SEARCH

SEARCH BY CITATION

Keywords:

  • winter sprite;
  • winter thunderstorm;
  • winter lightning;
  • sprite morphology;
  • continuing current

Abstract

  1. Top of page
  2. Abstract
  3. 1. Introduction
  4. 2. Observations
  5. 3. Observational Results
  6. 4. Discussion
  7. 5. Summary
  8. Acknowledgments
  9. References
  10. Supporting Information

[1] VHF, LF, and ELF lightning events, thunderstorms, and surface electric fields related to sprites were observed simultaneously during the winter of 2004/2005 in Hokuriku, Japan. The analysis of these observations enables us to investigate the relationship among sprites, lightning characteristics, and thunderstorm structure just before sprite genesis. Typical winter sprite parent thunderstorms had a mesoscale cloud area with small, embedded convective cells. Positive charges responsible for sprites tend to reside in the upper part of the thunderstorms; only a few positive charges were assumed to be located in the lower part. The total amount of positive charges removed by a sprite-producing flash from the upper and lower parts of the thunderstorms were estimated to be approximately 100 C and as large as 300∼400 C, respectively. Active thunderstorms with lightning accompanied by transient currents tended to generate simple sprites; more complex sprites were excited by lightning with continuing currents, which were generated by a few active thunderstorms and thunderstorms with precipitating stratiform clouds. VHF sources related to sprites can be found in the range of 5 to 72 km. The range of displacement between a sprite element and the corresponding positive cloud-to-ground lightning discharge or the first VHF source was 6∼30 km, and the bottom of the sprite bodies was located between 66 and 74 km. On the basis of these results, we deduced that the complexity of sprite morphology might be attributed to the differences in lightning characteristics.

1. Introduction

  1. Top of page
  2. Abstract
  3. 1. Introduction
  4. 2. Observations
  5. 3. Observational Results
  6. 4. Discussion
  7. 5. Summary
  8. Acknowledgments
  9. References
  10. Supporting Information

[2] Transient luminous events (TLEs) between the stratosphere and lower ionosphere in the altitude range of 10∼100 km occur above thunderstorms with active lightning discharges. The most commonly observed phenomenon of TLEs is a red sprite. This phenomenon is induced mainly by a positive cloud-to-ground lightning discharge (+CG) with a large charge moment change (Qds) [e.g., Lyons, 1996; Williams, 1998] and its duration is between 5 and 300 ms [Rodger, 1999]. It is well known that the occurrence rate of +CGs in winter is higher than that in summer and that the +CGs occasionally remove a large amount of positive charge from the thunderstorm [cf. Rakov and Uman, 2003]. Therefore, excitation of winter sprites above the Sea of Japan had been predicted since many +CGs are detected there [Brook et al., 1982; Michimoto, 1993]. The first winter sprites were observed there in 1998 [Fukunishi et al., 1999], a decade after the first imaging of summer continental sprites [Franz et al., 1990]. Since then, Japanese winter sprite campaigns have been conducted extensively using various instruments (high-sensitivity CCD camera, C-band radar, two-dimensional (2-D) cloud-to-ground lightning mapping system, and meteorological satellite) [Hobara et al., 2001, 2006; Takahashi et al., 2003; Hayakawa et al., 2004; Suzuki et al., 2006a, 2006b].

[3] The morphological features of winter sprites in Japan are found to be very different from those in summer, although most of those sprites exhibit very complex morphological shapes in summer (e.g., carrot-, jelly fish–, or plume-like), they are very simple (column-like) in structure in winter [Matsudo et al., 2007; Myokei et al., 2009]. Recently, winter sprites in the eastern Mediterranean Sea were also detected by Ganot et al. [2007] and Greenberg et al. [2007], who pointed out that the morphological distribution is somewhat different from the Japanese winter sprites reported by Hobara et al. [2001] and Hayakawa et al. [2004].

[4] Sprites in the U.S. High Plains tend to cluster over mesoscale convective systems (MCSs) with a radar echo area (>10 dBZ) of ∼20,000 km2 and with a cloud area (by infrared (IR) imagery of the GOES-8 satellite) of about 120,000 km2 [Lyons et al., 2003]. It was, however, reported that even much smaller thunderstorms (2500∼7500 km2) are sometimes able to generate sprites [Neubert et al., 2001]. The cloud area of sprite-producing winter thunderstorms was first investigated by the Geostationary Meteorological Satellite-5 (GMS-5) around Japan [Adachi et al., 2005]. It is reported that winter thunderstorm systems have average cloud top temperatures ranging from –25 to –40°C and average horizontal areas of 8500∼40,500 km2 at the –20°C isotherm. Their results suggested that winter sprite parent thunderstorms are usually lower and smaller than MCSs in the U.S. High Plains. It is additionally found that the parent thunderstorms, including active radar echo cells, are also very small (cell size is about several tens of kilometers) [Hayakawa et al., 2004, 2005, 2007; Suzuki et al., 2006a, 2006b]. On the basis of the analysis of four different winter storms with sprites above the eastern Mediterranean, Ganot et al. [2007] have concluded that the areas of active convective part were 6000∼30,000 km2, and their average vertical extents were from 5 to 7 km, reaching a cloud top height of 8∼9 km (T ∼ –40°C).

[5] Comprehensive observations of whole lightning activity and electrical properties were made by several different lightning detection systems (a very high frequency (VHF) and low-frequency (LF) lightning detection system (Surveillance et Alerte Foudre par Interférométrie Radioélectrique (SAFIR)), an LF ground flash detection system (Impact) and ELF/VLF receivers) [Suzuki et al., 2006b; van der Velde et al., 2006; Matsudo et al., 2007]. van der Velde et al. [2006] suggested direct evidence for the relationship between sprite morphology and intracloud (IC) lightning discharge associated with ground discharge: IC lightning discharge activity may play an important role in the generation of sprites and their morphology. Furthermore, IC lightning and CGs in association with sprites in MCSs were observed by the VHF lightning location system LMA (lightning mapping array) and the NLDN during the Severe Thunderstorm Electrification and Precipitation Study (STEPS) 2000 [Lyons et al., 2003]. The estimated altitude of the sprite triggering positive charge varied between 2.0 and 5.0 km with an average height of 4.1 km, and the altitude range corresponded to the radar bright band in association with the melting layer (near the 0°C isotherm). Lyons et al. [2003] concluded that the sprite-producing lightning discharges may be generated by the melting layer/bright band charge production mechanism in MCS stratiform precipitation regions. On the basis of the radar observations, the winter sprite parent thunderstorms have a vertically limited and flat-topped stratiform at altitudes of less than 6 km; this is much lower than summer sprite parent thunderstorms [Hayakawa et al., 2004; Williams and Yair, 2006]. However, several researchers reported that no bright band echoes, or only weak and temporal echoes, could be detected in the stratiform region of winter thunderstorms with sprites [Hayakawa et al., 2004; Suzuki et al., 2006b]. Williams and Yair [2006], such observational differences can probably be attributed to the fact that the cloud base is very often colder than 0°C. Furthermore, on the basis of the review of previous studies on the electrical structure inside winter sprite-producing thunderstorms, they proposed two possible mechanisms for the generation of +CG lightning: (1) formation of large areas of positive charges near the 0°C isotherm and (2) tilting of the upper, positively charged cloud top by strong wind shear (tilted dipole). They suggested an additional study by means of a combination of various instruments in order to study which of the two mechanisms is responsible for the generation of the strong positive ground flash. A further significant difference between winter and summer sprites is that the former have a much larger time delay after their parent lightning (peak occurrence time of 60∼90 ms in winter and 0∼30 ms in summer) [Matsudo et al., 2009; Huang et al., 1999]. The mechanism of this difference is extremely poorly understood, and this might be one of the key problems in understanding the sprite generation mechanism [Asano et al., 2009b].

[6] For summer sprite and parent +CG, Lyons [1996] reported that many of the sprites occurred within a distance of 50 km, and at an average of ∼40 km, from the parent +CG and that the bottom of the sprite elements was an average of ∼50 km in height [São Sabbas et al., 2003; Lyons, 1996]. On the other hand, Takahashi et al. [2003] triangulated sprites using two cameras and determined that the bottom of winter sprites is distributed in the altitude range of 68∼80 km. The lateral shift of winter sprites from their production of lightning is not well studied, but Matsudo et al. [2009] have found a similar lateral shift on the order of several tens of kilometers. As mentioned above, very significant properties of winter sprites as compared to those of summer ones have emerged from previous studies: (1) much simpler shapes, (2) much smaller scale of their parent thunderstorms, (3) much larger time delay, etc. However, the relationship among winter sprites, sprite-producing lightning, and thunderstorm structure is also poorly understood. In order to clarify these issues, we have been conducting winter sprite campaigns since 1999 [Hobara et al., 2001; Hayakawa et al., 2004, 2007]. In this paper, we present comprehensive studies of winter sprites, their parent thunderstorm structures, and electrifications by means of results of simultaneous observations performed using highly sensitive cameras, an ELF detector, a field mill network, a C-band radar, and the SAFIR VHF/LF lightning location system. Then, we discuss several winter sprites and their parent thunderstorms from both meteorological and electrical viewpoints.

