A method for interpreting thunderstorm activity in time and space domains

Abstract

The present invention discloses a method for explaining thunderstorm activities in the spatiotemporal domain. The method obtains time-series atmospheric electric field data to construct an atmospheric electric field feature set, and obtains time-series membership through fuzzy C-means clustering; according to the time-series changes of the time-series membership that characterizes thunderstorms and non-thunderstorms, a semantic interpretation of weather transitions is performed on the atmospheric electric field changes in the time domain. The acquired time-series atmospheric electric field data is used to calculate the spherical coordinates of thunderstorm cloud point charges, and the distance and elevation angle in the spherical coordinates form a data set. Through fuzzy C-means clustering, the membership that characterizes the distance is obtained, and the point charge movement is semantically interpreted to obtain an explanation of thunderstorm activities in the spatiotemporal domain. The present invention realizes the explanation of weather transitions in the time domain and the explanation of thunderstorm cloud point charge movement in the space domain, thereby ensuring the interpretability of thunderstorm activities in the spatiotemporal domain.

Description

Technical Field

[0001] The invention belongs to the field of lightning detection and relates to a method for interpreting thunderstorm activities in time and space domains. Background Art

[0002] As the global ecosystem deteriorates, thunderstorms and their associated meteorological disasters become more frequent. Whether from the perspective of reducing economic losses or ensuring personal safety, research on thunderstorm detection methods is particularly urgent.

[0003] The atmospheric electric field instrument observes the electrostatic field, and the obtained ground atmospheric electric field data has high resolution and can reflect the development and evolution of regional thunderstorm clouds. The traditional time domain methods such as setting the electric field amplitude threshold, capturing fast-changing jitter or polarity reversal characteristics are used to monitor thunderstorm activities in real time and provide multi-level alarms. However, these studies only roughly and vaguely evaluate and judge the general activities of thunderstorms from the perspective of atmospheric electric field characteristics in the time domain, without considering the presentation and interpretability of thunderstorm activities in the spatial domain. Summary of the invention

[0004] In view of the above problems, the present invention uses an algorithm with interpretability to interpret thunderstorm activity based on the atmospheric electric field, proposes a method for interpreting thunderstorm activity in the spatiotemporal domain, and applies it to lightning detection to further improve the interpretability of lightning detection.

[0005] The technical solution of the present invention is to provide a method for interpreting thunderstorm activity in time and space domains, comprising the following processes:

[0006] The time series atmospheric electric field data are obtained to construct the atmospheric electric field feature set, and the time series membership is obtained through fuzzy C-means clustering. According to the time series changes of the time series membership that characterizes thunderstorms and non-thunderstorms, the changes in the atmospheric electric field in the time domain are semantically interpreted as weather changes.

[0007] The acquired time-series atmospheric electric field data are used to calculate the spherical coordinates of the point charge of the thunderstorm cloud. The distance and elevation angle in the spherical coordinates are combined into a data set. The membership representing the distance is obtained through fuzzy C-means clustering, and the motion of the point charge is semantically interpreted to obtain the interpretation of thunderstorm activity in the spatiotemporal domain.

[0008] Beneficial effects of the present invention:

[0009] 1. The present invention obtains the dominant membership of the time-series weather based on the properties of the atmospheric electric field, thereby explaining the weather transition in the time domain and ensuring the interpretability of thunderstorms in the time domain.

[0010] 2. The present invention obtains the temporal distance dominant membership based on the point charge distance and azimuth attributes to interpret the point charge movement of thunderstorm clouds in the spatial domain, thereby ensuring the interpretability of thunderstorms in the spatial domain and promoting the interpretation of thunderstorm activities in the spatiotemporal domain. BRIEF DESCRIPTION OF THE DRAWINGS

[0011] Figure 1 Flowchart of the method for interpreting thunderstorm activity in the spatial and temporal domains;

[0012] Figure 2 Ez data for November 4 and December 20, 2019;

[0013] Figure 3 Results for mean, standard deviation and threshold;

[0014] Figure 4 The weather dominant membership results from 00:00 to 02:00 on December 20, 2019;

[0015] Figure 5 It is a point charge location model for thunderstorm clouds based on a three-dimensional atmospheric electric field instrument. DETAILED DESCRIPTION

[0016] The present invention is further explained below in conjunction with the accompanying drawings and specific embodiments.

