An enhancement of the ionospheric sporadic-e layer in response to negative polarity cloud-to-ground lightning

Transcription

1 GEOPHYSICAL RESEARCH LETTERS, VOL. 35, L05815, doi: /2007gl031909, 2008 An enhancement of the ionospheric sporadic-e layer in response to negative polarity cloud-to-ground lightning C. J. Davis 1 and Kin-Hing Lo 1 Received 2 September 2007; revised 5 February 2008; accepted 13 February 2008; published 14 March [1] Lightning data, collected using a Boltek Storm Tracker system installed at Chilton, UK, were used to investigate the mean response of the ionospheric sporadic-e layer to lightning strokes in a superposed epoch study. The lightning detector can discriminate between positive and negative lightning strokes and between cloud-to-ground (CG) and inter-cloud (IC) lightning. Superposed epoch studies carried out separately using these subsets of lightning strokes as trigger events have revealed that the dominant cause of the observed ionospheric enhancement in the Es layer is negative cloud-to-ground lightning. Citation: Davis, C. J., and K.-H. Lo (2008), An enhancement of the ionospheric sporadic-e layer in response to negative polarity cloud-toground lightning, Geophys. Res. Lett., 35, L05815, doi: / 2007GL Introduction [2] Previous studies of the influence of lightning on the ionospheric sporadic-e (Es) layer [Davis and Johnson, 2005; Johnson and Davis, 2006] revealed that the peak and background electron concentration were significantly enhanced in response to lightning. These studies used lightning data from the Arrival Time Difference (ATD) system of the UK Meteorological Office [Lee, 1989]. This European system can determine the position of each lightning stroke to within 5 km over the UK. It is most sensitive to cloud-to-ground (CG) lightning and information about the polarity of each stroke is not recorded. In order to investigate the effects of different lightning types on the ionosphere, a Boltek Storm Tracker ( [Kandalgaonkar et al., 2006] was installed at Chilton and has been collecting data since The Boltek system detects lightning by measuring low frequency (<500 khz) radio signals produced by lightning s electrical discharge. It uses a crossed-loop antenna to determine the direction of the lightning and the signal strength to calculate an approximate distance for the lightning. While less sophisticated than the ATD system, with a greater uncertainty in the exact location of lightning, this detector has the advantage that it detects both CG and inter-cloud (IC) lightning and can discriminate between positive and negative strokes from characteristics in the signal waveform. [3] While sensitivity thresholds within both the ATD and Boltek systems will favour detection of the most powerful 1 Space Science and Technology Department, Rutherford Appleton Laboratory, Chilton, UK. Copyright 2008 by the American Geophysical Union /08/2007GL lightning strokes, neither system can measure the peak current associated with lightning. [4] Cross calibration of the positional data from the ATD and Boltek systems has been attempted but it is difficult to compare directly the results for each individual stroke. We are confident that, on average, the position of storms measured by both systems is consistent. [5] Routine hourly measurements of the ionosphere are made at Chilton, UK. A Digisonde [Bibl and Reinisch, 1978] transmits shortwave radio pulses in the range 1 20 MHz through a sweep that takes around three minutes. One such sweep is made at the beginning of every hour. The peak electron concentration of each ionospheric layer is determined from the peak radio frequency that each layer is able to reflect vertically. This critical frequency is related to the electron concentration by the formula f c = 8.98 p Ne where f c is the critical frequency of the layer (in Hz) and Ne is the peak electron concentration (m 3 ). The approximate height of the layer is determined by the timeof-flight of the signal. [6] Unlike other ionospheric layers, which result from the photo-ionisation of atmospheric gases, the Es layer is thin and patchy because the plasma results from the ionisation of metallic atoms of meteoric origin [Whitehead, 1989]. These layers are ionised by sunlight and are therefore dominated by diurnal and seasonal trends. Their formation is also modulated by the availability of metallic material and the presence of wind-shear which preferentially moves these long-lived (many hours) ions [Plane et al., 2003] along geomagnetic field lines, forming them into thin layers [e.g., Whitehead, 1989]. The densest patches of ionisation within an Es layer are characterised by the critical frequency of the layer, foes. Because of the patchy nature of Es however, it is useful to estimate the density in the weakest parts of the layer too. This is done via the parameter fbes, the minimum frequency able to penetrate the Es layer and be reflected from one of the more intense ionospheric layers above. The approximate height of an Es layer, h Es, is between 100 and 120 km. The apparent reduction in h Es (1 km) in response to lightning observed by Davis and Johnson [2005] is not reported here since the response seen in the present (smaller) data set was not statistically significant. 