Intelligent tornado identification method and system

Abstract

The invention provides a tornado intelligent identification method and system, and the method comprises the steps: carrying out the feature extraction of obtained multi-source meteorological data, obtaining meteorological feature parameters, carrying out the tornado risk identification based on the meteorological feature parameters through employing a grading early warning rule, and generating a tornado identification result. And precise identification and early warning of the tornado are realized.

Description

Technical Field

[0001] This invention relates to the field of meteorological disaster monitoring and early warning, specifically to a method and system for intelligent tornado identification. Background Technology

[0002] Tornadoes are violently rotating convective weather phenomena characterized by their small scale, short lifespan, sudden onset, and immense destructive power. Their rapid formation and destructive nature pose a serious threat to people's lives, property, and social stability. Therefore, achieving accurate and automatic identification of tornadoes is a core challenge and urgent need in the field of meteorological disaster prevention and mitigation.

[0003] Currently, tornado monitoring and early warning mainly rely on Doppler weather radar and tornado occurrence probability prediction models based on meteorological environmental parameters. Radar identifies mesocyclones and vortices by analyzing reflectivity factors and radial velocity products, constituting the primary technical means for current early warning. However, radar detection is affected by the Earth's curvature and topographic obstruction, resulting in observation blind spots in the lower atmosphere, while tornadoes typically form in the near-surface layer. Furthermore, algorithms rely on empirical thresholds, limiting their ability to identify weak vortex signals and non-supercell tornadoes, leading to both false alarms and missed detections.

[0004] Geostationary meteorological satellites can provide continuous monitoring with high spatiotemporal resolution. Through cloud top brightness temperature, particle data, and microphysical characteristics, they effectively capture the macroscopic dynamic development of strong convective clouds, such as their upward movement and cooling rate, offering significant advantages in terms of wide coverage and lack of geographical limitations. Lightning location systems can monitor the frequency and spatial distribution of lightning in real time, unaffected by terrain obstruction. They can reflect the electrification process and intensity of microphysical activity within storms, providing independent information for assessing the intensity evolution of convective systems. However, the application of satellite and lightning data in tornado identification is still in an auxiliary stage and has not yet achieved deep integration with radar data. Furthermore, satellites cannot directly observe low-level vortices, and the correspondence between lightning and tornadoes is not unique, resulting in limited effectiveness when used alone. Currently, there is a lack of effective methods to integrate the low-level vortex detection capabilities of radar, the macroscopic dynamic monitoring advantages of satellites, and the indicative role of lightning's microphysical processes.

[0005] Therefore, there is an urgent need to develop an intelligent identification method that can deeply integrate multi-source observation data, break through the dependence of existing technologies on a single data source, and achieve more accurate and earlier automatic identification and early warning of tornadoes. Summary of the Invention

[0006] To address the problem that existing technologies rely on a single data source and cannot deeply integrate multi-source observation data, this invention proposes a tornado intelligent identification method and system.

[0007] Firstly, a method for intelligent tornado identification is provided, including: Meteorological feature parameters are obtained by extracting features from the acquired multi-source meteorological data. These meteorological feature parameters include radar reflectivity features, satellite overshoot cloud top center features, lightning surge features, and lightning hole features. Based on the meteorological characteristic parameters, a graded early warning rule is used to identify tornado risks and generate tornado identification results.

[0008] Secondly, a tornado intelligent recognition system is provided, including: The extraction module is used to extract meteorological feature parameters based on the acquired multi-source meteorological data. The meteorological feature parameters include radar reflectivity features, satellite overshoot cloud top center features, lightning surge features, and lightning hole features. The identification module is used to identify tornado risks based on the meteorological characteristic parameters and adopt graded early warning rules, and generate tornado identification results.

[0009] In another aspect, this application also provides an electronic device, comprising: at least one processor and a memory; the memory and the processor are connected via a bus; The memory is used to store one or more programs; When the one or more programs are executed by the at least one processor, a tornado intelligent recognition method as described above is implemented.

[0010] In another aspect, this application also provides a computer-readable storage medium having a computer program stored thereon, which, when executed, implements a tornado intelligent identification method as described above.

[0011] Compared with the prior art, the beneficial effects of the present invention are as follows: This invention provides a method and system for intelligent tornado identification. The method extracts meteorological feature parameters from multi-source meteorological data, and uses hierarchical early warning rules based on the meteorological feature parameters to identify tornado risks and generate tornado identification results, thereby achieving accurate identification and early warning of tornadoes.

[0012] This invention achieves accurate identification and early warning of tornadoes through multi-source data fusion and multi-feature combination recognition algorithms. Compared with existing technologies, it has the following significant advantages: high identification accuracy, effectively overcoming the limitations of a single data source through multi-source feature fusion; low false alarm rate, effectively distinguishing between tornadoes and non-tornado strong convection through joint multi-feature identification; strong adaptability, the system can adapt to different geographical environments and weather conditions, especially maintaining good identification capabilities even in radar detection blind spots; and high degree of automation, achieving full automation from data acquisition to early warning issuance, reducing reliance on human experience. Attached Figure Description

[0013] Figure 1 This is a flowchart of the intelligent tornado recognition method of the present invention; Figure 2 This is a schematic diagram of the process for extracting the number of overshoot cloud top centers in the intelligent tornado identification method of the present invention; Figure 3 This is a time series diagram of the number of OT centers in the intelligent tornado identification method of the present invention; Figure 4 This is a schematic diagram of the lightning surge recognition result of the tornado intelligent recognition method of the present invention; Figure 5 This is a schematic diagram of lightning holes and radar superimposed on the tornado intelligent identification method of the present invention; Figure 6 This is a schematic diagram of the tornado intelligent recognition system of the present invention; Figure 7 This is a schematic diagram of an electronic device structure according to the present invention. Detailed Implementation