[0017] Embodiment 1:

[0018] The present invention provides a method for interpreting thunderstorm activity in time and space domains, such as Figure 1 As shown, the following steps are included:

[0019] S1: Implementation of temporal weather change interpretation, specifically:

[0020] Thunderstorm activity is detected in a refined manner by dividing the atmospheric electric field signal. Considering that a thunderstorm activity lasts for about 2 hours, the division interval is set to 30 minutes. The sampling frequency of the atmospheric electric field instrument is 1Hz. Let the time series atmospheric electric field signal be x(n), n is the signal length, and n / 1800 atmospheric electric field signal components {x1(n), x2(n),..., xn / 1800(n)} are divided into equal intervals of 30 minutes (that is, 1800 sample points). For each component xi(n), i=1,2,...,n / 1800, the eigenvalues ​​of the components are combined to form a feature data set, which is further input into the fuzzy C-means clustering algorithm to obtain clustering and membership results.

[0021] The mean, standard deviation and threshold of the atmospheric electric field signal in thunderstorm weather are significantly greater than those in sunny weather. The vertical atmospheric electric field Ez data of a sunny day (00:00 to 02:00 on November 4, 2019) and a thunderstorm (00:00 to 02:00 on December 20, 2019) were selected respectively. Figure 2 ), to intuitively show the numerical differences of the three features under different weather conditions.

[0022] Figure 2In the above figure, the maximum amplitude of Ez on a sunny day does not exceed 2kV / m, while the maximum amplitude of Ez in a thunderstorm is close to 9kV / m. The former is much smaller than the latter. Compared with the smaller fluctuation of Ez on a sunny day, the Ez in a thunderstorm fluctuates violently. According to the electric field signal division criteria described in the first paragraph of step 1, a set of eigenvalues ​​is generated every 30 minutes. This produces the following Figure 3 The eigenvalue results are shown.

[0023] Figure 3 The solid and dotted lines in the figure correspond to the characteristic value results of sunny days and thunderstorms, respectively. The characteristic values ​​of thunderstorm weather are significantly greater than those of sunny weather, especially the mean and threshold. This intuitively shows the difference in the changes of Ez data under different weather conditions. Therefore, the mean, standard deviation and threshold are used to construct a characteristic data set and used as the input of the fuzzy C-means clustering algorithm to obtain the weather dominant membership used to explain the weather transition based on the atmospheric electric field. Considering that thunderstorm activity has a time attribute, time is also added to the data set.

[0024] The weather in a certain period of time is fuzzy and can be regarded as thunderstorm-dominated or non-thunderstorm-dominated weather. For example, if a certain weather is evaluated as thunderstorm-dominated weather, then the corresponding atmospheric electric field data carries thunderstorm attributes. The fuzzy C-means clustering algorithm can convert the atmospheric electric field attributes into fuzzy sets and use multiple membership degrees to characterize multiple dominant weather information to achieve weather fuzzification and weather transition interpretation.

[0025] Assume that the domain X = {non-thunderstorm dominant, thunderstorm dominant}, and the fuzzy set At represents the weather conditions in the time period t; among them, non-thunderstorm dominant µ1 represents the degree of dominance of non-thunderstorm weather in the time period, abbreviated as MD1, and thunderstorm dominant µ2 represents the degree of dominance of thunderstorm weather in the time period, called MD2. It can be inferred that the MD1 obtained using the electric field data of sunny days is greater than MD2, while the MD2 obtained using the electric field data of thunderstorm weather is greater than MD1.

