Lightning multi-parameter space-time analysis method based on clustering tracking, thunderstorm risk early warning method and system

Abstract

The invention discloses a lightning multi-parameter space-time analysis method based on clustering tracking and a thunderstorm risk early warning method and system, and relates to the technical field of severe convective weather early warning. The method comprises the following steps: collecting and analyzing lightning data; data preprocessing and space-time collaborative clustering: clustering lightning belonging to the same thunderstorm monomer in the current time window by using optimized clustering parameters; cross-period target tracking and association: constructing a lightning cluster tracking chain of a continuous time window; performing multi-factor dynamic feature extraction, namely capturing real-time dynamic feature parameters on the associated cluster sequence; and constructing a hazard index model and carrying out risk early warning. According to the method, dynamic tracking, strength evaluation and danger grading of thunderstorm activities can be realized by utilizing lightweight and easily acquired lightning positioning data, the method has the characteristics of light calculation load and high real-time performance, and objective and quantitative technical support can be provided for refined short-term and temporary early warning decision making.

Description

Technical Field

[0001] This invention relates to the field of severe convective weather early warning technology, and in particular to a lightning multi-parameter spatiotemporal analysis method based on cluster tracking, a thunderstorm risk early warning method and system. Background Technology

[0002] Thunderstorms are a common type of severe convective weather, often accompanied by lightning. Thunderstorms that produce a large amount of lightning can lead to various hazardous weather events such as heavy rain, hail, strong winds, and tornadoes, which can easily cause casualties and property damage.

[0003] Lightning activity triggered by charge separation within thunderstorm clouds (lightning originates within thunderstorm clouds; ice particles, supercooled water, and graupel particles collide and become charged, with larger particles sinking and smaller particles rising, eventually triggering lightning when the electric field strength exceeds the air breakdown threshold) can essentially serve as an "electrical signal" closely coupled with the life cycle of thunderstorms. Its spatiotemporal distribution characteristics are closely related to the generation, intensification, dissipation, and structural evolution of thunderstorms, making it a highly significant dynamic indicator of thunderstorm development. Existing research shows that lightning activity characteristics are closely related to the generation and dissipation processes of thunderstorms; high-density lightning areas highly overlap with strong radar echo bands; and the movement direction of lightning clusters can indicate the migration trend of strong convective cells. Features such as the initial lightning time and the ratio of cloud-to-ground lightning during thunderstorm processes show a good temporal correspondence with radar echo evolution and hazardous weather events such as hail and tornadoes. Incorporating these lightning characteristics into the analysis system can significantly improve the accuracy of identifying the development characteristics of thunderstorm movement trajectories and intensity changes. Therefore, using lightning data to assess lightning hazards and study the patterns of thunderstorm activity helps to explain and predict the evolution and development of thunderstorms, providing a practical reference for thunderstorm identification and forecasting.

[0004] However, current meteorological lightning data is typically distributed as static, discrete points. This necessitates manual analysis by forecasters to determine the correlation between lightning distribution and convective activity, a time-consuming and subjective process that compromises objectivity and consistency, failing to meet the objectivity and real-time requirements of short-term warnings. Furthermore, traditional lightning products usually only display basic information such as lightning density, polarity, and geographic location, lacking in-depth analysis of the spatiotemporal evolution of lightning activity. Moreover, the high complexity and environmental variability of severe convective weather systems lead to significant differences in lightning activity patterns across different regions and weather systems. For example, studies in one region have found that the proportion and frequency of positive ground lightning are higher during hailstorms than during rainstorms, and the peak lightning jump signal and frequency occur before hail. In another region, lightning jumps generally occur before the peak wind speed. Therefore, effectively integrating the spatiotemporal distribution characteristics of lightning with multi-parameter coupling effects to construct a dynamically adaptive quantitative model is a crucial issue in lightning hazard assessment.

[0005] In recent years, the development of machine learning algorithms has provided new solutions to the aforementioned problems. Among these, density-based clustering methods (such as the classic algorithm DBSCAN, short for Density-Based Spatial Clustering) have been employed. Clustering of Applications with Noise (DNR), a density-based clustering method with noise, can effectively identify lightning clusters and remove noise in high-density lightning regions, laying the foundation for multi-dimensional analysis. In practice, the DBSCAN clustering results replace the 2σ lightning leap algorithm based on radar echoes. By setting a minimum number of clusters and a neighborhood radius, the correspondence between clusters and strong convective cells reaches 88.1%. Furthermore, existing technologies also utilize the DBSCAN algorithm to cluster WWLLN lightning data in certain regions, establishing thunderstorm datasets and providing quantitative support for regional thunderstorm activity characteristic analysis. These research results based on lightning clustering provide a key technical path for constructing lightning hazard models.

[0006] However, existing clustering algorithms have the following drawbacks when performing clustering analysis on lightning data: 1) The DBSCAN clustering algorithm typically sets a fixed threshold, and its poor adaptability to the neighborhood radius and minimum number of neighborhood points (two key parameters) leads to problems such as overcrowding in some seasons and missed detections in others, reducing the accuracy of the analysis. Accordingly, some existing algorithms have proposed dynamically adjusting the threshold using radar echo intensity (i.e., providing a dynamic threshold). However, this adjustment relies on radar network data, and when radar observations are obstructed by terrain or product delays, the timeliness and reliability of early warnings will significantly decrease.

[0007] 2) Existing clustering algorithms mostly focus on single-period analysis, lacking cross-period lightning cluster tracking mechanisms and failing to identify the dynamic characteristics of individual thunderstorm cells. It should be noted that some existing algorithms have also proposed spatiotemporal clustering schemes based on deep learning to predict lightning. For example, Chinese patent ZL202011617345.3 discloses a lightning prediction method based on a spatiotemporal sequence clustering algorithm and an LSTM neural network, including: obtaining the Eps value and the lightning center for each time slice using a DBSCAN density clustering algorithm improved based on lightning prediction, according to the changes in the latitude and longitude of the lightning center; and predicting the geographical location of the lightning center in the next time slice using an LSTM neural network. Specifically, it divides the lightning data within the predicted time period into time slices and selects the lightning data for the predicted time period; the data includes the latitude and longitude of the lightning occurrence, lightning intensity, lightning slope, and the time of lightning occurrence; the data is divided into time slices; it is assumed that the movement of the lightning center between the divided time slices is abrupt, that is, the lightning center remains stationary within the divided time period. This scheme can automatically calculate the cluster radius of density clustering DBSCAN, and the LSTM neural network can predict the latitude and longitude of lightning centers with small errors and high accuracy. However, the above scheme does not incorporate the morphological evolution of thunderstorms into the evaluation system, resulting in insufficient accuracy in analysis and prediction.