[0014] This invention provides an intelligent tornado identification method and system. By utilizing multi-source data fusion analysis technology, combining lightning activity characteristics (including lightning surges and lightning holes) and satellite overshoot cloud top (OT) characteristics, it processes lightning, radar, and satellite data with high spatiotemporal resolution, and then extracts key identification indicators of the tornado parent thunderstorm. The multi-source monitoring information is integrated into a unified early warning decision framework for efficient and accurate tornado identification. It can accurately capture the spatial structure characteristics, temporal evolution patterns, and intensity development trends of tornado storms, thereby achieving early identification and accurate judgment of tornadoes. It has high practicality and reliability.

[0015] To better understand the present invention, the following description, in conjunction with the accompanying drawings and embodiments, will further illustrate the content of the present invention.

[0016] Example 1: A tornado intelligent identification method, such as Figure 1 As shown, it includes: Step 1: Based on the acquired multi-source meteorological data, feature extraction is performed to obtain meteorological feature parameters, which include radar reflectivity features, satellite overshoot cloud top center features, lightning surge features, and lightning hole features. Step 2: Based on the meteorological characteristic parameters, use graded early warning rules to identify tornado risks and generate tornado identification results.

[0017] The multi-source meteorological data includes radar data, satellite data, and lightning data.

[0018] In this embodiment, before extracting meteorological feature data from the acquired multi-source meteorological data, quality control processing can be performed on the multi-source meteorological data to remove noise and ambiguity issues, making the acquired data more accurate. Specifically, this includes: The radar data from the acquired multi-source meteorological data were subjected to velocity ambiguity processing using a dual PRF algorithm, and the processed radar data were subjected to ground clutter suppression using a clutter identification algorithm to obtain suppressed lightning data. Radiometric calibration and geometric processing are performed on satellite data from multi-source meteorological data to obtain processed satellite data. Radiometric calibration involves converting raw count values ​​into brightness temperature values, and geometric correction involves projecting the data onto geographic coordinate coefficients. The lightning data from the acquired multi-source meteorological data were subjected to positioning accuracy control and deduplication processing. Positioning accuracy control involved removing lightning records with fewer than the preset number of participating stations, while deduplication was performed using a spatiotemporal window method.

[0019] In one specific embodiment, the radar data uses the reflectivity of a weather radar network. Subsequently, quality control is performed on the radar data. First, velocity deblurring is performed using a dual PRF algorithm to address velocity ambiguity; then, ground clutter suppression is performed using a clutter identification algorithm based on a terrain database.

[0020] Satellite data acquisition was performed using the Himawari-9 satellite. Radiometric calibration and geometric correction were then applied to the satellite data: radiometric calibration converted the raw count values ​​into brightness temperature values ​​using officially provided calibration coefficients; geometric correction used satellite attitude and orbital parameters to project the data onto a geographic coordinate system, achieving pixel-level positioning accuracy.

[0021] Lightning data is acquired from the State Grid Lightning Monitoring Center via a dedicated line, and the data format is txt. Location accuracy control and deduplication are performed on the lightning data: first, lightning records with fewer than M (M is generally set to 3) are removed; then, a spatiotemporal window method is used to remove duplicate lightning records.

[0022] In this embodiment, during the process of extracting meteorological feature parameters from multi-source meteorological data in step 1, the radar data, satellite data, and lightning data from the multi-source meteorological data are processed specifically to extract the characteristic features of tornadoes, thereby achieving accurate identification of tornadoes in the subsequent process. Specifically, this includes: The acquired multi-source meteorological data is subjected to quality control processing to obtain processed multi-source meteorological data. Based on the processed radar data from the processed multi-source meteorological data, storm cells that meet the preset intensity and preset area thresholds are identified on the combined reflectivity factor map, and radar reflectivity characteristics are determined based on the storm cells. The brightness temperature difference value is calculated based on the water vapor channel and infrared window channel in the processed satellite data from the processed multi-source meteorological data, and multiple target OT centers are identified and marked as satellite overshoot cloud top center features based on the brightness temperature difference value. A density-based clustering algorithm is used to spatially cluster lightning data in the processed multi-source meteorological data to obtain multiple lightning clusters representing different thunderstorm cells. Based on the lightning clusters, the lightning frequency jump signal is identified to obtain lightning jump characteristics. Lightning data is aggregated according to a preset time window to generate lightning distribution data. The lightning distribution data is then overlaid with radar composite reflectivity products and OT centers within the same time window to generate an overlay map. Spatial regions in the overlay map where the radar composite reflectivity meets a preset reflectivity and lightning activity meets a preset activity sparsity condition are identified as lightning void features.

[0023] In this embodiment, during the processing of radar data, storm cells on the combined reflectivity factor map are identified, and strong updraft regions and morphological characteristics are determined based on the storm cells, thereby determining the radar reflectivity characteristics. Specifically, this includes: Calculate the geometric center of the pixels inside the storm cell that exceed a preset reflectivity threshold; The storm cell was subjected to morphological feature identification to determine the hook-shaped echo region; The geometric center and the hook-shaped echo region are used as radar reflectivity features.

