix of the observation error covariance matrix is ​​used to weight the observation bias term.

[0058] In a preferred embodiment, the three-dimensional meteorological field obtained from multi-source observations / fusion is continuously injected into a low-air meteorological numerical model, such as the WRF-LES model, within a very short time window. The model's state variables and key parameters are continuously corrected through an assimilation algorithm. The specific method is as follows:

[0059] First, custom observation operators are developed for lidar radial wind speed, microwave radiometer brightness temperature, and UAV observations, converting actual observed physical quantities into model variables, such as the lidar radial wind speed operator:

[0060] ;

[0061] Where θ is the azimuth angle and ϕ is the elevation angle. For instrument noise, This represents the radial wind speed value actually observed by the wind-measuring lidar, where u, v, and w represent the three orthogonal components.

[0062] Then, a rapid update cyclic assimilation is performed. In the WRF-LES mode, a 5-minute assimilation time window is set, and the fused 3D wind field is injected with a weight of 0.7; the temperature and humidity profile is injected with a weight of 0.3. The boundary layer height parameter is optimized to ensure that the RMSE of the low-level wind speed output by the model is ≤1.2m / s, achieving minute-level dynamic updates of the meteorological field from 0 to 1500m in the low atmosphere, supporting short-term warnings and dynamic airspace decision-making. The system configuration is shown in the table below:

[0063] Module Parameter settings Technical basis model WRF-LES (500m horizontal grid, 30 layers vertically) Balancing computational efficiency and turbulence resolution requirements Assimilation Time Window 5 minutes (background field + observation increment updated simultaneously) Balancing timeliness and data coverage Assimilation methods Hybrid EnVAR (Ensemble Variational Hybrid Assimilation) Combining the advantages of ensemble forecasting and variational methods Collection members 20 Balance between computational resources and error covariance Localization radius 3km (horizontal), 200m (vertical) Adaptation to boundary layer process scale

[0064] Finally, real-time feedback correction is performed. The short-term forecasts output by the assimilated model are dynamically compared with real-time observations to trigger error backpropagation learning in the QC module, forming a closed-loop optimization.

[0065] In a preferred embodiment, the method for generating short-term low-altitude meteorological forecast fields by inputting a three-dimensional dynamic field into a boundary layer numerical model and employing a fast-update cyclic assimilation technique is as follows:

[0066] In the boundary layer numerical model, an assimilation time window oriented towards the low altitude is constructed; an ensemble variational hybrid assimilation framework is adopted to construct the model initial field for assimilation processing.

[0067] Within the assimilation time window, the three-dimensional dynamic field is injected as an observation increment into the boundary layer numerical model to update the model state; the state update process is used to correct the spatial distribution and vertical structure of low-altitude meteorological elements in the model.

[0068] During the assimilation update process, the boundary layer parameters are updated synchronously. After completing one assimilation update, the next time window is entered based on the updated mode state, and the assimilation update process is repeated.

[0069] After completing at least one rapid update cycle assimilation process, forecast calculations are performed based on the updated boundary layer numerical model state; and a short-term low-altitude meteorological forecast field is generated within the target protection area.

[0070] In this embodiment, the initial model field refers to the initial state field of the boundary layer numerical model at the beginning of a certain assimilation time window, which serves as the starting point for assimilation and forecast integration within that time window. The initial model field is constructed by using the model state updated after assimilation in the previous time window as the background field, combined with the observation constraints formed by the three-dimensional dynamic field.

