Reflection height of daytime tweek atmospherics during the solar eclipse of 22 July 2009

JOURNAL OF GEOPHYSICAL RESEARCH, VOL. 117,, doi:10.1029/2012ja018151, 2012 Reflection height of daytime tweek atmospherics during the solar eclipse of 22 July 2009 Hiroyo Ohya, 1 Fuminori Tsuchiya, 2 Hiroyuki Nakata, 1 Kazuo Shiokawa, 3 Yoshizumi Miyoshi, 3 Kozo Yamashita, 4 and Yukihiro Takahashi 5 Received 24 July 2012; revised 11 September 2012; accepted 3 October 2012; published 14 November 2012. [1] We report multipoint observations of daytime tweek atmospherics during the solar eclipse of 22 July 2009. Sixteen and sixty-three tweek atmospherics were observed at Moshiri and Kagoshima, Japan, where the magnitudes of the solar eclipse were 0.458 and 0.966, respectively. This was the first observation of tweek atmospherics during a low-magnitude eclipse (0.458). The average and standard deviation of the reflection height were 94.9 13.7 km at Moshiri and 87.2 12.9 km at Kagoshima. The reflection height at Moshiri was almost the same as that for normal nighttime conditions in July (96.7 12.6 km) in spite of the low magnitude of the eclipse. The reflection height at Kagoshima seems be divided into two parts: propagation across the total solar eclipse path and propagation in the partial solar eclipse path. During the eclipse, we also observed the phase variation in the LF transmitter signals. The average change in the phase delay of the LF signals was 109 for the paths that crossed the eclipse path and 27 for the paths that did not cross the eclipse path. Assuming a normal daytime height for LF waves of 65 km, a ray tracing analysis indicates that the variations in phase correspond to a height increase of 5 6 km for the paths across the eclipse and 1 2 km for partial eclipse paths. The wide range of estimated tweek reflection heights at Kagoshima also suggests a difference in electron density in the lower ionosphere between total and partial solar eclipses. Citation: Ohya, H., F. Tsuchiya, H. Nakata, K. Shiokawa, Y. Miyoshi, K. Yamashita, and Y. Takahashi (2012), Reflection height of daytime tweek atmospherics during the solar eclipse of 22 July 2009, J. Geophys. Res., 117,, doi:10.1029/2012ja018151. 1. Introduction [2] Tweek atmospherics (1.5 10.0 khz) are very low frequency (VLF)/extremely low frequency (ELF) waves originating from lightning discharges. The tweeks propagate in the Earth-ionosphere waveguide. Tweeks can be observed only at nighttime owing to strong attenuation in the daytime ionosphere. However, a few researchers have reported observations of tweeks during daytime solar eclipses [Burton and Boardman, 1933; Reeve and Rycroft, 1972; Singh et al., 2011]. Burton and Boardman [1933] first reported observations of tweeks in the USA during the solar eclipse of 1 Graduate School of Engineering, Chiba University, Chiba, Japan. 2 Graduate School of Science, Tohoku University, Sendai, Japan. 3 Solar-Terrestrial Environment Laboratory, Nagoya University, Nagoya, Japan. 4 Department of Electrical Engineering, Salesian Polytechnic, Tokyo, Japan. 5 Graduate School of Science, Hokkaido University, Sapporo, Japan. Corresponding author: H. Ohya, Graduate School of Engineering, Chiba University, 1-33 Yayoi-cho, Inage-ku, Chiba 263-8522, Japan. (ohya@faculty.chiba-u.jp) 2012. American Geophysical Union. All Rights Reserved. 0148-0227/12/2012JA018151 31 August 1932. The cut-off frequency of the tweeks changed from 2300 to 1800 Hz around the time of maximum totality. The minimum peak of the cut-off frequency occurred 7 min after totality. The change in the cut-off frequency corresponded to a change in reflection height from 65 to 83 km. Reeve and Rycroft [1972] reported that tweek reflection height varied from 69 km at the beginning of the solar eclipse (14:30 LT) of 7 March 1970 to 76 km at totality (15:35 LT), on the basis of observations in Newfoundland, Canada (47.75 N, 52.72 W). At the end of the eclipse (16:30 LT), the reflection height recovered down to 69 km. Singh et al. [2011], from observations at Allahabad and Nainital, India, during the 22 July 2009 solar eclipse, reported that tweek reflection height rose from 89 km to about 91 92 km at maximum totality and then fell to 87 km at the end of eclipse. Such daytime tweek occurrences are the result of a decrease in electron density in the D-region ionosphere due to the solar eclipse. [3] Guha et al. [20110] reported from intensity of VLF sferics during two solar eclipses of 22 July 2009, and 15 January 2010 that the intensity at six discrete frequencies (4, 7, 9, 11, 14, and 16.5 khz) increased (the average, 4.5 db) from mean normal level during both solar eclipses. They calculated that the virtual reflection height (H ) of D-region 1of9

