High time resolution Schumann resonance (SR) records are analyzed at a midlatitude (Nagycenk, 47.6°N, 16.7°E, Hungary) and a north polar (Hornsund, 77°N, 15.5°E, Spitsbergen) station from the point of view of the day-night asymmetry of the Earth-ionosphere cavity. The vertical electric field component, EZ, at Nagycenk in quasi-minute time resolution exhibits jump-like increases of SR amplitudes between the local ionospheric and surface sunrise times and sharp decreases between the local surface and ionospheric sunset times. These amplitude variations depend on frequency, increase with increasing mode number, and occur simultaneously in the three SR modes studied here. The duration of the sharp frequency-dependent amplitude changes is generally less than 30 min. The accurate timing (“clock-like accuracy”) of these sharp SR amplitude variations of about 12–25% in the local sunrise/sunset periods and their frequency dependence make these changes distinguishable from the amplitude variations related to the lightning source properties and strongly suggest an ionospheric origin for these sharp amplitude variations. The signature of the day-night asymmetry of the Earth-ionosphere cavity can also be found at Hornsund in the two short spring and autumn periods with alternating day and night periods every day, in the form of an enhanced day-night contrast of the SR amplitudes with consistent frequency dependence for the first three SR modes.
 Schumann resonances are electromagnetic oscillations of the Earth-ionosphere cavity [Schumann, 1952]. They are excited by lightning strokes and maintained by worldwide thunderstorm activity. In this way, the global lightning activity can be monitored by measuring SR intensities, as numerous studies have shown [Raemer, 1961; Balser and Wagner, 1962; Pierce, 1963; Galejs, 1972; Ogawa et al., 1969; Polk, 1969; Bliokh et al., 1980; Heckman et al., 1998; Nickolaenko et al., 1998]. According to the SR theory [Wait, 1996], the SR intensities at a given location depend not only on the lightning source parameters (current moments) but also on the cavity properties (ionospheric D region conditions) as well as the source-observer angular distance. All these parameters have temporal variations. To deduce intrinsic lightning activity from SR records or even to monitor planetary temperature [Williams, 1992] and atmospheric water vapor [Price, 2000], the SR intensity variations stemming from changing upper boundary conditions in the cavity and the source-observer distance (SOD) need to be carefully considered [Williams and Sátori, 2004].
 The day-night change of the conductivity profile is the well-known nonuniformity of the Earth-ionosphere cavity [Nickolaenko and Hayakawa, 2002; Smith et al., 2004]. In previous theoretical and observational work, no general consensus has developed yet in the SR community on the influence of the day-night asymmetry on SR parameters, and its distinction from a proximity effect of the lightning source [Cavazos et al., 1996]. Indeed, there is a difficulty in separating two superimposed effects, both caused by the Sun: (1) the enhancement of SR intensity in the thinner, daylight side of the cavity [Sentman and Fraser, 1991], and (2) the enhancement in intensity on the daylight side because of the destabilizing effects of surface heating by sunlight on lightning activity [Cavazos et al, 1996]. Earlier observations drew attention to this conundrum without addressing it explicitly. For example, Keefe et al.  found better agreement between the SR intensity records measured simultaneously in two distant stations if they were presented in local time. Sentman and Fraser  focused on the day-night asymmetry and deduced an average local time variation of the height of the ionospheric D region from cumulative SR intensities recorded simultaneously at two widely separated stations. Their approach, though targeting the day-night asymmetry of the Earth-ionosphere cavity, really involved both effects simultaneously. Pechony and Price  claim that such kind of SR intensity variation can be produced in a uniform cavity, too, due to changing source-receiver geometry in case of two stations. Nickolaenko and Hayakawa  concluded theoretically that the amplitude variations to be expected at the sr/ss terminator line are practically undetectable. Pechony et al.  inferred that the observed SR field variations are governed primarily by the variations in the source intensity and source-receiver geometry and that the effect of the day-night asymmetry in the ionosphere is secondary. Heckman  reported on distinct SR intensity variations of ionospheric origin observed at local sunrise/sunset in SR records at Rhode Island with 12-min time resolution. Melnikov et al.  presented observational evidence for the sr/ss terminator effect based on multistation SR observations with hourly time resolution. Steep increase/decrease of SR amplitudes appeared in the local sr/ss hours. Even the small differences between the lengths of day during the year at two stations with different latitudes were identified by the amplitude variations. A steep increase/decrease of SR amplitudes has also been observed at local sr/ss for the first four modes of EZ, at Modra, Slovakia [Ondrášková et al., 2007]. Both Melnikov et al.  and Ondrášková et al.  have found that the day-night contrast of the SR amplitudes adjacent to the local terminator lines are more pronounced in higher-order SR modes and more pronounced in the morning than in the evening, and stronger in winter than in summer. An approximate 20–25% day-to-night increase in the lower characteristic altitude (the lower of two heights of maximum dissipation in the ionospheric D region) was shown within the lower SR frequency range both theoretically (on the basis of the parameter's physical model and representative daytime and nighttime conductivity profiles for the lower ionospheric D region), and experimentally (from systematic observations of the background electromagnetic signal in the SR frequency range [Greifinger et al., 2005]). Three-dimensional finite difference time domain modeling of SR parameters by Yang and Pasko  describes jump-like Er field variations at the sr/ss terminator lines attributed to the changes in ionospheric height between daytime and nighttime.