2. Observations

  1. Top of page
  2. Abstract
  3. 1. Introduction
  4. 2. Observations
  5. 3. Observational Results
  6. 4. Discussion
  7. 5. Summary
  8. Acknowledgments
  9. References
  10. Supporting Information

2.1. Observational Instruments

[7] The 2004/2005 winter sprite campaign was conducted during the period of December 2004 through May 2005. Various instruments were simultaneously in operation during the campaign. Figure 1 shows the deployment location of our observational instruments. First, optical instruments have been installed at Shimizu (geographic coordinates: 33.99°N, 138.51°E) and Chofu (35.66°N, 139.54°E) as indicated in Figure 1 (bottom right inset). These camera systems were aimed at the Noto peninsula in the Hokuriku region on the west coast of Japan, and distances between the Noto peninsula and the optical observatories were about several hundred kilometers. Sprites were captured by a low-light-level CCD camera at Chofu and a CCD camera with an image intensifier at Shimizu. The captured video images with a GPS time stamp were recorded by an S-VHS video recorder and stored on the hard disk of a PC [Matsudo et al., 2007]. The frame integration time was 33.3 ms, and the accuracy of the time of the event was ±16.7 ms. An accurate azimuth of CCD cameras was determined by using the star field in the video frames, and sprite elements in the video frame were localized by triangulation with two CCD cameras. Second, the operating SAFIR lightning detection system in the Hokuriku region consists of three VHF/LF sensors with a short base line of several tens of kilometers as indicated in the upper left inset as stars [Suzuki et al., 2006a, 2006b]. Each sensor detects the direction of the VHF lightning radiation source associated with either IC or CG by interferometry. Two-dimensional horizontal VHF sources are provided by a triangulation method using three sensors. Our SAFIR can localize a maximum of 150 VHF sources in 1 s with a time resolution of 100 μs. The location accuracy was less than 1 km for VHF lightning sources associated with the analyzed sprite events. The polarity of the CG and peak current are estimated automatically by the waveforms detected by its LF receivers. The system can detect lightning discharges in a range of about a few hundred kilometers from each sensor. The conventional weather radar is operated in the C-band (5.3 GHz) with a beam width of 1.3°, and its range is 400 km. The radar has functions that digitize and store the Constant Altitude Plan Position Indicator (CAPPI) data (horizontal interval 1 km and vertical interval 0.5 km) every 5 min. The electric field mills are deployed at 27 observational sites in the Hokuriku region as shown in Figure 1 (top left insert); they are of the all-weather type. The measurement error of field mills is estimated to be less than 13%, and their dynamic range is ±400 V/cm. The data from the field mills are digitized into 9 bits every 90 s and are transmitted to the central site. From the polarity convention with the field mills, the positive values are indicative of a dominant positive charge overhead. The 2-D polarity distribution map of field mill network data acquired just before and 90 s before the sprite genesis is used to determine the vertical location of positive charges in the sprite parent thunderstorms. Finally, the electromagnetic field in the ELF band associated with sprites was measured by induction coils and a capacitive antenna at Moshiri (44.22° N, 142.16° E) in Hokkaido (see Figure 1, middle). The vertical component of the electric field (Ez) and the two horizontal components of the magnetic field (Bx, By) have been observed since 1997 [Hobara et al., 2000; Ando et al., 2005]. The sampling rate of our present analog-digital converter is 4 kHz. The vertical charge moment change (Qds) was estimated by the impulsive method and direct method from the observed ELF waveform [Huang et al., 1999]. All these instruments were synchronized with a GPS clock. More details about our instruments are given by Hobara et al. [2000, 2001], Ando et al. [2005], Suzuki et al. [2006b], and Matsudo et al. [2007]. The IR image of the GOES-9 satellite was used to investigate the distribution of clouds and cloud top temperatures, and the upper-air sounding data at Wajima—37.39° N, 136.90° E, located in the northern part of the Noto peninsula—were also analyzed.

image

Figure 1. Location of observatories. (top left) The deployment locations of instruments in Hokuriku area of Japan. The open squares indicate electric field mills and the stars indicate the locations of operating SAFIR VHF/LF antennas. The number above each field mill is the site number. (middle) The plus is an ELF observatory at Moshiri. (bottom right) The diamonds show the locations of optical observatories (Chofu and Shimizu), and the field of view from each optical observatory is also shown.

Download figure to PowerPoint

2.2. Positive Charge Estimation

[8] The location of the positive charge reservoir responsible for sprite-producing +CGs in winter thunderstorms is still poorly understood. Williams and Yair [2006] proposed two possible thunderstorm electrical structures for the +CG lightning from the previous studies on winter lightning and thunderstorms. One is a tilted dipole and another is an inverted dipole. Hayakawa et al. [2004] estimated the amount of positive charge (Q) removed from the sprite parent thunderstorm by means of the Qds value estimated using the ELF data. They only assumed an upper positive charge reservoir of the tilted dipole model, and the positive charge height (ds) was fixed to 5 km. However, the assumption for ds was not satisfactory, because they did not have any data to identify whether or not each analyzed thunderstorm was of tilted dipole. To clarify this issue, we investigated the extent and direction of development of the VHF sources and also examined the 2-D distribution map of surface electric field (E-field) polarity on the basis of charge structure models of thunderstorms proposed by Williams and Yair [2006]. If we can obtain a ds value that is more suitable than their assumption, our Q value estimation would be better than the estimation by Hayakawa et al. [2004].

[9] The VHF radiation magnitude associated with negative leader propagation has been reported to be stronger than that with positive leader propagation [Mazur et al., 1998; Shao et al., 1999; Rison et al., 1999; Thomas et al., 2001]. Therefore, most of the VHF sources observed by the SAFIR would also correspond to the negative leader propagating in the positively charged region. In particular, we found that all the charges removed from the sprite parent thunderstorm in this study would be positive, as has been reported by Matsudo et al. [2007]. Unfortunately, we cannot estimate the vertical location of positive charges directly and precisely, because we have only 2-D VHF lightning source locations and surface E-field values with low time and spatial resolution. However, if we assume some probable thunderstorm charge structures, we can deduce the relative vertical location of positive charges in the thunderstorm (i.e., we can distinguish whether the positive charge is located in the upper or lower part of the thunderstorm). We consider five cases of relative positive charge location in the parent thunderstorms: positive charge is located (case a) at the upper convective core (lower negative), (case b) in the anvil-like stratiform cloud (positive charge only located in the thunderstorm anvil), (case c) in the lower precipitating stratiform cloud (upper negative), (case d) at the lower convective core (upper negative), and (case e) in the upper precipitating stratiform cloud (lower negative). The surface E-field polarity patterns under five supposed thunderstorms are compared with that under the observed thunderstorm. The optimal charge structure from the comparison is assumed to be the thunderstorm charge structure and is used to estimate the Q value in association with sprites.

[10] Ishii et al. [2003] investigated the vertical location of the positive charge layer by means of VHF lightning source mapping and then indicated that the twin peaks of VHF lightning source altitude associated with the positive charge layer are located at about –30°C and –10°C in temperature height. Williams and Yair [2006] described that the positive charge layer is frequently observed above the radar bright band where the in situ temperature is <0°C, and they indicated that the altitudes of the upper and lower positive charge layers in the winter thunderstorms are 4∼6 km and 1∼2 km, respectively. Here, we have no observed altitude data for charge structure in the thunderstorm, so the altitudes of the positive charge in the thunderstorm are deduced from the relationship between charge polarity and temperature obtained in previous studies [Takahashi, 1978, 1984, 1999]. On the basis of their results, we assume the positive charge height ds and then calculate the Q value using the Qds value. If the positive charge is deduced to be in the upper part of the thunderstorm (i.e., case a, b, or e is selected), the ds value is assumed to be at the 20 dBZ echo-top height, which would be the upper limit of a graupel. On the other hand, if the positive charge is located in the lower part of the thunderstorm (i.e., case c or d), the –5°C height from the upper air sounding data at Wajima is assumed to be the ds value. The –5°C height would be an intermediate height between the –10°C height, where the charge polarity of particles reverses from negative to positive, and the 0°C height near the bottom of the radar bright band at approximately ground level on a winter night at Hokuriku. This method for estimating the altitude of positive charges may be somewhat speculative in the sense that there is an insufficient number of E-field sites beneath the thunderstorm, but it would be possible to estimate the amount of positive charge removed that is on the order of 100 C if our observed winter thunderstorms were to be consistent with the supposed thunderstorm charge structures.

3. Observational Results

  1. Top of page
  2. Abstract
  3. 1. Introduction
  4. 2. Observations
  5. 3. Observational Results
  6. 4. Discussion
  7. 5. Summary
  8. Acknowledgments
  9. References
  10. Supporting Information

3.1. Sprite Elements and Lightning

[11] According to our summary [Matsudo et al., 2007], 23 sprite events were captured during the 2004/2005 winter sprite campaign, and 13 of 23 sprites were detected by our instruments. We analyzed those 13 sprites and their parent thunderstorms, and succeeded in triangulation of the sprite location by using two CCD cameras for only 4 of the 13 events. Table 1 lists the observational characteristics of each sprite event. Seven events of a single sprite or group of simpler column-like sprites (sprites 2, 3, 4, 14, 16, 17, and 18) and six sprite events, including complex sprites; that is, a single unusual sprite (such as a carrot) or combination of column and unusual sprites (sprites 5, 12, 13, 15, 22, and 23), were observed. Locations of sprite elements and their bottom altitude in some optical events were estimated for the four (sprites 3, 17, 18 and 23) cases (see event number followed by letter “t” in Table 1).Three of four sprites (sprites 3, 17 and 18) had simple column shapes, and one (sprite 23) was a combination of two simple column and two noncolumn shapes (intermediate shape between carrot and column). The altitude range of the bottom of the sprite body (the brightest part of the sprite element) in four particular cases, as estimated by optical observation with triangulation, ranged from 66 to 74 km, with an average altitude of 69.4 km. The first attempt was then made for the relative shift between winter sprites and their sprite-producing lightning. The horizontal distance between the sprite-producing CG point or the first VHF source and each sprite element was found to be in the range of 6 to 30 km with an average of 19 km.

Table 1. Characteristics of Sprites Observed Above Hokuriku Area During 2004/2005 Wintera
Sprite EventDateTime, UTCNumberSprite TypeAltitude, kmAverage Altitude, kmDisplacement, kmAverage Displacement, km
  • a

    Number of sprite elements, their occurrence date and time, information on sprite element, morphology, and relative distance between the sprite elements and lightning location are shown. A “t” following the sprite event number indicates events for which triangulation was successfully achieved using two CCD cameras. The letters Co, V, and Ca indicate column, V-shaped, and carrot sprites, respectively.