[0008] On the other hand, when constructing lightning hazard models based on lightning data, due to the numerous hazard-causing factors of lightning, the assessment of lightning hazard severity typically considers multiple dimensions of features (factors), such as lightning intensity, ground flash density, lightning current steepness, and percentage of small currents. A multi-dimensional lightning hazard intensity model is then constructed based on the weights of each feature. However, in practical applications, it has been found that existing multi-dimensional lightning hazard intensity models are still insufficient in terms of refined, quantitative, and continuous monitoring. This results in assessment accuracy failing to meet the actual needs of precise disaster early warning and hinders timely and effective decision support in the face of complex and variable thunderstorm weather. Summary of the Invention

[0009] The purpose of this invention is to overcome the shortcomings of existing technologies and provide a multi-parameter spatiotemporal analysis method for lightning based on clustering tracking, as well as a thunderstorm risk early warning method and system. This invention can utilize lightweight and easily obtainable lightning location data to achieve dynamic tracking, intensity assessment, and hazard classification of thunderstorm activity. It features low computational load and strong real-time performance, providing objective and quantitative technical support for refined short-term early warning decision-making.

[0010] To achieve the above objectives, the present invention provides the following technical solution: A multi-parameter spatiotemporal analysis method for lightning based on clustering tracking includes the following steps: Lightning data collection and analysis: Collect and analyze lightning data within a defined range to obtain information on multiple lightning elements; Data preprocessing and clustering: Historical data is selected based on a set range, and clustering thresholds are determined according to seasonal statistical parameters. Spatiotemporal co-clustering is then performed to obtain lightning cluster information. Specifically, a seasonal function is constructed based on the seasonal division criteria of the target area and the statistically obtained clustering parameters for different seasons. This seasonal function is used to establish a mapping relationship between seasons and the optimal clustering parameter configuration. After preliminary segmentation of the lightning data through a fixed time window, the aforementioned seasonal function is called. Based on the occurrence time in the data, the current season is determined, and the optimal clustering parameter configuration corresponding to the current season is identified as the clustering parameter combination required by the clustering algorithm. This clustering parameter combination is then used to cluster lightning events belonging to the same thunderstorm cell within the current time window into clusters. Cross-time period target tracking and association: Calculate the cluster center coordinates of lightning clusters based on clustering, and perform cross-time period target tracking, association, and identification; Multi-factor dynamic feature extraction: Based on lightning clusters, dynamic feature parameters are captured, and the features are multiple feature factors related to the hazard of lightning.

[0011] Furthermore, the lightning data is data that has passed the basic quality control process, covering the latitude and longitude of the lightning, the time of occurrence, the intensity, the peak current, and the lightning type parameters; the latitude and longitude grid is converted into a projected coordinate system so that the unit of the neighborhood radius ε is a preset unit and meets the distance measurement accuracy standard; During clustering, the DBSCAN clustering algorithm is used to perform spatiotemporal co-clustering of lightning latitude and longitude points to generate lightning clusters containing the coordinates, timestamps and intensity attribute features of all lightning points. The minimum convex hull of each cluster is calculated to determine the clustering influence range and complete the cluster boundary division, thereby realizing the clustering and cluster boundary delineation of lightning points within the same thunderstorm cell.

[0012] Furthermore, after completing the DBSCAN clustering of lightning points within the current time window and generating several lightning clusters, the geometric center of each lightning cluster is calculated. For the i-th lightning cluster, its geometric center is set to (x... i , y i ), x i , y i The calculation formula is as follows: ; ; Where m is the total number of lightning points within the cluster, (x j , y j Let be the coordinates of the j-th lightning point, where j = 1, 2, ..., m.

[0013] Furthermore, cross-time period target tracking, association, and identification are performed through a spatiotemporal joint matching mechanism, including the following steps: The coordinates of the cluster center of the lightning cluster are obtained, and cross-time period matching of the lightning cluster is performed based on the preset spatiotemporal matching conditions. The spatiotemporal matching conditions include time continuity conditions and spatial proximity conditions. The time continuity condition is used to require that the tracking has physical coherence within a continuous time step in the time dimension. The spatial proximity condition is used as the criterion that the Euclidean distance between the cluster centers is less than a preset threshold. If the current lightning cluster and the historical lightning cluster meet the preset spatiotemporal matching conditions, the lightning cluster number from the previous moment will be used. If the preset spatiotemporal matching conditions are not met, a new number will be assigned to it.

[0014] Furthermore, by using the identification mechanism of thunderstorm cell morphological evolution, dynamic association and tracking of the same thunderstorm cell lightning clusters across time periods are carried out, and a lightning cluster tracking chain with continuous time windows is constructed; in the lightning cluster tracking chain, each lightning cluster establishes a spatiotemporal association based on the number inheritance rule and through a unique number. The thunderstorm cells include splitting and merging morphologies, and the steps for identifying splitting and merging morphologies are as follows: When two or more independent lightning clusters are detected at the current moment, and both meet the matching conditions with the same parent cluster at the previous moment, it is determined that a thunderstorm cell split has occurred. At this time, the split lightning sub-clusters are identified by adding preset symbols (such as letters) after the number. The order of adding symbols is automatically generated according to the spatial orientation of the sub-clusters relative to the parent cluster. When a single lightning cluster is detected at the current moment and meets the matching conditions with multiple candidate clusters at the previous moment, it is determined that a thunderstorm single-cell merger has occurred. At this time, the historical cluster with the highest spatial correlation among all candidate clusters is obtained as the main inheriting cluster, and its number is directly reused.

[0015] Furthermore, based on the constructed lightning cluster tracing chain, dynamic feature parameters of the same lightning cluster are captured based on the cluster number. Furthermore, when extracting dynamic feature parameters, when a thunderstorm cell splitting event occurs, the sub-clusters are treated as independent new clusters to calculate dynamic features; when a thunderstorm cell merging event occurs, the dynamic feature parameters are calculated based on the candidate cluster with the highest correlation before merging, so that the features of the merging process are continuous.

