Unusually high frequency natural VLF radio emissions observed during daytime in Northern Finland

Geomagnetic field variations and electromagnetic waves of different frequencies are ever present in the Earth’s environment in which the Earth’s fauna and flora have evolved and live. These waves are a very useful tool for studying and exploring the physics of plasma processes occurring in the magnetosphere and ionosphere. Here we present ground-based observations of natural electromagnetic emissions of magnetospheric origin at very low frequency (VLF, 3–30 kHz), which are neither heard nor seen in their spectrograms because they are hidden by strong impulsive signals (sferics) originating in lightning discharges. After filtering out the sferics, peculiar emissions are revealed in these digital recordings, made in Northern Finland, at unusually high frequencies in the VLF band. These recently revealed emissions, which are observed for several hours almost every day in winter, contain short (∼1–3 min) burst-like structures at frequencies above 4–6 kHz, even up to 15 kHz; fine structure on the 1 s time scale is also prevalent. It seems that these whistler mode emissions are generated deep inside the magnetosphere, but the detailed nature, generation region and propagation behaviour of these newly discovered high latitude VLF emissions remain unknown; however, further research on them may shed new light on wave-particle interactions occurring in the Earth’s radiation belts.


Introduction
It is generally accepted that the Earth's magnetic field (MF) is one of the necessary conditions for the existence of the biosphere and for human life to emerge and develop on Earth. The invisible magnetic shell which protects the Earth from harmful cosmic rays is the magnetopause, the outer boundary of the magnetosphere, and many other geomagnetic flux tubes. Changes of solar and geomagnetic activity can influence different processes operating in the Earth's atmosphere, including the climate, the biosphere (e.g., Adey 1993, Chibisov et al 1995, and even human health (e.g., Watanabe et al 1994, Gurfinkel et al 1995, Palmer et al 2006, Kleimenova et al 2007, Babayev et al 2012. Thus, the geomagnetic field, with its essential feature of variations having different time scales and waves occurring over a wide range of frequencies, is an important factor in determining the environment of the Earth. These waves act as a specific interface between the interplanetary medium and the magnetosphere, transferring the energy input from the solar wind and the varying Interplanetary MF into the magnetosphere. Moreover, these waves are a very effective remote sensing tool for studying the physics of plasma processes operating in the Earth's magnetosphere and ionosphere. Here we pay especial attention to natural electromagnetic waves in the audio frequency range, which are termed very low frequency (VLF, 3-30 kHz) waves (Helliwell 1965). A particular type of these waves called atmospherics (sferics) originate in lightning discharges (Volland 1995) and propagate thousands of km in the Earth-ionosphere waveguide. Other types of wave, such as broadband unstructured auroral hiss (Gurnett 1966) and discrete rising frequency signals known as chorus, originate as plasma instabilities within the ionospheric or magnetospheric plasma itself. These whistler mode emissions are guided to the Earth's surface by geomagnetic flux tubes; they are common wave phenomena in the Earth's environment. These waves may be observed on the ground or aboard satellites in the ionosphere or magnetosphere (for example, see Tsurutani et al 2012).
Chorus is generated via a cyclotron resonance of the energetic electrons travelling in the opposite direction to the wave through the inner magnetosphere (Trakhtengerts and Rycroft 2008). These electromagnetic waves play a controlling role in the dynamics of Earth's radiation belts, known as the van Allen belts (e.g., Trakhtengerts 1963, Kennel and Petschek 1966, Rycroft 1991, Meredith et al 2001, Thorne et al 2013. It is also important to mention that in the last two decades some authors (e.g., Hayakawa et al 1996, Molchanov and Hayakawa 1998, Němec et al 2009, Rozhnoi et al 2009, Boudjada et al 2010, Písa et al 2013 have used VLF observations to study precursor effects of earthquakes, an active field of research today. Although more than 50 years have passed since the classical work by Helliwell (1965), and despite significant successes of many different ground-based and satellite observations, the full nature and behaviour of different VLF waves is still not fully understood. Many naturally occurring VLF waves at higher frequencies (above 4-6 kHz) could not be studied because strong atmospherics (sferics) hide all such waves. To study these waves, we have to apply special digital programmes which filter out the strong impulsive sferics. That process uncovers completely new types of high frequency daytime VLF emissions with various unusual spectral structures that have never been seen before. The aim of this paper is to present, and briefly discuss, the spectral characteristics of some of these new natural electromagnetic emissions of magnetospheric origin having frequencies higher than 4 kHz.