2. Method [7] There is evidence [Johnson and Davis, 2006, and references therein] for multiple mechanisms by which lightning can enhance the ionospheric ionisation. Mechanisms involving electrical discharge (such as sprites) and infrasonic waves would enhance the ionosphere directly above the storm while gravity waves launched by lightning [Chimonas, 1971] have a very small vertical component and L of6

2 Figure 1. Results from superposed epoch analyses, from 150 to +150 hours, for two lightning detection systems. The mean response to lightning in each panel is shown as a black line while the error-in-the-mean is shown in grey. Figures 1a 1c show the ionospheric response to (predominantly CG) lightning events measured with the ATD system. The distribution of lightning events influencing these average responses is shown in Figure 1c. Figures 1d and 1e show the ionospheric response to CG lightning events measured with the Boltek lightning detector. The distribution of CG lightning events influencing these average responses is shown in Figure 1f. The associated distribution of contaminating IC events is shown in Figure 1g. The ionospheric response to IC lightning measured with the Boltek system is shown in Figures 1h 1i. The distribution of IC lightning events is shown in Figure 1j. The associated distribution of contaminating CG events is shown in Figure 1k. The ATD system is unable to differentiate between lightning types and so there is no figure equivalent to Figures 1g and 1k for this system. would need to propagate several hundred kilometres horizontally before reaching ionospheric altitudes. [8] In order to differentiate between mechanisms in the current study, the influence of gravity waves was minimised by only considering lightning data from within a horizontal radius of 100 km of the ionospheric measurement. While this threshold is somewhat arbitrary, it was chosen in order to study the effect of lightning on the local ionosphere while preserving a reasonable number of events. These lightning data were subdivided according to type and polarity and each subset was used in a superposed epoch study on the ionospheric Es data. In a superposed epoch study, lightning events are identified and the average ionospheric variability at these times is calculated. In this way, the ionospheric fluctuations caused by randomly distributed mechanisms (such as changes in solar intensity or geomagnetic activity) cancel out while any repeated response, however small, will be reinforced in the final average. In the ionosphere, metal ions created by lightning persist for long enough to be concentrated into thin Es layers by the action of wind-shear over several hours [Whitehead, 1989], enabling the ionospheric response to be observed at the much longer timescales of the hourly ionospheric measurements. As a result of this and the lack of any observable relationship between the number of lightning strokes and the magnitude of the ionospheric response [Davis and Johnson, 2005], a lightning event was defined as any hour in which there was one or more lightning strokes. Davis and Johnson [2005] revealed that the peak ionospheric response occurred after a few hours (for foes and h Es) and up to 30 hours (for fbes), with a possible diurnal recurrence (due to the distribution of lightning events). In the current study, we therefore chose to examine the ionospheric response for a period of 150 hours either side of the lightning events. [9] Most lightning in the UK occurs on summer afternoons and is therefore not evenly distributed in time. To prevent the average ionospheric response being dominated by the diurnal and seasonal trends in the data, thirty-day running monthly medians were calculated from the foes and fbes values for each hour of the day (the resolution of our ionospheric data). These represent the smooth background variability of the parameters. Subtracting the appropriate median values from foes and fbes creates dfoes and dfbes. By considering these parameters, the study becomes sensi- 2of6

