Statistical study of auroral roar emissions observed at South Pole Station

Transcription

1 JOURNAL OF GEOPHYSICAL RESEARCH, VOL. 107, NO. A7, 1140, /2001JA000319, 2002 Statistical study of auroral roar emissions observed at South Pole Station J. LaBelle Department of Physics and Astronomy, Dartmouth College, Hanover, New Hampshire, USA A. T. Weatherwax Institute for Physical Science and Technology, University of Maryland, College Park, Maryland, USA Received 27 September 2001; revised 30 December 2001; accepted 2 January 2002; published 26 July [1] Auroral roar is a band-limited natural auroral radio emission observed at ground level in the auroral zone at frequencies near 3 MHz and 4.5 MHz, close to 2 and 3 times the electron gyrofrequency. Using semiautomated data analysis techniques, we generate a database of frequencies, amplitudes, and universal times of 2f ce -auroral roar emissions observed at South Pole Station (74 invariant latitude) during These data confirm that auroral roar emissions are observed at ground level only during times when the ionosphere is not illuminated, and magnetic local time plays a significant role in controlling the occurrence rates of auroral roar emissions observed at ground level. Auroral roar occurrence rates and the ranges of auroral roar amplitudes and frequencies increase during times of high geomagnetic activity. The distribution of auroral roar frequencies is bimodal for large-amplitude events, which we interpret as evidence that a substantial fraction of the auroral roar emissions observed at ground level originate in the topside ionosphere. Finally, superposed epoch analyses of occurrences of auroral roar relative to substorm onsets inferred from magnetometer and VLF data show that auroral roar emissions occur favorably preceding such events rather than after them, consistent with previous case study and anecdotal evidence. INDEX TERMS: 2407 Ionosphere: Auroral ionosphere (2704); 2471 Ionosphere: Plasma waves and instabilities; 2704 Magnetospheric Physics: Auroral phenomena (2407); 7847 Space Plasma Physics: Radiation processes; KEYWORDS: plasma waves, radio emissions, auroral roar, plasma instabilities, auroral waves, auroral physics 1. Introduction [2] LaBelle and Treumann [2002] present an extensive review of both theoretical and observational aspects of auroral roar emissions and related phenomena. Auroral roar is a band-limited natural emission observed at ground level in the auroral zone at frequencies near 3 MHz and 4.5 MHz, close to two and three times the electron gyrofrequency ( f ce ) at F-region altitudes [Kellogg and Monson, 1979, 1984; Weatherwax et al., 1993; Hughes and LaBelle, 1998]. The center frequency of auroral roar varies linearly with the magnitude of the geomagnetic field at the observatory, as expected if the emission is associated with the electron cyclotron harmonics [Hughes and LaBelle, 1998]. If generated exactly at 2f ce and 3f ce, the emissions originate near 275 km on the F-region bottomside [e.g., Hughes and LaBelle, 1998], although Hughes et al. [2001] show evidence that sources on the topside of the F-region also occur. Using satellite-borne topside-sounder receivers, James et al. [1974] observe 2- and 4-MHz auroral emissions which may represent the same phenomena as the 2f ce - and 3f ce -auroral roar observed at ground level. The ground level auroral roar Copyright 2002 by the American Geophysical Union /02/2001JA emissions are left-hand polarized with respect to the magnetic field [Shepherd et al., 1997], implying that they propagate in the LO-mode in the ionosphere although they need not originate in the LO mode. They are characterized by fine structures consisting of discrete tones increasing or decreasing in frequency with time, similar to the fine structure of Auroral Kilometric Radiation (AKR), VLF chorus, or structured Jovian decametric radiations [LaBelle et al., 1995; Shepherd et al., 1998b]. The typical amplitude of auroral roar at ground level is a few tens of mv/m, implying a few Watts for the power emitted, assuming an isotropic radiator at 275 km altitude. The duration of the emissions ranges from minutes to hours. The occurrence rate approaches 50% of nighttime hours during moderate geomagnetic activity at geomagnetic latitudes near 75, just poleward of the auroral zone [Hughes and LaBelle, 1998]. The latitude of the most intense auroral roar observed at ground level roughly tracks the large-scale north-south motion of the poleward edge of the auroral electrojet currents inferred from ground level magnetometer data [Shepherd et al., 1998a]. Dynamic spectral representations show that the bottom edge of the auroral roar emission sometimes exhibits modulations at 7 11 Hz, a phenomenon called flickering auroral roar and possibly related to flickering aurora [Hughes and LaBelle, 2001a]. SIA 17-1