[0071] In this embodiment, inputting the three-dimensional dynamic field into the boundary layer numerical model and performing rapid update cyclic assimilation processing refers to using the three-dimensional dynamic field obtained by fusing multi-source observations as a real-time constraint on the numerical model state. This continuously corrects the distribution and evolution of low-altitude meteorological elements in the model within a short timescale, thereby obtaining short-term low-altitude meteorological forecasts that are closer to reality. Specifically, firstly, the three-dimensional dynamic field is used as the observation input and combined with the background field of the boundary layer numerical model to construct the initial model field for assimilation processing. The three-dimensional dynamic field includes at least wind, temperature, and humidity fields, ensuring that the initial model state reflects the three-dimensional spatial structure of the low-altitude meteorological conditions at the current moment. Subsequently, a time window for low-altitude assimilation is set in the boundary layer numerical model. The length of the time window is shorter than the assimilation time interval of conventional mesoscale numerical models to accommodate the rapid changes in low-altitude meteorological elements within a short period. Within the assimilation time window, the three-dimensional dynamic field is injected into the boundary layer numerical model as an observation increment. The model state is updated through an assimilation algorithm, causing the low-altitude meteorological elements such as wind, temperature, and humidity fields in the model to converge towards the observation constraints in terms of spatial distribution and vertical structure. A weighted correction of the model state is performed using state updates, incorporating multi-source observation information to correct for low-altitude structure while maintaining the consistency of the numerical model's dynamics and thermodynamics. During the assimilation update process, boundary layer parameters related to the low-altitude atmosphere are also updated synchronously, allowing key parameters such as boundary layer height and turbulence diffusion intensity to dynamically adjust according to observational constraints, thereby improving the model's ability to characterize the low-altitude boundary layer structure and turbulence features. After each assimilation update, the updated model state is used as the new background field for the next assimilation time window, and the above assimilation update process is repeated. This rolling loop continuously introduces the latest three-dimensional dynamic field, achieving rapid update and cyclic assimilation. The final updated boundary layer numerical model state is used for forecast calculations to generate short-term low-altitude meteorological forecasts for the target protection area.

[0072] By using the above methods, systematic biases in air-wind field, temperature and humidity structure, and boundary layer characteristics can be significantly reduced, making the obtained short-term forecast results more suitable for low-altitude operation safety assurance and risk identification application scenarios.

[0073] In a preferred embodiment, the threat factors extracted from the short-term forecast field include one or more of the following factors:

[0074] Ice accumulation index threat factor, eddy dissipation rate threat factor, low visibility index threat factor, and wind shear index threat factor.

[0075] Risk identification conditions include preset threshold conditions and / or probability identification conditions constructed based on historical data. This can refer to extracting risk factors in the following ways, and then using these risk factors to identify various risks:

[0076] Icing Index (ISI) Factor: By integrating microwave radiometer liquid water content (LWC), UAV flight speed (V), exposure time t, temperature function f(T), critical icing thickness τ, and correction coefficient k, an icing index model is constructed, and the simplified formula is as follows:

[0077] .

[0078] The eddy current dissipation rate factor, obtained by inverting turbulence intensity through lidar spectral width, has the following core formula:

[0079] ;

[0080] Where σ is the Doppler spectral width, σ0 is the inherent spectral width caused by instrument noise and molecular thermal motion, k is an empirical constant, and V eff To effectively scan the average wind speed within the volume, ∆t is the time resolution, and EDR is the eddy current dissipation rate factor.

[0081] Low Visibility Index (LVI) Factor: Combines ground visibility observation, lidar backscattering coefficient, and cloud radar fog height, and outputs a 100-meter grid visibility estimate through random forest regression.

[0082] Wind shear index (WSI) factor: Modeled using a hybrid approach combining lidar vertical wind gradient (∂V / ∂z) and turbulent kinetic energy (TKE); where lidar is used to measure horizontal wind speed (V) at vertical height z-layers, and the gradient is calculated using the central difference method.

[0083] ;

[0084] The standard deviation (σ) of three-dimensional wind speed fluctuations using lidar u , σ v , σ w Calculate turbulent kinetic energy:

[0085] ;

[0086] Finally, the wind shear index is calculated by weighted fusion of vertical wind gradient and turbulent kinetic energy.

[0087] ,

[0088] Where α and β are weighting coefficients.

[0089] Taking drone application scenarios as an example, the method for outputting protection information after risk identification can be as follows: First, threat factors are extracted, and the eddy dissipation rate (EDR) is calculated using the 3D meteorological field and the lidar spectral width. For example, in hilly areas, the wind speed fluctuation spectral width σ = 0.8 m / s, and the calculated EDR = 0.45 m² / s³. The wind shear index (WSI) is calculated by fusing the lidar vertical gradient and the turbulent kinetic energy of the meteorological tower. For example, at a certain point, ∂V / ∂z = 4.3 m / s per 100 meters, TKE = 1.2 m² / s², and the output WSI = 3.2, indicating a red risk. Then, the meteorological forecast field and environmental background are provided to the data engine using the rapidly updated and assimilated WRF-LES model. The spatiotemporal graph convolutional network ST-GCN is used, with observation stations as nodes, spatial correlation of equipment (such as the topographic relief and distance attenuation coefficient between two stations as edge weights), and multidimensional risk factor time series as features, to predict the risk probability of the grid in the next 10 minutes. For example, the turbulent risk probability in hilly areas reaches 85%. Finally, it is determined whether a warning is triggered. For example, when WSI≥3 or EDR≥0.4, it is defined as a red risk, and control instructions for no-fly and emergency avoidance are issued, and obstacle avoidance vectors are sent to the drone.