[0024] In one specific embodiment, the feature extraction process during radar data processing includes: (1) Reflectivity analysis: Identify discrete storm cells that meet certain intensity and area thresholds on the combined reflectivity factor map.

[0025] (2) For each storm cell, a high reflectivity threshold is set to identify the core strength inside the storm cell.

[0026] (3) Core geometric center calculation: Calculate the geometric center coordinates of all pixels within the unit that exceed the above reflectivity threshold. This center point represents the core position of the unit in a region of strong updraft.

[0027] (4) Hook echo identification: Hook echo identification depends on its typical morphological features, namely a sharp, curved weak echo area extending downstream from a main high reflectivity factor nucleus.

[0028] In this embodiment, during satellite data processing and target OT center marking, the brightness-temperature difference value is calculated through the water vapor channel and the infrared window channel. This brightness-temperature difference value is then used to filter shallow convection regions, and further, based on these shallow convection regions, the target OT center is accurately identified. Specifically, this includes: Based on the brightness temperature difference value, the region where the brightness temperature difference value is greater than the preset brightness temperature difference threshold is determined as the shallow convection region; Closed brightness temperature contour lines are extracted at preset fixed intervals within the shallow convection region, and the region located within half of the contour lines is identified as a potential OT center region. Pixels whose infrared window brightness temperature value is lower than the brightness temperature value of adjacent pixels within a preset range in the potential OT center area are identified as candidate OT centers. For each candidate OT center, the innermost closed brightness temperature contour line containing the center is determined, wherein the area enclosed by the innermost closed brightness temperature contour line satisfies the preset spatial coverage condition, and has a preset number of pixels distributed in multiple directions with the candidate OT center as the origin. Calculate the average infrared window brightness temperature of all pixels in the anvil cloud region within the innermost closed brightness temperature contour line as the average anvil cloud temperature. The anvil cloud region is the pixel region between the innermost closed brightness temperature contour line and the innermost closed brightness temperature contour line. Calculate the difference between the infrared window brightness temperature value of the candidate OT center and the average anvil cloud temperature, and mark the candidate OT centers whose difference satisfies the preset gradient threshold as target OT centers; The identified and marked target OT center is used as the satellite overshoot cloud top center feature.

[0029] In one specific embodiment, the process of extracting the number of overshoot cloud top centers and determining the characteristics of satellite overshoot cloud top centers includes: (1) Data acquisition and preprocessing: Acquire satellite multi-channel remote sensing data, including brightness temperature data of the water vapor channel (channel 8, WV) and the infrared window channel (channel 13, IRW). Taking a satellite image at a specified time as an example, calculate the brightness temperature difference (BTD) between the water vapor channel and the infrared window channel, i.e., BTD = WV - IRW.

[0030] (2) Preliminary screening of shallow convection regions: Set a BTD threshold and retain only regions with BTD greater than the threshold to preliminarily screen out pixel regions that may have shallow convection.

[0031] (3) Extraction of closed brightness temperature contour lines: In the area that has been preliminarily screened, a series of closed brightness temperature contour lines are extracted at fixed intervals.

[0032] (4) Identification of potential OT center areas: Potential OT centers are more likely to appear in the colder half-side isoline areas. Specifically, the areas located within the half-side isoline are identified as potential OT center areas.

[0033] (5) Candidate OT center pixel identification: In the potential OT center area, each pixel is checked one by one. If the infrared window area brightness temperature value of a certain pixel is lower than the brightness temperature values ​​of its 8 neighboring pixels, then the pixel is identified as a candidate OT center.

[0034] (6) Determination of the innermost closed contour line (Ca): For each candidate OT center, find the innermost closed brightness temperature contour line containing the center, denoted as Ca. The area enclosed by Ca is required to contain at least 12 pixels in 16 directions with a radius of 8km from the candidate center as the origin.

[0035] (7) Definition of the anvil cloud region and calculation of the average anvil cloud temperature: The pixel region located between Ca-1 (i.e. the inner layer of contour line immediately adjacent to Ca) and Ca is defined as the anvil cloud region. The average infrared window brightness temperature of all pixels in this region is calculated as the average anvil cloud temperature.

[0036] (8) Final determination of OT center: Compare the infrared window brightness temperature of the candidate OT center pixel with the average anvil cloud temperature obtained in step (7). If the brightness temperature of the candidate center is lower than the average anvil cloud temperature by a specified threshold, the candidate center is considered to have sufficient brightness temperature gradient and is finally determined to be the OT center.

[0037] (9) Output identification results: Mark and output the number of all OT centers that meet the conditions, and complete the identification of the number of OT centers in the tornado mother storm at this moment.

[0038] In this embodiment, during the processing of lightning data to obtain lightning surge characteristics, a lightning cell tracking algorithm is used to identify and track the movement trajectory of lightning cells to obtain a ground flash frequency sequence. Then, based on the ground flash frequency sequence, lightning frequency surge signals are identified to obtain lightning surge characteristics, which specifically include: Based on the lightning cluster, a thunderstorm cell tracking algorithm is used to identify and track the movement trajectory of a specific thunderstorm cell corresponding to the storm system; The frequency of ground flashes within a preset frequency range is determined based on the movement trajectory, and a ground flash frequency sequence is generated based on the ground flash frequency. The lightning frequency change rate sequence is calculated based on the lightning frequency sequence, and the standard deviation of the lightning frequency change rate sequence is calculated based on the lightning frequency change rate sequence and a preset statistical backtracking time window. A dynamic judgment threshold is set based on the standard deviation of the lightning frequency change rate sequence, and a lightning surge signal is generated as a lightning surge feature when the lightning frequency change rate at the current moment exceeds the dynamic judgment threshold.