[0016] Furthermore, based on the constructed lightning cluster tracking chain, the captured characteristic factors include the lightning intensity I, intensity increment ΔI, lightning jump ΔN, number of lightning strikes N, and positive ground lightning ratio R. pg At this point, a feature vector containing the aforementioned feature factors is generated for each lightning cluster sequence, thereby enabling real-time capture and quantification of multi-dimensional feature factors during the dynamic evolution of lightning clusters. Wherein, the lightning intensity I is the maximum absolute value of the peak lightning current within a single lightning cluster, i.e., I = max Nt |I k |; where I k Nt represents the current intensity of the k-th lightning strike, in kA; Nt is the total number of lightning strikes within the cluster. The intensity increment ΔI is the difference between the maximum lightning intensity of the same lightning cluster at the current moment and the maximum lightning intensity at the previous moment, i.e., ΔI = |I| current |-|I previous |; where I current I represents the maximum current intensity of the lightning cluster at the current moment, in kA. previous The maximum current intensity of the same lightning cluster at the previous moment, in kA; The lightning surge ΔN is the difference between the total number of lightning strikes in the current time window and the total number of lightning strikes in the previous time window within the same lightning cluster, i.e., ΔN = N. current - N previous ; where N current N represents the total number of lightning strikes within the current time window of the lightning cluster. previous This represents the total number of lightning strikes in the previous time window; The lightning quantity N is the total number of lightning strikes within a single lightning cluster, measured in seconds. The positive ground flash ratio R pg R is the ratio of the number of positive ground flashes within a single lightning cluster to the total number of ground flashes. pg = N pg / N cg ; where N pg N represents the number of positive ground flashes occurring within a lightning cluster, expressed in times. cg This represents the total number of positive and negative ground flashes, expressed in flashes.

[0017] The present invention also provides a method for early warning of thunderstorm risk, the method comprising the following steps: Lightning spatiotemporal analysis steps: The spatiotemporal analysis of lightning data is performed using the method described in any one of claims 1-7; Hazard index model construction and risk warning steps: By weighted calculation of multiple feature factors and historical statistics to divide thresholds, risk values ​​and hazard levels are quantified, and the spatiotemporal distribution characteristics and trends of lightning activity are transformed into objective parameters and graded warnings.

[0018] Furthermore, a quantitative lightning hazard index (LHI) calculation model is constructed based on the feature parameters of multiple extracted feature factors. For any lightning cluster, the normalized calculation formula for its lightning hazard index (LHI) is as follows: ; Among them, N represents the number of lightning strikes in a lightning cluster, I represents the lightning intensity of the lightning cluster, △I represents the intensity increment, and R pg represents the positive cloud-to-ground lightning ratio, and △N represents the lightning jump; ref(·) represents the local reference value of each characteristic factor; w q represents the weight coefficient of each characteristic factor, q = 1, 2, 3, 4, 5, and w1 + w2 + w3 + w4 + w5 = 1.

[0019] Furthermore, the hierarchical early warning includes the steps of: Comparing the lightning hazard index LHI with the early warning threshold S1; When LHI < S1, no alarm is given; When LHI ≥ S1, enter the risk level classification and alarm process; the risk level classification and alarm adopts a three-level threshold system. Based on the preset risk thresholds S2 and S3, when the value of LHI satisfies S1 ≤ LHI < S2, it is determined as a low risk, and a low risk prompt is given to the area within the lightning clustering range; when the value of LHI satisfies S2 ≤ LHI < S3, it is determined as a medium risk, and a medium risk prompt is given to the area within the lightning clustering range; when the value of LHI satisfies LHI ≥ S3, it is determined as a high risk, and a high risk prompt is given to the area within the lightning clustering range; And, if the value of LHI satisfies the medium risk and both the lightning number and intensity increase simultaneously, the level is raised to a high risk.

[0020] The present invention also provides a thunderstorm risk early warning system, and the system includes: A data collection and analysis module, which is used to collect lightning data within a set range and analyze it to obtain multiple lightning element information; A data preprocessing and clustering module, which is used to select historical data based on a set range, statistically calculate parameters by season, determine the clustering threshold, perform spatio-temporal collaborative clustering, and obtain lightning cluster information; among them, a season function is constructed according to the season division standard of the location of the target area to be analyzed and the statistically obtained clustering parameters of different seasons, and the season function is used to establish a mapping relationship between the season and the optimal clustering parameter configuration; and, after initially dividing the lightning data through a fixed time window, the前述 season function is called, and after determining the current season based on the occurrence time in the data, the optimal clustering parameter configuration corresponding to the current season is determined as the clustering parameter combination required for the clustering algorithm, and the lightning within the current time window belonging to the same thunderstorm cell is clustered into clusters using the clustering parameter combination; A cross-period target tracking and association module, which is used to calculate the cluster center coordinates based on the clustered lightning clusters, perform cross-period target tracking, association and identification; A multi-factor dynamic feature extraction module is used to capture dynamic feature parameters based on lightning clusters, wherein the features are multiple feature factors related to the hazard of lightning. The hazard index model construction and risk warning module is used to quantify risk values ​​and hazard levels by weighted calculation of multiple feature factors and historical statistical threshold division, and to transform the spatiotemporal distribution characteristics and trends of lightning activity into objective parameters and graded early warnings.

[0021] Compared with the prior art, this invention, by adopting the above technical solution, has the following advantages and positive effects: This invention can use lightweight and easily obtainable lightning location data to realize dynamic tracking, intensity assessment and hazard classification of thunderstorm activity. It has the characteristics of light computational load and strong real-time performance, and can provide objective and quantitative technical support for refined short-term early warning decision-making.

[0022] On the one hand, by introducing a seasonal function to achieve seasonal adaptation, the core clustering parameters can be dynamically adjusted according to the date, ensuring that the algorithm maintains the accuracy of individual cluster identification in different seasons and enhancing the universality of the solution. Compared with existing clustering algorithms, this invention considers the significant differences in lightning activity in different seasons and dynamically adjusts the combination of clustering parameters (such as neighborhood radius ε, minimum number of cluster points, etc.) by date, so that the algorithm can maintain the accuracy of thunderstorm individual cluster identification under different climatic backgrounds.

[0023] On the other hand, a cross-time period target tracking mechanism for lightning clusters was constructed. By calculating the coordinates of the lightning cluster center and based on the spatiotemporal joint matching mechanism—including temporal continuity and spatial proximity matching—a lightning cluster tracking chain with continuous time windows was constructed, realizing cross-time period association and tracking of lightning clusters.