[0090] In a preferred embodiment, the method for outputting graded early warning instructions by using threat factor values ​​calculated from short-term forecast fields as input, combining them with a risk probability map generated by a spatiotemporal graph convolutional network, and then performing fusion analysis using a dynamic alarm rule base is as follows:

[0091] Using a spatiotemporal graph convolutional network, with observation stations as nodes, spatial correlation of equipment as edge weights, and time series of the threat factors in multiple dimensions as features, the network outputs the risk probability of a 100-meter grid in the next few minutes.

[0092] A dynamic alarm rule base is constructed. Threat factor values ​​are input into the dynamic alarm rule base and matched with several alarm rules. The alarm rules that are successfully matched are set as the alarm strategies to be triggered.

[0093] The warning area is transformed into a dynamic digital geofence, which is displayed by overlaying a route heatmap. When an alarm policy is triggered, the control command corresponding to the alarm policy is automatically pushed to the digital geofence that caused the alarm policy to be triggered.

[0094] In this embodiment, a graph structure is constructed using observation stations as nodes and spatial correlation of equipment as edge weights. Essentially, this explicitly encodes the spatial propagation or influence relationships between stations into the model. Multiple threat factors, such as turbulence, visibility, and icing, are input as node features in time-series format. ST-GCN simultaneously captures the spatial correlation between stations and the dynamic patterns of threat evolution over time, thereby outputting a 100-meter grid risk probability map for the next few minutes. This embodiment introduces a dynamic alarm rule base as a business-controllable interpretation and constraint mechanism. The approach involves inputting threat factor values ​​into the rule base and matching them with multiple alarm rules; once a rule is satisfied, the corresponding rule is set as the triggered alarm strategy. This solution transforms the warning area into a dynamic digital geofence and displays it through overlay of flight route heatmaps, ensuring that command personnel can intuitively see which key flight routes are covered by the risk area. When an alarm strategy is triggered, the system pushes the corresponding control instructions to the fence that caused the trigger, thereby achieving precise spatial delivery and minimal impact control, avoiding unnecessary restrictions on unrelated airspace or aircraft.

[0095] In a preferred embodiment, the dynamic alarm rule base includes: when WSI≥3 or EDR≥0.4, it is defined as a red risk, and control instructions for no-fly and emergency avoidance are issued; when 2≤WSI≤3 and LVI≤1km, it is defined as a yellow risk, and control instructions for speed limit and altitude lock are issued; when ISI≥60%, it is defined as a blue risk, and control instructions for avoiding cold clouds are issued.

[0096] In this embodiment, by setting configurable thresholds and combinations of conditions for key threat factors, complex meteorological risks are transformed into actionable tiered policy outputs. The core lies in binding the indicator system with action strategies, thus making the system output of which control actions should be taken for which state more clearly and specifically. The advantages include consistent model output, continuous optimization of business strategies based on operational experience, regulatory requirements, and seasonal climate differences, and clear traceability for each alarm, facilitating auditing.

[0097] In a preferred embodiment, the airspace capacity of each grid cell is assessed in real time, and high-risk airspace is automatically closed. Based on an avoidance algorithm, the method for dynamically generating or adjusting three-dimensional flight paths for affected aircraft is as follows: An airspace capacity assessment matrix is ​​constructed, meteorological elements are linked to airspace levels, and a density attenuation coefficient corresponding to turbulence intensity is set; the target airspace is divided into several 1km×1km×100m micro-airspace management units, and meteorological safety scores are calculated in real time. :

[0098] ;