Table 1. Summary of LF Transmitter and Receiver Details Station Latitude Longitude Transmitter/Receiver Frequency (khz) BPC 34.5 N 115.8 E Transmitter 68.5 Fukushima 37.4 N 140.9 E Transmitter 40.0 Saga 33.5 N 130.2 E Transmitter 60.0 Rikubetsu 43.5 N 143.8 E Receiver - Tainan 23.1 N 120.2 E Receiver - Zao 38.1 N 140.53 E Receiver - height-electron density model [Cummer et al., 1998] increased by 4.85 km during the solar eclipse of 22 July, 2009. [4] Several studies using VLF/LF/MF transmitter signals have also examined the lower ionosphere during solar eclipses [e.g., Bracewell, 1952; Crary and Schneible, 1965; Lynn, 1981; Abraham et al., 1998; Clilverd et al., 2001; Guha et al., 2010; De et al., 2011]. Bracewell [1952] first reported solar eclipse effects on VLF transmitter signals, sent from Rugby (16 khz) to Cambridge in the UK. Crary and Schneible [1965] and Lynn [1981] reported increases in amplitude of 1 3 db and decreases in phase of 20 100. Reflection height rose by 3 11 km. The peak of amplitude occurred 2 7 min before maximum solar obscuration, while the maximum phase changes occurred 1 7 min after maximum solar obscuration. Abraham et al. [1998] reported using ionospheric radio wave absorption (2.5 MHz, A1 technique) that the absorption decreased by 18 db at the totality during the solar eclipse of 24 October 1995. Clilverd et al. [2001] found that the amplitude showed positive changes for path lengths <2,000 km and negative changes for path lengths >10,000 km using 17 paths (16 24 khz). The phase showed negative shifts during the eclipse regardless of the path length. The effective ionospheric height and scale height parameters, H and b,ofwait s model were 79 km and 0.5 km 1, respectively, at maximum totality, whereas H and b were 71 km and 0.43 km 1 for usual daytime conditions [Wait and Spies, 1964]. Guha et al. [2010] showed a decrease (3.2 db) in the amplitude of VLF transmitter signals (18.2 khz) in India at the peak of the total solar eclipse of 22 July 2009. De et al. [2011] also reported a decrease in the amplitude of VLF/LF transmitter signals from Australia (19.8 khz) and Japan (40 khz) to India during the same solar eclipse. [5] Rocket measurements also have been performed during solar eclipses [Ulwick, 1972; Mechtly et al., 1969, 1972]. For rocket measurements at Cassino, Brazil, during an eclipse of 12 November 1966, H and b were estimated to be 80 km and 0.5 km 1, respectively [Ulwick, 1972]. Mechtly et al. [1969] reported, based on Langmuir probe observations on board a rocket experiment during a solar eclipse of 12 November 1966 that the electron densities at 82 km were 800, 400, 300, and 100 cm 3 at full sun (two hours after totality), at 2.5% visibility of the solar disk, at second contact (the total eclipse starts), and at third contact (the total eclipse ends), respectively. Mechtly et al. [1972], based on observations at Wallops Island, USA, reported that reflection height changed by 8 km and that electron density in the lower ionosphere decreased for 1 2 min after totality. [6] Ionozondes have also been adopted to investigate ionospheric effects associated with solar eclipses. Chandra et al. [2007] reported that a reduction of about 20% in the lowest frequency of ordinary mode (f min ) of ionograms was observed over Ahmedabad, India on the solar eclipse day of 11 August, 1999. The reduction in the f min indicates a decrease in D-region ionization. [7] In this study, we investigated the D- and lower E region ionospheric response during the solar eclipse of 22 July 2009 using tweek atmospherics observed at two stations, one having high and one having low solar eclipse magnitude. The tweek observation sites were Moshiri (44.37 N, 142.27 E; magnitude of eclipse: 0.458) and Kagoshima (31.48 N, 130.72 E; magnitude of eclipse: 0.966), Japan. Until this study, tweeks had not been observed at stations at which the eclipse magnitude was low. So far, the lowest magnitude of eclipse during which tweeks were observed was 0.85 [Singh et al., 2011]. We also observed the phase of LF transmitter signals both across and not across the eclipse path. The LF transmitter and receiver stations are shown in Table 1. Using the tweeks and LF transmitter signals, we documented the D- and lower E region responses during a low partial solar eclipse. 2. Observation Systems [8] The total solar eclipse occurred on 22 July 2009. Figure 1 shows our observation sites, propagation paths of LF transmitter signals, and the solar eclipse path. The solid lines in the figure indicate the paths of the LF transmitter signals that propagated across the solar eclipse path; the dotted lines show those that did not propagate across it. The path of the Moon s umbral shadow began in India at 00:53 UT and crossed China. After leaving China, the eclipse path crossed the Kita-ioh Islands (25.42 N, 141.27 E), Japan, at 01:56 UT (10:56 LT, LT = UT + 9 h) and then curved southeast over the Pacific Ocean. [9] From 21 to 23 July 2009, VLF radio waves including tweeks were continuously observed at Kagoshima, Japan, from 00:00 UT (09:00 LT) to 04:00 UT (13:00 LT) to investigate solar eclipse effects, while a routine observation mode was performed at Moshiri. For the routine observation mode, only the 2-min intervals at 20 22 min and 50 52 min of every hour were recorded. Tweeks usually cannot be observed in the daytime because of strong absorption of VLF waves caused by solar ionization of the ionosphere. The appearance of tweeks at Moshiri and Kagoshima during the daytime solar eclipse indicates a decrease in the electron density in the D- and lower E regions. [10] Tweeks were received with two-turn triangular loop antennas. The antenna at Kagoshima