A method for manually observing the daily duration of thunderstorms based on lightning detection data

Abstract

This invention relates to the field of meteorological observation technology and discloses a method for manually observing the continuation of thunderstorm days based on lightning detection data. First, data quality is controlled to improve the quality of thunderstorm and lightning data, resulting in more accurate and reliable numbers of thunderstorm and lightning days. Then, a correlation model is established between total lightning days and ground-to-ground lightning days to calculate the number of total lightning days in previous years. Next, a correlation model is established between the number of thunderstorm days and the number of total lightning days to identify the optimal matching radius for thunderstorm and total lightning days. The number of total lightning days within the optimal matching radius is then determined as the manually observed number of thunderstorm days in the central region. The total lightning days obtained using this method show good consistency with manually observed thunderstorm days. The average annual number of total lightning days at five meteorological stations in Wuhan is identified as 0.4 days more than the average annual number of thunderstorm days observed manually, with relative errors ranging from 2.8% to 14.1%. The relative error between the average number of thunderstorm days and the average number of total lightning days from 2007 to 2012 is 1.4%.

Description

Technical Field

[0001] This invention relates to the field of meteorological observation technology, specifically to a method for manually observing the duration of thunderstorm days based on lightning detection data. Background Technology

[0002] Thunderstorms are small- to medium-scale weather processes accompanied by various weather phenomena, with lightning being the most fundamental characteristic. Lightning generates high voltage, large current, and intense electromagnetic radiation, directly or indirectly impacting human life and infrastructure. With rapid economic and social development and the widespread application of microelectronics, the dangers and losses caused by lightning have become increasingly severe, making it one of the ten most serious natural disasters.

[0003] The electrification mechanism of thunderstorms, the occurrence, development, spatiotemporal distribution characteristics of lightning, and lightning protection technologies have always been important research topics in the field of atmospheric electricity. Thunderstorm observation, as one of the observation operations of meteorological stations, has accumulated more than 50 years of data. However, due to the limitations of low station network density and low spatiotemporal resolution, as well as the actual development of automated observation, according to the requirements of my country's meteorological departments, meteorological stations at all levels have canceled manual observation of thunderstorms since 2014. Nevertheless, the number of days of manually observed thunderstorms is an important basis for lightning protection classification and lightning protection engineering design in standards. On the basis of automated meteorological observation, researching the method of continuing the number of days of manually observed thunderstorms has significant engineering implications.

[0004] For example, Chinese patent CN 117805939 B discloses a method to replace manual observation of thunderstorm days. This patent establishes a linear model and a power model of the annual average cloud-to-ground lightning density and the number of manually observed thunderstorm days after quality control of lightning location data, and completes the consistency analysis of lightning location data and thunderstorm days. Then, it verifies the reliability of lightning location data to replace the number of manually observed thunderstorm days, thus proving the feasibility of lightning location data to replace the number of manually observed thunderstorm days.

[0005] However, in practical applications, considering that the observation of thunderstorm days at meteorological stations is determined by whether the human ear can hear thunder, which includes both ground lightning and cloud lightning, there is an error in retrieving thunderstorm days solely through ground lightning. Therefore, this application proposes a method for manually observing the continuation of thunderstorm days based on lightning detection data to solve the above problems. Summary of the Invention

[0006] (a) Technical problems to be solved

[0007] To address the shortcomings of existing technologies, this invention provides a method for manually observing the continuity of thunderstorm days based on lightning detection data. This method first controls data quality to improve the quality of thunderstorm and lightning data, resulting in more accurate and reliable numbers of thunderstorm and lightning days. Then, it establishes a correlation model between total lightning days and ground lightning days to calculate the number of total lightning days in previous years. Next, it establishes a correlation model between thunderstorm days and total lightning days to identify the optimal matching radius for thunderstorm and total lightning days. The number of total lightning days within the optimal matching radius is then determined as the number of manually observed thunderstorm days in the central region.

[0008] (II) Technical Solution

[0009] To achieve the above objectives, the present invention provides the following technical solution: a method for manually observing the daily duration of thunderstorms based on lightning detection data, comprising the following steps:

[0010] S1. Data Quality Control:

[0011] For a specific meteorological station, data quality control is performed on its thunderstorm days and lightning days within a specific radius. In the first quality control, the data of each thunderstorm day and lightning day are compared one by one. If the two match successfully, the thunderstorm day and lightning day are considered to be real and valid data.