[0099] Where TIavg is the average turbulence intensity within the cell and Vis is the visibility; when calculating the meteorological safety score When the value is below 60, the micro-spatial management unit is automatically shut down; it receives real-time wind field grid data and spatial polygon sets of turbulence hazard zones, and uses Dijkstra-A* hybrid path planning to solve the objective function:

[0100] ;

[0101] Where, ‖ΔP‖ represents the magnitude of the displacement vector of the path, V wind For wind speed vectors, TI is the unit vector of the flight direction. zone For turbulence intensity, I in_turbulence The indicator function is α, and β are weighting coefficients. The route is optimized by minimizing the cost function (Cost), generating a continuous three-dimensional waypoint sequence to avoid high-risk micro-airspace management units. The micro-airspace management unit status database is updated every few minutes. Flight plans passing through areas with a meteorological safety score no greater than 70 are replanned, and structured instructions containing new waypoints, altitude layers, and effective times are issued to aircraft via air-to-ground communication links. The system detects whether flight plans cross red alert zones; if high risk exists, a meteorological avoidance algorithm is invoked to generate new routes, which are then issued and updated in real-time via the command and control link to maintain airspace occupancy status. In low-risk scenarios, the cruise altitude layer is optimized based on the wind field.

[0102] In this embodiment, the passability of grid cells is first assessed in real time, then high-risk airspace is automatically closed. Subsequently, a path planning algorithm is invoked to generate or adjust three-dimensional routes for affected aircraft, and structured instructions are issued and updated on a fixed cycle to form a closed-loop dynamic control and avoidance system. This embodiment receives real-time wind field grid data and a set of spatial polygons representing turbulent hazard zones. Dijkstra-A* hybrid path planning is used to solve the objective function, which considers both wind speed drag and turbulence crossing penalties. The displacement vector magnitude reflects the flight cost, the wind speed vector and the flight direction unit vector reflect the impact of headwinds and tailwinds on efficiency, and the turbulence intensity and indicator function impose additional penalties for crossing hazard zones. The optimization result is to select a more economical, tailwind-friendly, and turbulence-free three-dimensional path point sequence while avoiding high-risk micro-airspaces, achieving a controllable trade-off between safety and efficiency.

[0103] Figure 3 An internal structural diagram of a computer device in one embodiment is shown. Specifically, this computer device may be... Figure 1 Computer equipment 120. (e.g.) Figure 3As shown, the computer device includes a processor, memory, network interface, input device, and display screen connected via a system bus. The memory includes non-volatile storage media and internal memory. The non-volatile storage media stores an operating system and may also store a computer program that, when executed by the processor, enables the processor to implement low-altitude weather protection methods. The internal memory may also store a computer program that, when executed by the processor, enables the processor to implement low-altitude weather protection methods.

[0104] Those skilled in the art will understand that Figure 3 The structure shown is merely a block diagram of a portion of the structure related to the present application and does not constitute a limitation on the computer device to which the present application is applied. Specific computer devices may include more or fewer components than those shown in the figure, or combine certain components, or have different component arrangements.

[0105] In one embodiment, the low-altitude meteorological protection method provided in this application can be implemented as a computer program, which can be implemented as follows: Figure 3 It runs on the device shown.

[0106] In one embodiment, a low-altitude weather protection system is provided, the system including a memory and a processor, the memory storing a computer program that, when executed by the processor, causes the processor to perform the steps of the low-altitude weather protection method as described above.

[0107] In this embodiment, the system can be a computer hardware system that executes its corresponding methods when running. For a description of the low-altitude weather protection method, please refer to the above text; it will not be repeated here.

[0108] It should be understood that although the steps in the flowcharts of the various embodiments of this application are shown sequentially according to the arrows, these steps are not necessarily executed in the order indicated by the arrows. Unless explicitly stated herein, there is no strict order restriction on the execution of these steps, and they can be executed in other orders. Moreover, at least some steps in each embodiment may include multiple sub-steps or multiple stages. These sub-steps or stages are not necessarily completed at the same time, but can be executed at different times. The execution order of these sub-steps or stages is not necessarily sequential, but can be performed alternately or in turn with other steps or at least a portion of the sub-steps or stages of other steps.

[0109] The technical features of the above embodiments can be combined in any way. For the sake of brevity, not all possible combinations of the technical features in the above embodiments are described. However, as long as there is no contradiction in the combination of these technical features, they should be considered to be within the scope of this specification.

