Radio Signals Associate with the Trailing jet (the Continuing Current)75

Figure 4-2 (b), Figure 4-3 (b), and Figure 4-4 (b) show that, in the ULF data, long continuing current waveform follows the FDJ-associated surge current for the GJs. As tabulated in Table 4-1, the peak current moment of the continuing current for both the tree-like and the tree-carrot-like GJs has a large peak current moment (>27 kA-km and

>39 kA-km, respectively) as inferred from the ULF sferics, while that for the carrot-like GJs is relatively small (<27 kA-km). The combined CMCs of the surge current and the continuing current in the FDJ and trailing jet stages are also tabulated in Table 4-1 and both show the same trend. The tree-carrot-like GJs have a larger combined charge moment change (>3166 C-km) comparing to that of the carrot-like GJs (<3118 C-km).

Due to the slow-varying nature of the continuing current flowing at this stage, it is believed that the distinct signals embedded in the spectrograms, Figure 4-2 (c), Figure 4-3 (c), and Figure 4-4 (c), are coincidental signals from non-associate discharge processes and the currents associate with the trailing jets radiated no ELF/VLF band emissions. The spectrograms for the ULF band signals are also examined (not shown) and the major frequencies of their emissions are found to stay below 20 Hz.

76

4.5 Discussion: GJ Morphology and the Underlying Discharge Parameters

The cloud-top emerging height for these negative GJs is ~15 kilometers [Su et al., 2003; Chou et al., 2011]. Since the termination height of these GJs is at ~90 kilometers (see Section Chapter 14.2), the discharge channel length for the FDJs is ~75 kilometers [Cummer et al., 2009]. Thus with the current moment being 60-159 kA-km, the surge current for the tree-like GJs are computed to be 0.8-2.1 kA, which is generally larger than the 0.73 kA reported in Cummer et al. [2009]. However, if the estimated channel length shortens or lengthens by 5 kilometers (~7%), the inferred surge current would vary by

~10%.

As depicted in Figure 4-2 (a), the luminous duration of the FDJ for tree-like GJs is brief and lasts no more than one to four image fields (less than 67 milliseconds; see Table 4-1). It may be that for this form of negative GJs the surge current moment at the

GJ-ionosphere contact is large enough to cause a sufficiently high degree of ionization at the upper trunk of GJs. Therefore, the upper discharge channel is practically shorted out and causes the local ionosphere boundary drops down to ~50 kilometers [Kuo et al., 2009]

in less than one image field (less than 17 milliseconds). Hence the discharge channel for the trailing continuing current is ~35 kilometers in length. With this channel length, the continuing currents are inferred to range between 0.77 kA and 2.26 kA; the large currents are consistent with the bright trailing jets seen in the optical band data; Figure 4-2 (a).

The surge current moments of carrot-like and tree-carrot-like GJs are 16 kA-km to 36 kA-km, which are substantially lower than those for tree-like GJs. With the same assumed channel length (~75 kilometers), the corresponding surge currents at the

77

GJ-ionosphere contact range between 0.16 kA and 0.48 kA for these two forms of GJs.

The small surge current moments could explain why the FDJs for these two categories of GJs are relatively dim and mingling with non-luminous gaps as well as diffusive crowns as they first formed; Figure 4-3 (a) and Figure 4-4 (a). Besides, their further

developments are remarkably different. As for the carrot-like GJs, the trailing continuing current moments are the smallest (0.33-0.36 kA) among the trio, and this may explain why the dim upper trunks of the FDJs soon fade away and the trailing jets are also dim.

The tree-carrot-like GJs have a larger continuing current moments (39-71 kA-km) comparing to those for the carrot-like GJs, thus it may imply that, during the

transformation from carrot-like to tree-like GJs, a higher number of streamers have been generated, causing the upper trunk to brighten up and eventually produce clearly

discernible trailing jets. Also in contrast with the tree-like GJs, the local ionosphere boundary height for the carrot-like and the tree-carrot-like GJs remains un-affected for a notably longer period (5-9 image fields; 67-150 milliseconds). It is possible that, for smaller surge currents, longer time is needed to build up enough ionization at the upper trunk and cause the local ionosphere boundary to drop.

The aforementioned results seem to imply that the magnitude of the surge current is a crucial factor in determining the form of the negative GJs. If a negative GJ starts with a sufficiently large surge current moment, >60 kA-km, for the events analyzed in this work, a tree-like GJ is formed; at the same time the ionization at the upper trunk is sufficiently high to short out the discharge channel and to lower the local ionosphere boundary.