[0026] The time series data set obtained by dividing the components of the time series atmospheric electric field signal is input into the fuzzy C-means clustering algorithm, and the time series weather dominant membership {MD1j, MD2j, ..., MD n / 1800 j}, i = 1, 2, ..., n / 1800, j = 1 or 2; the corresponding time period is {T1, T2, ..., T n / 1800}; MD in time period Ti i 1 if less than MD i 2, then the weather at this time is dominated by thunderstorms, otherwise it is dominated by non-thunderstorms; time period T i+1 MD i+1 1 if less than MD i+12, then the weather at this time is still dominated by thunderstorms, and the weather has not changed; otherwise, it is dominated by non-thunderstorms, and the weather changes from thunderstorms to non-thunderstorms. More detailed semantics of weather change interpretation based on weather dominance membership are shown in Table 1, which effectively enhances the objectivity and efficiency of thunderstorm interpretation, thereby ensuring the real-time nature of thunderstorm detection. The pre-membership in Table 1 represents the membership obtained in the previous time period.

[0027] Figure 4 Shows the use of Figure 3 The weather dominant membership (MD1j to MD4j) obtained by adding the time attribute to the middle dashed line eigenvalue. Taking this membership result as an example, based on Table 1, the weather transition from 00:00 to 02:00 on December 20, 2019 is explained.

[0028] Figure 4 In the table, the weather dominance membership experienced a change from MD11 greater than MD12 to MD21 greater than MD22 to MD31 less than MD32 and then to MD41 less than MD42, which means that the weather has experienced a transition from non-thunderstorm dominance to thunderstorm dominance. Combined with Table 1, a standard weather transition explanation is quickly given: from 00:00 to 00:30, the weather is dominated by non-thunderstorms; from 00:30 to 01:00, the weather has no change and is still dominated by non-thunderstorms; from 01:00 to 01:30, the weather changes from non-thunderstorm dominance to thunderstorm dominance; from 01:30 to 02:00, the weather has no change and is still dominated by thunderstorms. These changes in membership reflect the thunderstorm events from non-thunderstorms to thunderstorms. This conclusion is consistent with the law of weather development and proves the effectiveness of using temporal membership to explain weather transitions. This provides reliable weather dominance information for the subsequent interpretation of thunderstorm activities in the spatial domain.

[0029] Table 1. Semantics for explaining weather changes

[0030]

[0031] S2: A method for interpreting thunderstorm activity in the spatiotemporal domain is proposed, which is as follows:

[0032] Figure 5 A positioning model of the thunderstorm cloud point charge q based on a three-dimensional atmospheric electric field instrument is presented. The three-dimensional atmospheric electric field instrument (that is, N) is located at the origin of the coordinate system (0,0,0).

[0033] The horizontal electric field components Ex and Ey of the two atmospheric electric fields and the vertical electric field component Ez constitute the three-dimensional atmospheric electric field. The positive semi-axes of the X-axis and the Y-axis are oriented toward the north and the east, respectively. S (r, α, β) and S (x, y, z) represent the spherical coordinates and rectangular coordinates of q, respectively. r represents the distance from S to N.

[0034] The localization formula for a point charge q is:

[0035] (1)

[0036] in, , ; is the charge value of charge q, , They are respectively the dielectric constant of air and the dielectric constant of the ground where the atmospheric electric field instrument is located.

[0037] Thunderstorm activity requires a time-domain explanation based on the change of the atmospheric electric field, and also requires an explanation in the spatial domain. Thunderstorm activity in the spatial domain is reflected by the movement of point charges. This spatial movement relative to the atmospheric electric field meter is mainly manifested as a change in the distance from the electric field meter, that is, Figure 5 Therefore, the distance from the point charge to the electric field meter is taken as one of the parameters required for this explanation.

[0038] In addition, the elevation angle β of a point charge in the coordinate system changes regularly as the point charge moves. For example, when a point charge approaches the electric field meter from a long distance, its elevation angle changes from small to large. When the point charge moves away from the electric field meter, its elevation angle changes from large to small. Therefore, the elevation angle of the point charge is used as another parameter required to explain thunderstorm activity in the spatial domain.