[0024] On the other hand, a tracking mechanism based on thunderstorm cell splitting and merging events is introduced: Thunderstorm cell splitting (using a suffix number when multiple sub-clusters derive from the same parent cluster) and merging (inheriting the parent cluster number when the current cluster matches multiple historical clusters) are determined through spatiotemporal matching conditions. The calculation logic of feature parameters is dynamically adjusted based on the splitting / merging events of thunderstorm cells, achieving accurate tracking and dynamic parameter calibration of thunderstorm morphological evolution. In this way, the dynamic features such as the movement trajectory, splitting, and merging of thunderstorm cells can be captured in real time, enriching the dynamic feature capture capability.

[0025] On the other hand, a multi-factor dynamic feature extraction module is proposed, which can capture dynamic feature parameters such as the number of lightnings, lightning intensity, intensity increment, positive ground lightning ratio, and lightning jump in lightning clusters in real time, providing a quantitative basis for lightning hazard assessment.

[0026] On the other hand, this invention improves existing multi-factor lightning hazard intensity models, optimizes the selection of multi-factors related to lightning hazard, establishes a Lightning Hazard Index (LHI) model, and constructs a multi-factor weight allocation mechanism. Furthermore, it combines a three-level threshold system to achieve continuous graded early warning of lightning risk, providing objective quantitative support for short-term decision-making. The thunderstorm risk early warning product provided by this invention is independent of radar echoes, relying only on lightweight lightning data, making it particularly suitable for operational real-time early warning scenarios. Attached Figure Description

[0027] Figure 1 A flowchart illustrating the thunderstorm risk warning method provided in this embodiment of the invention.

[0028] Figure 2 This is an example diagram showing the results of lightning cluster tracking and lightning surge in a target area, provided in an embodiment of the present invention.

[0029] Figure 3 This is an example diagram showing the results of lightning cluster tracking and LHI and risk level calculation for a target area on a certain day of a certain year, provided as an embodiment of the present invention. Detailed Implementation

[0030] The following detailed description, in conjunction with the accompanying drawings and specific embodiments, provides a more comprehensive overview of the lightning multi-parameter spatiotemporal analysis method, thunderstorm risk early warning method, and system disclosed in this invention. It should be noted that the technical features or combinations of technical features described in the following embodiments should not be considered isolated; they can be combined to achieve better technical results. In the accompanying drawings of the following embodiments, the same reference numerals in each drawing represent the same features or components, which can be applied to different embodiments. Therefore, once an item is defined in one drawing, it does not need to be further discussed in subsequent drawings.

[0031] It should be noted that the structures, proportions, sizes, etc., illustrated in the accompanying drawings are merely for illustrative purposes and to aid those skilled in the art in understanding and reading the invention. They are not intended to limit the conditions under which the invention can be implemented. Any modifications to the structure, changes in proportions, or adjustments to size, provided they do not affect the effectiveness or purpose of the invention, should fall within the scope of the technical content disclosed in the invention. The scope of the preferred embodiments of the present invention includes other implementations, wherein functions may be performed not in the order stated or discussed, including substantially simultaneously or in reverse order, depending on the functions involved. This should be understood by those skilled in the art to which the embodiments of the present invention pertain.

[0032] Techniques, methods, and apparatus known to those skilled in the art may not be discussed in detail, but where appropriate, such techniques, methods, and apparatus should be considered part of the specification. In all examples shown and discussed herein, any specific values ​​should be interpreted as merely exemplary and not as limitations. Therefore, other examples of exemplary embodiments may have different values. Example

[0033] To address the shortcomings of existing technologies, this invention proposes a multi-parameter spatiotemporal analysis method for lightning based on clustering tracking, which possesses seasonal adaptation and dynamic tracking capabilities. Then, based on the aforementioned clustering tracking method for multi-parameter spatiotemporal analysis of lightning, a complete thunderstorm risk early warning model is constructed.

[0034] Specifically, this method introduces a seasonal adaptive function into the clustering algorithm to dynamically adjust core parameters, enabling the algorithm to adapt to the differences in lightning activity seasonally (e.g., between summer and spring / autumn), improving the identification accuracy of thunderstorm cells under different seasonal backgrounds and enhancing the model's universality under complex climatic conditions. Simultaneously, through a spatiotemporal collaborative clustering algorithm, static discrete lightning data is aggregated into dynamically correlated thunderstorm cell clusters, and a cross-time period tracking algorithm is constructed to automatically identify the correlation between lightning clusters at different times, improving analysis efficiency and the objectivity of results, providing an automated technical path for real-time monitoring of thunderstorm activity. This step can draw on the improved DBSCAN algorithm's efficient search mechanism for lightning data neighborhoods, combined with spatiotemporal correlations, to achieve dynamic aggregation and tracking of thunderstorm cell clusters, reducing manual intervention and improving analysis efficiency. Furthermore, through cross-time period target tracking and multi-factor dynamic feature extraction technology, dynamic information such as the geometric center migration, splitting, merging, and evolution of lightning clusters is captured in real time. Combined with a hazard index model, a multi-factor weight allocation mechanism is constructed to achieve quantitative and continuous monitoring of thunderstorm activity intensity and risk level classification, providing key technical support for refined short-term early warning. Furthermore, based on the correlation between lightning characteristic parameters and thunderstorm activity in existing studies, the lightning hazard model is optimized by incorporating characteristic factors such as lightning quantity, intensity, positive ground lightning ratio, intensity surge, and quantity surge into the hazard index weighting model. By tracking the dynamic changes of each factor across time periods, a quantitative assessment of thunderstorm activity intensity can be achieved, providing more accurate technical support for short-term early warning.

[0035] The core of the above scheme includes: introducing a seasonal function into the clustering algorithm, which dynamically adjusts the core clustering parameters according to the date, ensuring that the algorithm maintains the accuracy of single-unit identification under different climatic backgrounds, effectively reducing the impact of the significant seasonality of lightning activity (high frequency and small area in summer versus low frequency and large area in spring and autumn) on the algorithm's universality; spatiotemporal collaborative clustering, using specific clustering parameters to cluster lightning belonging to the same thunderstorm unit within the current time window; cross-time period target tracking, calculating the geometric center of lightning clusters and constructing a lightning cluster tracking chain for continuous time windows; multi-factor dynamic feature extraction, capturing real-time dynamic feature parameters on the associated cluster sequence; hazard index model construction and risk warning: establishing a multi-factor weight allocation mechanism, calculating the Lightning Hazard Index (LHI), and setting a threshold to achieve continuous classification of risk levels. This invention achieves full-process automation from data processing to risk warning. Compared with existing technologies, this invention can achieve automatic tracking, intensity assessment, and hazard classification of thunderstorm activity independently of radar echoes by utilizing lightweight and easily accessible lightning location data.

