3.1. Time Domain
 Statistical analysis of the ELF signal in the time domain reveals that the magnetic field data follow the normal distribution very well, between 1% and 99%. Figure 2a shows the distribution of ELF amplitudes for the Hns component (Figure 1a). In blue are the measured counts for specific magnetic field amplitude, while in red is the normal (Gaussian) distribution fit where its parameters μ (mean) and σ (variance) were estimated on the basis of the observed data. Figure 2b shows a normal probability plot. The plot has the sample data displayed with pluses. On the plot is superimposed a line joining the first and third quartiles of the data. This line is extrapolated out to the ends of the sample to help evaluate the linearity of the data. For a normal distribution the plot will appear linear. The curvature introduced at the extremes shows that the most intense events deviate from the normal distribution.
Figure 2. (a) Histogram of Hns in blue and the normal distribution (μ and σ were estimated) in red. (b) Normal probability plot.
Download figure to PowerPoint
 To further study only the intense ELF transients (Q bursts), with values 5σ above the background noise (∼15 pT), thousands of files were analyzed (hundreds of hours of raw data). It was found that the number of intense events obeys the Poisson distribution (Figure 3), thus the parameter of the Poisson distribution λ was estimated for every 5-min file. The λ parameter is the mean number of intense events that arrive in a 5-min interval and enables us to estimate how often such event occurs (Figure 3a, blue). To examine the validity of the Poisson distribution a random vector was generated (named “Theory”) and plotted in red.
Figure 3. (a) Distribution of the number of intense events per file for Hns in blue and a random vector distributed by estimated Poisson parameter λ in red. (b) Quantile-quantile plot of measured and theoretical vectors.
Download figure to PowerPoint
 Figure 3b shows the sample quantiles of intense events versus theoretical quantiles from the Poisson distribution (blue versus red data points in Figure 3a). Since the plot is almost linear, our assumption regarding the distribution is justified. The λ parameter was estimated for every file with 99.9% confidence intervals in order to evaluate the goodness of the fit. The mean and variance was also evaluated for each data file and found to be highly correlated (for an ideal Poisson distribution the mean and variance are equal). Files with more than 60 intense events occurred when there were high levels of local man-made noise or close thunderstorms, which leads to the excursion from the Poisson distribution.
 For each 5-min file we have estimated the normal distribution parameter σ and calculated the number of intense events per file. Data files were analyzed every hour from February until October 2004 and a total number of 4320 files were processed. Hence, for every hour, some 180 files were analyzed to obtain the diurnal behavior of these parameters. Figure 4 presents hourly averaged diurnal variations in σ and the number of transients, for both Hns and Hew.
 Each horizontal magnetic field sensor is sensitive to waves arriving from directions normal to its axis. Therefore, because of the position of the observation station in Mitzpe-Ramon (30°N, 34°E), the major regions of thunderstorm activities are well separated (see Figure 5). Furthermore, waves from Southeast Asia arrive from the east and are detected primarily by the north-south magnetic induction coil, with maximum activity around 0800 UT. From the Americas the EM waves come primarily from the west, with maximum activity around 2000 UT. Waves from Africa arrive primarily from the south and therefore are detected by the east-west magnetic induction coil, with the maximum activity around 1400 UT.
Figure 5. ELF observation station in Mitzpe-Ramon (MR) and the general direction to the main global lightning activity centers: Southeast Asia, central Africa, North America, and South America.
Download figure to PowerPoint
 The normal distribution parameter σ is determined by the variance of the background noise level: increased lightning activity leads to increased values of the σ parameter. In Figure 4a, the diurnal change of the σ parameter for Hns shows three maxima at 0800–0900, 1200–1300 and 1700–1800 UT which marked with arrows, representing the three major “hot spots.” The African chimney is less obvious since the north-south coil is sensitive mainly to the other two regions. The Asian chimney looks more powerful than the American chimney, although only slightly larger.
 Figure 4b presents the diurnal change in the number of intense events above 15 pT. In Figure 4b the Asian center dominates the number of ELF transients, with Africa following, and the Americas in third place. The mean number of intense events arrived to the Hns sensor every 5 min during 0800 UT is ∼32 while during 2000 UT is ∼24. Since the north-south coil is sensitive to pulses arriving from Asia and America we can deduce that an intense event occur every ∼10 and ∼12 s, respectively and disregard from the number for Africa during 1400 UT since this coil is not sensitive to pulses arrive almost in parallel to its axis. The values of σ show moderate variability compared to the number of intense events. While the values of σ at 0800 UT are 1% greater than its values at 2000 UT, the number of intense events is more prominent by 30%.
 In Figures 4c and 4d, the diurnal change of the σ parameter and number of intense events above 15 pT are given for the east-west magnetic component (Hew). There is one clear maximum at 1300–1400 UT due to the sensitivity of the east-west sensor to signals arriving from Africa. The mean number of intense events arriving from Africa to the Hew sensor every 5 min around 1400 UT is ∼44, so we can deduce that an intense event occur every ∼7 s. Here we disregard the number for Asia and America since this coil is not sensitive to pulses arrived from its directions.
 Once more the values of σ show moderate variability compared to the number of intense events. While the values of σ at 1400 UT are 30% greater than its values at other hours during the day, the number of intense events at 1400 UT is 230% greater than at other times of the day.
 The source current amplitudes are frequently represented by lognormal distributions. The probability that the peak current of a return stroke exceeds the median current is given by [Galejs, 1972]
where m is the median value of log . The physical meaning of our results is that the tail of the distribution is more sensitive to the global lightning activity than the mean values. While variations during the day are minor for the common amplitudes which occur often, the number of intense discharges which occur rarely changes significantly. We should also mention that the low sampling frequency does not allow us to distinguish between every two adjacent lightning discharges since its mean duration is about 10 ms (assuming uniform global rate of 100 flashes per second) and our system time period is 25 ms (limited by instrumental bandwidth 40Hz), resulting in interference between numerous EM waves arriving simultaneously.
3.2. Frequency Domain
 Figure 6 presents the observed dynamic spectrum in the ELF band below 30 Hz, showing the diurnal changes of the Schumann resonance modes at 8, 14 and 20 Hz, both for Hns and Hew. In Figure 7 are plotted the diurnal behavior of each mode shown in Figure 6. In Figure 7a, the fundamental SR mode at 8Hz for Hns shows two prominent peaks at 0800 and 2000 UT produced by lightning in Southeast Asia and the Americas, respectively. The Asian chimney appears to be more powerful than the American chimney. The maximum is wider at 0800 UT than at 1400 UT, implying that the quality factor of the Earth-ionosphere cavity at 0800 UT is lower than at 2000 UT. In Figure 7b (Hns second mode) the most powerful peak is at 1400 UT relative to the two other peaks. Although the north-south induction coil is less sensitive to signals arriving from the direction closely normal to its axis (Africa), the modal structure of the waveguide attenuates the second SR mode at 14Hz when the source region and the observation station are separated by 10,000 km. Figure 7c presents the diurnal change of the third SR mode. Here too, Africa dominates the diurnal cycle of the Hns component. In Figures 7d–7f we present the three SR modes for Hew. The most active thunderstorm center on the globe can clearly be seen at 1400 UT because of its natural dominance, together with the sensitivity of the east-west induction coil to signals arriving from Africa.
Figure 7. Diurnal changes of Schumann resonance modes based on ELF measurements in Israel for Hns at (a) 8 Hz, (b) 14 Hz, and (c) 20 Hz and for Hew at (d) 8 Hz, (e) 14 Hz, and (f) 20 Hz.
Download figure to PowerPoint