225 Dec 20041832:50.065Co    
3(t) 1835:18.565Co68.3, 69, 68.768.78, 14, 2716
4 1837:19.522Co    
531 Dec 20041443:261V    
1210 Jan 20051421:35.893Co+V    
13 1633:19.791V    
1411 Jan 20051037:59.361Co    
15 1330:06.698Co+V    
16 1400:32.692Co    
17(t) 1405:31.853Co70.4, 72.2, 73.872.118, 27, 1520
18(t) 1513:22.952Co66.1, 70.668.46, 2817
22 1651:54.4619Co+Ca    
23(t) 1658:00.034Co+V68.9, 68.4, 68.9, 66.968.330, 26, 269, 23
Average  4  69.4 19

3.2. Meteorological Conditions

[12] Except for event 5, the sprite events in the Hokuriku region were generated during cold air mass advection. For example, the prevailing wind from the upper air sounding was westerly above the Wajima upper-air station on 11 January 2005 [Suzuki et al., 2006b]. We could not determine any shear in the wind direction between the surface and subsidence inversion, but we observed moderate wind speed shear of about 10∼20 kt/km near the ground surface and cloud top. The upper level wind direction corresponds almost completely to that of radar echo movement (east). The altitude of –10°C, where the charge separation is very active, was about 1.5 km, and the altitude of 0°C associated with the bright band was very close to the ground (360 m above MSL). Convective available potential energy (CAPE) was often less than 50 J/kg or could not be calculated, because of a low or absent level of free convection (LFC). Sea surface temperature (SST) was 14∼16°C near the Hokuriku coast when sprites occurred. This meteorological condition is often found in the winter monsoon season (midwinter) around the Hokuriku region. Michimoto [1993] reported that single (one) flash lightning events (lightning activity is very weak but a large amount of positive charge is occasionally lowered to the ground by CGs) occur under such a condition. The IR images of the day for the four events stated above were observed by the GOES-9 satellite, and a typical winter cloud pattern (cloud streets) with cold air mass advection appeared over the Sea of Japan in the IR images. Thunderstorm echo cells are often observed at the mature and dissipating stages in cold advection clouds.

3.3. Sprite Morphology, Thunderstorm, and Lightning Characteristics

[13] Table 2 shows that the characteristics of thunderstorms associated with 13 sprites were investigated by the radar reflectivity at 20 dBZ and the IR images. When the sprites occurred, the 20 dBZ echo area at a height of 1.6 km associated with the –10°C isotherm was in the range of 100 to 8500 km2. Most of the sprite parent thunderstorms studied (8 out of 10 storms) have echo areas of less than 1500 km2 above 20 dBZ, with an average of, at most, 1300 km2. In particular, two thunderstorms accompanied by the most frequent sprite events were found to cause 2∼3 sprite events during a short interval of about 3∼5 min. One of them generated three sprite events (storm 1) and another generated two sprite events (storm 7). Moreover, the height of the top of the radar echo at 20 dBZ of the parent thunderstorms was found to be from 3 to 5 km, and the average echo top was 4.1 km. The cloud top temperature of the parent thunderstorms as determined by the GOES-9 satellite reached approximately –25°∼–35°C, with an average temperature of –31°C. The cloud top height above the Hokuriku region is much lower than that above other regions. The cloud area related to the sprite events at the –20°C isotherm was more than 2000 km2, and some exceeded 10,000 km2. A winter thunderstorm system with sprites seems to be a large thunderstorm system similar to a summer MCS.

Table 2. Characteristics of Thunderstorms With Sprite Events Observed in Hokuriku Areaa
StormSprite Event20 dBZ Echo Area, km220 dBZ Echo Top, kmMaximum Cloud Top, °CCloud Area, km2
  • a

    Here radar echo areas (>20 dBZ) and echo tops were analyzed by radar reflectivity at 1.6 km altitude ground level. Maximum cloud top and cloud area were observed GOES-9 satellite. A “t” following the sprite event number indicates events for which triangulation was successfully achieved using two CCD cameras.

12250, 4503.6, 5.1−287625
 3(t)4503.6−287625
 44503.6−287625
2582504.6−398300
3124004.6−3411,125
4138004.6−3411,125
51411253.6−272900
6151003.6−262275
7164504.1−293025
 17(t)3004.1−293025
818(t)15005.1−3113,200
92212503.1−286050
1023(t)26004.1−2913,475
Average 13134.1−307490

[14] Table 3 shows characteristics of lightning discharges with sprite events observed by SAFIR and location of VHF source clusters (VHF clusters) mapped in the thunderstorms. CGs and VHF clusters associated with sprites were analyzed in terms of their polarities and extent. Sprite events were not always associated with CGs or a positive CG, and some of them were accompanied by several VHF clusters. The major axis for the VHF clusters with sprites had ranged from 5 to 72 km in length with an average horizontal length of 25 km. These VHF clusters were mapped near the convective cells and on the stratiform cloud area but they did not develop to the west (upwind) side of the convective cells. Furthermore, relative locations between the VHF clusters and the thunderstorm (obtained by the radar echo and IR image) and the direction of development of the clusters in the thunderstorm were compared for 13 clusters associated with sprites (see Tables 1 and 3). Then, we analyzed the relationship between the complexity of the sprite morphology and lightning locations inside the sprite parent thunderstorms. Seven VHF clusters with simple sprites (sprites 2, 3, 4, 14, 16, 17, and 18) were mapped between the well-developed cell and the nonprecipitating cloud part with an anvil-like stratiform structure (without a detectable echo under the cloud part). Four VHF clusters associated with complex sprites were located in the precipitating stratiform cloud (sprites 5 and 23) or near the dissipating convective cells, and they evolved into a stratiform structure (sprites 15 and 22). The other two clusters were generated above the anvil-like stratiform cloud with embedded mature cells; however, they were associated with a complex morphology (sprites 12 and 13).

Table 3. Characteristics of Lightning Discharges With Sprite Events Observed by SAFIR and Location of VHF Source Clusters Mapped in the Thunderstormsa
StormSprite EventCGs Peak Current, kAVHF Source Clusters
NumberLength, kmLightning Inside Thunderstorms
  • a

    A “t” following the sprite event number indicates events for which triangulation was successfully achieved using two CCD cameras.

12217.2, −47.7272, 8mature cell - anvil-like stratiform cloud
 3(t)−37.0126mature cell - anvil-like stratiform cloud
 4−167.9126mature cell - anvil-like stratiform cloud
25−48.8, −11.7, −15.2, 10.0121mature cell and precipitating stratiform cloud
312−11.8236, 17anvil-like stratiform cloud - mature cell
4134.9, −8.3118mature cell - anvil-like stratiform cloud
5145.7125mature cell - anvil-like stratiform cloud
615None140anvil-like stratiform cloud - dissipating cell
71636.3, 14.315in the vicinity of mature cell
 17(t)43.4, −22.0110in the vicinity of mature cell
818(t)−21.6123mature cell - anvil-like stratiform cloud
922None145dissipating cell - anvil-like stratiform cloud
1023(t)None219, 13precipitating stratiform cloud - mature cell
Average +CG: 47.4; -CG: −39.2 25 

[15] We also analyzed the relationship between ELF waveforms and the complexity of sprites. Figure 2 shows typical examples of an ELF waveform with simple sprites (Figure 2a) and with more complex sprites (Figure 2b). As a result, we found that the morphological complexity of sprites may be characterized by the ELF waveform for the sprite-producing lightning. That is, simple ELF waveforms with a relatively large single peak and small perturbations were associated with simpler sprites (sprites 2, 3, 4, 16, 17, and 18) whereas, complex ELF waveforms with relatively large multiple peaks were associated with more complex sprites (sprites 5, 12, 13, 15, 22, and 23). However, sprite 14, categorized as a simple sprite by its morphology, was induced by lightning with a complex ELF waveform. This sprite was the brightest one of our sprites. Its duration was relatively long (3 frames), and its morphology seems to have slightly complex morphology and some differences in comparison with ordinary column sprites. Therefore, sprite 14 can be categorized as a complex sprite. Nevertheless, the complexity of ELF waveforms was in good agreement with the morphological complexity of sprites.

image

Figure 2. Typical examples of ELF waveform associated with (a) simple sprites and (b) complex sprites. The dashed lines in the waveforms represent the first time a sprite event is detected with GPS time-synchronized CCD cameras.

Download figure to PowerPoint

[16] Here, we consider the relationship among complexity of sprite morphology, thunderstorm types, and lightning characteristics with reference to ELF waveforms. The active thunderstorms with anvil-like stratiform cloud generated 9 sprite events (sprites 2, 3, 4, 12, 13, 14, 16, 17, and 18). Six out of these nine events (sprites 2, 3, 4, 16, 17, and 18) were simple sprites with simple ELF waveforms, but the other three (sprites 12, 13, and 14) included complex sprites and had complex ELF waveforms. Moreover, thunderstorms with a precipitating stratiform cloud or dissipating thunderstorms generated 4 sprite events (sprites 5, 15, 22, and 23) associated with complex ELF waveforms. These results may imply that the active thunderstorms, which generate lightning with transient current, produced simple sprites, but some active thunderstorms, which generate lightning with continuing current, induced complex sprites. The thunderstorms with precipitating stratiform cloud, which generate lightning with continuing current, might produce complex sprites. It is said that the continuing current flows in a dissipating thunderstorm or trailing stratiform precipitation region and it may relate to emission of summer sprites [e.g., Lyons, 1996; Mazur et al., 1998]. The complexity in ELF waveforms would be caused by the continuing current, and then, it may play an important role in the complexity of a sprite.