was 10 20 m in size and the one at Moshiri was 43 60 m. Tweek signals were amplified by a pre-amplifier and a main amplifier. The analog tweek signals were digitized using a 16-bit A/D converter with a 20-kHz sampling frequency. The signals (0 10 khz) were digitally recorded on hard disks. Tweek reflection height and propagation distance were analyzed using an automated procedure of spectral fitting to estimate the cut-off frequency [Ohya et al., 2006, 2011]. The source location of the tweeks could not be estimated because only one magnetic component was observed at Kagoshima. In general, however, 78% of all lightning occurs within 30 latitude [Christian et al., 2003], so many of the tweeks are expected to have occurred in the equatorial region and propagated across the eclipse path. [11] The amplitude and phase of LF transmitter signals were received at Rikubetsu and Zao in Japan and Tainan in 2of9

Figure 1. Map of the VLF/LF stations used in this study. The transmitter sites of LF signals are FUK, SAG, and BPC stations. The receiver sites are RIK, ZAO, and TNN stations. The solid blue lines indicate the paths of LF transmitter signals that propagated across the solar eclipse path; the dotted blue lines show those that did not propagate across it. Tweeks were observed at MSR and KAG stations. The time on the path of the total solar eclipse (the red lines) is in UT. Taiwan (Table 1). Figure 1 shows a map of the propagation paths of the LF transmitter signals. Solid and dotted lines indicate paths transverse and paths not transverse to the eclipse path, respectively. The frequencies of the LF transmitter signals were 40.0 khz (wavelength = 7.5 km) from Fukushima (FUK) and 60.0 khz (5.0 km) from Saga (SAG) in Japan and 68.5 khz (4.4 km) from China (BPC). We observed the amplitude and phase of these signals with monopole antennas of about 2 m length. The data were amplified by a main amplifier and then digitized using a 16-bit A/D converter with a 200-kHz sampling frequency. The waveform data were converted to a spectrum by FFT and amplitudes and phases at transmitter frequencies are recorded on hard disks with a time resolution of 0.1 s. The amplitude showed the composition of direct and reflected waves. In this study, however, we do not discuss amplitude because the phase of LF transmitter signals is more sensitive than amplitude for studying vertical variation in the ionosphere [Davies, 1969]. The JJY time code (the call sign of a LF time signal radio station in Japan) was adopted for adjusting the time. The Tainan station is one of the Asia VLF Observation Network (AVON) system sites [Ohya et al., 2010]. 3. Results 3.1. Tweek Atmospherics [12] Figures 2a and 2b show examples of dynamic spectra of tweek atmospherics observed during the solar eclipse at Moshiri (00:51:27 UT) and Kagoshima (01:33:24 UT), Japan, respectively. Clear decreasing frequency trends, which are the typical signature of tweeks, are seen in these events. The white dots in the spectra show power peaks that were used for spectral fitting to determine the reflection height [Ohya et al., 2008]. The durations of the tweek signals shown in Figures 2a and 2b were 5.4 ms and 8.0 ms, respectively. The durations of the tweek signals at Moshiri and Kagoshima during the solar eclipse ranged from 5.7 to 21.1 ms and 6.4 19.3 ms, respectively, which is much shorter than the usual nighttime duration (about 50 ms). [13] Figures 3a and 3b show the reflection heights of the tweeks obtained at Moshiri and Kagoshima, Japan, during the solar eclipse, respectively. Sixteen and sixty-three tweek atmospherics were observed at Moshiri and Kagoshima, where the magnitudes of the solar eclipse were 0.458 and 0.966, respectively. This was the first observation of tweek atmospherics at a low eclipse magnitude (0.458). Tweeks were continuously observed at Kagoshima during the solar eclipse. No tweeks were simultaneously received at both sites. Tweeks were not observed on both 21 and 23 July 2009 at both Moshiri and Kagoshima during these local times. In Figure 3a, three solid vertical lines indicate the beginning time (01:07:51 UT), peak time (02:10:52 UT), and end time (03:13:43 UT) of the solar eclipse. The average and standard deviation of the tweek reflection height through the solar eclipse was 94.9 13.7 km at Moshiri. The reflection height at Moshiri was almost the same as that for normal nighttime in July (96.7 12.6 km) in spite of the low magnitude of the eclipse. At 47 min before the beginning of the solar eclipse at Moshiri, the first tweek was observed. The last tweek was observed 97 min after the end time of the solar eclipse. 3of9