[0012] A second quality control process is performed on the remaining lightning and thunderstorm data. The lightning and thunderstorm data is overlaid and compared with radar data. If the data meets the identification threshold, the lightning and thunderstorm day is considered to be real and valid data. If the data does not meet the identification threshold, it is identified as false data. Similarly, a second quality control process is performed on the remaining thunderstorm data. The thunderstorm data is overlaid and compared with radar data. If the data meets the identification threshold, the thunderstorm day is considered to be real and valid data. If the data does not meet the identification threshold, it is identified as false data.

[0013] Data quality control is performed on all-flash lightning days within a specific radius area. All-flash lightning days are overlaid and compared with radar data. If the identification threshold is met, the all-flash lightning days are considered to be real and valid data. If the identification threshold is not met, they are identified as false data.

[0014] S2. Compare the actual number of days with total lightning strikes and the number of days with ground lightning strikes, and establish a correlation model between days with total lightning strikes and days with ground lightning strikes;

[0015] S3. Based on the correlation model between total lightning days and ground lightning days, calculate the number of total lightning days per year, month, and day for different meteorological stations in different radii of spherical spatial regions in previous years.

[0016] S4. Based on S3, the number of days with total lightning strikes per year, month, and day for different meteorological stations with different radii in spherical spatial regions over the years is calculated. A correlation model between the number of thunderstorm days and the number of days with total lightning strikes is established under different radius sizes after data quality control. The optimal matching radius for thunderstorm days and days with total lightning strikes is identified. That is, the number of days with total lightning strikes within the optimal matching radius is the number of artificially observed thunderstorm days in the central region.

[0017] (III) Beneficial Effects

[0018] Compared with existing technologies, this invention provides a method for manually observing the daily duration of thunderstorms based on lightning detection data, which has the following beneficial effects:

[0019] This method for manually observing thunderstorm days based on lightning detection data first controls data quality to improve the quality of thunderstorm and lightning data, resulting in more accurate and reliable numbers of thunderstorm and lightning days. Then, it establishes a correlation model between total lightning days and ground lightning days to calculate the number of total lightning days in previous years. Next, it establishes a correlation model between thunderstorm days and total lightning days to identify the optimal matching radius for thunderstorm and total lightning days. The number of total lightning days within the optimal matching radius is then determined as the manually observed thunderstorm days for that central region. The total lightning days obtained using this method show good consistency with manually observed thunderstorm days. It identifies that the average annual number of total lightning days at five meteorological stations in Wuhan is 0.4 days more than the average annual number of thunderstorm days observed manually, with relative errors ranging from 2.8% to 14.1%. The relative error between the average thunderstorm days and the average annual number of total lightning days from 2007 to 2012 is 1.4%. Attached Figure Description

[0020] Figure 1 This is a flowchart of the data quality control process of the present invention;

[0021] Figure 2 This is a graph showing the relationship between radar echo intensity and ADTD two-dimensional monitoring lightning distribution according to the present invention.

[0022] Figure 3 This is a schematic diagram showing the proportion of lightning at different distances from the center of the strong echo in this invention. Detailed Implementation

[0023] The technical solutions of the present invention will be clearly and completely described below with reference to the embodiments of the present invention. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0024] The relevant terms involved in this invention are explained as follows:

[0025] Thunderstorm days: At a meteorological station, hearing thunder once or more in a day is counted as one thunderstorm day; the total number of thunderstorm days in a year is counted as the annual thunderstorm days.

[0026] Lightning days: Within a specific area, one or more lightning strikes within a day are counted as one lightning day; the total number of lightning days in a year is counted as the annual lightning days.

[0027] Lightning location system: A system that uses technical means and methods to determine the time, location, polarity and other lightning parameters of a lightning strike. The lightning location system established in Hubei Province in 2006 can monitor the occurrence of ground lightning in real time, and the lightning location system upgraded in 2014 can monitor the occurrence of cloud lightning and ground lightning (i.e., total lightning monitoring) in real time.

[0028] Data period: Thunderstorm day data is from 2007 to 2013; ground flash monitoring data is from 2007 to 2024; total flash monitoring data is from 2015 to 2024.