 The purpose of this paper is to provide new observations that enable quantification of the day-night asymmetry of the Earth-ionosphere cavity using SR data of quasi-minute time resolution. It is hoped that this work will motivate further theoretical studies.
 The vertical electric and horizontal magnetic field components have been measured in the SR frequency range using a ball antenna and two induction coils at Nagycenk (NCK), Hungary [Sátori et al., 1996] and Hornsund (HRN), Spitsbergen [Neska and Sátori, 2006] and in West Greenwich, Rhode Island, USA [Heckman et al., 1998]. This study is based primarily on the long-term data set (from May 1993 to February 2006) of the vertical electric field component at NCK and the HNS and HEW magnetic field components measured at HRN from September 2004 to September 2005. The HRN station lies above the Arctic circle, and thereby provides special (full-day and full-night, as well as mixed) conditions from the point of view of the day-night asymmetry. Segments of the ELF field record in the SR frequency band, each 40 s long, are processed with the spectral technique of complex demodulation [Sátori et al., 1996]. Sampling and processing alternate each other. This method makes possible about 85 estimates of SR peak frequencies and amplitudes for the first three SR modes in both field components in an hour. A compromise had to be found between the stabilization of the spectral estimates and a time resolution sufficiently short to define the sharp terminator transitions. A smoothing process was applied here to stabilize the spectral estimates. The averages of spectral amplitudes were computed by complex demodulation [Sátori et al., 1996] from SR records in five consecutive time windows (of 40 s), with a repetition of this process, with time windows shifted one by one. This means that every sixth SR amplitude can be considered an independent one and the effective time resolution becomes about 3 min.
 Data are shown in universal time (UT) that differs by only one hour from the local time (LT) at both NCK and HRN. This presentation form allows the possibility to point to both local time and universal time-dependent variations on the same plots.
 High time resolution SR amplitude records have been analyzed both for individual days and as an average of numerous days focusing on the local sunrise and sunset periods at NCK. Figure 1 shows the relative diurnal amplitude variations of the EZ field component for the first three modes with respect to the local midnight levels recorded on 10 January 1996. Amplitude fluctuations occur throughout the day but it is clearly seen that frequency-dependent jump-like amplitude variations increasing with increasing mode number are built in to the diurnal records simultaneously in the three modes between the local ionopsheric and surface sunrise (sri and srs), and they disappear between the times of local surface and ionospheric sunset (sss and ssi). The vertical solid lines indicate the proper local ionospheric and surface sunrise/sunset times. The local ionospheric sunrise at about 100 km height precedes the local surface sunrise by about 45 min, and the local ionospheric sunset at about 100 km height follows the local surface sunset also by about 45 min. This 45 min time difference is considered as an average and is reasonable time interval from the point of view of this study. The percentage variations indicated in Figure 1 correspond to the relative amplitude variations between the ionospheric and surface sunrises as well as between the local surface and ionospheric sunset times.
Figure 2 shows two close-ups of the amplitude variations of the third SR mode for the sunrise period on 8 August 2005 (Figure 2a) and for the sunset interval on 14 October 2005 (Figure 2b) to illustrate how the relative (percentage) amplitude variations are computed in the sr/ss intervals. The mean SR amplitudes are determined for the 20-min interval prior to ionospheric sunrise (time point defined above) and for 20-min period centered on the surface sunrise. The procedure is similar for sunset but the 20-min interval just after ionospheric sunset is considered. Two additional dashed time markers are shown in Figure 2. They show that the steepest part of the amplitude increase ends well before (by about 15–20 min) the local surface sunrise and the steepest part of the amplitude decrease starts later (again by about 15–20 min) than the local surface sunset. In general, the duration of the sharp frequency-dependent amplitude increase/decrease is confined to an interval of less than 30 min in the sunrise/sunset periods of 45 min marked off by the local surface and ionospheric sunrise/sunset times at ∼100 km height.
