Experimental and simulative study of characteristics of "site error" inherent to magnetic direction-finder (DF) lightning locating technique

Abstract

For lightning research and other applications, such as forest-fire detection, management of electric power lines and airplane protection etc, lightning locating systems have been developed. To locate a lightning striking position, time-of-arrival (TOA) and gated wideband magnetic direction-finding (DF) technologies are mostly utilized. One of the major factors of DF performance is its "site error". Firstly, the properties of the "site error" as a function of the source azimuth have been examined. It is found that for a given DF, its "site error" as a function of source azimuth appears as a complicated waveform with a dominant sinusoidal cycle of either 360° or 180°. A different DF has a different "site error" pattern but this pattern is timely invariant unless the DF's site environment is changed. Secondly, with the magnetic field waveforms recorded by the two crossed magnetic loops of a broadband direction finder (DF), characteristics of "site error" in determination of lightning stroke direction have been investigated in frequency domain in bands of 100 Hz to 600 kHz. For a given lightning stroke, it is found that the source directions determined by ratios of signals from the two crossed magnetic loops vary at different frequencies. The variation of the source directions versus frequencies for a stroke usually shows a fluctuation with some sharp mono-polar and bi-polar pulses superposed on a relative flat line. Such a fluctuation in the source directions determined is usually attributed to the "site error" in literature. Thirdly, theoretical interpretation and modeling of such characteristics of the "site error" have been attempted based on an electromagnetic dipole model. It shows that the mono-polar shape fluctuations of "site error" in frequency domain are due to re-excitation of lightning signal by 'electric-dipole-wise' objects near the DF, while the bipolar ones are due to re-excitation by 'magnetic-dipole-wise' objects near the DF. The proposed model can also interpret well the azimuthal properties of the 'site error' reported in literature. Furthermore, possible approaches for making "site error" correction have been discussed. Fourthly, although lightning location network (LLN) has been widely used all over the world, its performance is still constrained by the "site error" as long as the direction-finder technique is deployed. Based on lightning data from a regional LLN consisted of 25 DF/TOA combined sensors, a method for "site error" estimation and correction has been proposed and practiced. By comparing the lightning locations reported by at least 4 sensors between DF and TOA techniques, the spatial and seasonal signatures of "site error" for individual sensors are found and discussed. The signatures found are well consistent with those in literature. The "site error" obtained are then used to correct and improve the accuracy of lightning locations reported by only 2 sensors. Results show that the proposed "site error" correction method could significantly improve the location accuracy of the LLN. Finally, an improved approach for locating close lightning strokes based on single station with broadband DF technique has been proposed and practiced. In the approach, a lightning stroke is modeled with an electrical dipole carrying current components in very low frequency and low frequency (VLF/LF) bands, as wavelengths in these frequency bands are much longer than the effective length of the channel of a lightning stroke. For a close lightning stroke, the ratio of the spectra of the vertical electrical field and horizontal magnetic field at ground is theoretically a function of the frequency and the distance to the stroke. The distance of the stroke can then be obtained by fitting the theoretical function with the observed data. The approach was examined by applying it to broadband VLF/LF electrical and magnetic field waveforms observed simultaneously at one station for several strokes in a range of about 10 to 50 km. Furthermore, a prototypal single-station lightning location system (S-LLS), which can be analogized to a modified VLF/LF broadband DF programmed with our proposed lightning stroke distance determining approach, has been built up and tested. Comparisons of individual stroke locations with the local LLN show that the S- LLS has a good location accuracy of about 0.1 - 4 km for close lightning strokes in ranges of 15 to 60 km, but has a poor location accuracy of about 12.4 - 26 km for distant lightning strokes in ranges of 80 to 130 km.