[0029] Considering that meteorological stations determine whether a thunderstorm day is a thunderstorm day by whether the human ear can hear thunder (including both ground lightning and cloud lightning), relying solely on ground lightning to invert thunderstorm days introduces errors. This invention proposes a method for manually observing thunderstorm day continuation based on lightning detection data. First, it controls data quality to improve the quality of thunderstorm and lightning data, obtaining more accurate and reliable numbers of thunderstorm and lightning days. Then, it establishes a correlation model between all-flash thunderstorm days and ground lightning thunderstorm days, calculating the number of all-flash thunderstorm days in previous years. Next, it establishes a correlation model between thunderstorm days and all-flash thunderstorm days, identifying the optimal matching radius for thunderstorm and all-flash thunderstorm days. The number of all-flash thunderstorm days within the optimal matching radius is then the number of manually observed thunderstorm days in that central region.

[0030] Data quality control is based on the correlation characteristics between manual thunderstorm observations, lightning monitoring, and radar echo intensity changes. Following the general approach of "internal consistency checks of correlated meteorological data within the same spatiotemporal interval," and combining literature review and case analysis, a comprehensive selection of consistency thresholds for multi-source data in severe convective weather processes is undertaken. A multi-level lightning data quality control technology framework based on multi-source data is then designed, such as... Figure 1 As shown.

[0031] It should be noted that domestic and foreign researchers have found that thunder is usually only heard within a range of 8-15km from lightning, and in good cases it can reach 20km. Therefore, the threshold for matching thunderstorm days and lightning days is 20km, that is, the situation in a spatial area with a radius of 20km centered on the meteorological station is statistically analyzed.

[0032] The process for selecting the radar echo intensity identification threshold is as follows:

[0033] Firstly, regarding spatial spacing, the system was selected with a positioning error more than three times that of the lightning location system. Research literature indicates that, taking Hubei Province as an example, the theoretical positioning accuracy of the lightning location system deployed by the Hubei meteorological department is approximately 500 meters. We collected some lightning disaster cases since the availability of ground flash return stroke monitoring data, and randomly selected nine cases with relatively well-organized data. By comparing and analyzing the actual time and location of the lightning disasters with the monitoring results, we verified the positioning accuracy of the monitoring system. Table 1 presents relevant data for eight lightning disaster cases in Hubei Province. Table 1 is as follows:

[0034] Table 1. Location data of 8 lightning strike incidents in Hubei Province

[0035] serial number Place nature of damage Actual location of lightning strike Positioning the counterattack Straight-line distance deviation (m) Time of occurrence <![CDATA[1 # ]]> An oil storage depot Direct lightning strike (building damaged) E 114.6010°N 30.5520° E 114.6017°N 30.5539° 222 Around 6:00 AM on June 23, 2008 <![CDATA[2 # ]]> Henao Primary School Lightning detection (equipment malfunction) E 114.271°N 30.070° E 114.2668°N 30.0751° 697 Around 3:00 AM on June 23, 2008 <![CDATA[3 # ]]> Batan Road Gas Station Lightning detection (equipment malfunction) E 114.2801°N 30.4912° E 114.287°N 30.4896° 684 August 4, 2012, 21:30 <![CDATA[4 # ]]> Network Center of a University in Wuhan Lightning detection (equipment malfunction) E 114.4062°N 30.5124° E 114.4003°N 30.5118° 569 Around 9:00 AM on September 2, 2012 <![CDATA[5 # ]]> Tanziling Scenic Area Secondary Platform Direct lightning strike (personnel casualties) E 111.0183°N 30.8346° E 111.0072°N 30.8426° 1383 2:00 PM, July 16, 2013 <![CDATA[6 # ]]> Jiangxia Toll Station on Beijing-Zhuhai Expressway Lightning detection (equipment malfunction) E 114.3116°N 30.3096° E 114.3189°N 30.3125° 771 April 3, 2015, 20-22 <![CDATA[7 # ]]> A mechanical and electrical equipment engineering company Lightning detection (equipment malfunction) E 114.3054°N 30.4761° E 114.3041°N 30.4737° 295 May 18, 2018 <![CDATA[8 # ]]> A university in Xinzhou District Direct lightning strike (personnel casualties) E 114.5458°N 30.7166° E 114.5476°N 30.7199° 405 18:40 on August 31, 2018

[0036] As can be seen from Table 1, the location errors of the above 8 lightning strike accidents range from 222 to 1383 m. According to the probability attributes of ground flash events, the scale of the target area should be appropriately enlarged. Referring to the calculation of the equivalent receiving area in lightning protection engineering design, it is generally more than 3 times. Therefore, the spatial interval identification threshold between lightning data and radar echo used in this method is set to 5 km. Similarly, the spatial interval identification threshold between thunderstorm data and radar echo is increased by 5 km on the basis of 20 km, that is, it is set to 25 km.