3.4. Positive Charge Reservoirs

[17] Table 4 shows that summary of positive charge removed from the sprite parent thunderstorms. The vertical location of positive charges in the thunderstorms were deduced by the 2-D map of E-field polarity pattern of the field mill network data together with other observation data (radar, SAFIR, GOES-9 satellite) and thunderstorm charge structure [e.g., Kitagawa and Michimoto, 1994; Williams and Yair, 2006]. However, about half of the positive charge locations could not be deduced, because no field mill sites and/or no E-field data were observed under the parent thunderstorms, and/or no ELF data were observed. This attempt was made for 7 of 13 sprite-causative positive charges. As a result, the majority (5 of 7 cases: sprites 3, 4, 14, 16, and 17) of positive charges were deduced in the upper part of the thunderstorms, whereas the remaining (2 of 7 cases: sprites 5 and 23) were considered in the lower part of the thunderstorms. The VHF clusters associated with the upper positive charge in the thunderstorm were mapped in the vicinity of an active small convective cell (2 of 5 cases: sprites 16 and 17) or in the anvil-like stratiform cloud outside the convective cells (3 of 5 cases: sprites 3, 4, and 14). Here, the important point is that a small winter thunderstorm (e.g., storm 7) could generate several sprite events and would have an ordinary dipole electrical structure (upper positive and lower negative). These special events would be generated by the removal of upper positive charge in the vicinity of the convective cell. Therefore, some winter thunderstorms, which are slightly larger than an isolated summer thunderstorm, would be able to generate a large amount of positive charges associated with sprites. On the other hand, we also considered that some sprite-causative positive charges possibly reside in the lower part of the precipitating stratiform cloud. We have already calculated the Qds values in our previous paper [Matsudo et al., 2007]. In this study, we attempted to estimate the amount of positive charges (Q) removed from the parent thunderstorms on the basis of the charge structures assumed as shown in section 2.2 and the Qds values. We assumed the ds values from the vertical locations of positive charge and calculated the Q values removed from the parent thunderstorms. Consequently, we found that the range of Q values removed was roughly from 100∼400 C with an average of about 200 C. The estimated upper positive charge values were on the order of 100 C. The lower positive charge of 300∼400 C would be neutralized by sprite-producing lightning discharges. The positive charge values removed from the lower part of the parent thunderstorms seemed to be larger than those from the upper part. However, Q values estimated in the both the upper and the lower parts would be sufficient to generate sprites.

Table 4. Summary of Positive Charge Removed From the Sprite Parent Thunderstormsa
StormEventPositive Charge LocationQds, C kmds, kmQ, C
  • a

    Relative positive charge locations in the thunderstorms were deduced by a combination analysis among field mill network, radar, and meteorological satellite. Charge moment changes (Qds) were estimated by the ELF data observed at Moshiri. The amounts of positive charges removed from the parent thunderstorms were estimated roughly by Qds estimated and ds postulated by thunderstorm models and radar echo-top height. A “t” following the sprite event number indicates events for which triangulation was successfully achieved using two CCD cameras.

  • b

    Numbers indicate ds and Q values if we assume that positive charges were located at the upper part of the sprite-parent thunderstorms.

12    
 3(t)anvil-like stratiform cloud (upper)4523.6126
 4anvil-like stratiform cloud (upper)4223.6117
25precipitating stratiform cloud (lower)5801.4 (4.6b)414 (126b)
312 469  
413 713  
514anvil-like stratiform cloud (upper)2913.681
615 613  
716convective cell (upper)5164.1126
 17(t)convective cell (upper)4374.1107
818(t) 197  
922 540  
1023(t)precipitating stratiform cloud (lower)3891.1 (4.1b)354 (95b)
  average4683.1189
  average of upper positive charges 3.8 (4.0b)111 (111b)
  average of lower positive charges 1.3384

3.5. Case Studies

[18] Figure 3 shows all lightning locations (+CGs and the first VHF source without +CGs) associated with sprite events observed on 11 January 2005. We found that sprite-producing lightning discharges (the +CGs and the first VHF sources) occurred over the land, and six thunderstorm produced sprites there on the same day. In this section, we present the analysis results for two sprite parent thunderstorms observed during our campaign with respect to the development of thunderstorms, their lightning production, the amount of positive charge removed, and sprite locations.

image

Figure 3. Locations of +CGs and first VHF sources without +CGs associated with sprite events on 11 January 2005. The number with “S” corresponds to the sprite event number given in Table 1. The pluses and diamonds indicate the +CG and the first VHF source, respectively.

Download figure to PowerPoint

3.5.1. Small Active Thunderstorm With Two Simple Sprite Events (Sprites 16 and 17)

[19] Two sprite events (sprites 16 and 17) were produced by the same thunderstorm on 11 January 2005. The sprites were observed during a short interval of about 5 min around 1400 UTC; one of the sprites appeared at 1400:32.69 UTC and another at 1405:31.85 UTC. Many thunderstorm cells were found to have developed and decayed over the Sea of Japan on the same day. The first radar echo in the parent thunderstorm developed over the Sea of Japan about 10 km west of the Hokuriku coast at 1320∼1325 UTC. The thunderstorm moved eastward with developing radar echo area and reached the Hokuriku coast at 1335∼1340 UTC. After it crossed over the Hokuriku coast, the radar reflectivity area was enlarged, and a strong echo exceeding 30 dBZ appeared between 1345 and 1420 UTC. The maximum echo top reached a height of 4.1 km and was maintained during this period. The two sprites were generated when an active convective cell was developing in the parent thunderstorm. The thunderstorm had generated no lightning before the sprite-producing lightning occurred, and the only lightning generated by the thunderstorm was two discharges with the sprite events. The SAFIR LF sensors detected two +CGs for sprite 16 and both +CG and –CG for sprite 17, but we found that both sprites were produced by positive charge removal with large amplitude of the ELF waveform. We show an analysis for the thunderstorm and associated lightning only with sprite 17 because we found similar characteristics for sprites 16 and 17. The parent thunderstorm would correspond to charge structure case a (see section 2.2), and for these sprites, we succeeded in triangulation of the location of sprite elements. We present the details for sprite 17 below.

[20] Figure 4a shows the relative locations of sprite elements (letters “I”) for sprite 17, VHF sources (dots), CG flashes (diamond), and E-field values superimposed on the image of radar reflectivity at an altitude of 1.6 km. The E-field value was acquired less than 90 s before the sprite genesis, for which there were three column sprite elements (17–1∼17–3). The distance between any sprite element and the +CG point is more than 15 km (17–1: 18 km, 17–2: 27 km, and 17–3: 15 km), and these sprite elements were distributed on the west side of the sprite-producing lightning.

image

Figure 4. (a) Thunderstorm and electrical activities related to sprite 17 at 1405:31.85 UTC on 11 January 2005. Location of sprite elements, the radar echo image (CAPPI image at 1.6 km MSL at 1408:16 UTC on 11 January 2005), the values of static electric field (within 90 s before the sprite event), and lightning localized by the SAFIR (1405:31–1405:32 UTC). The entire area is enclosed in the central rectangle in Figure 4b. The Letter “I” denotes the location of the sprite element for sprite 17, estimated by the triangulation method using sprite images from Shimizu and Chofu observatories. The number under letter “I” is the sprite element number corresponding to the data in Table 1. The location of lightning data is shown (individual VHF sources: small dots; sprite-producing CGs: diamonds). The arrow represents lightning associated with +CG and the dashed line with an arrow represents that with –CG. The small square below the site number is the location of the field mill site, and the plus or minus in the small square is the polarity of dominant charge above the field mill. The number below the square with the sign is the value of the electric field in the unit of V/cm. (b) IR image captured by GOES-9 satellite at 1500 UTC on 11 January 2005. The white contour lines indicate the IR cloud temperature in degrees centigrade. A white square in the top right insert indicates the Hokuriku region (the area of Figure 4b), and the central black rectangle corresponds to the entire area of Figure 4a. The letter “I” denotes the locations of sprite element. The number next to this letter is the sprite element number corresponding the data in to Table 1. The dots indicate the distribution of VHF lightning sources association with the sprite 17. The arrow represents the parent thunderstorms associated with the thunderstorm echo in Figure 4a.

Download figure to PowerPoint

[21] The sprite parent thunderstorm contained a well-developed convective cell (core) with a strong echo greater than 32 dBZ, where the extent of the convective core was, at most, ∼5 × 5 km. The echo area with radar reflectivity above 20 dBZ in the parent thunderstorm is about 20 × 15 km (300 km2), which is distinctly smaller than that in other regions [e.g., Stanley, 2000; Lyons et al., 2003; van der Velde et al., 2006]. It seems that the extent is the same as, or a little larger than, the summer-isolated thunderstorm.

[22] The sprite-producing CGs and their related VHF cluster are located near the south edge of the 20 dBZ contour. The VHF cluster is only developed in the vicinity of the convective cell. The horizontal extent of the cluster is smaller than 5 × 10 km (50 km2), and it is very compact as compared to the other sprite-producing lightning in this study.

[23] The VHF source started inside the convective cell above the negative E-field region and developed to the relatively weaker radar reflectivity area that was located on the downstream (east) side of the convective cell in the parent thunderstorms. When the sprites were generated, the +CG detected by the SAFIR first occurred in the vicinity of the convective core under the negative surface E-field region. Following the +CG, the –CG were mapped at the same point. The surface E-field tended to be dominant with a strong negative value under and in the vicinity of the parent convective echo cell, but the strong positive values tended to be observed around the strong negative dominant region. Therefore, this sprite parent thunderstorm was anticipated having an ordinary dipole in the electrical structure (upper positive and lower negative). Several sprite events were probably produced by the removal of the upper positive charge in the active convective cell. If we assume that the sprite-causative positive charge was located at an altitude of 4.1 km (see Table 4 and section 2.2) in the upper part of the parent thunderstorms, the values of positive charges removed from the parent thunderstorm were approximately 100 C as estimated by the Qds value. This fact shows that some small winter thunderstorms can generate a large amount of positive charges associated with sprites even if the echo size is very small (at most several tens of kilometers).