[0036] The specific implementation details of the present invention are described in detail below with reference to the accompanying drawings.

[0037] Referring to Figure 1, the thunderstorm risk warning method mainly includes the following steps: S100, Lightning Data Collection and Analysis: Collection and analysis of lightning data within a defined range to obtain information on multiple lightning elements. These lightning elements include latitude and longitude, intensity, and lightning type.

[0038] S200, Data Preprocessing and Clustering: Based on historical data selected within a set range, clustering thresholds are determined according to seasonal statistical parameters, and spatiotemporal collaborative clustering is performed to obtain lightning cluster information.

[0039] In the data preprocessing and clustering process, a seasonal function needs to be constructed based on the seasonal division criteria of the target area and the statistically obtained clustering parameters for different seasons. This seasonal function is used to establish a mapping relationship between seasons and the optimal clustering parameter configuration. Furthermore, after initially dividing the lightning data through a fixed time window, the aforementioned seasonal function is called. Based on the occurrence time in the data, the current season is determined, and the optimal clustering parameter configuration corresponding to the current season is identified as the clustering parameter combination required by the clustering algorithm. This clustering parameter combination is then used to cluster lightning bolts belonging to the same thunderstorm cell within the current time window into clusters.

[0040] Taking a target area (or study area) located in the Northern Hemisphere as an example, based on radar echo data from specific cases, the core clustering parameters for different seasons within the study area can be dynamically adjusted. A seasonal function is then constructed according to the Northern Hemisphere seasonal division standard. This seasonal function establishes a mapping relationship between seasons and the optimal clustering parameter configuration. As a typical example, the seasonal function might have a season identifier parameter `season(str)`, supporting 'spring', 'summer', 'autumn', and 'winter'. After obtaining the occurrence time (date) information of the current lightning data, the corresponding seasonal parameters can be determined. Then, calling the seasonal function retrieves the optimal clustering parameter configuration for that season. Subsequently, the obtained optimal clustering parameter configuration is applied to the target model. As the date changes and the seasons alternate, the above process is repeated to achieve adaptive dynamic adjustment of the clustering parameters. The seasonal function implements a parameter adaptive adjustment mechanism based on seasonal characteristics. By establishing a mapping relationship between seasons and optimal model parameters, it can automatically select the most suitable parameter configuration according to the current season, thereby improving the model's performance under different seasonal conditions.

[0041] S300, Cross-Time Target Tracking and Association: Based on clustering, lightning clusters calculate cluster center coordinates and perform cross-time target tracking, association, and identification through a spatiotemporal joint matching mechanism.

[0042] S400, multi-factor dynamic feature extraction: acquires feature parameters such as moving speed, lightning intensity, intensity increment, lightning jump, number of lightning strikes, and positive ground lightning ratio. These features are characteristic factors related to the hazard of lightning and are used to construct hazard models.

[0043] S500, Hazard Index Model Construction and Risk Warning Module: Through dynamic multi-factor weighted calculation and historical statistical threshold division, it quantifies risk values ​​and hazard levels, transforming the spatiotemporal distribution characteristics and trends of lightning activity into objective parameters and graded warnings.

[0044] Correspondingly, there are modules for data collection and analysis, data preprocessing and clustering, cross-time target tracking and association, multi-factor dynamic feature extraction, and hazard index model construction and risk warning.

[0045] In this embodiment, the ADTD (Advanced Direction and Time of Arrival Detecting system) lightning location data input to the data preprocessing and clustering module is lightning data that has passed basic quality control processes such as status data inspection and location data verification, covering key parameters such as latitude and longitude, occurrence time, peak current, and lightning type. The latitude and longitude grid is converted to a projected coordinate system to ensure that the neighborhood radius ε is in km and the distance measurement accuracy meets the standard. Then, after preliminary segmentation of the lightning data using a fixed time window, the aforementioned seasonal function is called to determine the current season based on the occurrence time in the data, thereby determining the required combination of clustering parameters. The DBSCAN clustering algorithm is used to perform spatiotemporal co-clustering of lightning latitude and longitude points, generating lightning clusters containing attributes such as the coordinates, timestamps, and intensity of all lightning points. Subsequently, the minimum convex hull of each cluster is calculated to determine the clustering influence range and complete the cluster boundary delineation, realizing the clustering of lightning points within the same thunderstorm cell and the delineation of cluster boundaries.

[0046] In step S300, after completing the DBSCAN clustering of lightning points within the current time window and generating several lightning clusters, it is necessary to calculate the geometric center of each lightning cluster.

[0047] For the i-th lightning cluster, let its geometric center be (x i , y i ), x i , y i The calculation formula is as follows: ; ; Where m is the total number of lightning points within the cluster, (x j , y j Let be the coordinates of the j-th lightning point, where j = 1, 2, ..., m.

[0048] After completing the geometric center calculation, the target is tracked, associated and identified across time periods through a spatiotemporal joint matching mechanism.

[0049] The cross-time period matching process follows preset spatiotemporal matching conditions, including temporal continuity and spatial proximity conditions. The temporal continuity condition requires physical coherence in tracking within consecutive time steps. The spatial proximity condition uses a preset threshold as the criterion for judgment. Based on these conditions, matching is performed. If the lightning cluster at the current moment meets the spatiotemporal matching conditions with a cluster at a historical moment, the lightning cluster number from the previous moment is used; otherwise, a new number is assigned.

[0050] After cross-time period matching, the dynamic tracking process begins. Specifically, a mechanism for identifying the morphological evolution of thunderstorm cells is introduced to dynamically associate and track the same thunderstorm cell lightning cluster across different time periods.

[0051] The identification mechanism for the morphological evolution of thunderstorm cells specifically includes the judgment logic for the transformation of two key morphological phases: splitting and merging.

[0052] Split identification mechanism: When two or more independent lightning clusters are detected at the current moment and both meet the matching conditions of the same parent cluster at the previous moment, it is determined that a thunderstorm cell split has occurred. At this time, a specific identification method is used, which adds a preset symbol after the number - such as letters (e.g., "N1a" "N1b") to identify the split lightning sub-clusters. The letter order can be automatically generated according to the spatial orientation of the sub-cluster relative to the parent cluster.