[0039] In one specific embodiment, the process of processing lightning data to obtain lightning surge characteristics includes: (1) Lightning data preprocessing and density clustering: Acquire lightning detection data for the target area; divide the lightning location data within each time window into an independent dataset at 1-minute intervals; perform spatial clustering analysis on the lightning data for each minute using the density-based DBSCAN clustering algorithm. Set an appropriate neighborhood radius parameter eps to form the minimum number of points min_samples required for clustering. Through this step, discrete lightning points are aggregated into lightning clusters representing different thunderstorm cells, and the cluster center coordinates of each lightning cluster are calculated.

[0040] (2) Thunderstorm trajectory tracking and ground flash frequency sequence construction: Based on the lightning cluster centers obtained in step (1) every minute, a thunderstorm cell tracking algorithm is used to identify and track the movement trajectory of a specific thunderstorm cell corresponding to the storm system; the frequency of ground flashes of the target thunderstorm cell in each minute is counted. To smooth out random fluctuations and highlight trends, the moving average of the ground flash frequency over adjacent 2 minutes is calculated to form a ground flash frequency sequence F that reflects the intensity of lightning activity of the thunderstorm cell over time, denoted as F=[f1,f2,...,f n ], where f n This represents the average number of flashes during the nth 2-minute time window.

[0041] (3) Calculation of the rate of change of ground flash frequency: Based on the lightning flash frequency sequence F obtained in step S2, the rate of change of lightning flash frequency at adjacent time points in the sequence is calculated to obtain the lightning flash frequency change rate sequence D; specifically, the rate of change D at the i-th time point... i The calculation formula is: D i =(F i –F i-1 ) / 2 where F i and F i-1 These represent the average flash frequency over 2 minutes at time i and i-1, respectively.

[0042] (4) Identification of lightning frequency surge signals: This step is crucial for identifying key early warning signals: First, a statistical backtracking time window is set, which is 2 to 12 minutes before the current identification time, for a total of 10 minutes (i.e., 5 data points at 2-minute intervals). Then, the standard deviation σ of the lightning frequency change rate sequence D within this retrospective time window is calculated to characterize the normal fluctuation level of lightning activity changes during this period. Finally, a judgment threshold is set, and the rate of change of the current lightning frequency is compared with the threshold. If the condition is met: the rate of change of the lightning frequency is greater than the threshold, then it is determined that a significant "lightning frequency surge" event has occurred at the current moment. This event is an important indication signal that the tornado parent thunderstorm may be intensifying or about to generate a tornado.

[0043] In one specific embodiment, the process of processing lightning data to obtain lightning void characteristics includes: (1) Data preparation and preprocessing: Acquire radar combined reflectivity data and lightning location data. Ensure that the radar and lightning data have been accurately registered in time and space.

[0044] (2) Spatiotemporal aggregation of lightning data: The continuously acquired CG lightning location data are aggregated according to a preset time window (preferably 6 minutes) to generate a lightning density or frequency distribution map within that time period.

[0045] (3) Radar reflectivity and lightning data overlay analysis: The aggregated lightning distribution data obtained in step (2) is spatially overlaid with the radar composite reflectivity product and OT center within the same time window to generate a lightning-reflectivity-OT center overlay map.

[0046] (4) Identification of lightning void areas: In the overlay map, identify spatial areas with high radar composite reflectivity but significantly sparse or absent lightning activity, and preliminarily determine such areas as potential "lightning voids".

[0047] In this embodiment, after obtaining meteorological feature parameters based on the aforementioned step 1 in step 2, a tiered early warning rule is used to identify tornado risk. This method achieves accurate tornado identification through combinations of different meteorological feature parameters, specifically including: When multiple lightning surge signals occur within a preset time interval and the number of target OT centers exceeds a preset data threshold, a Level 3 tornado identification and warning will be triggered. When a Level 3 tornado identification and warning is triggered and lightning hole characteristics are detected, a Level 2 tornado identification and warning will be triggered. If a Level II tornado identification and warning is triggered and radar reflectivity characteristics are detected, a Level I tornado identification and warning will be triggered.

[0048] In one specific embodiment, tornado risk identification specifically includes: Establish tornado recognition rules based on feature combinations: Primary identification: When a certain number of lightning frequency spikes and the number of thunderstorm OT centers exceeding a certain threshold are detected simultaneously within 30 minutes, a Level 3 (general) warning is triggered; Intermediate identification: If lightning hole features are detected simultaneously based on primary identification, a secondary (severe) warning is triggered; Advanced identification: Based on intermediate identification, if the radar detects hook-shaped echo area features, a Level 1 (more severe) warning is triggered.

[0049] Furthermore, a three-level early warning product is generated based on the identification results: Level 3 Warning (Attention): Meets the initial identification criteria; Level 2 Warning (Early Warning): Meets the criteria for intermediate-level identification; Level 1 Warning (Emergency): Meets advanced identification criteria; The early warning product includes information such as the potential areas and times of tornadoes, encapsulated in XML format. A WebGIS platform is provided to display the identification results and early warning information in real time. Example 2: The following is a detailed description of the specific application of the intelligent tornado recognition method of the present invention, which includes the following steps: Step 1: Data Acquisition and Quality Control The radar data used was the reflectivity of a weather radar network. Quality control was then implemented on the radar data. First, velocity deblurring was performed using a dual PRF algorithm to address velocity ambiguity. Next, ground clutter suppression was performed using a clutter identification algorithm based on a terrain database.