[24] Figure 4b shows the cloud top temperature estimated by the IR images from the GOES-9 satellite at 1500 UTC. The letter “I” indicates locations of sprite elements, and the dots represent VHF sources estimated by the SAFIR. The black square in Figure 4b indicates the same area as that shown in Figure 4a. The IR image at 1500 UTC was captured at about 1430∼1435 UTC in the Hokuriku region, about 25∼30 min after the sprite genesis. When the IR image at 1500 UTC was captured, the parent thunderstorm (arrow in Figure 4b) had already passed through the location of the VHF cluster, and so it was located on the east side of the VHF cluster. The cloud top temperature associated with the parent thunderstorm was –29°C (at 4.2 km), and the extent of the cloud area was approximately 3000 km2. We were able to find some small isolated convective echo cells embedded in the cloud area.

[25] Figure 5 shows a comparison of time series data of different parameters for the sprite-producing lightning with those for sprite 17. It shows the vertical component of electric field (Ez) of an ELF waveform observed at Moshiri, the peak current and polarity of CG flashes obtained by the SAFIR, and the number of VHF sources counted every 1 ms. An example of the bottom of a sprite body is indicated by the white arrow under “17–1” in the sprite photograph in Figure 5. Time delay between the peak of ELF waveform and sprite 17 was within one video field (∼17 ms) and the duration of sprites was within 2 video frames (∼66 ms). The ELF waveform associated with sprite 17 has a large transient peak and a distinct slow tail-like signature. The risetime of the large transient is ∼1 ms, and the duration of the transient and the slow tail-like signature is within 200 ms. At the same time, numerous VHF sources (more than 90 points) corresponding to very active IC lightning, were detected by the SAFIR VHF sensor. The time series of VHF sources were in good agreement with the risetime in the ELF waveform. Similar characteristics in terms of ELF waveforms, VHF activities, delay of sprites with respect to the peak of ELF waveform, and the duration of sprites were also obtained for another sprite (sprite 16).

image

Figure 5. (top right inserts) Sprite images taken from Shimizu, ELF waveform observed at Moshiri, and lightning activities detected by the SAFIR VHF/LF sensors before and after the sprite 17. The white arrow under the sprite element 17–1 is an example of the bottom of a sprite body. (top) The displayed ELF waveform shows the vertical component of the electric field (Ez). (bottom) The first panel is the expanded ELF waveform. The arrows in the ELF waveform indicate the observation time of sprites. The dashed lines with a horizontal arrow indicate the duration of one video frame. The second panel shows the time sequence of CGs polarity discriminated by the SAFIR LF sensor (plus, upward; minus, downward). The vertical unit indicates a peak current (kA), and (third panel) the histogram shows lightning VHF sources detected by the SAFIR VHF sensors. Time delay from lightning pulses detected by the SAFIR to ELF peak is about 3 ms.

Download figure to PowerPoint

3.5.2. Thunderstorm With a Precipitating Stratiform Cloud Producing a Complex Sprite Event (Sprite 23)

[26] Sprite 23 was generated at 1658:00.03 UTC on 11 January 2005. This event included two dim columns and two unusual sprites that had intermediate shape (V-shaped) between a column and a carrot (four sprite elements; 23–1∼23–4). Intermediate sprites that are V-shaped occasionally appeared at the same time as column sprites [Matsudo et al., 2007; Myokei et al., 2009]. We succeeded in triangulation of sprite locations. The horizontal distance between any sprite element and the first VHF source was more than 25 km west of the first VHF source (23–1: 30 km, 23–2: 26 km, and 23–2: 26 km), except for one sprite element (23–4: 9 km). The average distance was 19 km.

[27] The parent thunderstorm with the intermediate sprite (sprite 23) was maintained for a relatively longer time (>2 h) than the common winter thunderstorm (<1 h). The thunderstorm with sprite 23 entered our radar range at about 1500 UTC. Several convective cells were merged into a thunderstorm band and the length of its band reached more than 50 km. With developing radar reflectivity, the thunderstorm band was moving toward the Hokuriku coast. The convective cells in the band were in the mature stage over the Sea of Japan but the thunderstorm band began to decay near the Hokuriku coast. Most of the convective cells in the band weakened almost simultaneously. The thunderstorm was composed of a group of dissipating convective cells. This was followed by the band area of dissipating convection turning into a stratiform precipitation area (indicated by a relatively larger weak echo area). However, a few small convective cells remained and were embedded in the larger weak echo area of the stratiform precipitation. This thunderstorm band generated only three lightning discharges over inland Hokuriku during its lifetime. Two lightning discharges occurred near the Hokuriku coast just before the sprite genesis, followed by one lightning discharge more than 10 km inland from the coast.

[28] Figures 6a and 6b are the same as Figures 4a and 4b, but for sprite 23. The maximum echo-top height at 20 dBZ was still high and reached an altitude of 4.1 km, but the convective cell was rapidly getting smaller after it crossed the Hokuriku coast, becoming no more than several square kilometers (at most 2 × 2 km2) in the horizontal scale (Figure 6a). A precipitating stratiform (weak echo) area in the thunderstorm was extended horizontally, and the depth of the weak echo was about 3 km. The echo area exceeding 20dBZ at 1.6 km in the thunderstorm was about 65 × 40 km (2600 km2). However, we could not find any distinct bright band just above the 0°C isotherm in the lower part of the stratiform precipitation region.

image

Figure 6. (a) Same as Figure 4a but for sprite 23 at 1658:00.03 UTC on 11 January 2005. Letters “I” and “V” represent the locations of column and V-shaped sprites, respectively, estimated by the triangulation method using sprite images from Shimizu and Chofu observatories. The number next to these letters is the sprite element number corresponding to Table 1. (b) Same as Figure 4b but for event 23 at 1658:00.03 UTC on 11 January 2005. The IR images were captured by GOES-9 satellite at 1800 UTC on 11 January 2005.

Download figure to PowerPoint

[29] The IR image at 1800 UTC was captured at about 1735∼1740 UTC in the Hokuriku region, about 35∼40 min after sprite genesis (Figure 6b). The maximum top of the stratiform cloud area that covered the precipitating stratiform area reached approximately –30°C (4.3 km), and the extent of the cloud area was larger than 10,000 km2 (Table 2). The weak echo area with stratiform precipitation was contained under this large stratiform cloud area.

[30] Figure 7 is the same as Figure 5, but for sprite 23. The risetime to the largest transient on the ELF waveform was ∼10 ms, and the duration of the ELF transient and the ELF slow tail-like signature was about 500 ms. Moreover, the overall ELF waveform of sprite 23 was very complex and noisy, and it included small negative bipolar pulses (although some positive bipolar pulses did appear). It also showed that a larger positive charge was removed from the parent thunderstorm. In addition, the time delay of the ELF peak to the sprite was longer than that of the other sprites in this study. These observational results would be indicative of the continuous charge transfer with continuing current. As shown in Figure 7, two groups of VHF clusters were detected during a short duration (<100 ms). The first VHF cluster, i.e., VHF sources with more than 80 points, associated with a very small ELF peak appeared about 70 ms before sprite 23, but the second VHF cluster included numerous VHF sources of approximately 100 points during a slow rise and small amplitude of the ELF transient. It is considered that the first VHF cluster was able to contribute little to the sprite emission because the ELF peak was too small to produce a sprite. Therefore, the second VHF cluster associated with the larger ELF peak was probably more important for the generation of these sprites as judged from the time sequences among VHF sources, peak of the ELF waveform, and occurrence time of sprites shown in Figure 7. However, another VHF cluster may also have some effect for this sprite. Sprite 23 accompanied no causative CGs with the SAFIR LF sensor, but many VHF sources were detected by the SAFIR. The second VHF cluster associated with sprite 23 traveled almost within the precipitating stratiform cloud (Figure 6). The VHF clusters were located near the coast, and they developed within the stratiform echo area. The horizontal extent of VHF cluster was approximately 20 km and its width was several kilometers. The first VHF source started from the precipitating stratiform cloud and then the last VHF source nearly reached the small convective echo cell embedded in the precipitating stratiform cloud.

image

Figure 7. Same as Figure 5 but for sprite 23 at 1658:00.03 UTC. An enlarged sprite picture (top right insert) was taken from Tokyo. Picture courtesy of Koji ITO (http://usjma.jp/∼kaminari/index.html).

Download figure to PowerPoint

[31] Positive surface E-fields (EFM 4 and EFM 5) tended to be dominant under the precipitating stratiform cloud, while the negative E-field (EFM 9) tended to concentrate in the vicinity of a small convective echo in the same precipitating stratiform cloud. EFM 5, which was below VHF cluster, indicated a much smaller positive value (+25 V/cm) than EFM 4 (+242 V/cm) under the same precipitating stratiform cloud. In contrast, EFM 9 indicated a strong negative (−271 V/cm) value under the small convective cell embedded in the same precipitating stratiform cloud. Here we considered why the positive values under EFM 4 and 5 were very different, though both of them were located under the same precipitating stratiform cloud. One of the possibilities is that the smaller positive value above EFM 5 may be indicative of the presence of negative charge above the positive charge in the precipitating stratiform cloud. So, we considered that the positive charge layer might be located inside the lower part of precipitating stratiform cloud. Then, if we assume that sprite-causative positive charges reside in the lower part of the precipitating stratiform cloud, the altitude of positive charge in the winter thunderstorm is assumed to be 1.1 km (–5°C) (Table 4). The value of estimated positive charge was approximately 300 C. However, we can also see large positive E-field values (EFM 7 and 8) under the outer side of the stratiform (weak) echo area. This could indicate that there would be a positive charge layer in the upper part of the thunderstorm (anvil-like stratiform cloud). Therefore, if we assume that the altitude of the upper positive charge layer is 4.1 km (echo top relation to sprite 23 in Tables 2 and 4), the value of positive charge is calculated to be about 100 C. Both values of positive charge calculated in the upper and lower parts of the thunderstorms would be sufficient for sprite generation.