3 Figure 2. Superposed epoch analyses for (a d) CG lightning events excluding IC lightning events within ten hours and (e h) IC lightning events excluding CG lightning events within ten hours. The remaining number of contaminating events is small and more evenly distributed (Figures 2d and 2h). tive to deviations from the diurnal and seasonal trends rather than to the trends themselves. 3. Results 3.1. Ionospheric Response to ATD and Boltek Lightning Data [10] Figure 1 shows the ionospheric changes seen in response to lightning data from the ATD system (Figures 1a 1c) and the Boltek lightning detector (Figures 1d 1k). While an ionospheric response to the ATD data has already been demonstrated, it was desirable to repeat this study for the more recent data in order to make a comparison with observations made with the Boltek system. The ATD system has been designed to detect CG strokes preferentially. The Boltek system was used to study the ionospheric response to CG (Figures 1d 1g) and IC (Figures 1h and 1k) events separately. Both systems detect a significant enhancement due to CG lightning in both dfoes (Figures 1a and 1d) and dfbes (Figures 1b and 1e) around four hours after lightning. These responses are significant at the 99.9% level (0.1% probability of occurring by chance) for dfoes and dfbes using ATD events (Figures 1a and 1b) and significant at the 98.5% and 96% levels, respectively, for dfoes and dfbes using the Boltek CG lightning data (Figures 1d and 1e). The response to IC lightning is less clear, with peaks observed around 4 hours after the lightning trigger in dfoes significant at the 98.1% level and dfbes significant at the 97% level. The observed increase can be as much as 0.5 MHz above the background levels, equivalent to electrons m 3. [11] The ATD system recorded 1291 events between January 2000 and September 2005 while the Boltek system recorded 435 CG events and 560 IC events between January 2005 and March As fewer events are seen with the Boltek system, the distribution of lightning observed for this system is comparatively noisy. As with the previous study, the diurnal nature of storms can be seen in the wings of the distribution of total number of lightning events (most apparent in Figure 1c for which there is the most events). [12] A complication in trying to determine the response to different types of lightning is that, if lightning types are closely linked, a study of one may be contaminated if the occurrence of the other is not randomly distributed in time. This is the case when trying to separate the response to IC and CG lightning. Figure 1f shows the distribution of trigger events for the Boltek CG study while Figure 1g shows the associated distribution of IC strokes within this study. As the distribution of IC events is not randomly distributed in time (with 306 IC events at time zero compared with 435 CG events) there is an ambiguity as to which type of lightning is responsible for the observed ionospheric response. The same argument applies to contamination of the IC study by CG events (Figures 1j and 1k) where 560 IC events are associated with 306 CG-events at time zero. The ionospheric response to CG lightning is much clearer than the response to IC lightning despite having a higher fraction of contaminating events. This suggests that, even if IC lightning were generating an ionospheric response, it is much weaker than that associated with CG lightning. This would be expected since the peak currents associated with IC lightning are, in general, substantially less than for CG lightning [e.g., MacGorman and Rust, 1998]. [13] In order to separate the effects of the two lightning types, the two superposed epoch studies were repeated, with the occurrence of the contaminating events excluded from within ten hours of the trigger event. Ten hours was chosen as it encompasses the previously observed ionospheric response time while preserving a sufficiently large subset of the data. The results of this study are shown in Figure 2. 3of6