[0039] The fuzzy C-means clustering algorithm is introduced again for the interpretation of point charge motion. According to the above analysis, it is feasible to use the dataset consisting of distance and elevation angle as the input of fuzzy C-means clustering. Since the point charge motion has a temporal attribute, this attribute is also added to the dataset. Before making an interpretation, three key points need to be considered in advance: 1) The spatial domain thunderstorm interpretation in non-thunderstorm events is unnecessary; 2) The thunderstorm interpretation needs to maintain the temporal consistency of the time domain and the spatial domain to ensure a strong spatiotemporal correlation; 3) The magnitude difference between the point charge distance and the elevation angle and time.

[0040] For the first point, the atmospheric electric field component with thunderstorm dominant properties is used to calculate the distance r and elevation angle β using equation (1). For the second point, since the time interval in the time domain is 30 minutes, the time interval in the space domain is also set to 30 minutes. The time sequence in the space domain is consistent with that in the time domain, which is {T1, T2, ..., T n / 1800 Here, the results of distance and elevation angle are divided into 30-minute time series, and the divided data are averaged to reduce the negative impact of the fluctuation of distance and elevation angle data caused by abnormal point charges on the qualitative thunderstorm interpretation. For the third point, the distance is generally in the order of tens to hundreds of meters, while the elevation angle is between 0 and 90 degrees. The distance, elevation angle and time are normalized to [0,1] to reduce the impact of the magnitude difference of the three on the interpretation of thunderstorm activity.

[0041] Determine the order of operations from the first point (distance and elevation calculation) to the second point (average calculation after segmentation), and then to the third point (normalization). The distance, elevation, and time obtained after sequential processing form a data set and input it into the fuzzy C-means clustering algorithm.

[0042] Assume that the domain X = {long distance dominant, short distance dominant}, the fuzzy set At represents the point charge motion condition in the time period t, and the distance dominant membership threshold reflecting the point charge motion is set to 0.5. If it is less than the threshold, it is short distance dominant, otherwise it is long distance dominant. The long distance dominant µ3 represents the degree of dominance of the long distance on the point charge motion in the time period, abbreviated as MD3, and the short distance dominant µ4 represents the degree of dominance of the short distance on the point charge motion in the time period, called MD4. It can be inferred that the MD3 obtained using the distant point charge coordinate data is greater than MD4, while the MD4 obtained using the nearby point charge coordinate information is greater than MD3.

[0043] A certain period of time T i MD i 3 If less than MD i 4, then the distance at this time is dominated by the short distance, and the point charge is near the atmospheric electric field instrument; otherwise, it is dominated by the long distance, and the point charge is not near the atmospheric electric field instrument. Time period T i+1 MD i+1 3 If less than MD i+1 4. At this time, the distance is still dominated by the short distance, and its movement does not change; on the contrary, it is dominated by the long distance, and the point charge moves away.

[0044] The semantics of point charge motion interpretation based on distance-dominated membership is shown in Table 2. The pre-membership in Table 2 represents the membership obtained in the previous time period. Based on Table 2, the spatial domain thunderstorm activity interpretation is performed. So far, combining Tables 1 and 2, the spatial and temporal domain thunderstorm activity interpretation can be realized.

[0045] Table 2 Semantics for explaining the motion of a point charge

[0046]

Claims

1. A method for interpreting thunderstorm activity in the spatiotemporal domain, characterized in that: The process includes: Acquire the time series atmospheric electric field data to construct the atmospheric electric field feature set, and obtain the time series membership through fuzzy C-means clustering; according to the time series changes of the time series membership that characterize thunderstorms and non-thunderstorms, the semantic interpretation of weather changes in the atmospheric electric field changes in the time domain is performed; The acquired time-series atmospheric electric field data are used to calculate the spherical coordinates of thunderstorm cloud point charges. The distance and elevation angle in the spherical coordinates form a data set. The membership degree representing the distance is obtained through fuzzy C-means clustering. The motion of the point charge is semantically interpreted to obtain the thunderstorm activity interpretation in the spatiotemporal domain. The specific process is as follows: The positioning model of thunderstorm cloud point charge q based on the three-dimensional atmospheric electric field instrument, the three-dimensional atmospheric electric field instrument N is located at the origin of the coordinate system (0,0,0); The horizontal electric field components Ex and Ey of the two atmospheric electric fields and the vertical electric field component Ez constitute the three-dimensional atmospheric electric field; the positive semi-axes of the X-axis and the Y-axis are oriented toward the north and the east respectively; S(r, α, β) and S(x, y, z) represent the spherical coordinates and rectangular coordinates of q respectively; r represents the distance from S to N; The localization formula for a point charge q is: in, q' is the charge value of charge q, ε1 and ε2 are the dielectric constant of air and the dielectric constant of the ground where the atmospheric electric field instrument is located, respectively; The distance, elevation angle and time attributes form a data set and perform fuzzy C-means clustering: Let the domain X = {long distance dominant, short distance dominant}, the fuzzy set At represents the point charge motion condition in the time period t, and the distance dominant membership threshold reflecting the point charge motion is set to 0.