[15] Figures 4a and 4b show the number of automatically analyzed tweeks per every 2-min interval during the solar eclipse at Moshiri and Kagoshima, respectively. Note that measurements were performed continuously at Kagoshima but were only taken during the 2-min intervals of 20 22 min and 50 52 min of every hour at Moshiri. The 2-min interval observation is normal operation, and the intervals were chosen because of huge data amount (about 5 MB per two minutes). We did not perform continuous observations at Moshiri, because we did not expect that tweeks could be received at such low eclipse magnitude. The vertical lines are the same as those in Figures 3a and 3b. At Moshiri, the number of tweeks was zero at peak eclipse time, with peaks occurring around the beginning and end times of the eclipse. On the other hand, sharp peaks in the number of tweeks occurred near the time of the total eclipse at Kagoshima. 3.2. LF Transmitter Signals [16] Figure 5 shows variations in the phase of LF transmitter signals received at Tainan (TNN) from (a) BPC, China Figure 2. Waveforms and dynamic spectra of tweek atmospherics observed during the solar eclipse (a) at Moshiri (00:51:27 UT) and (b) at Kagoshima (01:33:24 UT), Japan. [14] In Figure 3b, a solid vertical line indicates the peak time (01:57:45 UT) of the solar eclipse at Kagoshima. The beginning and end times were 00:37:31 UT and 03:21:20 UT, respectively. Tweek signals were also observed before and after the solar eclipse at Kagoshima, but these could not be analyzed because of a poorer signal-to-noise ratio than occurred at Moshiri. The average and standard deviation of the tweek reflection height through the solar eclipse at Kagoshima were 89.2 12.2 km. The average and standard deviation of normal July nighttime reflection height are 96.7 12.6 km, obtained from tweek observations at Kagoshima in 1976 2010 during magnetically quiet times [Ohya et al., 2011]. The tweek reflection height for Kagoshima during the eclipse was much lower than this value. At 01:00 01:40 UT, tweek reflection heights were higher than 100 km, although they were estimated in a wide range of 75 130 km at 01:40 02:20 UT at around the peak eclipse time. Figure 3. Reflection heights of all tweek atmospherics observed at (a) Moshiri and (b) Kagoshima, Japan, during the solar eclipse. The three solid vertical lines in Figure 3a indicate the beginning time (01:07:51 UT), peak time (02:10:52 UT), and end time (03:13:43 UT) of the solar eclipse. The solid vertical line in Figure 3b indicates the peak time (01:57:45 UT) of the solar eclipse at Kagoshima. 4of9

Figure 4. Number of automatically analyzed tweeks per 2-min interval at (a) Moshiri and (b) Kagoshima during the solar eclipse. The vertical lines are the same as in Figures 3a and 3b. (68.5 khz), (b) Saga (SAG, 60.0 khz), and (c) Fukushima (FUK, 40.0 khz). Black, red, and blue lines indicate the phases of LF transmitter signals on a day before the solar eclipse (20 July 2009), the day of the solar eclipse (22 July), and the day after the solar eclipse (23 July), respectively. The SAG transmitter signal did not operate on 21 July. The two vertical lines in each panel indicate the start and end times of the total solar eclipse over each propagation path. All paths propagate across the eclipse path. The propagation distances of (a), (b), and (c) are 1333.6 km, 1511.8 km, and 2520.2 km, respectively. For all propagation paths, large variations in phase (average: 109 ) were seen only on the day of the solar eclipse. The variations in phase were seen from the western propagation path (BPC TNN) to the eastern path (FUK TNN), according to the motion of the Moon s umbral shadow. The velocity of the peak from the BPC TNN path to the FUK TNN path was estimated to be 0.996 km/s, which was consistent with the velocity of the Moon s umbral shadow (1.0 km/s) [Espenak and Anderson, 2008]. As shown in Figure 5a, time differences existed between the peak of the phase and the eclipse maximum. The phase of BPC TNN started to change at about 00:45 UT (09:45 LT) and then peaked at 01:38 UT (10:38 LT). The maximum difference in the phase was about 135 at 01:38 UT. The peak time of the solar eclipse over the BPC TNN path was from 01:28 UT (10:28 LT) to 01:33 UT (10:33 LT). The difference in the peak times was about 7 10 min. As seen in Figure 5b, the maximum difference in the phase for the SAG TNN signals was 127, and the delay of the peak time of the LF phase was ~6 min. In Figure 5c, the LF phase variation in FUK TNN is similar to that of the SAG TNN phase path in Figure 5b. The maximum difference in the phase was 65 at 02:00 UT (11:00 LT). The totality time of the solar eclipse along the propagation path was from 01:50 UT (10:50 LT) to 01:57 UT (10:57 LT). The time difference between the phase peak and the total solar eclipse was 3 10 min. Accordingly, the differences between the peak time for the LF phase and the total solar eclipse were ~10 min for the BPC TNN and FUK TNN paths. [17] Figure 6 shows the variations in the phase of the LF transmitter signals received at Rikubetsu (RIK), Japan, that did not propagate across the solar eclipse path from (a) BPC, (b) SAG, and (c) FUK. In the same way as for Figure 5, black, red, and blue lines indicate the phases of LF transmitter signals on the day before the solar eclipse (20 July 2009), on the day of the solar eclipse (22 July), and on the day after the solar eclipse (23 July), respectively. The lengths of the paths were (a) 2607.4 km, (b) 1620.7 km, and (c) 720.5 km, respectively. Variations in the phase (average: 27 ) for these paths were observed but were much smaller than those of the paths that propagated across the eclipse path. As shown in Figure 6a, the phase started to vary at about 01:00 UT (10:00 LT) and then peaked at about 01:52 UT (10:52 LT). The maximum difference in the phase reached 32. Then, the phase began and continued to recover. The phase of the SAG RIK transmitter signals (Figure 6b) started to vary at about 01:15 UT (10:15 LT) and then peaked at about Figure 5. Variations in the phase of LF transmitter signals that propagated across the solar eclipse path for the (a) BPC TNN, (b) SAG TNN, and (c) FUK TNN paths. The black, red, and blue lines show variations in the phase on 20, 22, and 23 July 2009, respectively. The three vertical lines indicate the beginning, peak, and end time of the solar eclipse over each path. 5of9