[0053] Merging and identification mechanism: If a single lightning cluster at the current moment meets the matching conditions with multiple candidate clusters at the previous moment, it is determined that a thunderstorm single-cell merging has occurred. At this time, the historical cluster with the highest spatial correlation between this cluster and all candidate clusters is used as the main inheriting cluster, and its number is directly used to realize the dynamic association and tracking of the same thunderstorm single-cell lightning cluster across time periods.

[0054] During target tracking, each lightning cluster establishes a spatiotemporal association based on a unique number according to the numbering inheritance rule.

[0055] In step S400, the multi-factor dynamic feature extraction module captures dynamic feature parameters of the same lightning cluster based on the lightning cluster number. The features are pre-configured feature factors related to lightning hazard, including the lightning intensity I, intensity increment ΔI, lightning jump ΔN, number of lightning strikes N, and positive ground flash ratio R of the lightning cluster. pg .

[0056] The lightning intensity I is the maximum absolute value of the peak lightning current within a single lightning cluster, i.e., I = max Nt |I k |; where I k denoted as , where is the current intensity of the k-th lightning strike, expressed in kA; and Nt is the total number of lightning strikes within the cluster. The lightning intensity represents the energy release of a single lightning discharge, is related to charge exchange within the thunderstorm cloud, and is a core parameter for assessing potential destructiveness.

[0057] The intensity increment ΔI is the difference between the maximum lightning intensity of the same lightning cluster at the current moment and the maximum lightning intensity at the previous moment, i.e., ΔI = |I| current |-|I previous |

[0058] Among them, I current I represents the maximum current intensity of the lightning cluster at the current moment, in kA.previous The value represents the maximum current intensity of the same lightning cluster at the previous moment, expressed in kA. This intensity increment visually reflects the change in the degree of charge transfer within the thunderstorm cloud over time. A positive increment corresponds to the convective development phase, indicating an increase in charge energy within a single lightning cluster, while a negative increment reflects the decrease in energy within a single lightning cluster.

[0059] The lightning surge ΔN is the difference between the total number of lightning strikes in the current time window and the total number of lightning strikes in the previous time window within the same lightning cluster, i.e., ΔN = N. current - N previous .

[0060] Where, N current N represents the total number of lightning strikes within the current time window of the lightning cluster. previous This represents the total number of lightning strikes in the previous time window. A surge in the number of lightning strikes can reflect changes in lightning activity over time. A surge in the number of lightning strikes usually corresponds to a stage in the development of a thunderstorm, while a sharp decrease generally indicates the decay of individual lightning strikes.

[0061] The lightning count N refers to the total number of lightning strikes within a single lightning cluster, measured in seconds. The lightning count is a direct measure of thunderstorm activity intensity; high-density lightning areas typically correspond to areas of strong convection development and can help determine the maturity of individual thunderstorm cells.

[0062] The positive ground flash ratio R pg R is the ratio of the number of positive ground flashes within a single lightning cluster to the total number of ground flashes. pg = N pg / N cg .

[0063] Where, N pg N represents the number of positive ground flashes occurring within a lightning cluster, expressed in times. cg This represents the total number of positive and negative ground flashes, expressed in terms of flashes. Positive ground flashes are typically much more intense than negative ground flashes and are more likely to cause ground-based disasters. A high proportion of positive ground flashes is usually a warning signal of severe weather.

[0064] In this embodiment, when extracting the aforementioned dynamic feature parameters, when a thunderstorm cell splitting event occurs, the sub-clusters are treated as independent new clusters for calculating dynamic features. This is because splitting causes a sharp decrease in the initial lightning count of the sub-clusters relative to the parent cluster, avoiding interference from historical data on the features of the new clusters. When a thunderstorm cell merging event occurs, the dynamic features are calculated based on the candidate cluster with the highest correlation before merging, ensuring the continuity of features during the merging process.

[0065] Based on the constructed tracking chain, a feature vector containing the aforementioned characteristic factors, namely lightning density, lightning intensity, intensity increment, positive cloud-to-ground flash ratio, and lightning jump, etc., is generated for each lightning cluster sequence. In this way, the real-time capture and quantitative expression of multi-dimensional characteristics in the dynamic evolution process of lightning clusters can be achieved, providing comprehensive characteristic parameter support for the subsequent calculation of the total hazard value and risk level assessment. See Figure 2 As shown, the lightning cluster tracking and lightning jump results in a certain research area are exemplified.

[0066] In step S500, the hazard index model construction and risk warning module constructs a quantitative evaluation system based on the multi-factor characteristic parameters of lightning clustering tracking, and realizes the precise classification of lightning risks by integrating lightning activity intensity, charge structure, and dynamic change characteristics.

[0067] Specifically, a quantitative lightning hazard index LHI calculation model is constructed based on the characteristic parameters of the aforementioned 5 characteristic factors extracted.

[0068] For any lightning cluster, the normalized calculation formula of its lightning hazard index LHI is as follows: ; where, ref(·) represents the local reference value of each characteristic factor; w q represents the weight coefficient of each characteristic factor, q = 1, 2, 3, 4, 5; and w1 + w2 + w3 + w4 + w5 = 1.

[0069] After calculating the lightning hazard index, enter the risk classification and warning process. Specifically, the classification warning includes: comparing the lightning hazard index LHI with a preset warning threshold S1; when LHI < S1, no alarm is made; when LHI ≥ S1, enter the risk level classification and alarm process.

[0070] Preferably, the risk level classification is based on the total hazard value (LHI) to set a three-level threshold system, based on the preset risk thresholds S2 and S3: When S1 ≤ LHI < S2, it is determined as a low risk, and a low risk prompt is given to the area within the lightning clustering range.

[0071] When S2 ≤ LHI < S3, it is determined as a medium risk, and a medium risk prompt is given to the area within the lightning clustering range.

[0072] When LHI ≥ S3, it is determined as a high risk, and a high risk prompt is given to the area within the lightning clustering range.

[0073] In addition, if the value of LHI meets the medium risk and both the lightning quantity and intensity increase simultaneously, the level is raised to a high risk.