Claims

1. A method for providing low-altitude meteorological support, characterized in that, The method includes: S10: Acquire multi-source observation data of the target area from ground-based, air-based, and space-based sources; perform spatiotemporal benchmark unification and multi-level quality control on the multi-source observation data to generate a standardized observation dataset; S20: Perform data fusion on standardized data to generate a three-dimensional dynamic field with a resolution of 100 meters; S30: Input the three-dimensional dynamic field into the boundary layer numerical model and use the fast update cyclic assimilation technique to generate a short-term forecast field for low-altitude weather. S40: Using the threat factor values ​​calculated from the short-term forecast field as input, combined with the risk probability map generated by the spatiotemporal graph convolutional network, and after fusion analysis by the dynamic alarm rule base, outputs graded early warning instructions; S50: Maps the risk probability map and graded early warning instructions to a digital airspace grid to obtain airspace status information; assesses the passability of each grid cell in real time and automatically closes high-risk airspace; and dynamically generates or adjusts three-dimensional routes for affected aircraft based on avoidance algorithms. S60: Through multiple platforms, it releases integrated risk probability maps and airspace status information in real time, completing the closed loop of protection services.

2. The low-altitude meteorological support method according to claim 1, characterized in that, The method for unifying the spatiotemporal reference and performing multi-level quality control on the multi-source observation data to generate a standardized observation dataset is as follows: The timestamps of the multi-source observation data are corrected based on a unified time reference; for observation data with different time resolutions, interpolation or resampling is performed according to a preset time step to unify the time reference of the multi-source observation data. Based on a unified geographic coordinate system, the multi-source observation data are spatially mapped. The observation data is converted into a unified vertical height layer structure to unify the spatial reference of multi-source observation data; By removing obvious anomalies, distortions, and / or physical meanings from the data processed through time and space benchmark unification, a standardized observation dataset is obtained.

3. The low-altitude meteorological support method according to claim 1, characterized in that, The method for generating a 3D dynamic field with a 100-meter resolution by fusing standardized data is as follows: By integrating data from microwave radiometers, lidar, and radiosondes, and using a variational optimization algorithm based on physical constraints, an integrated vertical profile of temperature, humidity, and wind at the minute level is generated within an altitude range of 0-1500m. The integrated vertical profile is unified into a 200m×200m×50m spatial grid, and Kriging interpolation is used to map discrete points to the spatial grid to form a three-dimensional dynamic field.

4. The low-altitude meteorological support method according to claim 1, characterized in that, The method for generating short-term low-altitude meteorological forecast fields by inputting a three-dimensional dynamic field into a boundary layer numerical model and employing a fast-update cyclic assimilation technique is as follows: In the boundary layer numerical model, an assimilation time window oriented towards the low altitude is constructed; an ensemble variational hybrid assimilation framework is adopted to construct the model initial field for assimilation processing. Within the assimilation time window, the three-dimensional dynamic field is injected as an observation increment into the boundary layer numerical model to update the model state; the state update process is used to correct the spatial distribution and vertical structure of low-altitude meteorological elements in the model. During the assimilation update process, the boundary layer parameters are updated synchronously. After completing one assimilation update, the next time window is entered based on the updated mode state, and the assimilation update process is repeated. After completing at least one rapid update cycle assimilation process, forecast calculations are performed based on the updated boundary layer numerical model state; and a short-term low-altitude meteorological forecast field is generated within the target protection area.

5. A low-altitude meteorological support method according to claim 1, characterized in that, Threat factors extracted from the short-term forecast field include one or more of the following factors: Ice accumulation index threat factor, eddy dissipation rate threat factor, low visibility index threat factor, and wind shear index threat factor.