[0037] Secondly, regarding the radar echo intensity threshold, it can be determined comprehensively through typical case analysis and relevant literature results. Taking a typical process that occurred in Wuhan and surrounding areas in May 2023 as an example, a comprehensive analysis was conducted. From May 25th to May 27th, 2023, under the combined influence of an upper-level trough and warm and humid southwest airflow, Wuhan and surrounding areas experienced a heavy precipitation process, accompanied by severe convective weather such as thunderstorms, strong winds, and short-term heavy rainfall. Figure 2 As shown, Figure 2 The distribution of ground flash activity within 6 minutes before and after the superposition of Doppler radar echo intensity at 6 typical moments is presented.

[0038] As shown in the figure, there is a good correlation between the Doppler radar echo intensity at six typical moments and the distribution of lightning activity within 6 minutes before and after. Lightning mainly occurred in the center of strong echoes with an intensity above 35 dBz and its surrounding area. To further analyze the relationship between radar echo intensity and lightning activity distribution, the relationship between the strong echo center (>=35 dBz) and the lightning strike area during this process was statistically analyzed, as shown below. Figure 3 As shown, Figure 3 The graph shows the percentage of lightning occurring at different distances from strong radar echoes. As can be seen from the graph, 63.7% of lightning occurred within 1 km of the center of the strong echo, and 84.5% of lightning occurred within 5 km of the center of the strong echo.

[0039] In summary, based on the analysis of typical cases and relevant literature, 35 dBz is used as the threshold for judging the quality control of lightning data by radar echo intensity. If it is greater than this value, the lightning data observed at the same time is judged to be real; otherwise, it is judged to be false.

[0040] Based on this, in specific implementation, taking Hubei Province as an example, in 2006, the Hubei Provincial Meteorological Department planned and constructed a lightning monitoring system, which can monitor lightning parameters such as the time, latitude and longitude, and polarity of cloud-to-ground lightning in real time; in 2014, the lightning monitoring system was upgraded to add cloud-to-ground lightning monitoring function, which can monitor lightning parameters such as the time, latitude and longitude, and polarity of cloud-to-ground lightning in real time. From 2007 to 2013, the five meteorological stations in Wuhan had a total of 906 thunderstorm days, and the number of days with ground lightning within 20km of each meteorological station was 1959 days.

[0041] Statistics were collected from five meteorological stations on each thunderstorm day. The data for each thunderstorm day was compared one by one with the data for lightning strike days within a 20km radius of the stations. It was found that the time of each thunderstorm day was recorded as a lightning strike day, and all 906 thunderstorm days were determined to be valid data.

[0042] We statistically analyzed the number of lightning and thunderstorm days at five meteorological stations across various regions. We then compared the lightning and thunderstorm day data with the thunderstorm day data one by one. A total of 906 days were recorded as thunderstorm days with lightning strikes, indicating that after the first quality control, the accuracy rate of the lightning and thunderstorm day data was 46%. A second quality control was performed on the remaining 54% of the lightning and thunderstorm day data. This involved overlaying and comparing the lightning data with radar data for each lightning and thunderstorm day. If the identification threshold was met (the spatial interval between the lightning data and the radar echo was 5 km), and 35 dBz was used as the identification threshold for lightning data quality control based on radar echo intensity, then if the radar echo intensity within 5 km of a lightning strike was greater than 35 dBz, the lightning strike was identified as a real lightning strike, and that day was considered a real thunderstorm day.

[0043] The quality control results showed that, for 1,959 thunderstorm days, after the first quality control, 906 lightning-ground day data were retained, which were considered to have been manually observed by meteorological stations and were therefore considered to be genuine lightning-ground day data, with the effective lightning-ground day data accounting for 46%. In the second quality control, 835 lightning-ground days that were "snatched" out from those removed in the first quality control were found, which met the radar echo intensity identification threshold, increasing the effective lightning-ground day data to 88.9%.

[0044] To establish a correlation model between full-flash lightning days and ground-flash lightning days, lightning monitoring data (full-flash data) from Hubei Province, spanning 2015-2024, was used. Five meteorological stations in Wuhan were selected as the center, with different monitoring distances defined as radii R. Within a range of R from 1 to 20 km, quality control was performed on the annual full-flash lightning day data for each meteorological station at different monitoring radii R. Taking the full-flash lightning days within a 20 km radius of the Huangpi meteorological station in 2022 as an example, the statistical results show that in 2022, the number of days monitored as full-flash lightning days and ground-flash lightning days within a 20 km radius of the Huangpi meteorological station were 61 and 35, respectively. After overlay and comparison with radar data, a total of 44 days were monitored as full-flash lightning days and met the radar echo threshold, and a total of 32 days were monitored as ground-flash lightning days and met the radar echo threshold. This indicates that after quality control, the accuracy rate of the full-flash lightning day data was 72%, and the accuracy rate of the ground-flash lightning day data was 91%.