Whereas, if the surge current moment is relatively small, <36 kA-km, carrot-like or tree-carrot-like GJs are formed and the local ionosphere boundary remains un-affected for an extended period. The deciding factor separating the latter two forms of GJs appears

78

related to the magnitude of the trailing continuing current. If the follow-up continuing current moments are relatively large, >39 kA-km as inferred from the ULF sferics, the current will enhance the luminosity of the upper trunk of the FDJs and supplies a bright trailing jet in the case of the tree-carrot-like GJs. If the trailing continuing current moments are less than 27 kA-km, dim carrot-like GJs with very dim trailing jets are formed.

For the carrot-like and tree-carrot-like GJs, their morphology, the lower streamer region, the middle non-luminous gap and the upper diffuse top are very similar to those in a sprite. Pasko et al. [1998] and Pasko and Stenbaek-Nielsen [2002] proposed that the diffuse, transition and streamer regions in sprites result from the interplay of the dissociative attachment timescale, the ambient dielectric relaxation timescale, and the timescale for the development of an individual electron avalanche into a streamer. We believe that these timescales may have also played similar roles in affecting the spatial structures of the negative gigantic jets reported here.

4.6 Conclusion

Dozens of gigantic jets were observed to occur over Typhoon Lionrock (2010) in a six hour interval on the night of 31 August 2010. From the fourteen negative GJs analyzed in this work, a few important results are summarized:

 The complementary ULF and ELF/VLF band data for these GJs contain signals from the GJ initiation, the leading jet, the surge current, and the continuing current, which are respectively associated with the initiating lightning, the leading jet, the fully-developed jet, and the trailing jet stages of the optical band data.

79

 >80% of initiating lightning are found to have recognizable optical emissions. The absence of events is likely due to the obstruction of clouds and light pollution at the inner city observation site, since all the events were detected with signals from this process in the ELF/VLF data.

 Besides the previously recognized tree-like and carrot-like GJs, a third intermediate type called tree-carrot-like form was identified. The three forms of negative GJs are distinct in their morphology and in the luminous duration of the fully-developed jet and trailing jet stages.

 When the peak current moment of the surge currents exceeds 60 kA-km, tree-like GJs are produced with a bright but brief fully-developed jet stage. The peak current moment of the continuing current following the fully-developed jets appears to be the next factor separating the carrot-like and tree-carrot-like GJs. If the peak current moment is less than 25 kA-km, dim carrot-like GJs with very dim trailing jets are formed. The available data indicate that, if the peaks current moments are 39-71 kA-km, tree-carrot-like GJs are formed and the trailing jets are also relatively bright.

 Finally, the peak current moment-versus-charge moment change diagram for the initiating lightning seems to imply that different types of GJs have different occurrence trends. This feature suggests that the forms of negative GJs may have been determined at the initiating stage.

80

Chapter 5

Coordinated TLE Campaign using the ISUAL, the Lulin ULF and the Cingcao ELF/VLF Stations

5.1 Overview

Although coordinated studies of TLEs using optical imagers and low frequency band radio systems had been briefly discussed in Section 0 and some exemplified events had been analyzed in the previous chapters, coordinated studies of TLEs using

multi-platforms are far from trivial. First of all, suitable locations to install

ULF/ELF/VLF stations for long-term observations are limited, at least in Taiwan.

Meanwhile, optical observations of TLEs from ground sites highly depend on the local weather conditions, which can change in hours even in minutes. Hence, it is hard to compare the features of the TLEs recorded under different observational conditions.

However, from vintage points in space, ISUAL observations alleviate the constraints usually associate with ground TLE observations. ISUAL treks nearly along the same orbits everyday and observed the same area with a local time that shows only a small seasonal variation [Chern et al., 2003; Chen et al., 2008]. From June 2009 to June 2012, 7,276 pure elves, 415 halos (174 pure halos and 241 halo-with-elve events), 733 sprites

81

(451 pure sprites, 57 sprite-with-elve events, 146 sprite-with-halo events, and 79

sprite-with-halo-and-elve events), and 1,671 pure blue jets were concurrently recorded by ISUAL and the NCKU ULF/ELF/VLF systems. As reported by Chen et al. [2008], elve is the most dominated type of TLEs in the ISUAL dataset, but blue jet definitively is the second most frequently occurring TLE events after elve since the ISUAL observation in 2004. The large number of ISUAL recorded blue jets can be used to look for the associate sferics and to elucidate whether blue jets are related to narrow bipolar events (NBEs). In order to perform coordinated ULF/ELF/VLF and ISUAL studies of TLEs, one has to set up a few criteria to screen for the concurrent TLE events. Also one has to identify the typical sferic waveforms for each group of TLEs and to deduce the sferic detectability of the radio recording systems. With the concurrent events, the polarity distribution and other characteristics of sferic waveforms can be examined in detail. Finally, the analyses of signal signatures including the rise and decay time and their frequency spectra will be discussed in the following sections.