[0050] Data was acquired from the Himawari-9 satellite. Radiometric calibration and geometric correction were performed on the satellite data: radiometric calibration converted the raw count values ​​into brightness temperature values ​​using officially provided calibration coefficients; geometric correction used satellite attitude and orbital parameters to project the data onto a geographic coordinate system, achieving pixel-level positioning accuracy.

[0051] Lightning data is obtained from the State Grid Lightning Monitoring Center via a dedicated line, and the data format is txt. Location accuracy control and deduplication are performed on the lightning data: first, lightning records with fewer than 3 participating stations are removed; then, a spatiotemporal window method (1 second time window, 20 km spatial window) is used to remove duplicate lightning records.

[0052] Step 2: Radar Feature Extraction (1) Reflectance analysis: Identify discrete storm cells that meet certain intensity (e.g., reflectance ≥ 35 dBZ) and area threshold on the combined reflectance factor map; (2) For each storm cell, set a high reflectivity threshold (e.g., 50 dBZ) to identify the core strength inside the cell.

[0053] (3) Core geometric center calculation: Calculate the geometric center coordinates of all pixels within the unit that exceed the above threshold. This center point represents the core position of the unit in the region of strong updraft.

[0054] (4) Identify the hook-shaped echo region by observing its typical morphological features, namely a sharp, curved weak echo region extending downstream from a main high reflectivity factor nucleus (usually ≥50dBZ).

[0055] Step 3: Extracting the number of passes through the center of the cloud top. (1) Data Acquisition and Preprocessing: Acquired multi-channel remote sensing data from the Himawari-9 satellite at 14:10 on April 27, 2024, including the water vapor channel (channel 8, WV, Figure 2 a) and infrared window area channels (channel 13, IRW, Figure 2 (b) Brightness temperature data. Taking a satellite image at a specified time (e.g., 14:10) as an example, calculate the brightness temperature difference (BTD) between the water vapor channel and the infrared window channel. Figure 2 c), that is, BTD = WV - IRW.

[0056] (2) Preliminary screening of shallow convection regions: Set a BTD threshold (e.g. -1K) and retain only regions with BTD greater than the threshold to preliminarily screen out pixel regions that may have shallow convection.

[0057] (3) Extraction of closed brightness temperature contour lines: In the region that has been preliminarily screened (BTD>-1K), a series of closed brightness temperature contour lines are extracted at fixed intervals (e.g., 5K). For example, six layers of closed contour lines can be extracted in sequence: C1=190K (innermost layer), C2=195K, C3=200K, C4=205K, C5=210K and C6=215K (outermost layer).

[0058] (4) Identification of potential OT center regions: Potential OT centers are more likely to appear in the colder half-side isoline regions. Specifically, the region located within the C3=200K isoline is identified as the region where potential OT centers are located. Figure 2 d), where latitude represents latitude.

[0059] (5) Candidate OT center pixel identification: In the potential OT center area, each pixel is checked one by one. If the infrared window area brightness temperature value of a certain pixel is lower than the brightness temperature values ​​of its 8 neighboring pixels, then the pixel is identified as a candidate OT center.

[0060] (6) Determination of the innermost closed contour line (Ca): For each candidate OT center, find the innermost closed brightness temperature contour line containing the center, denoted as Ca. The area enclosed by Ca is required to contain at least 12 pixels in 16 directions with a radius of 8km from the candidate center as the origin.

[0061] (7) Definition of the anvil cloud region and calculation of the average anvil cloud temperature: The pixel region located between Ca-1 (i.e. the inner layer of contour line immediately adjacent to Ca) and Ca is defined as the anvil cloud region. The average infrared window brightness temperature of all pixels in this region is calculated as the average anvil cloud temperature.

[0062] (8) Final determination of OT center: Compare the infrared window brightness temperature of the candidate OT center pixel with the average anvil cloud temperature obtained in step (7). If the brightness temperature of the candidate center is lower than the average anvil cloud temperature by a specified threshold (e.g., 6.5K), the candidate center is considered to have sufficient brightness temperature gradient and is finally determined to be the OT center.

[0063] (9) Output identification results: Mark and output the number of all OT centers that meet the conditions, and complete the identification of the number of OT centers in the tornado mother storm at this moment. Three OT centers can be identified (e.g. Figure 2 e).

[0064] (10) Record the number of OT centers at each time step to form an OT center time series (e.g. Figure 3 (As shown).

[0065] Step 4: Identification of Lightning Increase Characteristics (1) Lightning data preprocessing and density clustering: Lightning detection data for the target area is acquired. Lightning location data within each time window is treated as an independent dataset, with a 1-minute time interval. The density-based DBSCAN clustering algorithm is used to perform spatial clustering analysis on the lightning data for each minute. The neighborhood radius parameter eps is set to 0.2°, and the minimum number of points required to form a cluster (min_samples) is set to 5. This step aggregates discrete lightning points into lightning clusters representing different thunderstorm cells, and the cluster center coordinates of each lightning cluster are calculated.