4 Figure 3. Superposed epoch analyses for (a d) negative and (e h) positive polarity CG lightning events. The number of independent events is dramatically lower for both studies but a clear response is still seen in dfoes (Figure 2a) from twenty two CG lightning events while the equivalent study with seventy four IC lightning events shows no response within ten hours either side of the event (Figure 2e). We conclude that CG lightning occurring within a horizontal range of 100 km is the dominant cause of the observed ionospheric response Ionospheric Response to Polarity of CG Lightning [14] CG lightning can be further categorised by its polarity. The CG lightning data were divided into two subsets, positive and negative, transferring + and charge from the cloud to ground, respectively, and each set was used in a superposed epoch study to determine the ionospheric response. [15] The response of Es to negative lightning events is shown in Figures 3a 3d. Both dfoes (Figure 3a) and dfbes (Figure 3b) show statistically significant enhancements at the 97.6% and 97.5% levels, respectively. The distribution of the 382 lightning events for this study is shown in Figure 3c. Once again, the recurrence of storms can be seen in the wings of the distribution. [16] The equivalent plots for a superposed epoch study using positive CG lightning events are shown in Figures 3e 3h. No significant response is seen in either dfoes or dfbes. With fewer positive lightning strokes (192) for this study, the resulting increase in background noise may be concealing a response. Since the horizontal limit of 100 km is somewhat arbitrary, we subsequently repeated the analysis for the positive lightning strokes, relaxing the maximum allowed distance of lightning from Chilton in 50 km steps up to 250 km but no significant response was seen in either dfoes or dfbes. [17] The distributions in Figures 3d and 3h show that lightning strokes of both polarities nearly always occur within the same hour. As with the coincidence of CG and IC events in the previous section, this makes it difficult to separate the effects of each type of lightning on the ionosphere. In order to minimise the effect of opposite polarity strokes, the original (within 100 km range) superposed epoch analysis was repeated but this time excluding any events in which there were strokes of the opposite polarity within ten hours of the trigger event. The results of this analysis are shown in Figures 4a 4d. For the superposed epoch analysis using negative CG strokes as the trigger event, the number of events was reduced from 382 to 89 but the resulting ionospheric response seen in dfoes and dfbes increased in significance to the 99.9% and 99.4% levels, respectively. When applied to the superposed epoch analysis for positive CG strokes, the same restriction resulted in all events being rejected (since there are far more negative CG strokes than positive). [18] While the smaller number of positive CG events means that no conclusions can be drawn about ionospheric enhancement by positive CG lightning strokes, it is apparent that a significant enhancement is seen in the ionosphere above a lightning storm in both dfoes and dfbes for negative CG lightning strokes. 4. Discussion [19] Rycroft [2006] summarises the potential mechanisms by which lightning may couple the atmosphere and the ionosphere. Johnson and Davis [2006] discussed these mechanisms in relation to the enhancement of the Es layer. Of these, sprites, elves, halo events and infrasonic waves would be expected to act vertically from the thundercloud to 4of6

5 can be expected in response to an EMP from horizontal (IC) lightning and this may provide a mechanism by which sprites could subsequently be triggered by positive CG lightning. [21] While it remains possible that the observed enhancement in Es at 100 km is associated with sprites, the current study suggests that this unlikely. Our results could be explained by the EMP associated with elves although we see a significant response to negative CG lightning only. It is unfortunate we lack information about the peak current associated with the lightning in our data. The peak current is associated with the radiation field amplitude which in turn determines the level of ionisation resulting from a lightning stroke. For a given peak current there are far more negative CG strokes than positive ones. Since no significant response is seen to the positive CG strokes, it may be that these are instead mislabelled negative IC strokes. In order to address these uncertainties, we will need to extend our work to include measurements of the peak current and simultaneous optical observations. [22] Davis and Johnson [2005] pointed out that the ionisation potential for metal atoms was considerably lower than for atmospheric gas species. We urge the modelling community to extend their work to consider ionisation of metallic species and the subsequent effects on the ionospheric Es layer at 100 km. [23] Acknowledgments. We would like to thank Paul Odhams for supplying the ATD data and Richard Henwood and Daniel Wilkins for setting up and maintaining the Boltek lightning detector at Chilton. Figure 