 The result of a statistical analysis of the relative SR amplitude variations is shown in Figure 3 on the basis of SR amplitude records with high time resolution observed at NCK for 120 days, mainly in 2005. The relative variations of the mean SR amplitudes between sri and srs, as well as betweeen sss and ssi, were computed for the first three modes in the case of 120 sunrise and 120 sunset periods. Then their averages were determined for each mode both at sr and ss. The mean relative amplitude variations between sri and srs, as well as between sss and ssi, increase with increasing frequency from 14% to 25% (between the first and third modes) and are slightly more pronounced at sunrise. A careful data selection from the standpoint of local weather conditions and/or other local disturbances necessarily preceded this statistical analysis. SR amplitudes with hourly time resolution were looked through without any “a priori” assumption for sr/ss effects in the high time resolution amplitudes. An equal number of days (10) was chosen from each month to avoid the oversampling of any month or season. At first, the statistical analysis above had been performed for two independent sub–data sets (two sets of 5 different days from each month) and they resulted in similar frequency-dependent amplitude variations as shown in Figure 3. The two independent but integrated data sets give assurance that the results in Figure 3 are representative.
 The slight departure between the sunrise and sunset values might be attributed to the different ionospheric chemistry at sr and ss. The ionization process is faster at ionospheric sunrise than the neutralization at ionospheric sunset in the ionospheric D region [Hargreaves, 1992]. The local ionospheric sunset and the diminishing African lightning activity in local afternoon (UT = LT − 1) very often overlap during the year at NCK. The superposition of these two effects adds to the difficulty in separating the amplitude variations of ionospheric origin at sunset from the amplitude variations due to the diminishing African lightning activity.
 The “clock-like accuracy” of the jump-like increase/decrease of SR amplitudes recorded at NCK in sr/ss periods is demonstrated for the third mode on three different days (2, 16, and 27 February 2006) in Figure 4. The top group of arrows indicates the surface sunrise (srs) and sunset (sss) times and the bottom group of arrows marks the ionospheric sunrise (sri) and sunset (ssi) times, for the corresponding days. It is seen that the more rapid transitions are more closely timed with the sunrise and sunset times at the ionosphere, in comparison with like times at the Earth's surface. The time differences between the local surface sunrises on 2 February (0619 UT) and 16 (0558 UT) as well as on 16 February (0558 UT) and 27 February (0538 UT) are 21 and 20 min, respectively. The time differences between the local surface sunsets on 2 February (1556 UT) and 16 February (1618 UT) as well as on 16 February (1618 UT) and 27 February (1635 UT) are 22 and 17 min, respectively. These time differences can be identified with some minute accuracy in the time sequences of the jump-like amplitude increases between the ionospheric and surface sunrises and jump-like amplitude decreases at the surface and ionospheric sunsets on the corresponding February days. The evolution in the behavior of sunrise/sunset times shown in Figure 4 as one progresses through the month of February is exactly what one expects on the basis of the changing length of day for the same time period. A meteorological explanation for such an accurate timing of consistent amplitude variations both at sunrise and sunset is difficult to construct. The tropical thunderstorm sources are well recognized not to turn on at sunrise, in general.
Figure 5a shows normalized SR amplitude records at sunrise hours on one winter day (1 January 2005) and on one summer day (23 June 2005). The short-dashed and long-dashed lines delimit the steepest phase of the amplitude increase in the sunrise intervals marked off by the local ionospheric and surface sunrise times (sri and srs). In the case of 1 January 2005 the short-dashed line coincides with the marker for sri. What is noteworthy here, and in general has been observed in other individual days and near sunset, too, is that the short-term (∼ minutes) amplitude fluctuations, characteristic of SR records with such high time resolution, are decreasing systematically during the steepest phase of the amplitude variations lasting about 20–30 min. This can be seen more clearly if the linear trends are removed from the sections of I, II, III as presented in Figure 5b. Then the variance of amplitudes is computed for the three sections. Figure 5c shows that the SR amplitudes have the smallest variance in the interval II. The length of the intervals I and III was arbitrarily chosen here, but it seems suitable to consider three quasi-equal time intervals of about 30–45 min for this analysis. This finding might be important for future work and may require additional research and closer attention from the ELF theoreticians and interpreters.