[0045] Using the above quality control methods, the number of total lightning days and ground lightning days monitored by the lightning location system in spherical spatial areas of different radii at 100m intervals were statistically analyzed at each of the five meteorological stations. The numbers of total lightning days and ground lightning days were then compared and analyzed. After fitting, a good linear relationship was found between the number of total lightning days and the number of ground lightning days, with a correlation coefficient of 0.9658, reaching a significance level of 0.001. The expression is as follows:

[0046] T 1 n = 1.2009T 2 n + 0.0444 (1)

[0047] In the formula, T 1 n For a day of total lightning, T 2 n It is the day of lightning and thunder.

[0048] It should be noted that the steps for identifying lightning days within spherical spatial regions of different radii are as follows:

[0049] Spatial coordinates of the meteorological station (LonA, LatA, h) A The spatial coordinates of the lightning are (LonB, LatB, h). B ), with a radius of R;

[0050] When [(LonB-LonA)] 2 +(LatB-LatA) 2 +(h B -h A ) 2 ] 1 / 2When h ≤ R, the lightning strike is included in the daily lightning dataset for that radius, i.e., it is included in the daily lightning dataset for all lightning strikes; when h B If the altitude is equal to the location, it indicates that the lightning is a ground lightning strike, and it will be included in the daily data set of ground lightning strikes. B If the altitude is greater than this location, it indicates that the lightning is cloud-to-cloud lightning and will not be included in the daily lightning data set.

[0051] Otherwise, when [(LonB-LonA)] 2 +(LatB-LatA) 2 +(h B -h A ) 2 ] 1 / 2 When the value is greater than R, the lightning strike is not included in the lightning days data set for that sphere radius.

[0052] The number of days with total lightning strikes at different radii of spherical spatial regions at different meteorological stations from 2007 to 2012 was calculated using the above method. Then, a correlation model was established between the number of thunderstorm days (after data quality control) and the calculated number of days with total lightning strikes. The optimal matching radius for thunderstorm days and days with total lightning strikes was identified, and the number of days with total lightning strikes within the optimal matching radius was determined to be the number of artificially observed thunderstorm days in that central region. The specific steps are as follows:

[0053] 1) Taking a single station as the center, combined with lightning monitoring data from 2007 to 2012, according to formula (1), the number of all-flash lightning days under different radius sizes is counted year by year, month by month, and day by day. Data sets of all-flash lightning days and thunderstorm days after data quality control are constructed. The annual mean square error, monthly mean square error and daily F1 score of the two sets of data are calculated respectively.

[0054] 2) By setting the weights of the annual and monthly mean square errors and the F1 score, the overall matching score S(i) under the current radius size is obtained;

[0055] 3) Determine the optimal matching size between thunderstorm days and all-flash lightning days for multiple meteorological stations. By calculating the average of the comprehensive scores of multiple meteorological stations under different radius sizes, the radius size corresponding to the highest average score is the optimal matching size between thunderstorm days and all-flash lightning days for multiple meteorological stations.

[0056] 4) When selecting the optimal matching radius size around the meteorological station, the matching between manually observed thunderstorm days and lightning monitoring full-flash thunderstorm days is optimal. That is, within a certain area, the number of full-flash thunderstorm days within the area with the lightning positioning system monitoring size as the optimal matching radius can be used to replace the number of manually observed thunderstorm days.

[0057] For a single meteorological station, the annual mean square error is calculated based on a specific matching radius size, as shown in the following formula:

[0058] (2-1)

[0059] In the formula, MSE y T represents the annual mean square error. d (i) represents the number of thunderstorm days in year i; T n (i) represents the number of days with total lightning in year i; n is the number of years, which is 6.

[0060] The formula for calculating the monthly mean squared error is as follows:

[0061] (2-2)

[0062] In the formula, MSE m T represents the monthly mean square error. d (i) represents the number of thunderstorm days in the i-th month; T n (i) represents the number of days with total lightning in the i-th month; n is the number of months, which is 72.