[0066] (2) Thunderstorm trajectory tracking and ground flash frequency sequence construction: Based on the lightning cluster centers obtained in step (1) every minute, a thunderstorm cell tracking algorithm is used to identify and track the movement trajectory of a specific thunderstorm cell corresponding to the storm system; the frequency of ground flashes of the target thunderstorm cell in each minute is counted. To smooth out random fluctuations and highlight trends, the moving average of the ground flash frequency over adjacent 2 minutes is calculated to form a ground flash frequency sequence F that reflects the intensity of lightning activity of the thunderstorm cell over time, denoted as F=[f1,f2,...,f n ], where f n This represents the average number of flashes during the nth 2-minute time window.

[0067] (3) Calculation of the rate of change of ground flash frequency: Based on the lightning frequency sequence F obtained in step S2, the rate of change of lightning frequency at adjacent time points in the sequence is calculated to obtain the lightning frequency change rate sequence D; specifically, the formula for calculating the rate of change Di at the i-th time point is: Di = (Fi – Fi-1) / 2 Where Fi and Fi-1 represent the average 2-minute flash frequency at time i and i-1, respectively.

[0068] (4) Identification of lightning frequency surge signals: This step is crucial for identifying key early warning signals: First, a statistical backtracking time window is set, which is 2 to 12 minutes before the current identification time, for a total of 10 minutes (i.e., 5 data points at 2-minute intervals). Then, the standard deviation σ of the lightning frequency change rate sequence D within this retrospective time window is calculated to characterize the normal fluctuation level of lightning activity changes during this period. Finally, a judgment threshold is set, which is twice σ (2σ); the current lightning frequency change rate D_current is compared with the threshold; if the condition D_current>2σ is met, then a significant "lightning frequency jump" event is determined to have occurred at the current moment. This event is an important indicator signal that the tornado parent thunderstorm may be intensifying or about to form a tornado. Sixteen lightning frequency jumps can be identified in this thunderstorm process (e.g., Figure 4 (As shown).

[0069] Step 5: Lightning Hole Feature Identification (1) Data preparation and preprocessing: Acquire radar combined reflectivity data and lightning location data. Ensure that the radar and lightning data have been accurately registered in time and space.

[0070] (2) Spatiotemporal aggregation of lightning data: The continuously acquired CG lightning location data are aggregated according to a preset time window (preferably 6 minutes) to generate a lightning density or frequency distribution map within that time period.

[0071] (3) Radar reflectivity and lightning data overlay analysis: The aggregated lightning distribution data obtained in step (2) is spatially overlaid with the radar composite reflectivity product within the same time window to generate a lightning-reflectivity overlay map.

[0072] (4) Identification of lightning void regions: In the overlay image, identify spatial regions with high radar composite reflectivity (usually ≥55dBZ) but significantly sparse or absent lightning activity. These regions are initially identified as potential "lightning voids" (e.g., Figure 5 (As shown).

[0073] Step 6: Intelligent Recognition and Classification Establish tornado recognition rules based on feature combinations: Primary identification: When a certain number of lightning frequency spikes (≥3) and the number of thunderstorm OT centers (≥5) are detected simultaneously within 30 minutes, a Level 3 (general) warning is triggered, such as... Figure 4 Three lightning surges were observed between 14:30 and 14:50. Figure 4 As can be seen in the data, the number of OT centers at 14:50 is greater than 5; Intermediate identification: Based on the primary identification, lightning hole features were detected simultaneously at 14:48 (as shown in the figure). Figure 5 As shown), a Level 2 (Severe) warning is triggered; Advanced recognition: Building upon intermediate recognition, hook-shaped echo region features were detected on the radar reflectivity at 14:48 (e.g., Figure 5 As shown), a Level 1 (more severe) warning is triggered.

[0074] Step 7: Early Warning Product Generation A three-level early warning product is generated based on the identification results: Level 3 Warning (Attention): Meets the initial identification criteria Level 2 Warning (Early Warning): Meets the criteria for intermediate identification. Level 1 Warning (Emergency): Meets advanced identification criteria The early warning product includes information such as the area and time where tornadoes may occur, and is encapsulated in XML format.

[0075] Provides a WebGIS platform to display identification results and early warning information in real time. Example 3: Based on the same inventive concept, this invention also provides a tornado intelligent recognition system, such as... Figure 6 As shown, it includes: The extraction module is used to extract meteorological feature parameters based on the acquired multi-source meteorological data. The meteorological feature parameters include radar reflectivity features, satellite overshoot cloud top center features, lightning surge features, and lightning hole features. The identification module is used to identify tornado risks based on the meteorological characteristic parameters and adopt graded early warning rules, and generate tornado identification results.

[0076] In one possible implementation, this system adopts a modular layered design, including five core parts: data acquisition layer, preprocessing layer, feature extraction layer, intelligent recognition layer, and early warning output layer.

[0077] (1) The data acquisition and quality control layer is responsible for the real-time access, acquisition, and quality control of multi-source meteorological data, mainly including: Radar data acquisition module: Connects to the next-generation weather radar (CINRAD), with a horizontal spatial resolution of approximately 1km and a temporal resolution of 6 minutes; Satellite data acquisition module: Receives real-time data from the Himawari-9 satellite via the satellite data receiving station, mainly using infrared channel (channel 13, 10.4μm) and water vapor channel (channel 8, 6.2μm) data, with a spatial resolution of 0.02° and a temporal resolution of 10 minutes; Lightning data acquisition module: Connects to the State Grid wide-area lightning monitoring network to acquire ground flash data in real time, including parameters such as time, latitude and longitude, lightning current amplitude, and polarity, with a time resolution of milliseconds.

[0078] After the data is accessed, quality control is carried out according to the rules.