we discuss possible causes and implications of the observed tweeks and LF phase variations. Figure 6. Variations in the phase of LF transmitter signals that did not propagate across the solar eclipse path for the (a) BPC RIK, (b) SAG RIK, and (c) FUK RIK paths. The black, red, and blue lines show variations in the phase on 20, 22, and 23 July 2009, respectively. 02:07 UT (11:07 LT). The maximum difference in phase was about 30. As shown in Figure 6c, the phase of FUK RIK started to vary at 01:10 UT (10:10 LT) and then peaked at 02:06 UT (11:06 LT). The maximum difference in the phase was 18. The results of the LF transmitter signals are summarized in Table 2. The variations in LF reflection height variations in Table 2 are discussed in Section 4. 4. Discussion [18] Both the appearance of tweeks and a phase delay of LF transmitter signals indicate a significant decrease in electron density (increase of reflection height) in the D- and lower E region ionosphere during the solar eclipse [e.g., Reeve and Rycroft, 1972; Clilverd et al., 2001]. We can investigate the difference of variations in height during the solar eclipse using tweek and LF transmitter signal observations. The tweeks reflect at slightly lower heights than those of LF transmitter signals, because the cut-off frequency of tweeks is lower than the LF transmitter signals. The propagation path lengths of both tweeks and LF transmitter signals are long, although the path location is different. So wide area of the lower ionosphere at low and midlatitudes can be monitored through tweeks and LF transmitter signals. Here 4.1. Tweek Atmospherics [19] Sixteen and sixty-three tweek atmospherics were observed at Moshiri and Kagoshima, Japan, where the magnitudes of the solar eclipse at the station were 0.458 and 0.966, respectively. This is the first observation of tweek atmospherics for such a low-magnitude eclipse (0.458). In previous studies, tweeks had been observed only at sites where the magnitude of the eclipse was greater than 0.85 [Burton and Boardman,1933; Reeve and Rycroft, 1972; Singh et al., 2011]. In addition, two and seven tweeks were analyzed at Moshiri in the 1 2hperiod before and after the solar eclipse, respectively. [20] We compared the occurrence time of the tweeks with lightning distribution data from the World Wide Lightning Location Network (WWLLN) [Lay et al., 2004]. Figures 7a and 7b show the lightning distribution detected by WWLLN at 00:20 00:22 UT and 04:50 04:52 UT, respectively. The 00:20 00:22 UT and 04:50 04:52 UT intervals are the time periods during which the first and last tweeks were observed at Moshiri, respectively. At those times, the partial eclipse occurred at sunrise (sunset). We could not estimate the propagation distance by fitting the frequency dispersion curves of these tweeks because the duration of the tweek signals was very short. The figures show that several lightning events occurred in and near the partial eclipse region. These lightning events are a possible origin of the tweeks, although the detection rate of WWLLN is ~10.3% [Abarca et al., 2010]. The present results indicate that even with a partial eclipse, the electron density in the D-region decreases sufficiently for tweeks to propagate over thousands of kilometers. [21] The average and standard deviation of the tweek reflection height through the solar eclipse were 94.9 km 13.7 km at Moshiri and 87.2 km 12.9 km at Kagoshima. These average reflection heights are in agreement with those of previous studies of tweeks during solar eclipses [Burton and Boardman, 1933; Reeve and Rycroft, 1972; Singh et al., 2011]. The normal July nighttime average and standard deviation tweek reflection height are 96.7 km 12.6 km at Kagoshima, according to 35-year statistics [Ohya et al., 2011]. [22] We estimated the electron density from the tweek reflection height based on the International Reference Ionosphere (IRI) 2007 model. The electron density n e (cm 3 )at the tweek reflection height can be estimated from the following equation: n e ¼ 1:241 10 8 cf H 2h ; ð1þ Table 2. Summary for Results of LF Transmitter Signals Transmitter-Receiver Frequency (khz) Distance (km) Maximum Obscuration (%) Phase (deg) Height (km) Time Delay (min) BPC-Tainan 68.5 1334 100 135 +5.885 7 10 Saga-Tainan 60.0 1512 100 127 +4.975 6 Fukushima-Tainan 40.0 2520 100 65 +2.060 3 10 BPC-Rikubetsu 68.5 2607 91 32 +0.720 - BPC-Zao 68.5 2247 91 60 +1.150 - Saga-Rikubetsu 60.0 1621 89 30 +1.150 - Saga-Zao 60.0 1061 89 38 +1.995 - Fukushima-Rikubetsu 40.0 721 62 18 +1.060-6of9