[0074] See Figure 3 As shown in the example, taking a hailstorm event in a target area on a certain day of a certain year as an example, the above technical solution is used to track and calculate the LHI index value of lightning clusters and classify their risk levels. Based on the lightning disaster occurrence conditions in the target area, the following parameters are analyzed: lightning quantity N, lightning intensity I, intensity increment ΔI, and positive ground lightning ratio R. pg The local weighting coefficients w1, w2, w3, w4, and w5 for the lightning surge ΔN are 0.3, 0.2, 0.1, 0.2, and 0.2, respectively; based on local data, the lightning quantity N, lightning intensity I, intensity increment ΔI, and positive ground lightning ratio R are... pg With the local reference values ​​for the lightning jump ΔN being 120, 100, 1, 100, and 100 respectively, the formula for calculating the LHI index is as follows:

[0075] If the configured early warning threshold S1=0, the first risk threshold S2=30, and the second risk threshold S3=60, then: When 0 ≤ LHI < 30, it is considered low risk.

[0076] When 30 ≤ LHI < 60, it is considered medium risk.

[0077] When LHI ≥ 60, it is considered high risk.

[0078] When the LHI value meets the medium risk requirement, if both the number and intensity of lightning surge simultaneously, the risk level will be raised to high risk.

[0079] The results show that the algorithm identified and tracked four lightning clusters N1, N2, N3, and N4 within the designated area. After calculation, the LHI scores and risk levels of the four clusters were obtained. (See also...) Figure 3 As shown. Furthermore, it can also output the LHI values ​​of the aforementioned four lightning clusters N1, N2, N3, and N4 over time, as needed.

[0080] It should be noted that the values ​​of the warning threshold S1, the first risk threshold S2, and the second risk threshold S3 configured above are examples and not limitations. When setting them specifically, those skilled in the art can make adaptive adjustments according to the regional seasonal characteristics, the ratio relationship, and the requirements for false alarm hits.

[0081] The above scheme transforms the spatiotemporal distribution and electrical characteristics of lightning activity into quantitative risk values ​​through dynamic multi-factor weighted calculation. This not only solves the subjectivity problem of traditional manual analysis, but also enables continuous monitoring of thunderstorm intensity through dynamic thresholds.

[0082] In the above description, the disclosure of this invention is not intended to limit itself to these aspects. Rather, within the scope of the objectives of this disclosure, components can be selectively and operationally combined in any number. Furthermore, terms such as “comprising,” “encompassing,” and “having” should be interpreted by default as inclusive or open-ended, rather than exclusive or closed, unless explicitly defined as such. All technical, scientific, or other terms are to be understood by those skilled in the art, unless defined as such. Public terms found in dictionaries should not be interpreted in the context of the relevant technical documents in an overly idealistic or impractical manner, unless explicitly defined as such in this disclosure. Any modifications or alterations made by those skilled in the art based on the foregoing disclosure are within the scope of the claims.

Claims

1. A multi-parameter spatiotemporal analysis method for lightning based on clustering tracking, characterized in that... Including steps, Lightning data collection and analysis: Collect and analyze lightning data within a defined range to obtain information on multiple lightning elements; Data preprocessing and clustering: Historical data is selected based on a set range, and clustering thresholds are determined according to seasonal statistical parameters. Spatiotemporal co-clustering is then performed to obtain lightning cluster information. Specifically, a seasonal function is constructed based on the seasonal division criteria of the target area and the statistically obtained clustering parameters for different seasons. This seasonal function is used to establish a mapping relationship between seasons and the optimal clustering parameter configuration. After preliminary segmentation of the lightning data through a fixed time window, the aforementioned seasonal function is called. Based on the occurrence time in the data, the current season is determined, and the optimal clustering parameter configuration corresponding to the current season is identified as the clustering parameter combination required by the clustering algorithm. This clustering parameter combination is then used to cluster lightning events belonging to the same thunderstorm cell within the current time window into clusters. Cross-time period target tracking and association: Calculate the cluster center coordinates of lightning clusters based on clustering, and perform cross-time period target tracking, association and identification; Multi-factor dynamic feature extraction: Based on lightning clusters, dynamic feature parameters are captured, and the features are multiple feature factors related to the hazard of lightning.

2. The method according to claim 1, characterized in that, The lightning data is data that has passed the basic quality control process and covers the latitude and longitude of the lightning, occurrence time, intensity, peak current and lightning type parameters; the latitude and longitude grid is converted into a projected coordinate system so that the unit of the neighborhood radius ε is a preset unit and meets the distance measurement accuracy standard; During clustering, the DBSCAN clustering algorithm is used to perform spatiotemporal co-clustering of lightning latitude and longitude points to generate lightning clusters containing the coordinates, timestamps and intensity attribute features of all lightning points. The minimum convex hull of each cluster is calculated to determine the clustering influence range and complete the cluster boundary division, thereby realizing the clustering and cluster boundary delineation of lightning points within the same thunderstorm cell.

3. The method according to claim 1, characterized in that, After completing the DBSCAN clustering of lightning points within the current time window and generating several lightning clusters, calculate the geometric center of each lightning cluster. For the i-th lightning cluster, let its geometric center be (x... i , y i ), x i , y i The calculation formula is as follows: ； ； Where m is the total number of lightning points within the cluster, (x j , y j Let be the coordinates of the j-th lightning point, where j = 1, 2, ..., m.

4. The method according to claim 1, characterized in that, Cross-time period target tracking, association, and identification are performed through a spatiotemporal joint matching mechanism, including the following steps: The coordinates of the cluster center of the lightning cluster are obtained, and cross-time period matching of the lightning cluster is performed based on the preset spatiotemporal matching conditions. The spatiotemporal matching conditions include time continuity conditions and spatial proximity conditions. The time continuity condition is used to require that the tracking has physical coherence within a continuous time step in the time dimension. The spatial proximity condition is used as the criterion that the Euclidean distance between the cluster centers is less than a preset threshold. If the current lightning cluster and the historical lightning cluster meet the preset spatiotemporal matching conditions, the lightning cluster number from the previous moment will be used. If the preset spatiotemporal matching conditions are not met, a new number will be assigned to it.