4. Superposed epoch analysis for negative polarity CG lightning events for which there are no positive polarity events within 10 hours. The number of positive polarity CG lightning strokes remaining in the data is comparatively small and has a much flatter distribution (Figure 4d). the ionosphere [Sentman et al., 2003; Rodger et al., 2001; Blanc, 1985]. [20] We have observed an enhancement in both the peak (dfoes) and the weakest (dfbes) electron concentration of the ionospheric sporadic E layer. Davis and Johnson [2005] interpreted this as evidence that the mechanism responsible was creating ionisation rather than redistributing existing ionisation (as waves would do). Sprites are electrical discharges between thunderclouds and the ionosphere. They act vertically but generally only extend to around 70 km in altitude (30 km below the Es layer) and are rarely seen over continental Europe [e.g., Neubert et al., 2005]. Boccippio et al. [1995] found that more than 80% of sprites were associated with positive CG strokes while Barrington-Leigh et al. [1999] have observed sprites triggered by negative lightning discharges. The disparity between the observed and expected number of sprites associated with negative CG lightning is discussed by Williams et al. [2007]. Elves extend to greater altitudes and modelling work shows that the associated EMP can cause a significant increase in electron concentration [Rodger et al., 2001]. Barrington- Leigh and Inan [1999] observed elves triggered by positive and negative ground flashes. Modelling work by Cho and Rycroft [2001] indicates that an ionospheric enhancement References Barrington-Leigh, C. P., and U. S. Inan (1999), Elves triggered by positive and negative lightning discharges, Geophys. Res. Lett., 26, Barrington-Leigh, C. P., U. S. Inan, M. Stanley, and S. A. Cummer (1999), Sprites triggered by negative lightning discharges, Geophys. Res. Lett., 26, Bibl, K., and B. W. Reinisch (1978), The universal digital ionosonde, Radio Sci., 13, Blanc, E. (1985), Observations in the upper atmosphere of infrasonic waves from natural or artificial sources: A summary, Ann. Geophys., 3, Boccippio, D. J., E. R. Williams, S. J. Heckman, W. A. Lyons, I. T. Baker, and R. Boldi (1995), Sprites, ELF transients, and positive ground strokes, Science, 269, Chimonas, G. (1971), Enhancement of sporadic E by horizontal transport within the layer, J. Geophys. Res., 76, Cho, M., and M. J. Rycroft (2001), Non-uniform ionisation of the upper atmosphere due to the electromagnetic pulse from a horizontal lightning discharge, J. Atmos. Terr. Phys., 63, Davis, C. J., and C. G. Johnson (2005), Lightning-induced intensification of the ionospheric sporadic-e layer, Nature, 435, Johnson, C. G., and C. J. Davis (2006), The location of lightning affecting the ionospheric sporadic-e layer as evidence for multiple enhancement mechanisms, Geophys. Res. Lett., 33, L07811, doi: / 2005GL Kandalgaonkar, S. S., M. I. R. Tinmaker, M. K. Kulkarni, A. Nath, and H. K. Trimbake (2006), Overview of initial performance of Boltek storm tracker: A lightning detector, J. Atmos. Electr. Jpn., 26(2), Lee, A. C. L. (1989), Ground truth confirmation and theoretical limits of an experimental VLF arrival time difference lightning flash location system, Q. J. R. Meteorol. Soc., 115, MacGorman, D. R., and W. D. Rust (1998), The Electrical Nature of Storms, Oxford Univ. Press, New York. Neubert, T., et al. (2005), Co-ordinated observations of transient luminous events during the EuroSprite2003 campaign, J. Atmos. Sol. Terr. Phys., 67, Plane, J. M. C., D. E. Self, T. Vondrak, and K. R. I. Woodcock (2003), Laboratory studies and modelling of mesospheric iron chemistry, Adv. Space Res., 32, of6

6 Rodger, C. J., M. G. Cho, M. A. Clilverd, and M. J. Rycroft (2001), Lower ionosphere modification by lightning-emp: Simulation of the night ionosphere over the United States, Geophys. Res. Lett., 28, Rycroft, M. J. (2006), Electrical processes coupling the atmosphere and ionosphere: An overview, J. Atmos. Terr. Phys., 68, Sentman, D. D., et al. (2003), Simultaneous observations of mesospheric gravity waves and sprites generated by a midwestern thunderstorm, J. Atmos. Sol. Terr. Phys., 65, Whitehead, J. D. (1989), Recent work on mid-latitude and equatorial sporadic-e, J. Atmos Terr. Phys., 51, Williams, E., E. Downes, R. Boldi, W. Lyons, and S. Heckman (2007), Polarity asymmetry of sprite-producing lightning: A paradox?, Radio Sci., 42, RS2S17, doi: /2006rs C. J. Davis and K.-H. Lo, Space Science and Technology Department, Rutherford Appleton Laboratory, Chilton OX14 5UA, UK. (c.j.davis@rl. ac.uk) 6of6