Figure 6 presents diurnal SR amplitude variations lined up from the first day to the last day of the year for (1) HRN (using observations from both 2004 and 2005), and for (2) NCK (1995). The x axis shows the day and the y axis the hours in UT (UT = LT − 1) at both stations. The color scale indicates the SR amplitudes. The HNS and HEW magnetic field components were selected at HRN and EZ for NCK as they are the records of highest quality at the two sites. The dashed and solid lines indicate the “full-night” as well as the “full-day” parts of the year at HRN in Figure 6a. The two observatories lie almost on the same longitude and also share the longitude of the center of the African lightning source region. This geometry means that the HNS field component is responsive to the lightning activity in the Americas and in Asia/maritime continent, and the HEW field component is responsive to the African lightning activity at HRN. In contrast, the EZ field component is isotropic in its response to lightning activity. SR amplitude records at both stations indicate the three main thunderstorm regions with maxima near 1000 UT for the Asia/maritime continent and near 1600 UT for Africa and between 2100 UT and 2300 UT in the case of the Americas. Noteworthy here is that the lightning centers in the Americas and in Asia/maritime continent are almost at equal angular distance from HRN and NCK. The HNS field component at HRN indicates the lightning activity in the Americas and Asia/maritime continent with similar strength depending on the seasons (Figure 6a). However, from September to December, America becomes the more pronounced lightning source. According to classical thunder day analysis [Whipple, 1929] and the OTD/LIS satellite observations [Christian et al., 2003], in general, the tropical lightning activity is stronger in South America than in the maritime continent. The great circle path from HRN to the maritime continent crosses more land (Eurasia) than the path to South America across the Atlantic Ocean. Consequently the electromagnetic signals coming from the dominantly land-related lightning activity along the Eurasian part of the great circle path can increase SR amplitudes/intensities from the direction of the maritime continent depending on the season even if the tropical lightning activity is less intense in the maritime continent than in South America.
 In case of the EZ field component at NCK, as shown in Figure 6b, there is a large contrast between the SR amplitudes measured at the time of the maximum lightning activity of Asia/maritime continent near 1000 UT and in the maximum lightning activity in the Americas in the late evening between 2100 UT and 2300 UT. The latter and very important American lightning region is very weakly manifest in the annual records (see Figure 6b), and it cannot be explained with the different land/ocean ratio along the great circle path in the direction of Asia/maritime continent and of Americas, as these conditions are the same for HRN where this large contrast in the SR records for these two lightning regions cannot be seen. There is a basic difference in the local conditions for the HRN and NCK observers. The HRN observer detects both the Asian/maritime continent and the American lightning sources under purely local nighttime, or purely local daytime conditions, during most of the year. In contrast, the maximum activity of the Asia/maritime continent lightning region is always observed under local daytime conditions at NCK, while the maximum activity of the American lightning source is always detected under local nighttime conditions at NCK, every day of the year. The SR amplitude records at NCK (see Figure 6b) exhibit a remarkable and regular variation, namely a sharp change of amplitudes at the sr/ss terminator lines indicating the different length of the local day over the year, as already shown in an eight-year integration of SR amplitude records at NCK by Melnikov et al. . These observations at NCK support an additional effect of ionospheric origin on SR amplitudes attributed to the day-night asymmetry of the Earth-ionosphere cavity. The missing terminator-related amplitude variation at HRN seems to be plausible, as the polar station is under daytime condition for about four months and under nighttime condition for another four-month period. There are only two shorter two-month intervals in spring and autumn when daytime and nighttime transitions occur every day, but with elongated sr/ss periods associated with the low elevation of the Sun.
 It has once again to be mentioned that the EZ field component is responsive to lightning activity from all directions. Nickolaenko et al.  have modeled the north-south annual migration of the three main lightning source regions using a source model based on OTD/LIS satellite observation and a uniform ionosphere model. The modeled SR amplitudes exhibit a similar but much less regular “eye shape” form in UT time, as compared to the very regular “eye shape” form in SR amplitudes in LT time marked off clearly by the local terminator lines in Figure 6b.