[0079] (2) The feature extraction layer is the key to tornado recognition, and includes three parallel processing modules: Radar feature extraction module: Based on radar data, identify feature parameters of high reflectivity areas; Satellite feature extraction module: Based on satellite infrared and water vapor channel data, it identifies overshoot cloud top features; Lightning feature extraction module: Based on lightning location data, calculate lightning frequency feature parameters and detect lightning surge and lightning void features.

[0080] (3) The intelligent identification layer adopts a multi-feature fusion strategy to identify tornado risks under different conditions of primary, intermediate and advanced based on feature combinations.

[0081] (4) The early warning output layer is responsible for generating early warning products: Early warning generation module: Automatically generates early warning products of different levels based on the recognition results. Visualization module: Provides a WebGIS platform to display identification results and early warning information in real time, and overlays them with images of lightning, radar or satellite to intuitively show the characteristics of tornadoes.

[0082] Example 4 like Figure 7 As shown, the present invention also provides an electronic device, which may be a computer device, a microcontroller device, a smart mobile device, etc. The electronic device in this embodiment may include a processor, a memory, a transceiver component, etc. The memory, processor, and transceiver component are connected via a bus; the memory can be used to store executable programs, and an exemplary executable program may include instructions; the processor is used to execute the instructions stored in the memory. The memory can also be used to store data, which can be accessed and / or modified when instructions are executed.

[0083] The processor may be a central processing unit (CPU), or it may be other general-purpose processors, digital signal processors (DSPs), application-specific integrated circuits (ASICs), field-programmable gate arrays (FPGAs), or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components, etc. It is the computing and control core of the terminal, and it is suitable for implementing one or more instructions. Specifically, it is suitable for loading and executing one or more instructions in the storage medium to implement the corresponding method flow or corresponding function, so as to implement the steps of the tornado intelligent recognition method in the above embodiments.

[0084] Example 5 Based on the same inventive concept, this invention also provides a readable storage medium, specifically an electronic device readable storage medium (Memory). This readable storage medium is a memory device within an electronic device used to store programs and data. It is understood that the storage medium here can include both built-in storage media within the electronic device and extended storage media supported by the electronic device. The storage medium provides storage space, which stores the terminal's operating system. Furthermore, this storage space also stores one or more instructions suitable for loading and execution by a processor. These instructions can be one or more executable programs (including program code). It should be noted that the storage medium here can be high-speed RAM or non-volatile memory, such as at least one disk storage device. Loading and executing one or more instructions stored in the storage medium by the processor can implement the steps of the tornado intelligent recognition method described in the above embodiments.

[0085] Those skilled in the art will understand that embodiments of the present invention can be provided as methods, systems, or computer program products. Therefore, the present invention can take the form of a completely hardware embodiment, a completely software embodiment, or an embodiment combining software and hardware aspects. Furthermore, the present invention can take the form of a computer program product embodied on one or more computer-usable storage media (including, but not limited to, disk storage, CD-ROM, optical storage, etc.) containing computer-usable program code.

[0086] This invention is described with reference to flowchart illustrations and / or block diagrams of methods, apparatus (systems), and computer program products according to embodiments of the invention. It will be understood that each block of the flowchart illustrations and / or block diagrams, and combinations of blocks in the flowchart illustrations and / or block diagrams, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, special-purpose computer, embedded processor, or other programmable data processing apparatus to produce a machine, such that the instructions, which execute via the processor of the computer or other programmable data processing apparatus, generate instructions for implementing the flowchart illustrations and / or block diagrams. Figure 1 One or more processes and / or boxes Figure 1 A device that provides the functions specified in one or more boxes.

[0087] These computer program instructions may also be stored in a computer-readable storage medium that can direct a computer or other programmable data processing device to function in a particular manner, such that the instructions stored in the computer-readable storage medium produce an article of manufacture including instruction means, which are implemented in a process Figure 1 One or more processes and / or boxes Figure 1 The function specified in one or more boxes.

[0088] These computer program instructions may also be loaded onto a computer or other programmable data processing equipment to cause a series of operational steps to be performed on the computer or other programmable equipment to produce a computer-implemented process, thereby providing instructions that execute on the computer or other programmable equipment for implementing the process. Figure 1 One or more processes and / or boxes Figure 1 The steps of the function specified in one or more boxes.

[0089] The above are merely embodiments of the present invention and are not intended to limit the present invention. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention are included within the scope of the claims of the present invention pending approval.

Claims

1. A tornado intelligent recognition method, characterized in that, The method comprises: feature extraction based on the obtained multi-source weather data to obtain weather feature parameters, wherein the weather feature parameters include radar reflectivity features, satellite overshooting top center features, lightning jump features, and lightning cavity features; tornado risk identification based on the weather feature parameters using hierarchical warning rules to generate a tornado identification result.

2. The method of claim 1, wherein, The feature extraction based on the obtained multi-source weather data to obtain weather feature parameters comprises: quality control processing of the obtained multi-source weather data to obtain processed multi-source weather data; based on the processed radar data in the processed multi-source weather data, identifying storm cells that meet preset intensity and area threshold values on a combined reflectivity factor map, and determining radar reflectivity features based on the storm cells; based on the water vapor channel and the infrared window channel in the processed satellite data in the processed multi-source weather data, calculating a brightness temperature difference value, and identifying and labeling a plurality of target OT centers as satellite overshooting top center features based on the brightness temperature difference value; using a density-based clustering algorithm to perform spatial clustering on lightning data in the processed multi-source weather data to obtain a plurality of lightning clusters representing different thunderstorm cells, and performing lightning frequency jump signal identification based on a change rate sequence of cloud-to-ground lightning frequency calculated from the lightning clusters to obtain lightning jump features; time-aggregating lightning data according to a preset time window to generate lightning distribution data, superimposing the lightning distribution data with radar composite reflectivity products and OT centers in the same time window to generate an overlay map, and identifying a spatial region in the overlay map where radar composite reflectivity meets a preset reflectivity and lightning activity meets a preset activity sparsity condition as a lightning cavity feature.

