| Lightning occurs in strong convective clouds or convective cloud systems,reflecting the severe weather of strong convection.The lightning location network can provide parameters such as the time,location,and even the intensity of lightning.The continuous and uninterrupted lightning location network provides an important reference for monitoring and early warning of strong convective weather and is of great significance for the investigation of lightning disasters and the study of the evolution of lightning.At present,the ground-based lightning location network in China cannot cover the domestic and surrounding areas effectively,especially the marine areas,due to the limitation of the detection range of a site.Therefore,it is of great significance to develop a long-range ground-based lightning location network to provide accurate lightning activity data and forecast lightning activity trends for sea vessels,air flights and the general public in these areas in real-time.This paper has carried out research work from the following aspects:(1)Investigation of VLF ground-based lightning location network;(2)Development of VLF ground-based lightning detection equipment;(3)VLF lightning waveform feature analysis and identification;(4)VLF lightning location algorithm and location error analysis(5)Evaluation of VLF lightning detection efficiency.The following results have been achieved:The advantages and disadvantages of VLF,LF and VHF lightning location network at home and abroad were compared and analyzed respectively,and the VLF frequency band with long signal propagation and small amplitude attenuation was selected as the working frequency band of long-range lightning location network.A VLF lightning electric field detector was developed,which integrated high-speed acquisition,storage and processing functions of VLF electric field signals.Trigger sampling mechanism was adopted to the device,and the data acquisition and processing module was separated based on the FPGA+ARM dual processing architecture to realize real-time signal processing.The device also integrated complete network communication and remote data transmission function based on the Linux operating system,which can transmit data at a gigabit bandwidth rate in a wired network or adaptively connect to 2G/3G/4G mobile wireless networks.At the same time,the use of solar power has greatly reduced the power supply noise of the equipment and improved the signal-to-noise ratio(SNR).In this paper,the waveform characteristics of the four VLF lightning signals have been analyzed and summarized from three aspects: spectrum,time-frequency and characteristic parameters.These signals are cloud-to-ground(CG)flash,cloud-to-cloud(CC)flash,Narrow bipolar events(NBEs)and ground-ionosphere dispersion waveform(sferic).The CG flash has a strong low-frequency component and has a peak near the frequency of 4 kHz.There is a strong high-frequency component at the peak of the CG flash.The rise time of CG flash is 8～12μs,the fall time is 8～50μs,the pulse width is18～60μs,and the SNR is 0.75～1.75.IC flash is a multi-pulse signal with stronger highfrequency components in rising and falling edges.Its time-frequency diagram is characterized by discreteness,large amplitude fluctuations,and the presence of multiple strong frequency components.The rise time of the IC flash is 10～22μs,the fall time is12～24μs,the pulse width is 24～46μs,and the SNR is 0.4～1.25.NBEs are isolated pulse with strong high-frequency components,and its time-frequency diagram shows that the frequency changes rapidly around the peak point.The rise time of NBEs is 8～16μs,the fall time is 12～30μs,the pulse width is 22～46μs,and the SNR is 3.5～7.0.The frequency component of sferic is mainly concentrated within 20 kHz,and there is a clear peak near the frequency of 10 kHz.Time-frequency analysis shows that the duration of sferic lowfrequency energy is positively correlated with the signal propagation distance.The longer the propagation distance,the longer the low-frequency duration.The rise time of sferic pulse is 16～22μs,the fall time is 14～24μs,the pulse width is 32～44μs,and the SNR is 0.57～1.33.A waveform library composed of four VLF lightning types was constructed and based on deep learning methods,a one-dimensional deep convolutional neural network model was designed for VLF lightning waveform classification.The model consisted of 10 convolutional layers,4 maximum pooling layers,and 1 global mean pooling layer,which were sequentially stacked and contain 649700 parameters.The classification performance of the model was evaluated based on the 5-fold cross-validation method.After testing,the model has a good recognition accuracy rate for the four types of sampled waveforms,and the overall classification accuracy rate is 98.57%.And the recognition rates of cloud ground flashes,cloud flashes and NBEs were: 98.33%,97.83%,and 98.40%.Established China’s first VLF lightning location network: the Asia-Pacific VLF lightning location network(APLLN).A complete VLF lightning location method and an improved time difference positioning algorithm were proposed to improve the location accuracy of VLF lightning.APLLN used the peak time of the signal Hilbert envelope as the lightning arrival time.APLLN was based on the spherical triangle time difference location algorithm and optimized by the improved Levenberg-Marquardt least-squares algorithm.Monte Carlo stochastic simulation method was used to evaluate the location accuracy of APLLN and the influence of the formation of the sites on the location accuracy.The results show that the detection accuracy within the APLLN is better than 2km and the linearly distributed site formation has the greatest influence on the location accuracy.A preliminary assessment of the location accuracy of APLLN was performed by using lightning strike accidents.The average location accuracy of the APLLN was 5 to 10 km.Based on the lightning detection data of APLLN in the past year,three thunderstorm processes were analyzed.The results show that APLLN can effectively monitor the thunderstorm activity process and predict the thunderstorm trend based on the density of lightning occurrences.Severe thunderstorms have significant agglomeration and dissipative characteristics.Thunderstorms in the Qinghai-Tibet Plateau mainly occur in summer,and the number of lightning occurrences in one thunderstorm is significantly less than in other areas.The statistical results show that the average number of lightning storms during the thunderstorm in the plateau is 11.38 times per minute.For the same strong thunderstorm process,the data of APLLN and the Chinese Academy of Sciences VLF/LF three-dimensional high-precision lightning location network were compared in this paper,and the relative detection efficiency and detection accuracy of APLLN were evaluated.The detection efficiency of APLLN for CG flashes and IC flashes is 63.55% and 55.34%,respectively.The average North-South location error is 5.43 km,the standard deviation is 7.84 km,the average East-West location error is 4.35 km,and the standard deviation is 4.82 km.Overall,the average location error is7.75 km and the standard deviation is 8.55 km.The detection efficiency of APLLN increases gradually with the increase of the lightning peak current and reaches the maximum detection efficiency around 50 kA,which is about 97.6%.The detection efficiency of APLLN for lightning less than 10 kA is 18.69%.Based on the historical data of APLLN over the past year,a map of lightning density distribution in the Asia-Pacific region from 2019 to 2020 was produced.Several regions with high lightning activity in the Asia-Pacific region are parts of Guangxi,Guizhou,and Guangdong provinces of China,and Bangladesh.These regions have an average annual lightning density of more than 80 times per square kilometer. |
PDF Full Text Request