 The analysis of the HRN observations is more subtle. SR amplitude variations due to the day-night asymmetry of the cavity can be expected in the polar records at HRN in the two transition seasons. Four periods, each 15 days long, were selected from the HEW record from the full-day (17–31 May 2005), full-night (2–9 and 23–29 January 2005) and the two transition seasons (16–30 September 2004 and 17–31 March 2005). The reason for the selection of the HEW field component is as follows: HEW is responsive to the African lightning activity which has maximum intensity near 1600 UT when HRN is under daytime condition in the selected time intervals for the two transition seasons. The preconception here was that the daytime amplitudes must preserve the amplitude increase built in at local ionospheric sunrise even if it is rather elongated in time at polar latitudes. Consequently the SR amplitudes are expected with higher relative importance at around 1600 UT in the two transition seasons compared to the “full-day” and “full-night” periods. Figure 7 shows the normalized mean diurnal SR amplitude variations in the four selected periods for each mode. The mean SR amplitude at the local midnight hour (2300–2400 UT corresponds to 0000–0100 LT at NCK) was chosen as the reference level for each SR mode. The relative importance of SR amplitudes has remarkably increased at around 1600 UT in the transition periods (17–31 March 2005 and 16–30 September 2004) when HRN is under mixed (daytime/nighttime) conditions. This increased day-night contrast of SR amplitudes can hardly be explained by the increased contrast of the daytime and nighttime lightning activity of the African thunder region in the March and September days studied here. More lightning can be expected in both daytime and nighttime hours in the hotter periods. The semiannual variation in tropical surface temperature can cause increased lightning activity in Africa with April and October maxima [Williams, 1994; Sátori and Zieger, 1996; Füllekrug and Fraser-Smith, 1997], but it does not significantly modify the day-night contrast of the lightning activity, as was checked by comparing the day-night contrast of the amplitudes of the HEW field component in the third part of October and in a January period of 10 days when already “full-night” conditions are present at HRN, and consequently the local ionospheric day-night transitions had ceased. The increased day-night contrast of the African peak activity in the two short transition seasons cannot be explained by the north-south migration of the African lightning source with respect to HRN, as it can hardly be expected that the peak activity of Africa and its nighttime activity have different north-south migration characteristics. The odd magnetic resonant modes (first and third here) have maximum amplitudes at the polar station (HRN) at about 90° angular distance from the African lightning center, and they are more pronounced in a relative sense here and in an absolute sense, too, than the even ones (second mode here) as it has a minimum amplitude at 90° angular distance. It should also be mentioned that the May (day-only) period was selected instead of a real summer month (also day-only) as the local artificial electromagnetic noise strongly increases because of anthropogenic activity at the Polish Polar Station at HRN during the summer months.
 In Figure 8, the average of normalized SR amplitudes computed from the two periods with day-night transitions (March, September) at 1400–1600 UT is compared to the average of normalized SR amplitudes determined from two other periods without day-night transitions (January, May) at 1400–1600 UT, for each mode separately. The relative importance of SR amplitudes is higher in the transition periods (March, September) and increases with increasing mode number. The third SR mode exhibits the largest relative amplitude variation. The frequency dependence of these SR amplitude ratios is in accordance with the frequency dependence of SR amplitudes in the ionospheric sr/ss intervals observed in EZ at the midlatitude station NCK.
Figure 9a focuses on the SR amplitude variations of the first three modes recorded with high time resolution at NCK in the sunrise period of 5 February 2005. Figure 9b shows the cumulative SR intensities of the first three modes in the first part of the same February day with hourly time resolution at NCK and HRN. The average of SR amplitudes recorded in the 20-min interval just prior to the ionospheric sunrise was considered as a reference level for each mode in Figure 9a, as was already illustrated in Figure 2. It can be clearly seen in Figure 9a that the frequency-dependent jump-like amplitude increases are intrinsic to the amplitude records between the times of ionospheric (sri) and surface sunrise (srs) as in Figure 1. The sharp amplitude increase is the highest for the third mode, but a pronounced increase can be seen even in the case of the first SR mode.