Data
Our results are based on the VLF observations made in Northern Finland at Kannuslehto, with the geographic coordinates (67.74°N, 26.27°E), and L∼5.5, where L is the distance measured in Earth radii (1R E =6378 km) from the centre of the planet to the equatorial crossing of the geomagnetic field line through the site. Several wintertime VLF campaigns (2006)(2007)(2008)(2009)(2010)(2011)(2012)(2013)(2014)(2015)(2016) have been carried out at this remote, low noise field site some 35 km North of the Sodankylä Geophysical Observatory, in the auroral zone. The VLF emissions were recorded digitally in the frequency band of 0.2-39 kHz by two orthogonal magnetic loop antennas oriented in the north-south and east-west directions. The threshold of the receiver sensitivity is about 0.1 fT, (i.e. ∼10 −14 nT 2 Hz −1 ). As a result of these very sensitive observations, some interesting new emissions typical of this high geomagnetic latitude, at frequencies below 3-4 kHz, have been recently reported (Manninen et al 2012(Manninen et al , 2014.

Results
Figures 1 and 2 demonstrate examples on different days of the 1 h total power spectrograms (frequencytime dynamic spectra) of such previously unknown daytime VLF emissions. The left panels show unfiltered data, including strong narrowband communication transmitter signals at frequencies above 11 kHz, and the right panels present the spectrograms after filtering. The three strong white high-frequency lines in the right panels are the removed radio transmitter traces; the bottom red band at frequencies <0.5 kHz is due to the local power line harmonic radiation (PLHR). Many very strong sferics hide-like a curtain -all signals with frequencies above ∼5 kHz.
The right panels show the same events after digital filtering out of the sferics, the narrowband transmitter signals and the PLHRs. Strong sferics are removed by a broadband digital filter (from 0.6 to 16 kHz), with properly rounded edges so as not to cause deleterious effects on the dynamic spectral analysis. Over successive intervals of 20 ms (the duration of the filter) the signal is reduced to zero, and corrections are applied to the remaining power estimate for the power loss occurring during these 20 ms intervals. The signals from several strong VLF transmitters at frequencies >10 kHz are removed by several narrow band digital Fourier transform filters tuned to the known transmitter frequencies. Similarly, the strongest PLHR occurring at, or near, multiple harmonics of the 50 Hz electrical grid system in Finland are removed by a number of digital notch filters tuned to the interfering frequencies (up to several kHz) which are measured over successive 5 s samples of the signal. A correction is then applied for the small power loss of the dynamic spectral signal at each of these frequencies which are filtered out. Typically, these filters are set to from 950 to 4050 Hz at every odd harmonic.
Different kinds of atypical high frequency VLF emissions are noted. Figure 1(a) shows 10-15 min long bursts of banded hiss-like emissions in the frequency range of ∼7-9 kHz with a sharp low frequency cut-off, looking rather like a bullet in the frequencytime plane. Figure 1(b) presents a narrow (5-6 kHz) band of very closely repeated risers about 1 min apart, Figure 1. The dynamic spectrograms (0-16 kHz) of non-filtered (left panels) and filtered (right panels) VLF data with (a) repeated bursts of the 8-10 kHz hiss with a very sudden ending, (b) a very narrow (5-6 kHz) band of quasi-periodic emissions with a smalltoothed rising tone structure, (c) a combination of hiss bursts at 6-9 kHz and subsequently of strong discrete 8-9 kHz signals of a few minutes duration, accompanied by typical chorus and hiss emissions at lower frequency (1-3 kHz), and (d) two bands of structured hiss emissions at 2-4 and 6-10 kHz, with a well-defined gap between 4 and 5 kHz. yet looking like a band of continuous hiss with a smalltoothed rising tone structure. Figures 1(c) and (d) exhibit the presence of relatively strong structured hiss-like bands of emissions, with a well-defined gap between 4 and 5 kHz. There were relatively strong structured hiss-like emissions at frequencies below ∼4 kHz and 'clouds' of hiss of unusually high frequency, between 6 and 10 kHz, of 10-30 min duration, followed by very short (∼1 min) bursts.