3. The method of claim 2, wherein, The determination of radar reflectivity features based on the storm cells comprises: calculating the geometric center of pixel points inside the storm cell that exceed a preset reflectivity threshold; performing morphological feature identification on the storm cell to determine a hook-shaped echo area; using the geometric center and the hook-shaped echo area as radar reflectivity features.

4. The method of claim 2, wherein, The identification and labeling of a plurality of target OT centers as satellite overshooting top center features based on the brightness temperature difference value comprises: determining a region where the brightness temperature difference value is greater than a preset brightness temperature difference threshold as a shallow convection region; extracting closed isotherms at a preset fixed interval within the shallow convection region, and identifying a region located within half of the isotherms as a potential OT center region; determining a pixel in the potential OT center region as a candidate OT center if the infrared window region brightness temperature value of the pixel is lower than the brightness temperature values of adjacent pixels in a preset range; for each candidate OT center, determining the innermost closed isotherm that contains the center, wherein the region enclosed by the innermost closed isotherm meets a preset spatial coverage condition and has a pixel distribution with a preset number of pixels in multiple directions with the candidate OT center as the origin; Calculate the average infrared window region brightness temperature of all pixels in the anvil cloud region of the innermost closed brightness temperature contour, as the average anvil cloud temperature, wherein the anvil cloud region is the pixel region between the inner one layer contour adjacent to the innermost closed brightness temperature contour and the innermost closed brightness temperature contour; Calculate the difference between the infrared window region brightness temperature value of the candidate OT center and the average anvil cloud temperature, and mark the candidate OT center whose difference meets the preset gradient threshold as the target OT center; Determine and mark the target OT center as the satellite overshooting cloud top center feature.

5. The method of claim 2, wherein, The lightning frequency jump signal identification is performed based on the calculated ground flash frequency change rate sequence of the lightning cluster, and a lightning jump feature is obtained, including: The moving track of a specific thunderstorm cell corresponding to the storm system is identified and tracked based on the lightning cluster using a thunderstorm cell tracking algorithm; The ground flash occurrence frequency within a preset frequency is determined based on the moving track, and a ground flash frequency sequence is generated based on the ground flash occurrence frequency; A ground flash frequency change rate sequence is calculated based on the ground flash frequency sequence, and a ground flash frequency change rate sequence standard deviation is calculated based on the ground flash frequency change rate sequence and a preset statistical backtracking time window; A dynamic judgment threshold is set based on the ground flash frequency change rate sequence standard deviation, and a lightning jump signal is generated as a lightning jump feature when the ground flash frequency change rate at the current time exceeds the dynamic judgment threshold.

6. The method of claim 1, wherein, The quality control processing is performed on the obtained multi-source meteorological data to obtain processed multi-source meteorological data, including: The radar data in the obtained multi-source meteorological data is subjected to speed ambiguity processing using a double PRF algorithm, and the processed radar data is subjected to ground clutter suppression using a clutter recognition algorithm to obtain suppressed lightning data; The satellite data in the obtained multi-source meteorological data is subjected to radiation calibration and geometric processing to obtain processed satellite data, wherein the radiation calibration is to convert the original count value to a brightness temperature value, and the geometric correction is to project the data to a geographic coordinate system; The lightning data in the obtained multi-source meteorological data is subjected to positioning accuracy control and de-duplication processing, wherein the positioning accuracy control is to remove lightning records with less than a preset number of stations participating in positioning, and the de-duplication processing is to process using a space-time window method.

7. The method of claim 1, wherein, The tornado risk identification is performed based on the meteorological feature parameters using a hierarchical early warning rule to generate a tornado identification result, including: When multiple lightning jump signals appear within a preset time interval and the number of target OT centers exceeds a preset data threshold, a level 3 tornado identification early warning is triggered; When the level 3 tornado identification early warning is triggered and the lightning void feature is detected, a level 2 tornado identification early warning is triggered; When the level 2 tornado identification early warning is triggered and the radar reflectivity feature is detected, a level 1 tornado identification early warning is triggered.

8. A tornado intelligent recognition system, characterized in that, The feature extraction module is configured to extract meteorological feature parameters based on the obtained multi-source meteorological data, wherein the meteorological feature parameters include radar reflectivity features, satellite overshooting cloud top center features, lightning jump features, and lightning void features. ​ The identification module is configured to identify the tornado risk based on the meteorological characteristic parameters by using a hierarchical early warning rule to generate a tornado identification result.

9. An electronic device, comprising: The method comprises: at least one processor and a memory; the memory and the processor are connected through a bus; the memory is configured to store one or more programs; when the one or more programs are executed by the at least one processor, the method for intelligently identifying tornadoes according to any one of claims 1 to 7 is implemented.

10. A readable storage medium, characterized by, The computer program product has an execution program stored thereon, and the execution program is executed to implement the method for intelligently identifying tornadoes according to any one of claims 1 to 7.

Images