5. The method according to claim 4, characterized in that, By identifying the morphological evolution of thunderstorm cells, dynamic correlation and tracking of lightning clusters of the same thunderstorm cell across time periods are carried out, and a lightning cluster tracking chain with continuous time windows is constructed. In the lightning cluster tracing chain, each lightning cluster establishes a spatiotemporal association based on a unique number inheritance rule; The thunderstorm cells include splitting and merging morphologies, and the steps for identifying splitting and merging morphologies are as follows: When two or more independent lightning clusters at the current moment are detected and all meet the matching conditions with the same parent cluster at the previous moment, it is determined that a thunderstorm cell has split; at this time, the split lightning sub-clusters are identified by adding a preset symbol after the number, and the order of adding the symbol is automatically generated according to the spatial orientation of the sub-cluster relative to the parent cluster; When a single lightning cluster at the current moment is detected and meets the matching conditions with multiple candidate clusters at the previous moment, it is determined that a thunderstorm cell has merged; at this time, the historical cluster with the highest spatial correlation degree among all candidate clusters of this lightning cluster is obtained as the main inheritance cluster, and its number is directly used.

6. The method according to claim 5, characterized in that, Based on the constructed lightning cluster tracking chain, the dynamic characteristic parameters of the same lightning cluster are captured based on the number of the lightning cluster; And, when extracting the dynamic characteristic parameters, when a thunderstorm cell split event occurs, the sub-clusters are regarded as independent new clusters to calculate the dynamic characteristics; When a thunderstorm cell merge event occurs, the dynamic characteristic parameters are calculated based on the candidate cluster with the largest correlation before the merge, so that the characteristics during the merge process are continuous.

7. The method according to claim 5, characterized in that, Based on the constructed lightning cluster tracking chain, the captured characteristic factors include the lightning intensity I, intensity increment ΔI, lightning jump ΔN, number of lightning strikes N, and positive ground lightning ratio R. pg At this point, a feature vector containing the aforementioned feature factors is generated for each lightning cluster sequence, thereby enabling real-time capture and quantification of multi-dimensional feature factors during the dynamic evolution of lightning clusters. Wherein, the lightning intensity I is the maximum absolute value of the peak lightning current within a single lightning cluster, i.e., I = max Nt |I k |; where I k Nt represents the current intensity of the k-th lightning strike, in kA; Nt is the total number of lightning strikes within the cluster. The intensity increment ΔI is the difference between the maximum lightning intensity of the same lightning cluster at the current moment and the maximum lightning intensity at the previous moment, i.e., ΔI = |I| current |-|I previous |; where I current I represents the maximum current intensity of the lightning cluster at the current moment, in kA. previous The maximum current intensity of the same lightning cluster at the previous moment, in kA; The lightning surge ΔN is the difference between the total number of lightning strikes in the current time window and the total number of lightning strikes in the previous time window within the same lightning cluster, i.e., ΔN = N. current - N previous ; where N current N represents the total number of lightning strikes within the current time window of the lightning cluster. previous This represents the total number of lightning strikes in the previous time window; The lightning quantity N is the total number of lightning in a single lightning cluster, and the unit is times; The positive ground flash ratio R pg R is the ratio of the number of positive ground flashes within a single lightning cluster to the total number of ground flashes. pg = N pg / N cg ; where N pg N represents the number of positive ground flashes occurring within a lightning cluster, expressed in times. cg This represents the total number of positive and negative ground flashes, expressed in flashes.

8. A method for early warning of thunderstorm risk, characterized in that... including, Lightning spatio-temporal analysis step: The spatio-temporal analysis of lightning data is carried out by using the method described in any one of claims 1-7; Hazard index model construction and risk warning step: By weighted calculation of multiple characteristic factors and historical statistical division of thresholds, the risk value and hazard level are quantified, and the spatio-temporal distribution characteristics and trends of lightning activities are transformed into objective parameters and graded warnings.

9. The method according to claim 8, characterized in that, Based on the characteristic parameters of multiple extracted characteristic factors, a quantitative lightning hazard index LHI calculation model is constructed. For any lightning cluster, the normalized calculation formula of its lightning hazard index LHI is as follows: ； Where N represents the number of lightning bolts in the lightning cluster, I represents the lightning intensity of the lightning cluster, ΔI represents the intensity increment, and R pg represents the positive ground lightning ratio, ΔN represents the lightning surge; ref(·) represents the local reference value of each characteristic factor; w q Let w1 represent the weight coefficients of each characteristic factor, q = 1, 2, 3, 4, 5, and w1 + w2 + w3 + w4 + w5 = 1.

10. The method according to claim 8 or 9, characterized in that, The graded warning includes the steps: Compare the lightning hazard index LHI with the warning threshold S1; When LHI < S1, no alarm is made; When LHI ≥ S1, enter the risk level division alarm process; the risk level division alarm adopts a three-level threshold system, based on the preset risk thresholds S2 and S3, When the value of LHI satisfies S1 ≤ LHI < S2, it is determined as low risk, and a low risk prompt is given to the area within the lightning clustering range; When the value of LHI satisfies S2 ≤ LHI < S3, it is determined as medium risk, and a medium risk prompt is given to the area within the lightning clustering range; When the value of LHI satisfies LHI ≥ S3, it is determined as high risk, and a high risk prompt is given to the area within the lightning clustering range; And, if the value of LHI satisfies medium risk and both the lightning quantity and intensity increase suddenly, the level is raised to high risk.

11. A thunderstorm risk early warning system, characterized in that... including: Data collection and analysis module, used to collect lightning data within a set range and analyze it to obtain multiple lightning element information; The data preprocessing and clustering module is used to select historical data based on a set range, determine the clustering threshold according to seasonal statistical parameters, and perform spatiotemporal co-clustering to obtain lightning cluster information. Specifically, it constructs a seasonal function based on the seasonal division criteria of the target area and the statistically obtained clustering parameters for different seasons. This seasonal function is used to establish a mapping relationship between seasons and the optimal clustering parameter configuration. Furthermore, after initially dividing the lightning data through a fixed time window, it calls the aforementioned seasonal function, determines the current season based on the occurrence time in the data, and identifies the optimal clustering parameter configuration corresponding to the current season as the clustering parameter combination required by the clustering algorithm. This clustering parameter combination is then used to cluster lightning belonging to the same thunderstorm cell within the current time window into clusters. The cross-time period target tracking and association module is used to calculate the cluster center coordinates of lightning clusters based on clustering, and to perform cross-time period target tracking, association and identification; A multi-factor dynamic feature extraction module is used to capture dynamic feature parameters based on lightning clusters, wherein the features are multiple feature factors related to the hazard of lightning. The hazard index model construction and risk warning module is used to quantify risk values ​​and hazard levels by weighted calculation of multiple feature factors and historical statistical threshold division, and to transform the spatiotemporal distribution characteristics and trends of lightning activity into objective parameters and graded early warnings.

Images