Moreover, this strange spectral world of differently structured high-frequency (>5 kHz) waves was only opened up after filtering out the sferics; to the best of our knowledge, such signals have not been reported before. The dynamic spectra of these high frequency VLF emissions, termed recently revealed emissions (RREs), are plotted in figure 2. We call these peculiar daytime events 'bird-emissions' since when they are played through a loudspeaker they sound like bird song. For the most part, these 'bird-emissions' were observed at frequencies above ∼5-6 kHz. On a 1 h spectrogram, these signals often resemble 'sticks' or 'wands'. Actually, they are mostly hiss bursts with durations of a few seconds. Three examples of such emissions are presented in figure 2: (a) randomly Figure 2. The same as in figure 1, but for emissions sounding like bird song: (a) randomly appearing short (lasting 1-3 min) discrete signals in the frequency range of about 6-10 kHz, (b) vertical 'sticks' over an unusually wide high-frequency range (∼5-12 kHz), and (c) signals of short duration and wide range of frequencies (from 4 to 15 kHz), separated by several minutes. The durations of the 'sticks' in (b) are several seconds; when falling in frequency they are likely to be whistlers or, when their frequency is rising, VLF emissions.
appearing short (lasting 1-3 min) discrete signals in the frequency range of about 6-10 kHz; (b) vertical 'sticks' over an unusually wide high-frequency range (∼5-12 kHz); and (c) signals of short duration and a wide range of frequencies (from 4 to 15 kHz), separated by several minutes.
The diurnal distribution in hourly intervals is given in figure 3 for each campaign. The numbers on the vertical axis show the normalised data calculated for each hour as the ratio of the number of the hours with these RREs in the given hour to the total number of the observation hours in each campaign. It can be seen that all campaigns demonstrate the same rather similar feature, namely a remarkable enhancement in the local daytime. Without exception, these new VLF emissions are observed during rather quiet geomagnetic conditions after moderate geomagnetic disturbances or in the late recovery phase of magnetic storms. Because these RREs are observed during the local winter, when it is dark at high latitudes throughout the day, they suffer the minimum amount of attenuation as they propagate through the ionosphere to the receiver on the ground.
We also have analysed the fine structure, on time scales of a few s or more, of these previously hidden VLF emissions. Usually the studied events exhibit a sequence of separate patches in the frequency-time domain, with a wide variety of spectral structures. Several individual examples of 3 min spectrograms of remarkably diverse signals in the frequency range 1-10 kHz from different campaigns are plotted in figure 4. Panel (a) demonstrates an example of the simultaneous generation of two separate 'patches' in the frequency-time domain, 1 min long, and at different frequencies in the 6-10 kHz band, with different onsets and dynamics of spectral structure. Such complicated wave features could indicate that the development of the magnetospheric plasma instability responsible for them is rather localised in time and space. Panels (b)-(c) show the strange frequency-time dynamic structure resembling several flying birds, with the emitted frequency rising with time. A combination of several short signals at different frequencies between 4 and 10 kHz is plotted in panel (c), where one can see the frequency of the signal increasing over about 30 s. In panel (d), the fine spectral structure of the quasi-periodic hiss-like emissions, shown in figure 1(b), is presented; hiss bursts of rising frequency with periodically repeated enhancements are seen. Panels (e) and (f) show two different bursts of waves lasting about 1 min; the overall frequency is increasing from 4 to 7 kHz (e) or decreasing from ∼8 to ∼5 kHz (f) as time progresses, and the strong emissions at frequencies below 2 kHz are typical chorus signals (Helliwell 1965) at this site. Because the wave frequency generated is directly proportional to the electron cyclotron frequency in the source region f ce (which is proportional to L −3 ), this observation may support the hypothesis of the generation source shifting inwards (e) or outwards (f) across the L-shells.