[0063] Conduct daily matching degree analysis, taking the observation results of 7 consecutive days as an example:

[0064] Assume the thunderstorm daily sequence is: T d = [0, 0, 1, 1, 0, 1, 0];

[0065] Assume the daily lightning sequence is: T n = [0, 1, 1, 0, 0, 0, 0].

[0066] Based on the daily comparison of the binary sequence, the following four categories can be statistically analyzed:

[0067] category definition Calculation formula Meaning Explanation hit a a = Σ(Tn·Td) The day was both a day of total lightning and a day of thunderstorms. False alarm b b = Σ[Tn·(1−Td )] The day was a day of total lightning, but not a day of thunderstorms. underreporting c c = Σ[(1−Tn)·Td] The day was a thunderstorm day, but not a day of full-fledged lightning. Correct negation d d = Σ[(1−Tn)·(1−Td )] The day was neither a day of full lightning nor a day of thunderstorms.

[0068] Then, the daily matching index F1 score under this sphere radius size is calculated. The formula for calculating the F1 score is as follows:

[0069] (2-3)

[0070] (2-4)

[0071] (2-4)

[0072] In the formula, F1 is the F1 score; Precision is the precision rate; Recall is the recall rate; a indicates that the day is both a day of all-lightning and a day of thunderstorms; b indicates that the day is a day of all-lightning and a day of thunderstorms, but not a day of thunderstorms; c indicates that the day is a day of thunderstorms, but not a day of all-lightning and lightning.

[0073] To eliminate differences in units and the data themselves, the annual mean square error, monthly mean square error, and F1 score data are normalized. The normalization method is shown in the following formula:

[0074] (2-6)

[0075] In the formula, D k r is the normalized value of the current k-th indicator. k The k-th index value; max k and min k These are the maximum and minimum values ​​for this indicator, respectively.

[0076] Then, by setting the weights of the annual and monthly mean square errors and the F1 score, the overall matching score S(i) for the current radius size is obtained, where i represents the current radius size. In this method, the weights of the annual and monthly mean square errors and the F1 score are all 1 / 3, that is, k1, k2, and k3 are all 1 / 3. The formula for calculating the overall score S(i) is as follows:

[0077] (2-7)

[0078] In the formula, the MSE / F1 ratios are normalized data.

[0079] Finally, the optimal matching size between the number of thunderstorm days and the number of all-flash lightning days for multiple meteorological stations is determined. This is achieved by calculating the average of the comprehensive scores of multiple meteorological stations under different radius sizes. Similarly, the radius size corresponding to the highest average score is the optimal matching size between the number of thunderstorm days and the number of all-flash lightning days for multiple meteorological stations.

[0080] To verify the reliability of the above method, the calculated annual average number of days with total lightning strikes was compared with the number of days with thunderstorms observed manually from 2007 to 2012, as shown in Table 2 below. Table 2 shows that the average annual number of days with total lightning strikes identified by the five meteorological stations in Wuhan was 0.4 days more than the average annual number of thunderstorms observed manually, with the relative error for each station ranging from 2.8% to 14.1%. The relative error between the average number of thunderstorm days and the average number of days with total lightning strikes from 2007 to 2012 was 1.4%.

[0081] Table 2. T data for Wuhan stations from 2007 to 2012 d With T n Comparative test

[0082] Serial Number Station Name <![CDATA[T d Thunderstorm day / d]]> <![CDATA[T n Thunderstorm Day / d]]> <![CDATA[T n / T d ]]> Relative error / % 1 Wuhan 26.7 29.2 1.094 9.4 2 Jiangxia 31.3 33.1 1.058 5.8 3 New Zealand 29 28.2 0.972 2.8 4 Huangpi 29.7 25.5 0.859 14.1 5 Caidian 26.8 29.3 1.093 9.3 6 average 28.7 29.1 1.014 1.4

[0083] Statistics from the 2013 lightning location system and manual thunderstorm observation data show that thunderstorm days were recorded only from February to September. Within a circular area of ​​8 km radius centered on five meteorological stations in Wuhan, 99.7% of the days identified from February to September were total lightning strikes. Therefore, the number of thunderstorm days identified by each meteorological station in 2013 was replaced by relevant data from February to September 2013, as shown in Table 3 below:

[0084] Table 3. T at Wuhan stations in 2013 d With T n Comparative test

[0085] Serial Number Station Name <![CDATA[T d Thunderstorm day / d]]> <![CDATA[T n Thunderstorm Day / d]]> <![CDATA[T n / T d ]]> Relative error / % 1 Wuhan 25 27.8 1.112 11.2 2 Jiangxia 26 28.7 1.104 10.4 3 New Zealand 33 27.2 0.8242 17.6 4 Huangpi 29 31.5 1.086 8.6 5 Caidian 33 26.9 0.8152 18.5 6 average 29.2 28.4 0.9726 2.7

[0086] Statistics from 2013 show that the average number of days with total lightning strikes per year at the five meteorological stations in Wuhan was 0.8 days less than the average number of days with thunderstorms per year as observed manually. The relative error for each station ranged from 8.6% to 18.5%, and the relative error between the number of thunderstorm days and the number of days with total lightning strikes per year in 2013 was 2.7%.

[0087] In summary, the total lightning days obtained using this method show good consistency with manually observed thunderstorm days.

[0088] Although embodiments of the invention have been shown and described, it will be understood by those skilled in the art that various changes, modifications, substitutions and alterations can be made to these embodiments without departing from the principles and spirit of the invention, the scope of which is defined by the appended claims and their equivalents.

Claims

1. A method for manually observing the daily duration of thunderstorms based on lightning detection data, characterized in that, Includes the following steps: S1. Data Quality Control: For a specific meteorological station, data quality control is performed on its thunderstorm days and lightning days within a specific radius. In the first quality control, the data of each thunderstorm day and lightning day are compared one by one. If the two match successfully, the thunderstorm day and lightning day are considered to be real and valid data. A second quality control process is performed on the remaining lightning and thunderstorm data. The lightning and thunderstorm data is overlaid and compared with radar data. If the data meets the identification threshold, the lightning and thunderstorm day is considered to be real and valid data. If the data does not meet the identification threshold, it is identified as false data. Similarly, a second quality control process is performed on the remaining thunderstorm data. The thunderstorm data is overlaid and compared with radar data. If the data meets the identification threshold, the thunderstorm day is considered to be real and valid data. If the data does not meet the identification threshold, it is identified as false data. Data quality control is performed on all-flash lightning days within a specific radius area. All-flash lightning days are overlaid and compared with radar data. If the identification threshold is met, the all-flash lightning days are considered to be real and valid data. If the identification threshold is not met, they are identified as false data. S2. Compare the actual number of days with total lightning strikes and the number of days with ground lightning strikes, and establish a correlation model between days with total lightning strikes and days with ground lightning strikes; S3. Based on the correlation model between total lightning days and ground lightning days, calculate the number of total lightning days per year, month, and day for different meteorological stations in different radii of spherical spatial regions in previous years. S4. Based on S3, the number of days with total lightning strikes per year, month, and day for different meteorological stations with different radii in spherical spatial regions over the years is calculated. A correlation model between the number of thunderstorm days and the number of days with total lightning strikes is established under different radius sizes after data quality control. The optimal matching radius for thunderstorm days and days with total lightning strikes is identified. That is, the number of days with total lightning strikes within the optimal matching radius is the number of artificially observed thunderstorm days in the central region.

2. The method for manually observing the daily duration of thunderstorms based on lightning detection data according to claim 1, characterized in that, The threshold for matching thunderstorm days and lightning days in the data quality control process is less than 20km, that is, the statistics are based on the spatial area with a radius of less than 20km centered on the meteorological station.

3. The method for manually observing the daily duration of thunderstorms based on lightning detection data according to claim 2, characterized in that, The spatial interval identification threshold between the thunderstorm data and the radar echo is increased by 5km based on 20km, that is, it is set to 25km. The identification threshold during the superposition and comparison of lightning data and radar data is as follows: the spatial interval between lightning data and radar echo is 5km, and 35dBz is used as the identification threshold for quality control of lightning data by radar echo intensity. That is to say, when the radar echo intensity within 5km of a certain lightning is greater than 35dBz, the lightning record is identified as a real lightning, and the day is also a real lightning day.

4. The method for manually observing the daily duration of thunderstorms based on lightning detection data according to claim 1, characterized in that, In step S2, the monitoring radius R is less than 20 km. In step S2, the number of days with total lightning strikes and the number of days with ground lightning strikes are compared and fitted. The number of days with total lightning strikes and the number of days with ground lightning strikes show a linear relationship with a correlation coefficient of 0.9658, reaching a significance level of 0.