[32] We can also see a train of many small pulses associated with discharges on the ELF waveform (Figure 7), and the pulse train on the waveform continues for a long time. The characteristics of the ELF waveform suggest that the continuing current occurs such as is the case of a stratiform precipitation region in summer MCS with sprites. The continuing current is often observed under a stratiform region or a dissipating cell [e.g., Lyons, 1996; Mazur et al., 1998]. In addition, by using a VHF lightning mapping system, Ishii et al. [2003] showed that several positive charges in the winter thunderstorm could be found at altitudes ranging from 1 to 2 km. Therefore, we deduced that the removal of positive charges associated with sprite 23 occurred in the lower part of the precipitating stratiform cloud. In the case of this thunderstorm, the assumed vertical location of positive charge may be a little uncertain because there are few E-field sites under the precipitating stratiform cloud and the VHF cluster. Nevertheless, as we have already mentioned above, the relationship among the location of VHF cluster, radar echo, field mill network data, and ELF waveform seems to indicate that sprite 23 was probably produced by the removal of positive charges residing within the lower part of the precipitating stratiform cloud.

4. Discussion

  1. Top of page
  2. Abstract
  3. 1. Introduction
  4. 2. Observations
  5. 3. Observational Results
  6. 4. Discussion
  7. 5. Summary
  8. Acknowledgments
  9. References
  10. Supporting Information

4.1. Winter Thunderstorm and Positive Charge Reservoir

[33] We noted the possibility that sprites can be also generated by winter thunderstorms with various structures as compared to the summer thunderstorms (MCSs) that produce sprites. That is, we found three possible locations of positive charge reservoir in the winter sprite parent thunderstorms: (1) upper convective core (e.g., sprite 17), (2) anvil-like stratiform cloud (e.g., sprite 14), and (3) lower precipitating stratiform cloud (e.g., sprite 23). Then, it seems that there would be more varieties of the relative vertical location of positive charge in the sprite parent thunderstorms than we have understood.

[34] Suzuki et al. [2006b] reported that winter sprite-causative positive charge would reside in the stratiform cloud on the basis of a case study of the parent thunderstorm (sprite 14). They pointed out that no detectable radar echo and remarkable bright band could be found inside the stratiform cloud associated with their sprite-producing discharge. Similar cloud structures to those found in winter thunderstorms were identified in the summer thunderstorm anvil by Marshall et al. [1989]. They indicated the possibility that a large amount of positive charge (thousands of coulombs in big thunderstorms) resides in the convective core and anvil. Stolzenburg and Thomas [2009] also described the stratiform region as “anvil type.” The cloud base for this type is above the melting level (no precipitation under the cloud base). Consequently, owing to the lack of charged precipitation particles, that is, lack of melting of a significant number of aggregates, there is no charged region near the 0°C isotherm. The “anvil-type” charge structure resembles that of typical nonprecipitating anvil clouds [Marshall et al., 1989]. In some cases (e.g., storm 5), there would be no or only weak precipitation under the large cloud area of our thunderstorms with sprites so that the base of the cloud area may sometimes be higher than the melting level. The structure of the large cloud area causative of winter sprites would be similar to that of an anvil of summer thunderstorms. Thereby, as Marshall et al. [1989] concluded, the large amount of positive charge generating our winter sprites probably also blew out, and was transported from the convective core of the cloud area with the anvil-like structure. On the other hand, a summer sprite-producing +CG often occurs under the stratiform precipitation region. It is thought that summer sprite-causative positive charges were located in the lower part of the stratiform precipitation region. A bright band in the stratiform precipitation region is a good indicator for the sprite producing summer thunderstorm. For example, Williams [1998] and Lyons et al. [2003] suggested that summer sprite-causative positive charges reside in the bright band associated with the 0°C isotherm in the lower part of the stratiform precipitation region. The altitude of the 0°C isotherm associated with the bright band process is important for the accumulation of a large amount of positive charges in the stratiform precipitation region of summer MCSs [Lyons et al., 2003]. The 0°C isotherm associated with a bright band is about 4∼5 km in height in the summer continental area, whereas the 0°C isotherm is very close (less than 500 m in height) to the ground in the winter in the Hokuriku region. We found that several winter thunderstorms have a stratiform cloud and the stratiform precipitation reached the ground (e.g., storm 10), but we could not detect a bright band. In these cases, it is possible that a thin radar bright band exists in the stratiform precipitation region near the sea or ground. Our observational results for the stratiform precipitation may suggest the following possibilities: the bright band cannot be detected by any operational radar because (1) the band is located at an extremely low altitude near the ground surface and/or (2) the depth of the bright band layer is too small to be detected. For resolving this issue, we refer to several other winter thunderstorms with no sprites observed by using some range height indicator (RHI) data from another x-band radar (data not shown). We could find a stratiform precipitation in a thunderstorm above the Sea of Japan that had a weak echo and was several kilometers in depth. We could also find a distinct bright band that reached the sea surface in the lower part of the stratiform precipitation, and its depth was several hundred meters (the bright band is located at a very low height, and its depth is extremely small.). Therefore, we assumed that our operational radar could not detect the bright band layer in the thunderstorm. Moreover, the stratiform precipitation with a distinct bright band that reached the ground for the inland winter thunderstorms is observed infrequently. One of the reasons why we could often find the distinct bright band in the lower part of the maritime thunderstorm but hardly observe the bright band above the inland region would be that the SST (∼14°C in this study) is much warmer than the land surface temperature (0∼2°C). That is, it would be difficult to maintain the bright band in the thunderstorm above the inland region because the 0°C isotherm would be too close to the ground for the charge accumulation and/or the relative humidity over the land surface would be much lower than that over the sea surface. The winter thunderstorm structure over the sea may be different from that over the inland region. Then, we assumed that sprite 23 was produced by a positive charge in the lower part of a thunderstorm with a precipitating stratiform structure because sprite 23 was generated by lightning that occurred near the coast.

[35] Figure 8 shows a schematic diagram of the sprite-producing winter thunderstorm structure under the stratiform cloud (large cloud area), charge structures, and +CG lightning channels deduced from the analysis of our combined instruments. Here, we have no data for charge structure in the upper and midheight area of the stratiform precipitation in the thunderstorm for Figure 8, region c. As the meteorological structure of the stratiform precipitation cloud in the winter thunderstorm is similar to that in the summer MCS [e.g., Lyons et al., 2003], we supposed that both charge structures are nearly the same. The charge polarity in the upper and middle of the stratiform precipitation was postulated by the relationship between charge polarity and temperature by Takahashi [1978, 1984, 1999]. Our sprite-causative positive charges tended to be located in the upper part of the thunderstorms (Figure 8, regions a and b), whereas the positive charges deduced in the lower part of the stratiform precipitation (Figure 8, region c) seem to be a minority in our sprite parent winter thunderstorms. It is said that the positive charge associated with sprites would reside in the lower part (near the bright band) of the stratiform precipitation region in the summer MCS [e.g., Williams, 1998; Lyons et al., 2003]. However, our observational results may suggest the possibility that the major vertical location of the positive charge removed from the winter thunderstorms with sprites may be different from that in the case of summer ones. That is, the primary contributing charging mechanism may be different. For the charge structures (Figure 8, regions a and b), any other charging process (not a bright band charging process) inside a thunderstorm would also play an important role in the generation of a large amount of positive charges. For example, the noninductive charging mechanism within the small convective cell in the winter thunderstorm may work effectively, and in some case actively, in this season.

image

Figure 8. Schematic diagram of morphology of cloud, charge structure, and cloud-to-ground lightning discharge for winter sprite-producing thunderstorms: region a, upper convective core; region b, anvil-like stratiform cloud; and region c, lower precipitating stratiform cloud. The dashed line represents an outline of the minimum radar reflectivity, and the shaded areas indicate larger radar reflectivity. The thin solid line represents the cloud boundary and the thick solid (dendritic) lines depict lightning channels. The plus and minus signs indicate the locations of positive and negative charges, respectively, conjectured by combined analysis with our instruments. The plus and minus signs with the asterisk show imaginary charge polarity deduced from the general relationship between charge polarity and temperature by Takahashi [1978, 1984, 1999].

Download figure to PowerPoint

[36] Now, why does a winter thunderstorm with low cloud height show inactive lightning activity even though a large amount of positive charge that is sufficient to generate a sprite would be accumulated inside the cloud? We need to consider this point. We can find that many of the convective cells in the Hokuriku region are short-lived, as pointed out by Kitagawa and Michimoto [1994]. Actually, our radar observations show that numerous short-lived, small, shallow convective cells develop one after another over the Sea of Japan, but most of them dissipate without lightning discharge. Nevertheless, anvil clouds of these many decaying convective cells are merged and form a larger anvil-like stratiform cloud area. From our observational results, the large cloud area is maintained for a long time without dissipation after these convective cells decay. Most of the short-lived convective cells without discharge may be able to generate only a small amount of charge, but a large amount of positive charge may be accumulated in the long-lived large cloud area because these numerous small convective cells may leave positive charge there. One of the reasons for the inactivity of winter lightning may be that convective cells, which can generate lightning discharge themselves and act as a trigger for the removal of large amount of positive charge, are difficult to develop under the winter monsoon regime when the atmosphere is very stable except for its lower part near the sea surface. That is, only a few convective cells with lightning discharge may be able to develop in winter as compared to in summer.