Such a rich collection of very varied dynamic spectra of these high-latitude VLF emissions (as is evident in figure 4) at frequencies above 5 kHz must certainly be the result of spatially and temporally changing localised regions of electron cyclotron instability in the magnetospheric plasma. The detailed nature of these complex processes is still poorly understood and these new results raise a number of questions to be attacked in further research. Francis et al (1983) have reported receiving most unusual discrete VLF emissions in the high VLF range, from 6 to 9 kHz, at Halley, Antarctica (L∼4.3). The signals were generally a succession of chirps, each rising in frequency within 1 s, and often exhibiting a minimum intensity near 7.5 kHz; they also exhibited a ∼4 min periodicity in occurrence. On the other hand, the frequency of chorus commonly observed at Halley during the local morning hours is ∼2-4 kHz. These signals were recorded on one day in June 1972 near local noon, during the recovery phase of a minor geomagnetic storm. Francis et al (1983) interpreted these emissions as being due to a cyclotron resonance instability occurring near the equatorial plane on a nearby L∼3.6 flux tube.
The interpretation of the minimum intensity of the signals at 7.5 kHz discussed by Francis et al (1983) is that this frequency is 0.5 f ce at the equator on the geomagnetic flux tube. Signals below this frequency are guided by enhancements of plasma density in magnetically field aligned ducts, whereas higher frequencies are in principle guided by field aligned depletions of plasma density (Smith et al 1960).
It must be mentioned that these unusual VLF emissions are not auroral hiss. Auroral hiss is a broadband and unstructured (in the frequency-time plane) signal which may last for several hours (generally between local noon and midnight) with very little temporal variation on time scales less than a minute (Sazhin et al 1993, LaBelle andTreumann 2002). Hiss between 3 and 9 kHz, and having multiple banded structures, has been observed at Porojärvi, also North of Sodankylä, during the local evening, and reported by Titova et al (2007).
The considered RREs could not be classified as the high-frequency VLF emissions recorded by (Francis et al 1983). There are several important differences between the RREs and the unusual VLF emissions at Halley Bay (Francis et al 1983). The duration of separate RREs is much longer and amounts ∼30-60 s or even more (as is seen in figure 4), as opposed to tenths of a second at Halley Bay (see figures 1 and 4 in Francis et al 1983). The RRE spectral structure is completely different (as one can see in some examples of the wave dynamic spectra presented in our figure 4), contrary to a series of short rising frequency emissions recorded at Halley Bay. Further, we have never seen the frequency gap between two high-frequency emission bands as was found at Halley Bay.
VLF observations made aboard satellites can provide useful complementary information. Titova et al (2015) have studied natural VLF signals recorded aboard the Van Allen Probe-A when it was in the night sector near the geomagnetic equator at L=3.0-4.2, on geomagnetic flux tubes passing through the area to the North West of Scotland. Simultaneously, the same signals, rising in frequency, typically from 2 to 5 kHz, were recorded on the ground at Kannuslehto (L∼5.5). These quasi-periodic discrete signals certainly have different characteristics from the RREs reported here. Recently, Němec et al (2016) have reported other quasi-periodic signals, occurring between 1.4 and 2.4 kHz (see their figures 2 and 3), observed both on the DEMETER spacecraft and on the ground at Kannuslehto. Again these have completely different shapes in the frequency-time domain from the RREs considered here. Further, Parrot et al (2015) have reported some new and unexplained VLF signals observed aboard the microsatellite DEMETER the topside ionosphere; however, all these were recorded at low and middle latitudes, but not in the auroral zone. The RREs reported here do not fit into any category of unexpected VLF radio emissions discussed by Parrot et al (2015).