5.2 Criteria to Look for the Associate ULF and ELF/VLF Sferics of ISUAL TLEs

To ensure the true TLE sferics can be identified, three criteria are employed to screen for the associate sferics.

(i.) Consistence in the arrival time: Using the trigger time and the inferred

geo-location of the TLEs from the ISUAL data, we are able to compute the arrival time of the associate sferics with millisecond accuracy. However, there is a

reported time drift of around 5 to 25 milliseconds in the ISUAL clock [Cummer et

al., 2006b]. In the further discussion in Appendix A, we will show how to resolve

82

this ambiguity completely. After the correction, the standard deviation between the arrival time of TLE and the associate ULF sferic is found to be only 1.36 milliseconds.

(ii.) Select high SNR sferics: Only sferics with peaks higher than three standard deviations above the background noise are selected as the potential candidates.

(iii.) Select sferics with the correct bearing angle: The deviation of the sferics’ bearing angle is restricted to be less than 10-degree in the ULF band and 20-degree in ELF/VLF band data. Our ULF bear angle restriction is tighter than those adopted in Greenberg et al. [2007] and Huang et al. [2011], and is on par with the

minimum bearing angle deviation may be incurred by the Earth-ionosphere cavity [Füllekrug and Sukhorukov, 1999].

5.2.1 Procedures to Identify the Associate ULF and ELF/VLF Sferics of the ISUAL TLEs

Figure 5-1. The block diagram for the procedures to identify the associate sferics of ISUAL TLEs.

83

Between June 2009 and June 2012, the Lulin ULF and Cingcao ELF/VLF radio stations have recorded sferics continuously. In the same period, ISUAL had observed 10,117 TLEs; for the event classifications see Section 5.1. To find the associate sferics of the ISUAL TLEs, the sferics and the ISUAL datasets are screened using the procedure depicted in Figure 5-1. In the first step of the process, after considering the ISUAL event time, the distance between the satellite and the TLE and the drift of the ISUAL onboard clock, a time window is selected to screen for the potential sferics. As a side note, when the ISUAL was triggered by lightning or TLEs, the true time of the causative discharges sometimes differs substantially from the event trigger. For this kind of events, the true event time can be corrected with the assist of the ISUAL SP signals and the correction in the event time may vary between -24 to 173.9 milliseconds (1.55 ms on average).

Furthermore, the event distance between the satellite (observer) and the TLE (target) ranges from around 2 to 5.5 Mm (~3 Mm on average). Since ISUAL is triggered by the influx of optical photons from lightning and TLEs, the large distance causes a time delay of 6.67 to 18.33 milliseconds. Finally, there was a systematic drift in the ISUAL onboard clock [Cummer et al., 2006b]. The time ambiguity due to the ISUAL clock ranges from 5 to 25 milliseconds (14 ms on average) and the correction to the ISUAL clock can be found in Appendix A. All in all, the true event time of the ISUAL TLEs can be found as following with millisecond accuracy.

84

As an illustrative example about what are involved in the first step of the procedure to find the associate sferics of TLEs, we compute the true event time for an elve with an event trigger time of 30 May 2010 18:19:32.721; see Figure 5-2. Relative to the ISUAL trigger time at 721 milliseconds (the red solid line), the photometric peak in the ISUAL SP1 (not show) has a 0.3 ms time delay. The separation between the satellite and the elve is ~3.2 Mm. The time drift in the ISUAL clock is ~21 milliseconds. With the information listed above, the true event time for this elve is derived to be at 689 ms (denoted by the black dotted line in Figure 5-2). Furthermore, the distance between the elve and our Lulin ULF station is ~5 Mm. The propagation speed of the ULF and ELF/VLF band sferics are taken to be 90% and 100% of the speed of light respectively. The estimated arrival time for the Lulin ULF sferics from the elve-inducing CG is at 708 milliseconds (the green dashed line). After all these corrections, the time difference between the ISUAL elve and the associate sferic’s ULF peak turns out to be only 0.51 milliseconds (see Figure 5-2).

Figure 5-2. An ISUAL elve with a trigger time of 30 May 2010 18:19:32.721. The red solid line (721 ms) and the black dotted line (689 ms) respectively represent the trigger time and the true event time of this elve.

The green dashed line denotes the estimated arrival time (708 ms) for the associate sferics of this elve at the Lulin ULF system in Taiwan. The derived event time and the time of the ULF peak differs only by 0.51 milliseconds.