[37] For summer thunderstorms, the positive charge reservoir would be located in the lower part of the stratiform precipitation region [e.g., Williams, 1998]. In addition, by analysis of 3-D VHF lightning mapping data, Lyons et al. [2003] showed that the average altitude of sprite-causative positive charges is 4.1 km above ground level (AGL), which correspond to the lower part of the stratiform precipitation region. They also reported that the total charges lowered by sprite-producing +CGs from the lower part of the stratiform precipitation region were on the order of 200 C. On the other hand, Williams and Yair [2006] proposed two possible charging structures (tilted dipole and inverted dipole) for the sprite-producing winter thunderstorm. In this paper, we show that their proposal would be consistent with our observational results. Most of our sprite-causative positive charge layers are found to be located in the upper part of the winter thunderstorms (assumed ds = ∼4 km), and some of them would be located within the limited area in the vicinity of upper convective cells. Furthermore, several positive charge layers are possibly located in the lower part of the thunderstorms (assumed ds = ∼1 km). Ishii et al. [2003] indicated that positive charges were sometimes removed from low altitudes of about 1∼2 km in winter thunderstorms. This would imply that positive charges could be removed from the low altitude of our thunderstorm. Therefore, our assumption for lower positive charge height would be acceptable. In our results, with the exception of only one case, most of the estimated Q values exceeded 100 C, and these values can be enough to generate sprites. Consequently, we confirm that our early assumption for sprite-causative positive charge height (ds = 5 km), which corresponds to the upper part of the winter thunderstorm of Hayakawa et al. [2004], is mostly suitable. The estimated total positive charge removed from the upper and lower parts of the parent thunderstorms were on the order of 100 C and from 300 to 400 C, respectively. If our assumption for positive charge height is suitable, a larger amount of positive charge may be required to be removed from the lower part of the parent thunderstorms. Hayakawa et al. [2004] described that the threshold of Qds was 200∼300 C km when winter sprite-producing +CG occurred. We showed that sufficient positive charges generating sprites would be removed from the parent thunderstorms even if the Qds values of winter lightning are much smaller than those of summer lightning. One of the reasons for this is probably that the positive charge height in the winter thunderstorm is much lower than that in the summer thunderstorm.

4.2. Sprite Location and Lightning Discharge

[38] Takahashi et al. [2003] estimated that the bottom of winter sprites is distributed in the altitude range of 68∼80 km. Hayakawa et al. [2004] showed that the height distribution of sprites is found to be in a range of 50 km to 90 km, and the average height of the lowest end of column sprites, except the tendril (i.e., the lowest end of the sprite body), is ∼72 km. These altitudes of the bottom of sprite bodies are almost the same as our observational results. Moreover, São Sabbas et al. [2003] analyzed 34 summer sprite events by means of triangulation of simultaneous optical measurements. They reported that approximately 2 out of 3 of sprites occurred within 50 km of their parent +CG. The maximum horizontal distance between any sprites and the parent +CG was ∼82 km with a mean of 40 km. Lyons [1996] also analyzed seven sprite events and showed that five of them had a lateral displacement between the sprite and parent +CG of within 50 km. The seven events had an average of 42 km, and the range of lateral displacement varied from 13 to 111 km. Wescott et al. [1998] analyzed two summer column sprites and reported that the lateral displacement was less than 50 km. One of the displacements was 20 km from the group of sprites, and the bottom altitude of the sprites was an average of 76.2 km. On the other hand, the lateral displacement of our four winter sprite events was a maximum of 30 km and an average of 19 km, and the bottom altitude of the sprite body was an average of 69 km. The bottom altitude is in good agreement with the corresponding results of Wescott et al. [1998]. Winter sprites and summer column sprites are very similar in this respect. However, the lateral displacement for summer sprites reported by Lyons [1996] and São Sabbas et al. [2003] is different from that for winter ones. That is, comparing our winter sprites to their summer ones, we feel that the lateral displacement for summer carrots (mostly 40∼50 km) and summer columns (within 50 km) would be slightly larger than that for winter sprites (∼20 km) on the basis of our analysis of a very small sample of column sprites. Stanley [2000] reported that sprites occurred primarily above the periphery of the most recent (approximately 200∼300 ms) section of the parent lightning discharge and suggested that sprites can thus supply a crude outline of the horizontal extent of charge removal. The most recent section of our sprite parent VHF lightning occurred less than 60 ms before the sprites did. Thus, all VHF lightning before and after sprite-producing lightning discharges could not be detected by our SAFIR, so we could not reconfirm the total extent of the lightning. Our sprites, however, also seem to be located roughly above the periphery of the parent VHF lightning, as Stanley [2000] mentioned.

[39] Asano et al. [2008, 2009a, 2009b] performed a computer simulation on the relationship between a sprite and its lightning discharge. Their simulation results also indicate that the horizontal offset between a sprite and the CG point may be determined by the length (extent) of a lightning channel. These results seem to be supported by several previous observational studies of the lateral displacement of winter sprites (this paper) and summer sprites [e.g., Lyons, 1996; São Sabbas et al., 2003]. However, one of our sprites (sprite 17) triangulated in three dimensions had a slightly large lateral displacement in spite of such a compact VHF cluster (<10 km in extent). We cannot explain clearly the reason behind this large lateral displacement. As one possible explanation, we need to consider the effect of a tilted lighting channel. For instance, if we assume that the horizontal length and vertical height are 5 km (lightning channel tilt is 45°), how does the displacement change? Additionally, when sprite 23 was generated, we found that two VHF clusters occurred within a short interval (50 ms) and removed positive charges from the same thunderstorm. If several lightning discharges occur and/or a lightning area is extended, how does the distribution of sprite elements change? Furthermore, the CGs and VHF sources for winter lightning cannot always be detected by the SAFIR even though sprites were observed (Table 3). Actually, the SAFIR could not observe CGs for 11/23 sprites and VHF sources for 8/23 sprites [Matsudo et al., 2007]. It is reported that winter lightning is often initiated by an upward leader, and it is also known that there is occasionally no return stroke. Such a case may not be detected by the SAFIR.

[40] It is agreed that summer sprites have a more bright and complex nature in their morphology, whereas winter sprites are dominated by a relatively faint and simple structure similar to a column shape [e.g., Hobara et al., 2001, 2006; Takahashi et al., 2003; Hayakawa et al., 2004, Matsudo et al., 2007]. Lyons [1996] pointed out that the +CG generating sprites are associated with an unusually large positive charge transfer, and the continuing current corresponds with intracloud “spider” or “dendritic” lightning known to accompany many strong +CGs. Ohkubo et al. [2005] investigated the relationship between observed Japanese winter sprites and their ability to produce a VLF/ELF lightning waveform. They suggested that an in-cloud discharge activity may play an important role in the generation mechanism of sprites. On the basis of 3-D computer simulations, Asano et al. [2008] found that heating by a temporal quasi-electrostatic (QE) field associated with a large amount of positive charge transfer would be fundamentally important for sprite genesis and sprite brightness, but they also pointed out that the horizontal location and fine structure (morphology) of sprites would be affected by the electromagnetic field radiated by the lightning channel [Asano et al. 2009b]. We investigated 13 sprite events and only 10 parent thunderstorms, so it seems to be a very small sample. However, we found that the complexity of ELF waveforms associated with sprites is somewhat related to the complexity of sprite morphology as mentioned in section 3.3. Some of the sprites (e.g., sprite 23), in particular, have obviously different characteristics. At least, in the two case studies we presented, there are many differences in the meteorological and electrical aspects. For example, column sprites (sprite 17) were produced by lightning in the vicinity of the convective cell, whereas a combination of column and V-shaped sprite (sprite 23) was generated by the lightning under the precipitating stratiform cloud. The ELF waveforms were also very different between sprites 17 and 23. Let us allow some speculative consideration. The amount of positive charges removed from the lower part of our thunderstorms with sprites may be larger than those removed from their upper part. These differences between sprites 17 and 23 imply that some of the possible factors that determine the complexity in the morphology of sprites may be affected by the characteristics of lightning (with or without continuing current, extent, shape, complexity, path, etc.). That is, the different characteristics of lightning may be attributed to the differences in meteorological characteristics inside the parent thunderstorms. Therefore, our results seems to support Asano et al. [2009a] who presented the relationship between lightning with continuing current and the optical characteristics of sprites by means of computer simulation.

4.3. Comparison of Winter Sprites and Their Parent Thunderstorms in Hokuriku With Those in Other Places

[41] Matsudo et al. [2007] have reported the morphology of winter sprites around the Hokuriku region of Japan. Sprites with a simpler structure (columnar and ball-like) were most dominant during our campaign; carrots were rarely observed, although V-shaped sprites were observed sometimes. Other Japanese groups also obtained similar results [Hobara et al., 2001; Takahashi et al., 2003; Hayakawa et al., 2004; Adachi et al., 2005]. Recently, winter sprites were observed in the eastern Mediterranean [Ganot et al. 2007; Greenberg et al., 2007; Yair et al., 2009]. It was indicated that about half of the sprites observed were carrots and the other half were columns. Matsudo et al. [2009] compared the morphologies of winter sprites in the two regions of the Sea of Japan and the Pacific Ocean near the east coast of Japan. It was found that column sprites were dominant for the winter sprites in both regions of Japan (about 60–70%), but the highly important point is that the occurrence rate of carrot sprites (including V-shaped sprites) in the Pacific Ocean (∼28%) was higher than that in the Sea of Japan (∼16%). The occurrence rate of carrot sprites in the eastern Mediterranean is higher than that of winter sprites around the Hokuriku region of Japan. Therefore, it may be concluded that the morphological characteristics for winter sprites above the eastern Mediterranean would be more similar to those above the Pacific Ocean rather than above the Hokuriku region. As Ganot et al. [2007] mentioned, the SST around the Hokuriku coast was colder than that in the eastern Mediterranean. Generally, the SST during winter in the Pacific Ocean seems to be warmer than that in the Sea of Japan around the Hokuriku region. We investigated the SSTs for the Pacific Ocean in the case of Matsudo et al. [2009] using the Japan Meteorological Agency (JMA) daily SST analysis. Then, the range, which was within the camera's field of view, of SSTs on the Pacific Ocean was between 15° and 20°C. The SSTs in the Pacific Ocean are more similar to those in the eastern Mediterranean Sea than those in the Sea of Japan around the Hokuriku region. In addition, we could not often calculate most of the CAPEs values at Wajima in the winter season, because of the absent LFC, but a few CAPE values we could calculate were less than 50 J/kg. The CAPEs at Wajima in the Hokuriku region are lower than those over the eastern Mediterranean (200∼400 J/kg) [Ganot et al., 2007].