Figure 7. Lightning distribution detected by WWLLN at (a) 00:20 00:22 UT and at (b) 04:50 04:52 UT. The 00:20 00:22 UT and 04:50 04:52 UT intervals are the 2-min periods during which the first and last tweeks were observed at Moshiri, respectively. where c is the light velocity, f H is the electron gyrofrequency, and h is the reflection height of tweeks for the first-order mode [Ohya et al., 2003]. Here we assumed f H = 1.1 MHz because we received tweeks from lightning discharges that occurred mainly in low-latitude and equatorial regions. The value of 1.1 MHz was taken from the International Geomagnetic Reference Field (IGRF) model. From the IRI (2007) model, the electron density at an altitude of 100 km at 02:00 UT is 1.0 10 5 cm 3. The average tweek reflection height of 87.2 km during the solar eclipse corresponds to an electron density of 23 cm 3 according to equation (1). If the shape of the electron density-altitude profile of the IRI (2007) model is assumed to be the same as that for normal nighttime, the estimated electron density would be 5.0 10 2 cm 3 at an altitude of 100 km. It is estimated to be 3.0 10 2 cm 3 from the normal nighttime reflection height. That is, from tweek analysis, the electron density in the lower E region is estimated to have decreased from 1.0 10 5 cm 3 to 5.0 10 2 cm 3 at an altitude of 100 km during the solar eclipse, but the electron density did not reach the normal nighttime level of 3.0 10 2 cm 3. [23] The average and standard deviation of the tweek reflection height through the solar eclipse were 94.9 km 13.7 km at Moshiri and 87.2 km 12.9 km at Kagoshima. The maximum reflection height at Moshiri (113.5 km) was also different from that at Kagoshima (128.9 km) during the solar eclipse. The difference in the average and maximum reflection heights between Moshiri and Kagoshima was caused by the difference in relative location between the tweek propagation path and the Moon s umbral shadow. We could not estimate the direction of tweek propagation in this research because only one magnetic component was measured at each site. However, the present result does show that the tweek reflection height depends not simply on the magnitude of the eclipse at the observation sites but also on the condition of the ionosphere along tweek propagation paths. [24] As seen in Figure 3, the standard deviation of the tweek reflection height for the solar eclipse was 13.7 km at Moshiri and 12.9 km at Kagoshima. This deviation was much larger than that found (1 2 km) in previous studies [Reeve and Rycroft, 1972; Singh et al., 2011]. However, the deviation was comparable with that for normal nighttime conditions (12.6 km) at Kagoshima, Japan. [25] The durations of the tweek signals at Moshiri and Kagoshima during the solar eclipse ranged from 5.7 to 21.1 ms and 6.4 19.3 ms, respectively, which is much shorter than the usual nighttime duration (about 50 ms). The short duration of tweek signals might be caused by the larger attenuation of tweek signals during the daytime, consistent with the above discussion of the electron density. [26] No tweeks were simultaneously received at Moshiri and Kagoshima. The main reason is that simultaneous observation time was short to be 4 min in one hour. Tweeks were continuously observed at Kagoshima, while a routine observation of 2-min per 30 min was performed at Moshiri. The other reason may be that the attenuation of tweek signals at Moshiri would be significantly larger than that at Kagoshima under the low partial solar eclipse. If the location of the lightning (the origin of tweeks) could be too close to Kagoshima, the characteristic dispersion of tweeks could not be seen at Kagoshima, due to short propagation distance, while it could be seen at Moshiri. Such situation may cause the tweeks that was received only at Moshiri, but not at Kagoshima. 4.2. LF Transmitter Signals [27] The LF transmitter signal propagates horizontally in the wave duct between the ground and the D-layer of the ionosphere [Davies, 1969]. On the basis of the observed phase variations in LF transmitter signals, we calculated the height variation of the duct by a geometrical method [Jacobson et al., 2009]. If we assume that the duct height before the solar eclipse was 65 km based on the IRI (2007) model, the changes in the height along BPC TNN (across the eclipse path), SAG TNN (across), SAG RIK (not across), and SAG ZAO (not across) would be +5.885 km, +4.975 km, +1.150 km, and +1.995 km, respectively. Therefore, the changes in height of the LF paths that crossed the eclipse path were 3 5 kmlarger than those that did not cross the eclipse path. [28] On 22 July, a magnetic storm (minimum Dst index: 78 nt) occurred 5 h after the solar eclipse at 07:00 UT. However, magnetic field variation was quiet during the 7of9