85

After ‘the fetch by the time window step’ creates the first dataset, the sferic is further screened by SNR of the pulse peak. Only the pulse with a peak value greater than three standard deviations of the background noise is regarded as a valid pulse. The standout pulses are further inspected by the author. Distant sferics (>10 Mm) that has an unusually large and sharp waveform is rejected, since they could be accident sferics from nearby sources. After the second step, only 4559, 638, 2657, and 1754 sferics are selected from the Lulin ULF, the Duke electric VLF, and the Quasar electric and magnetic ELF/VLF datasets, respectively.

Finally, the bearing angle of a potential sferics is derived using the Lissajous figure [Huang et al., 2011; Huang et al., 2012]. The angular deviation between the bearing angle of the sferic and the direction to the ISUAL TLE is computed. As the histograms in Figure 5-3 indicate, only the sferics with an angular deviation of less than 10 (20) degrees in the ULF (ELF/VLF) dataset is taken to be the associate sferics. After the first two steps of the ‘find the associate sferics’ procedures, the distribution of the ULF sferics appears to be congregated together; see Figure 5-3 (a). Around 93% and 40% of ULF sferics have a bearing angle deviation less than 10 degrees (the red dashed line) and one degree, respectively; see Figure 5-3 (a). The distribution of the Quasar magnetic ELF/VLF sferics is more scattered. Around 88%, 60% and 21% of events have bearing angle deviation below 20 degrees (the red dashed line), 10 degrees and 1 degree, respectively; see Figure 5-3 (b). Füllekrug and Sukhorukov [1999] reported that the Earth-ionosphere cavity can easily cause a bearing angle deviation of ten degrees [Greenberg et al., 2007, and references therein]. However, the relatively large scatter in the bearing angle of the ELF/VLF sferics may have been incurred by the relatively intense background noise at the suburban Cingcao site in the Tainan City. In short, after these three criteria were

86

applied to screen for the TLE-associated sferics in the NCKU ULF/ELF/VLF datasets, a found sferic usually carries a high confidence level. The resulting TLE-associated sferics dataset will be used in the following sections.

Figure 5-3. Histograms of the bearing angle deviations relative to the ISUAL TLEs in the Lulin ULF sferics and the Quasar ELF/VLF sferics. The red dashed lines denote the screen criterion in each dataset.

As a further check, the bearing angle deviations versus the discharging source distances and SNRs are explored; see Figure 5-4. No obvious relations are uncovered to reproduce the relation reported in Füllekrug and Sukhorukov [1999]. Finally, it is noted that the geo-location uncertainty in the ISUAL events will contribute significantly toward the bearing angle deviation, especially for the nearby events. As described in Chen et al.

[2008], the geo-location of a TLE is derived from the ISUAL Imager frame with the assistance of the satellite attitudinal information. Depending on the occurring location of the TLEs relative to the Earth limb, the geo-location uncertainty is 50 to 220 kilometers per pixel. When the ISUAL TLE-associated discharge is near to the radio recording station (e.g. 1 Mm from the electric discharge to the station), 100 kilometers in the geo-location uncertainty will result in a ~6-degree bearing angle deviation. This accounts for more than half of the bearing angle deviation we use to screen for the potential ULF sferics and it will become more and more serious if the sferics source locates even closer.

Therefore, the uncertainty in the geo-location of TLEs should be reduced first and the

87

analyses in the following sub-sections would prove to be worthwhile.

Figure 5-4. (a) and (b) The distribution of the bearing angle deviations versus the distances of the

discharging sources to the Lulin ULF and the Cingcao ELF/VLF stations, respectively. Only sferics with a source distance less than 11 Mm are shown. (c) and (d) The distribution of the bearing angle deviations versus the sferics’ SNR for the Lulin ULF and the Cingcao ELF/VLF datasets, respectively. Only sferics with SNR between 3 and 100 standard deviations are shown.

88

5.3 Waveforms of the TLE-associated Sferics

Representative waveforms of the sferics and the related discussions for a few basic groups of TLEs including pure elve, sprite, halo, blue jet, and their inducing discharges are presented here.

5.3.1 Waveforms of the Elve-associated Sferics

As mentioned in Section 1.2.5, elve can be initiated either by positive or negative lightning [Barrington-Leigh and Inan, 1999]. Here, representative sferics from negative (a) and positive (b) lightning that induce pure elve are shown in Figure 5-5. Because the ISUAL TLEs are dominated by elve [Chen et al., 2008], the number of sferics associate with elve is quite large in our ULF/ELF/VLF datasets. Most of the elve-associated sferics

As mentioned in Section 1.2.5, elve can be initiated either by positive or negative lightning [Barrington-Leigh and Inan, 1999]. Here, representative sferics from negative (a) and positive (b) lightning that induce pure elve are shown in Figure 5-5. Because the ISUAL TLEs are dominated by elve [Chen et al., 2008], the number of sferics associate with elve is quite large in our ULF/ELF/VLF datasets. Most of the elve-associated sferics