 The hourly averages of the cumulative SR intensity for the first three modes (summation of the squared amplitudes: EZ2 for NCK and Hϕ2 = HNS2 + HEW2 for HRN) were computed from the unnormalized SR amplitude records in the first half day of 5 February 2005 (Figure 9b). The purpose is to show how the amplitude variations of ionospheric origin exhibited at NCK (Figure 9a) can alias SR intensity variations attributed to changes of lightning activity. The hourly averages were normalized for each station by the hourly average of the cumulative intensity in the hour of the world lightning minimum at 0200–0300 UT, as presented by the thin solid line for NCK and by the thick solid line for HRN in Figure 9b. It can be seen that the cumulative intensities indicate the maximum lightning activity in the maritime continent (MC) at 0900–1000 UT in both the NCK and HRN records, but a remarkable difference can be seen in their magnitudes relative to their own reference levels. This difference cannot be explained by the different field components (EZ for NCK and Hϕ for HRN). In general, the use of cumulative intensity is to minimize the angular distance dependence of the different modes and field components. The MC is almost the same angular distance (about 90°) from NCK and HRN, only the plane of the great circle path is more tilted in the case of NCK but both great circle paths cross the Eurasian land region from the NCK and HRN observers to the MC. Considering both the cumulative intensity and the normalization by the cumulative intensity in the hour of global lightning minimum, the normalized cumulative intensities at NCK and HRN should exhibit more similarity in a quantitative sense, too, on the same February day.
 There is a basic difference between the local observing conditions at NCK and HRN in February. The sunrise terminator crosses the NCK observation site between 0500–0600 UT in 5 February 2005 while HRN is under full-night conditions on that day. The percentage amplitude variations between the level just before ionospheric sunrise and the level near surface sunrise shown in Figure 9a are as follows: 25% for the first mode, 31% for the second mode and 35% for the third mode. It can be clearly seen here that the frequency-dependent amplitude increases attributed to ionospheric origin are superimposed on the more gentle amplitude increases attributed to the lightning activity switching on in the MC region also in 0500–0600 UT. The frequency-dependent amplitude increase (higher than the annual average shown in Figure 3) starts only at 0540 UT, some 10 min later than the ionospheric sunrise at ∼ 100 km height. The percentage amplitude variations between the amplitude levels at 0540 UT and at the surface sunrise are as follows: 19% for the first mode, 24% for the second mode and 29% for the third mode. These percentage variations were considered to remove the amplitude variations attributed to ionospheric origin from the high time resolution records at NCK in 5 February 2005. The removal process means that the amplitudes measured after the jump-like amplitude increases are shifted downward and are fit to the amplitudes recorded just before the jump-like amplitude increases to approach the “full-night” condition at NCK. The daytime amplitudes of high time resolution (starting by about 10 min earlier than the local surface sunrise) were decreased with the amplitude values corresponding to the percentage variations given above for each mode in the sunrise interval. After the correction process, the hourly averages of the SR amplitudes were recomputed for each mode and then the squared amplitudes (SR intensity) were summed for the first three modes in the first half day of 5 February 2005 using again the cumulative intensity at 0200–0300 UT as a normalization value. The corrected cumulative SR intensity for NCK is presented by the dash-dotted line in Figure 9b. The corrected and renormalized SR intensities at NCK are similar with the cumulative intensities at HRN, even in a quantitative sense. It can be seen that the original and corrected cumulative SR intensities at NCK pull apart well between 0500 UT and 0600 UT in the hour of ionospheric sunrise, and considerable differences in intensity remain in the later hours. In case of the original record in hourly time resolution at NCK in Figure 9b, the very steep increase of the cumulative SR intensity in the sunrise hour can be the only signature of the ionospheric effect built in to the SR record at local ionospheric sunrise at NCK. Noteworthy here is that the use of uncorrected SR data with hourly time resolution could lead to either an overestimation, underestimation or even a total neglect of SR amplitude/intensity variations of ionospheric origin. Earlier study has shown that if corrections for the day-night asymmetry in the Earth-ionosphere cavity are not implemented, then the correct ordering of the intensities of the three major source regions will not be achieved [Williams and Sátori, 2004].