eclipse (Dst index: 4 nt). Because large phase variations in LF transmitter signals were not observed during the storm, it is unlikely that the observed phase variation was caused by the magnetic storm. [29] For the case in which the LF transmitter signals did not propagate across the eclipse path and when h = 65 km was assumed, the maximum increases of the reflection height for the BPC RIK, SAG RIK, and FUK RIK signals were calculated by ray tracing to be 0.72 km, 1.15 km, and 1.06 km, respectively. The height changes were smaller than those for the paths that propagated across the eclipse path. [30] As shown in Figure 5, the time delay of the phase peak of LF transmitter signals after totality was 2 10 min. Previous studies also showed a similar time delay of 4 7 min [Sen Gupta et al., 1980; Lynn, 1981; Mendes Da Costa et al., 1995]. We speculated that the following two factors might have caused the time delay: the electron production rate of the partial solar eclipse after totality and the electron loss rate by recombination. The electron production rate by solar ionization at 65 km was estimated to be ~1 cm 3 s 1 [Tohmatsu, 1990]. In an actual partial solar eclipse, the electron production rate would be much less than that for normal daytime conditions. We estimated recombination rates based on the equations of Sheehan and St.-Maurice [2004]. If the electron temperature at 65 km is assumed to be 220 K based on the Mass Spectrometer Incoherent Scatter 90 (MSIS90) model, the efficient recombination coefficients (a eff ) with O 2 + and NO + are 2.4 10 7 cm 3 s 1 and 4.3 10 7 cm 3 s 1, respectively. From the IRI (2007) model, the electron density at an altitude of 65 km at 02:00 UT is 7.7 10 1 cm 3. The loss rate by recombination is a eff n 2 e [Brekke, 1997], and the loss rates for O 2 + and NO + are 1.4 10 3 cm 3 s 1 and 2.5 10 3 cm 3 s 1, respectively. The balance between the production of electrons by solar ionization in the partial solar eclipse after totality and the electron loss by recombination could have caused the observed time delay of the LF transmitter phase after the total eclipse. 5. Conclusions [31] This paper reports multipoint observations of daytime tweek atmospherics during the solar eclipse of 22 July 2009. Sixteen and sixty-three tweek atmospherics were observed at Moshiri and Kagoshima, Japan, where the magnitude of the solar eclipse was 0.458 and 0.966, respectively. This was the first reported observation of tweek atmospherics for a lowmagnitude eclipse (0.458). The observed features are summarized as follows. [32] 1. The average and standard deviation of the reflection height at Moshiri and Kagoshima were 94.9 13.7 km and 87.2 12.9 km, respectively. The reflection height at Moshiri was almost the same as that for normal nighttime in July (96.7 12.6 km) in spite of the low magnitude of the eclipse. [33] 2. During the eclipse, we also observed the phase variation in the LF transmitter signals. The average change in the phase delay of the LF transmitter signals was 109 for the paths that crossed the eclipse path and 27 for the paths that did not cross the eclipse path. Time delays of 2 10 min occurred from the eclipse maximum to the peak of the phase variation. [34] Both the appearance of tweeks and the phase delay of the LF transmitter signals indicate a decrease in the electron density in the D- and lower E regions. Assuming a normal daytime height for the LF waves of 65 km, a ray tracing analysis indicates that the variations in the phase correspond to height increases of 5 6 km for the paths that cross the eclipse and 1 2 km for the partial eclipse paths. The wide range (75 130 km) of estimated tweek reflection heights at Kagoshima also suggests a difference in electron density in the D- and lower E region ionosphere between total and partial solar eclipses. The balance between electron production by solar ionization in the partial solar eclipse after totality and electron loss by recombination could cause the observed time delay of the LF transmitter phase after the total eclipse. [35] Acknowledgments. We are grateful to M. Satoh, Y. Katoh, Y. Hamaguchi, Y. Yamamoto, M. Sera and Y. Ikegami, of the Solar- Terrestrial Environment Laboratory (STEL), Nagoya University, Japan, for their technical support of the continuous VLF/ELF measurements during the solar eclipse. This work was supported by Project 2 and the cooperative research program of STEL. We thank A. B. Chen of National Cheng Kung University for the operational support of the Asia VLF Observation Network (AVON) project in Tainan, Taiwan. [36] Robert Lysak thanks the reviewers for their assistance in evaluating this paper. References Abarca, S. F., K. L. Corbosiero, and T. J. Galarneau Jr. (2010), An evaluation of the Worldwide Lightning Location Network (WWLLN) using the National Lightning Detection Network (NLDN) as ground truth, J. Geophys. Res., 115, D18206, doi:10.1029/2009jd013411. Abraham, S., S. K. Dhaka, N. Nath, and K. L. Baluja (1998), Ionospheric absorption on October 24, 1995 solar eclipse, Geophys. Res. Lett., 25(15), 2945 2947, doi:10.1029/98gl01781. Bracewell, R. N. (1952), Theory of formation of an ionospheric layer below E layer based on eclipse and solar flare effects at 16 kc/sec, J. Atmos. Terr. Phys., 2, 226 235, doi:10.1016/0021-9169(52)90033-0. Brekke, A. (1997), Physics of the Upper Polar Atmosphere, Praxis, Chichester, U. K. Burton, E. T., and E. M. Boardman (1933), Effects of solar eclipse on audio frequency atmospherics, Nature, 131, 81 82, doi:10.1038/131081a0. Chandra, H., S. Sharma, P. D. Lele, G. Rajaram, and A. Hanchinal (2007), Ionospheric measurements during the total solar eclipse of 11 August 