001. The expression is as follows: T 1 n = 1.2009T 2 n +0.0444 (1) In the formula, T 1 n For a day of total lightning, T 2 n It is the day of lightning and thunder.

5. The method for manually observing the daily duration of thunderstorms based on lightning detection data according to claim 4, characterized in that, The steps for identifying lightning days within spherical spatial regions of different radii are as follows: Spatial coordinates of the meteorological station (LonA, LatA, h) A The spatial coordinates of the lightning are (LonB, LatB, h). B ), with a radius of R; When [(LonB-LonA)] 2 +(LatB-LatA) 2 +(h B -h A ) 2 ] 1 / 2 When h ≤ R, the lightning strike is included in the daily lightning dataset for that radius, i.e., it is included in the daily lightning dataset for all lightning strikes; when h B If the altitude is equal to the location, it indicates that the lightning is a ground lightning strike, and it will be included in the daily data set of ground lightning strikes. B If the altitude is greater than this location, it indicates that the lightning is cloud-to-cloud lightning and will not be included in the daily lightning data set. Otherwise, when [(LonB-LonA)] 2 +(LatB-LatA) 2 +(h B -h A ) 2 ] 1 / 2 When the value is greater than R, the lightning strike is not included in the lightning days data set for that sphere radius.

6. The method for manually observing the daily duration of thunderstorms based on lightning detection data according to claim 5, characterized in that, The process of step S4 is as follows: 1) Taking a single station as the center, combined with lightning monitoring data from previous years, the number of all-flash lightning days under different radius sizes is statistically calculated according to formula (1) to construct the dataset of all-flash lightning days and thunderstorm days for each year, month, and day. The annual mean square error, monthly mean square error and daily F1 score of the two sets of data are calculated respectively. 2) By setting the weights of the annual and monthly mean square errors and the F1 score, the overall matching score S(i) under the current radius size is obtained; 3) Determine the optimal matching size between thunderstorm days and all-flash lightning days for multiple meteorological stations. By calculating the average of the comprehensive scores of multiple meteorological stations under different radius sizes, the radius size corresponding to the highest average score is the optimal matching size between thunderstorm days and all-flash lightning days for multiple meteorological stations. 4) When selecting the optimal matching radius size around the meteorological station, the matching between manually observed thunderstorm days and lightning monitoring full-flash thunderstorm days is optimal. That is, within a certain area, the number of full-flash thunderstorm days within the area with the lightning positioning system monitoring size as the optimal matching radius can be used to replace the number of manually observed thunderstorm days.

7. The method for manually observing the daily duration of thunderstorms based on lightning detection data according to claim 6, characterized in that, For a single meteorological station, based on a specific matching radius size, the annual mean square error is calculated using the following formula: （2-1） In the formula, MSE y T represents the annual mean square error. d (i) represents the number of thunderstorm days in year i; T n (i) represents the number of days with total lightning in year i; n represents the number of years.

8. The method for manually observing the daily duration of thunderstorms based on lightning detection data according to claim 7, characterized in that, The formula for calculating the monthly mean squared error is as follows: （2-2） In the formula, MSE m T represents the monthly mean square error. d (i) represents the number of thunderstorm days in the i-th month; T n (i) represents the number of days with total lightning in the i-th month; n represents the number of months.

9. A method for manually observing the daily duration of thunderstorms based on lightning detection data according to claim 8, characterized in that, The formula for calculating the F1 score is shown below: （2-3） （2-4） （2-4） In the formula, F1 is the F1 score; Precision is the accuracy rate; Recall is the recall rate; 'a' indicates that the day is both a day of all-lightning and thunderstorms and a day of thunderstorms; 'b' indicates that the day is a day of all-lightning and thunderstorms but not a day of thunderstorms; 'c' indicates that the day is a day of thunderstorms but not a day of all-lightning and thunderstorms. To eliminate differences in units and the data themselves, the annual mean square error, monthly mean square error, and F1 score data are normalized. The normalization method is shown in the following formula: （2-6） In the formula, D k r is the normalized value of the current k-th indicator. k The k-th index value; max k and min k These are the maximum and minimum values ​​for this indicator, respectively.

10. A method for manually observing the daily duration of thunderstorms based on lightning detection data according to claim 9, characterized in that, The formula for calculating the comprehensive score S(i) is as follows: （2-7） Where i represents the current radius size; the weights of the annual and monthly mean square errors and the F1 score are all 1 / 3, that is, k1, k2, and k3 are all 1 / 3; in the formula, the MSE / F1 indicators are normalized data.

Images