Discussion
It is very important to emphasise that the recently revealed VLF emissions (RREs) are observed at frequencies above 4-6 kHz, and up to 10 kHz or even higher. It is generally accepted that the typical VLF emissions observed on the ground are generated in the magnetosphere due to the electron cyclotron instability (Trakhtengerts and Rycroft 2008). The frequency band of these waves is controlled by the electron gyrofrequency f ce at the geomagnetic equator, where f ce is the frequency of electron gyrating around the geomagnetic field line. The f ce value is proportional to the local MF strength (Stix 1962); it is thus proportional to L −3 . Chorus observed near the magnetic equator close to L∼5.5 aboard the Van Allen Probe-A spacecraft (Santolik et al 2014) occurs at frequencies below 2 kHz, at <0.5 f ce .
VLF waves, which are generated near the magnetic equator in the magnetosphere, are guided by the geomagnetic field to the ionosphere. The ducted whistlermode wave guiding structure (Smith et al 1960) is similar to that in fibre optics when light is guided by dielectric glass filaments. Ducted VLF emissions reaching the ground have an upper cut-off frequency at half the equatorial electron gyrofrequency f ce . That is the maximum frequency which can propagate in a duct caused by a field-aligned enhancement of electron density (Carpenter 1968); this is a hard, theoretical limit which arises from magnetoionic theory. The second important fact is that only those waves, which penetrate through the ionosphere, can be recorded on the ground in the vicinity (typically ∼120-150 km) of the footprint of the ionospheric exit point of the wave.
It should be noted that the equatorial electron gyrofrequency f ce at L∼5.5 (i.e. the L value of our observation point Kannuslehto) is ∼5 kHz. Assuming that ducting occurs across the equatorial region at L∼5.5, this corresponds to an upper frequency limit of f 0.5 f ce , i.e. f 2.5 kHz. This is the theoretical limit above which we should not be able to detect emissions on the ground if they propagate roughly along the receiver's geomagnetic flux tube. However, the recently revealed VLF emissions are observed at much higher frequency, well above 4-5 kHz, often up to 10 kHz and sometimes up to 15 kHz. To the best of our knowledge, the RREs have not been observed by space-borne instrumentation. Further, it is difficult to explain how the waves with frequency higher than half of the electron gyrofrequency can be guided from the generation region at or near the geomagnetic equator to the ground.
Perhaps these waves are generated much deeper in the magnetosphere, at much lower L-values (L∼3.5) where the equatorial electron gyrofrequency f ce is significantly higher, by a factor of ∼4. Then a new question arises: how do these waves cross a rather large range of L-values to be observed at high latitudes? This challenging problem is to be tackled in another paper. The detailed nature, generation region and propagation behaviour of these newly discovered VLF emissions still remain unknown. However, further research on them may shed new light on wave-particle interactions occurring in the Earth's radiation belts.

Summary
Based on these VLF (3-30 kHz) radio observations made in Northern Finland, we have discovered many new and unexpected natural electromagnetic emissions of magnetospheric origin at frequencies higher than 4 kHz. Previously, these were neither heard nor seen in their spectrograms because they were hidden by strong impulsive signals (sferics) originating in low latitude lightning discharges. Only after filtering these sferics out were the peculiar VLF emissions discovered. These RREs are probably generated close to the geomagnetic equator yet deep in the magnetosphere, at a considerably lower L-value (L∼3.5) than that of the observation site (L∼5.5). The details of the mechanism of the generation and propagation of these newly discovered VLF emissions remain unknown.
Author contributions JM and TT designed the research, performed most of the data processing and analysis, TT designed the digital electronics receiver and the filters, and supervised the observations, JM, NK, and MR took the lead in writing the manuscript, LG downloaded and analysed geomagnetic data, and IS undertook a statistical analysis of the VLF data. All authors discussed the results and contributed to the editing of the manuscript.