1999, Earth Planets Space, 59, 59 64. Christian, H. J., et al. (2003), Global frequency and distribution of lightning as observed from space by the Optical Transient Detector, J. Geophys. Res., 108(D1), 4005, doi:10.1029/2002jd002347. Clilverd, M. A., C. J. Rodger, N. R. Thomson, J. Lichtenberger, P. Steinbach, P. Cannon, and M. J. Angling (2001), Total solar eclipse effects on VLF signals: Observations and modeling, Radio Sci., 36(4), 773 788, doi:10.1029/ 2000RS002395. Crary, J. H., and D. E. Schneible (1965), Effect of the eclipse of 20 July 1963 on VLF signals propagating over short paths, J. Res. Natl. Bur. Stand. U.S., Sect. D, 69, 947 957. Cummer, S. A., U. S. Inan, and T. F. Bell (1998), Ionospheric D region remote sensing using VLF radio atmospherics, Radio Sci., 33(6), 1781 1792, doi:10.1029/98rs02381. Davies, K. (1969), Ionospheric Radio Waves, Blaisdell, Waltham, Mass. De, B. K., S. S. De, B. Bandyopadhyay, P. Pal, R. Ali, S. Paul, and P. K. Goswami (2011), Effects of a solar eclipse on the propagation of VLF- LF signals: Observations and results, Terr. Atmos. Oceanic Sci., 22(4), 435 442, doi:10.3319/tao.2011.01.17.01(aa). Espenak, F., and J. Anderson (2008), Total solar eclipse of 22 July 2009, NASA Eclipse Rep., NASA/TP-2008-214169. Guha, A., B. K. De, R. Roy, and A. Choudhury (2010), Response of the equatorial lower ionosphere to the total solar eclipse of 22 July 2009 during sunrise transition period studied using VLF signal, J. Geophys. Res., 115, A11302, doi:10.1029/2009ja015101. Jacobson, A. R., X.-M. Shao, and R. Holzworth (2009), Full-wave reflection of lightning long-wave radio pulses from the ionospheric D region: Numerical model, J. Geophys. Res., 114, A03303, doi:10.1029/ 2008JA013642. Lay, E. H., R. H. Holzworth, C. J. Rodger, J. N. Thomas, O. Pinto Jr., and R. L. Dowden (2004), WWLL global lightning detection system: 8of9

Regional validation study in Brazil, Geophys. Res. Lett., 31, L03102, doi:10.1029/2003gl018882. Lynn, K. J. W. (1981), The total solar eclipse of 23 October 1976 observed at VLF, J. Atmos. Terr. Phys., 43(12), 1309 1316, doi:10.1016/0021-9169(81)90156-2. Mechtly, E. A., K. Seino, and L. G. Smith (1969), Lower ionosphere electron densities measured during the solar eclipse of November 12, 1966, Radio Sci., 4(4), 371 375, doi:10.1029/rs004i004p00371. Mechtly, E. A., C. F. Sechrist, and L. G. Smith (1972), Electron loss coefficients for the D-region of the ionosphere from rocket measurements during the eclipses of March 1970 and November 1966, J. Atmos. Terr. Phys., 34, 641 646, doi:10.1016/0021-9169(72)90151-1. Mendes Da Costa, A., N. M. Paes Leme, and L. Rizzo Piazza (1995), Lower ionosphere effect observed during the 30 June 1992 total solar eclipse, J. Atmos. Terr. Phys., 57, 13 17, doi:10.1016/0021-9169(93)e0021-z. Ohya, H., M. Nishino, Y. Murayama, and K. Igarashi (2003), Equivalent electron densities at reflection heights of tweek atmospherics in the lowmiddle latitude D-region ionosphere, Earth Planets Space, 55, 627 635. Ohya, H., M. Nishino, Y. Murayama, K. Igarashi, and A. Saito (2006), Using tweek atmospherics to measure the response of the low-middle latitude D-region ionosphere to a magnetic storm, J. Atmos. Sol. Terr. Phys., 68, 697 709, doi:10.1016/j.jastp.2005.10.014. Ohya, H., K. Shiokawa, and Y. Miyoshi (2008), Development of an automatic procedure to estimate the reflection height of tweek atmospherics, Earth Planets Space, 60, 837 843. Ohya, H., F. Tsuchiya, K. Yamashita, Y. Takahashi, and T. Adachi (2010), Tweek reflection height observed by VLF observation network system in Southeast Asia, paper presented at 2010 Asia-Pacific Radio Science Conference, Int. Union of Radio Sci., Toyama, Japan. Ohya, H., K. Shiokawa, and Y. Miyoshi (2011), Long-term variations in tweek reflection height in the D and lower E regions of the ionosphere, J. Geophys. Res., 116, A10322, doi:10.1029/2011ja016800. Reeve, C. D., and M. J. Rycroft (1972), The eclipsed lower ionosphere as investigated by natural very low frequency radio signals, J. Atmos. Terr. Phys., 34, 667 672, doi:10.1016/0021-9169(72)90154-7. Sen Gupta, A. S., G. K. Goel, and B. S. Mathur (1980), Effect of the 16 February 1980 solar eclipse on VLF propagation, J. Atmos. Terr. Phys., 42, 907 909, doi:10.1016/0021-9169(80)90107-5. Sheehan, C. H., and J. P. St.-Maurice (2004), Dissociative recombination of N 2 +,O 2 +, and NO + : Rate coefficients for ground state and vibrationally excited ions, J. Geophys. Res., 109, A03302, doi:10.1029/2003ja010132. Singh, R., B. Veenadhari, A. K. Maurya, M. B. Cohen, S. Kumar, R. Selvakumaran, P. Pant, A. K. Singh, and U. S. Inan (2011), D-region ionosphere response to the total solar eclipse of 22 July 2009 deduced from ELF-VLF tweek observations in the Indian sector, J. Geophys. Res., 116, A10301, doi:10.1029/2011ja016641. Tohmatsu, T. (1990), Compendium of Aeronomy, Terra Sci., Tokyo. Ulwick, J. C. (1972), Eclipse rocket measurements of charged particle concentrations, J. Atmos. Terr. Phys., 34, 659 665, doi:10.1016/0021-9169(72)90153-5. Wait, J. R., and K. P. Spies (1964), Characteristics of the Earth-ionosphere waveguide for VLF radio waves, technical note, 300 pp., Nat. Bur. of Stand., Gaithersburg, Md. 9of9