[42] Most of our thunderstorms had a small echo size and a low echo top (∼4 km), and cloud top temperature was –25∼–35°C (cloud top 4∼5 km). Takahashi et al. [2003] also reported the height of cloud top in the vicinity of sprite parent stroke to be 4.2∼6.6 km in the Hokuriku district and 5.0∼7.0 km in the Pacific Ocean. According to Ganot et al. [2007], eastern Mediterranean winter thunderstorms are characterized by convective cells embedded in a much larger stratiform cloud, which has a limited vertical dimension of 5∼7 km and a cloud top temperature of ∼–40°C. Therefore, it seems that the maximum height of cloud top range in the Hokuriku region is slightly lower than that in the eastern Mediterranean and that the altitude range of the cloud top of winter thunderstorms with sprites above the eastern Mediterranean resembles those above the Pacific Ocean more than those above the Hokuriku region. As we already mentioned [Suzuki et al., 2006b], we often found that several small convective cells in the winter sprite parent thunderstorm were embedded in the large stratiform cloud. Hayakawa et al. [2004] indicated the parent thunderstorm structure from a case study of radar data and described that the horizontal scale of winter sprite parent thunderstorm cells is on the order of 20∼30 km. The thunderstorm cells are embedded in the flat-topped stratiform precipitation cloud that is extensive horizontally but extends in height from 2 to ∼6 km only. Furthermore, their sprite-producing cloud was considerably smaller than a summer continental MCS. It was also smaller than the minimum scale of summer sprite clouds, but it was still an order of size larger than that of a typical isolated thunderstorm. Hayakawa et al. [2004] questioned whether MCSs are necessary to generate winter sprites. Their sprite cloud was considerably smaller than the MCS as defined by Mohr and Zipser [1996], and so they concluded that even non-MCSs in winter can generate sprites. Takahashi et al. [2003] also described that a winter sprite-producing thunderstorm cell size was ∼30 km from an analysis of IR satellite images. Such a winter thunderstorm structure in the eastern Mediterranean was also reported by Ganot et al. [2007], and so the structure may be common for the winter thunderstorms in the Hokuriku and other regions. Adachi et al. [2005], by studying sprite parent thunderstorm systems in the Japanese winter, reported that winter thunderstorm system sizes were much smaller than those of summer MCS producing sprites. Williams and Yair [2006] have described that winter thunderstorms in the Hokuriku region are still considered to be larger than ordinary summer isolated thunderstorms. Suzuki et al. [2006a, 2006b] also pointed out that the winter thunderstorm echo is much smaller than the summer MCSs, although winter thunderstorms have a very large cloud area equivalent to MCSs, and a sprite-causative positive charge reservoir tended to reside in the cloud area (anvil-like stratiform cloud). Winter sprite parent thunderstorms would not strictly correspond to the definition of MCSs provided by Mohr and Zipser [1996], even if the cloud area were very large, because an MCS is defined on the basis of the extent of the radar reflectivity area (>2000 km2). However, the extent of cloud area could be a satisfactory definition of MCS size if we are able to use cloud area instead of radar reflectivity area. Therefore, a winter sprite parent thunderstorm may be an “MCS-like thunderstorm system” with a low cloud top.

[43] The Qds values of our sprite-producing lightning were estimated to be between about 300 and 600 C km (peak value 500 C km) [Matsudo et al., 2007]. Greenberg et al. [2007] reported Qds values of winter sprite-producing discharge in the eastern Mediterranean Sea ranging from 600 to 2800 C km with a peak of around 1000 C km. These were larger than our results. The range of Qds with summer sprite-producing CG is between 100 and 3000 C km with an average of ∼1000 C km [e.g., Huang et al., 1999; Hu et al., 2002; Lyons et al., 2003]. Therefore, it seems that the Qds values in the eastern Mediterranean Sea are almost equal to those in the continental summer rather than those in midwinter in the Hokuriku region.

[44] Sprites and their parent thunderstorms in different locations have been compared in this section. We still do not understand why such simple sprites are predominant in winter, but one possible explanation might be related to thunderstorm electrical characteristics attributed to different meteorological conditions and SST. These environmental differences would affect the behavior of summer thunderstorm electrification over land, in the eastern Mediterranean, Pacific Ocean, and the Sea of Japan (the Hokuriku region). Consequently, it may possibly affect the major morphology of sprites in these regions.

5. Summary

  1. Top of page
  2. Abstract
  3. 1. Introduction
  4. 2. Observations
  5. 3. Observational Results
  6. 4. Discussion
  7. 5. Summary
  8. Acknowledgments
  9. References
  10. Supporting Information

[45] Simultaneous observations by means of various instruments (two low-light CCD cameras, a C-band radar, a VHF/LF lightning mapping system, a field mill network, and an ELF observation system) were conducted during the 2004/2005 winter sprite campaign in Hokuriku, Japan. On the basis of a combined analysis of the data from those instruments, we studied meteorological and electrical structures for winter thunderstorms and spatial characteristics for sprites and obtained the following conclusions:

[46] 1. The sprite parent thunderstorms had an echo area with a 20 dBZ contour from 100 to 8500 km2, and their echo-top heights were between 3 and 5 km. These thunderstorms were characterized by small convective cells embedded in a large cloud area and/or precipitating stratiform cloud. The large cloud area was equivalent to an MCS in scale, and so they could be said to be a kind of MCS. The most important point is that our winter thunderstorms could occasionally generate several sprite events even if their radar echo sizes were smaller than 500 km2 in extent.

[47] 2. A sprite-producing lightning discharge with continuing current accompanied by relatively large multiple peaks with ELF waveforms generated complex sprites, but that with transient current with a relatively large single peak seems to generate simple sprites. Several active thunderstorms generate nine sprite events. Six out of these nine, which were generated by lightning with transient current, had simple structures, but the other three, which were generated by lightning with continuing current, had a complex morphology. The thunderstorms with stratiform structures generated lightning with continuing current and produced four complex sprites. Therefore, continuing currents are likely to affect the complexity of sprite morphology.

[48] 3. The positive charges responsible for winter sprites (5 of 7 cases) tended to reside in the upper part of the parent thunderstorms (convective core and anvil-like stratiform cloud), whereas only a few positive charges (2 of 7 cases) were deduced to be located in their lower part (precipitating stratiform cloud). The positive charges removed from the upper part of thunderstorms were estimated to be on the order of 100 C. Several positive charges removed from their lower part were estimated to be as large as 300∼400 C. The Qds value with winter sprites was low because the same positive charge height ds in winter is much lower than that in summer. Therefore, the amount of positive charge (Q) removed was sufficient for the generation of winter sprites in spite of the low Qds values.

[49] 4. The extent of the VHF clusters related to sprites in the parent thunderstorms was in the range of 5 to 72 km with an average of 25 km. The average horizontal displacement between sprite elements and sprite-producing lightning was approximately 20 km, and the range of displacement was 6∼30 km. The average bottom altitude for the sprite body was approximately 70 km and was located between a height of 66 and 74 km.

[50] We presented and discussed meteorological aspects and indirect 3-D electrical structures for winter sprite parent thunderstorms by means of an interpretation of various 2-D data sets. We obtained reasonable results for the parent thunderstorm characteristics using 2-D data for thunderstorm electrification, but we need to reconfirm whether the 3-D interpretation of the 2-D data is valid. That is, charge distributions and lightning phenomena in 3-D in the sprite parent thunderstorms are not well understood, and a further study of sprites and their related phenomena is under way. Future integrated studies, which include additional radar observations, 3-D lightning mapping, and in situ observations of the electric field and the particle types within the sprite parent thunderstorms, will enable us to confirm the results of the analysis of charge location presented in this paper. This will clarify the actual structure and electrification processes occurring within the sprite parent thunderstorm.

Acknowledgments

  1. Top of page
  2. Abstract
  3. 1. Introduction
  4. 2. Observations
  5. 3. Observational Results
  6. 4. Discussion
  7. 5. Summary
  8. Acknowledgments
  9. References
  10. Supporting Information

[51] The authors would like to thank T. Shimura of Air Weather Group, T. Hanada of Komatsu weather squadron, Y. Ando of the University of Electro-Communications, and Z.-I. Kawasaki and T. Morimoto of Osaka University for providing observational data, useful discussion, and valuable comments. We are thankful to T. Nagao of Tokai University for providing us with the optical field site at Shimizu. Thanks are also owed to Y. Ikegami and M. Sera of Solar-Terrestrial Environment Laboratory of Nagoya University at Moshiri, and K. Yamashita of the University of Electro-Communications, for maintaining the ELF measurements at Moshiri. Finally, we would like to thank the editor and four anonymous reviewers for their extremely useful suggestions and comments.

References

  1. Top of page
  2. Abstract
  3. 1. Introduction
  4. 2. Observations
  5. 3. Observational Results
  6. 4. Discussion
  7. 5. Summary
  8. Acknowledgments
  9. References
  10. Supporting Information

Supporting Information

  1. Top of page
  2. Abstract
  3. 1. Introduction
  4. 2. Observations
  5. 3. Observational Results
  6. 4. Discussion
  7. 5. Summary
  8. Acknowledgments
  9. References
  10. Supporting Information
FilenameFormatSizeDescription
jgrd16170-sup-0001-t01.txtplain text document1KTab-delimited Table 1.
jgrd16170-sup-0002-t02.txtplain text document1KTab-delimited Table 2.
jgrd16170-sup-0003-t03.txtplain text document1KTab-delimited Table 3.
jgrd16170-sup-0004-t04.txtplain text document1KTab-delimited Table 4.

Please note: Wiley Blackwell is not responsible for the content or functionality of any supporting information supplied by the authors. Any queries (other than missing content) should be directed to the corresponding author for the article.