5. If it is less than the threshold, it is short distance dominant, otherwise it is long distance dominant. Long distance dominant μ3 represents the degree of dominance of long distance on point charge motion in this time period, abbreviated as MD3, and short distance dominant μ4 represents the degree of dominance of short distance on point charge motion in this time period, called MD4; A certain period of time T i MD i 3 If less than MD i 4, then the distance at this time is dominated by the short distance, and the point charge is near the atmospheric electric field instrument; otherwise, it is dominated by the long distance, and the point charge is not near the atmospheric electric field instrument; time period T i+1 MD i+1 3 If less than MD i+1 4. At this time, the distance is still dominated by the short distance, and its motion does not change; on the contrary, it is dominated by the long distance, and the point charge moves away; The specific process of forming a data set with distance, elevation angle and time attributes is as follows: S1: Using the atmospheric electric field component with thunderstorm dominant properties, the distance r and elevation angle β are calculated using equation (1); S2: Set the time interval in the spatial domain to 30 minutes; the time sequence in the spatial domain is consistent with the time sequence in the time domain, which are {T1, T2, ..., T n / 1800 }, divide the distance and elevation results into 30-minute time series, and perform mean processing on the divided data; S3: Normalize the distance, elevation angle and time to [0,1].

2. A method for interpreting thunderstorm activity in time and space domain according to claim 1, characterized in that: The specific implementation process of the semantic interpretation of weather transformation of the atmospheric electric field changes in the time domain is as follows: Assume that the time series atmospheric electric field signal is x(n), n is the signal length, and divide the atmospheric electric field signal into n / 1800 components {x1(n), x2(n), ..., xn / 1800(n)} with equal intervals of 30 minutes; for each component xi(n), i = 1, 2, ..., n / 1800, a feature data set is formed by combining the eigenvalues ​​of the components; The feature data set is input into the fuzzy C-means clustering algorithm: let the domain X = {non-thunderstorm dominant, thunderstorm dominant}, the fuzzy set At represents the weather conditions in the time period t; among them, non-thunderstorm dominant μ1 represents the degree of dominance of non-thunderstorm in the weather of this time period, abbreviated as MD1, and thunderstorm dominant μ2 represents the degree of dominance of thunderstorm in the weather of this time period, called MD2; the temporal weather dominant membership {MD1j, MD2j, ..., MD n / 1800 j}, i = 1, 2, ..., n / 1800, j = 1 or 2; the corresponding time period is {T1, T2, ..., T n / 1800 }; MD in time period Ti i 1 if less than MD i 2, then the weather at this time is dominated by thunderstorms, otherwise it is dominated by non-thunderstorms; time period T i+1 MD i+1 1 if less than MD i+1 2. Then the weather at this time is still dominated by thunderstorms, and there is no change in the weather; otherwise, it is dominated by non-thunderstorms, and the weather changes from being dominated by thunderstorms to being dominated by non-thunderstorms.

3. A method for interpreting thunderstorm activity in spatiotemporal domain according to claim 2, characterized in that: The characteristic data set specifically comprises: calculating the mean, standard deviation and threshold of the atmospheric electric field signal component to form the corresponding characteristic value, and constructing the characteristic data set by the characteristic value and time attribute.

Images