 The mean diurnal cumulative SR intensity in hourly time resolution for the first three modes was considered for the EZ field component at NCK and for the HNS component in Rhode Island (RI) for the same January days (48) in four consecutive years (1994–1997) after a careful data selection from the point of view of local disturbances at both stations. Figure 10a shows that when the afternoon African lightning is maximum between 1500 UT and 1600 UT, RI is on the sunlit side of the Earth, while NCK is close to the evening terminator. The HNS field at RI is responsive to the African lightning activity because Africa is due east of the Rhode Island station along a great circle path. The peak activity of Africa is indicated at 1500–1600 UT (dashed arrow) while the EZ intensity at NCK shows it at 1400–1500 UT (solid arrow), as shown in Figure 10b. This presents an apparent contradiction, as the mean SR intensities should follow the peak lightning activity in Africa recorded on the same days at both stations. When the nighttime SR intensity in EZ at NCK is adjusted to the daytime record (dotted parts in Figure 10c) by removing the jump-like amplitude increases/decreases at sr/ss for each of the 48 January days using the high time resolution amplitude data at NCK, and then recomputing the average of the cumulative intensities, the time of the African peak activity coincides for both stations, as indicated by the common dash-dotted arrow. The surface sunset time at NCK on 1 January is 1515 UT and on 31 January is 1552 UT. The ionospheric sunset starts later by about 10–15 min. Consequently, both the daytime and nighttime conditions are mixed at 1500–1600 UT and even in the following hour for the selected January days. The correction process resulted in a 12% increase in the average of the cumulative intensities at 1500–1600 UT, 36% in 1600–1700 UT, 30% in 0500–0600 UT and 38% for the night hours. The longitudinal position of the African lightning center is very stable in January months on the basis of space-based observations with OTD/LIS (http://thunder.msfc.nasa.gov/research.html). So the one hour difference between the times of the African peak activity observed at NCK and RI (Figure 10b) cannot be explained by a systematic drift of the African lightning center with respect to either of the two observers, or by a systematic effect of local lightning activity on SR records at NCK or RI in January months.
 Two mechanisms, both involving the physical effects of solar radiation, are operating near solar terminator times to influence SR amplitude/intensity. Toward clarifying the manifestations of these two mechanisms in the observations reported here, it is important to distinguish other aspects of these two superimposed mechanisms. The effect of solar Lyman α radiation in causing lower D region ionization is essentially instantaneous and switches on at ionospheric sunrise and off at ionospheric sunset. Even the ion-chemical processes at sunrise are very fast near the lower characteristic height in the ionospheric D region because of the “photo-detachment of electrons from negative ions (accumulated during nighttime) by visible light” [Kazil, 2002, pp. 65–69]. This ion-chemical process takes about 10–30 min and massively increases the electron density at sunrise (private communication, M. Friedrich, 2006). Therefore it can account for the “clock-like accuracy” of the amplitude changes documented in Figures 1, 2, 4, 5a, and 9a in sharply defined sr/ss intervals delimited by ionospheric and surface sunrise as well as by surface and ionospheric sunset times. The frequency dependence of the amplitude variations at ionospheric sr/ss documented in Figure 3 is attributed to the frequency dependence of the lower characteristic altitude of the ionospheric D region for ELF waves [Greifinger et al., 2005]. The lower characteristic altitude increases with frequency and the contrast in daytime and nighttime height also increases with frequency because the contrast in the daytime and nighttime conductivity is increasing with altitude. On the basis of arguments pertaining to the conservation of Poynting flux in the SR waveguide [Sentman and Fraser, 1991] across the day-night boundary, the SR amplitudes should exhibit similar frequency-dependent responses to the variations in the lower characteristic altitude during sr and ss intervals.
 In contrast, the response of thunderstorm convection to sunlight is substantially less clock-like, and furthermore, is decidedly asymmetrical relatively to local noon. At sunrise, shallow convection can begin abruptly but not the kind of deep convection that drives SR with lightning activity. Thunderstorm convection generally does not begin until early afternoon (many hours after sunrise), when the cumulative surface heating in the morning hours is sufficient to trigger thresholds in Convective Available Potential Energy [Williams and Renno, 1993]. The sharp sunrise transitions in Figures 1, 2, 4, 5a, and 9a do not seem to find easy explanation in the meteorology of thunderstorm convection.
 At sunset, the cessation of local lightning activity is substantially more abrupt than at sunrise [Williams and Heckman, 1993], but still lacks the clock-wise accuracy evident in the analyses presented here. Indeed, lightning can often proceed after sunset, drawing on the stored energy in the atmosphere from afternoon solar heating. This general behavior is not favorable for an abrupt change at sunset, and the convective situation does not reproduce itself so similarly from day to day as is the case for the solar D region ionization.
 While lightning activity is predominant in daytime when solar heating is available to destabilize the atmosphere, this activity is decidedly asymmetrical about local solar noon [Williams and Heckman, 1993]. This asymmetry is readily apparent in Figure 6b that shows this asymmetry in daytime intensity between the remarkably symmetrical sunrise and sunset boundaries. This well-defined asymmetrical feature can be attributed only to meteorology and not to the day-night asymmetry of the Earth-ionosphere cavity.