6. A low-altitude meteorological support method according to claim 5, characterized in that, The method for obtaining the Ice Accumulation Index Threat Factor (ISI) is as follows: By integrating the liquid water content (LWC) from microwave radiometers, flight velocity (V), exposure time (t), temperature function (f(T)), critical icing thickness (τ), and correction coefficient (k), an icing index model is constructed: ； The turbulence intensity is inverted using lidar spectral width to obtain the value of the eddy dissipation rate threat factor (EDR): ； Where σ is the Doppler spectral width, σ0 is the inherent spectral width caused by instrument noise and molecular thermal motion, k is an empirical constant, and V eff The average wind speed within the effective scanning volume is given by ∆t, which represents the time resolution. The Low Visibility Index (LVI) threat factor is obtained as follows: By combining ground visibility observation, lidar backscattering coefficient, and cloud radar fog height data, the above data are processed based on the random forest regression algorithm to output a 100-meter grid visibility estimate. The method for obtaining the Wind Shear Index Threat Factor (WSI) is as follows: Hybrid modeling based on lidar vertical wind gradient and turbulent kinetic energy; This involves using lidar to create highly layered structures in the vertical direction. Measuring horizontal wind speed The gradient is calculated using the central difference method: ； Calculation of turbulent kinetic energy using the standard deviation of three-dimensional wind speed fluctuations from lidar: ； The wind shear index is calculated by weighted fusion of vertical wind gradient and turbulent kinetic energy. ； Where α and β are weighting coefficients.

7. A low-altitude meteorological support method according to claim 1, characterized in that, Using threat factor values ​​calculated from short-term forecast fields as input, combined with a risk probability map generated by a spatiotemporal graph convolutional network, and fused and analyzed through a dynamic alarm rule base, the method for outputting tiered early warning instructions is as follows: Using a spatiotemporal graph convolutional network, with observation stations as nodes, spatial correlation of equipment as edge weights, and time series of the threat factors in multiple dimensions as features, the network outputs the risk probability of a 100-meter grid in the next few minutes. A dynamic alarm rule base is constructed. Threat factor values ​​are input into the dynamic alarm rule base and matched with several alarm rules. The alarm rules that are successfully matched are set as the alarm strategies to be triggered. The warning area is transformed into a dynamic digital geofence, which is displayed by overlaying a route heatmap. When an alarm policy is triggered, the control command corresponding to the alarm policy is automatically pushed to the digital geofence that caused the alarm policy to be triggered.

8. A low-altitude meteorological support method according to claim 6, characterized in that, The dynamic alarm rule base includes: When WSI≥3 or EDR≥0.4, it is defined as a red risk, and control orders for no-fly and emergency avoidance are issued. When 2≤WSI≤3 and LVI≤1km, it is defined as a yellow risk, and speed limit and altitude lock control instructions are issued; When ISI ≥ 60%, it is defined as a blue risk, and a control order to avoid cold clouds is issued.

9. A low-altitude meteorological support method according to claim 1, characterized in that, The method involves real-time assessment of the traffic capacity of each grid cell and automatic closure of high-risk airspace; based on avoidance algorithms, the method for dynamically generating or adjusting 3D routes for affected aircraft is as follows: Link meteorological elements with traffic levels and set density attenuation coefficients corresponding to turbulence intensity; The target airspace is divided into several 1km×1km×100m micro-airspace management units, and meteorological safety scores are calculated in real time. : ； Where TIavg is the average turbulence intensity within the cell and Vis is the visibility; When calculating meteorological safety scores If the value is below 60, the micro-space management unit will be automatically shut down. The system receives real-time wind field grid data and a set of spatial polygons representing turbulent hazard zones, and uses Dijkstra-A* hybrid path planning to solve the objective function. ； Where, ‖ΔP‖ represents the magnitude of the displacement vector of the path, V wind For wind speed vectors, TI is the unit vector of the flight direction. zone For turbulence intensity, I in_turbulence The indicator function is α, and β are weighting coefficients. By minimizing the cost function, the route is optimized to generate a continuous three-dimensional waypoint sequence that avoids high-risk micro-airspace management units; The micro airspace management unit status database is updated every few minutes. Flight plans that pass through areas with a weather safety score of no more than 70 points are replanned, and structured instructions containing new waypoints, altitude layers, and effective times are issued to aircraft through the air-to-ground communication link. The system detects whether the flight plan crosses the red alert zone. If there is a high risk, it calls the weather avoidance algorithm to generate a new flight path and sends it out in real time through the command and control link to update the airspace occupancy status. In low-risk scenarios, it optimizes the cruise altitude layer according to the wind field.

10. A low-altitude meteorological support system, characterized in that, It includes a memory and a processor, wherein the memory stores a computer program that, when executed by the processor, causes the processor to perform the steps of the low-altitude weather protection method as described in any one of claims 1 to 9.

Images