 In analysis by Sentman and Fraser , both the day-night ionospheric asymmetry and the source proximity effect are involved simultaneously in the diurnal fits, but are not distinguished. It is interesting there that the sinusoid of best fit is phase shifted from local solar noon into the afternoon side, thereby supporting the idea that more than the simple day-night asymmetry of the ionosphere was affecting the results.
 The quantification of the regional variation in global lightning activity from a single station clearly depends on both variations in the characteristic heights of the ionosphere and the SOD effect, with the former at least as important as the latter in typical situations. Any correction for SOD when comparing intensities of different lightning sources with different SOD based on model calculations of SR power in terms of geometric properties [Sentman, 1996; Nickolaenko et al., 1998] should be done by first addressing the influence of the day-night asymmetry and then adjusting for the source-observer geometry.
 The method of the correction applied here and suggested for removing the distortion effect of the day-night asymmetry of ionospheric origin has some limitations. The percentage amplitude variations resulted in each mode (see Figure 3) based on NCK records are slightly different for sunrise and sunset periods for the reasons described in the paragraph . The same percentage correction values characteristic for each mode need to be used for both sunrise and sunset to avoid introducing discrepancy at the midnight interface on consecutive days. However, it should be mentioned that these correction parameters can be deduced for individual days if the time resolution as well as the quality of the SR record make it possible to perform a more sophisticated correction procedure. The reason is that the amplitude changes of ionospheric origin at sr/ss may have systematic temporal variations of ionospheric origin.
 The true variations in global lightning activity are always superimposed on the ionospheric effect in SR amplitudes/intensities, even in the short sr/ss periods. Consequently, the correction process can exaggerate the difference between the corrected and uncorrected records by up to 5% estimated from the gentle amplitude variations just before and after sunrise, as well as before and after the sunset time intervals in similarly short time periods. So the lower percentage values presented for the sunset time in Figure 3 seem more reasonable for the ionospheric corrections. The latter values are well in accordance with the percentage variations deduced at HRN (see Figure 8). It should also be mentioned that the correction value of about 25% for the third mode was consistent both at sr and ss at NCK (Figure 3) as well as at HRN (Figure 8). The amplitude changes associated with the day-night asymmetry still remain dominant in the short ionospheric sr/ss intervals before the correction.
 High time resolution SR amplitudes at Nagycenk, Hungary exhibit sharp frequency- dependent increases between the local ionospheric and surface sunrise as well as sharp decreases between the local surface and ionospheric sunset.
 The sharp amplitude increase always precedes the local surface sunrise and follows the local surface sunset by about 15–20 min in the sr/ss intervals of about 45 min.
 The duration of the sharp, frequency-dependent amplitude variations is less then 30 min.
 The relative SR amplitude variations at ionospheric sr/ss depend on the frequency and increase with increasing mode number from approximately 12% to 25% for the first three SR modes (8 Hz, 14 Hz, 20 Hz), considering the percentage values both at NCK (Figure 3) and HRN (Figure 8). These relative SR amplitude variations cumulate a mean SR intensity variation of about 30–40%.
 The “clock-like accuracy” of the jump-like SR amplitude/intensity variations of ionospheric origin has been revealed on a timescale of minutes (Figures 2, 4, 5a, and 9a).
 SR observations with hourly time resolution do not allow such fine distinctions.
 In high time resolution SR data, the amplitude/intensity variations supposed to be of ionospheric origin are easily distinguishable from the changes attributable to the switching on/off effect of the Asian and African thunderstorm centers, especially in the sunrise intervals when systematic changes in thunderstorm centers are less likely.
 The signature of the day-night asymmetry of the Earth-ionosphere cavity can also be found at HRN in the two short spring and autumn periods with alternating day and night periods every day, in the form of a consistent ranking of the relative amplitude variations for the first three SR modes in two-week intervals with and without local ionospheric day/night transitions.
 The high time resolution SR records yield a possibility for monitoring changes of ionospheric origin in the ionospheric sr/ss intervals.
 The neglect of SR amplitude variations of about 12–25% supposed to be of ionospheric origin in the local sr/ss intervals can lead to a misrepresentation of the lightning source intensity variations deduced from cumulative intensities of the first three SR modes by about 30–40% if the observer and the lightning source are in different (sunlit or night) sides of the cavity.
 This study was supported by a NATO Collaborative Linkage grant EST.CLG.980431 and OTKA grant NI 61013 from the Hungarian Science Research Fund. K. Ábrahám assisted with the SR data processing. E. Williams' participation was supported by a grant from the U.S. National Science Foundation on Schumann resonances (ATM-0003346).