2 SIA 17-2 LABELLE AND WEATHERWAX: AURORAL RADIO EMISSIONS AT SOUTH POLE STATION Figure 1. Typical profiles of the electron plasma frequency, upper hybrid frequency, and electron cyclotron harmonic frequencies as functions of altitude in the auroral ionosphere. Filled circles indicate where the matching conditions f uh =2f ce,3f ce occur. At these locations, strong electrostatic upper hybrid waves may be excited which may partially convert to LO mode and propagate to the ground where they are observed as auroral roar. [3] Figure 1 illustrates the leading theory put forth to explain auroral roar emissions. The highly non-maxwellian distribution of the kev auroral electrons, which often includes a positive slope in the reduced distribution function perpendicular to the magnetic field due to loss-cone and/or horseshoe features, can give rise to intense upper hybrid waves. The temporal growth rate maximizes for the condition that the upper hybrid frequency ( f uh ) matches a cyclotron harmonic [e.g, Kaufmann, 1980; Yoon et al., 1998a]. Figure 1 shows a typical profile of the plasma frequency, upper hybrid frequency, and cyclotron harmonic frequencies. Clearly, this matching condition implies localized regions indicated by filled circles in the figure. If intense electrostatic upper hybrid waves at these locations mode convert to LO waves and propagate to ground level, they would have the observed frequency characteristics of auroral roar emissions. The sources on the bottomside have straightforward access to ground level, and those from the topside might reach the ground as well, either by propagating through holes or gaps in the ionosphere or by scattering off of density irregularities. Gough and Urban [1983] observe directly with sounding rocket instruments modulations in auroral electrons at the condition f uh =2f ce with an amplitude high enough to explain ground level auroral roar observations, assuming a 1% conversion efficiency of electrostatic to electromagnetic waves. The upper hybrid waves near the matching condition are highly sensitive to refraction which could limit their resonance time with the electrons [e.g., Lotko and Maggs, 1981] but which may also allow them to become effectively trapped in density irregularities of appropriate depth and dimension (C. Carlson et al., Observations of intense electron bernstein wave emissions in an auroral plasma, unpublished manuscript, 1987). This wave trapping can involve discrete eigenmodes of the density cavity, which may provide an explanation for the auroral roar fine structures [Yoon et al., 2000]. The conversion mechanism for generation of LO-mode auroral roar from the electrostatic upper hybrid waves remains unknown. Several authors have suggested that the upper hybrid waves refract to the Ellis window condition, where direct conversion from the Langmuir/upper-hybrid mode to the LO mode occurs [e.g, Gough and Urban, 1983; Yoon et al., 1998b]. Other possibilities include linear conversion due to interaction with appropriate scale irregularities, analogous to a process analyzed for conversion of electromagnetic whistler modes to lower hybrid waves [Bell and Ngo, 1988, 1990], or nonlinear wave-wave interactions, though calculations based on measured upper hybrid wave intensities indicate that these are probably not efficient enough [e.g., Weatherwax et al., 1995]. Experiments by Shepherd et al. [1999] and Hughes and LaBelle [2001b] provide evidence that ground level auroral roar is associated with f uh =2f ce matching points in the ionosphere: direction of arrival measurements of auroral roar, combined with scanning incoherent scatter radar measurements of electron density, show that the roar comes from locations where the matching condition is met at the altitude where the local electron cyclotron frequency equals the observed frequency of the auroral roar emission. [4] For several years, Dartmouth College has operated an LF/MF/HF wave receiver at South Pole Station. In December 1998, the antenna for this system was moved about 1 km further away from the station, to a location 1.5 km distant from the station. This change significantly reduced station-generated interference below 5 MHz, making the South Pole receiver more sensitive than those at sites in the Northern hemisphere, increasing the number of auroral roar events observed, and allowing semiautomated analysis of the data. This analysis provides more complete statistical analysis of auroral roar emissions than previously achieved. However, auroral roar events near 3f ce are fairly rare at South Pole Station, so this paper focuses on 2f ce -auroral roar events, for which there are good statistics. 2. Data Presentation [5] The receiver used in this study is a programmable frequency receiver (PFR) described extensively in previous reports [e.g., Weatherwax et al., 1995; Shepherd et al., 1999]. Natural auroral emissions at MHz are detected with a 10 m 2 magnetic loop antenna and preamplifier deployed approximately 1.5 km from South Pole Station (74 invariant latitude). The magnetic loop is in the plane perpendicular to the Earth s surface and oriented so as to null electromagnetic interference originating at the station. The receiver, located in the station, has a dynamic range of 70 db and a bandwidth of 10 khz tunable between 50 khz and 5 MHz in a sequence controlled by a personal computer. In the standard mode, the frequency is increased monotonically from 50 to 5000 khz in 10 khz steps, repeated each 2 s. Data are collected every day starting at 1600 UT until 1000 UT the following day, with a 6-hour gap around local noon for housekeeping purposes. The resulting data are compressed, stored locally and transferred

3 LABELLE AND WEATHERWAX: AURORAL RADIO EMISSIONS AT SOUTH POLE STATION Figure 2. (a) MHz spectrogram recorded at South Pole Station on an active day (July 31 Aug 1, 1999), showing interference lines during UT and an auroral roar event occurring intermittently during UT. The grayscale represents root power pffiffiffiffiffiffi spectral densities ranging from 10 p 10 9 V/m Hz ffiffiffiffiffiffi (white pixels) to V/m Hz (black pixels). (b) Frequencies identified as auroral roar by a computer program, which falsely identifies some interference features as auroral roar. (c) Final auroral roar identification achieved by interactively modifying the spectrograms using a second computer program. In panels b c, black pixels indicate identification of auroral roar. to Dartmouth daily. A full year of compressed data fits on several CD-roms. [6] Figure 2 illustrates the analysis method. Figure 2a (top panel) shows a dynamic spectrogram of data recorded on an active day, July 31 August 1, The vertical (frequency) axis covers MHz, which includes the range of 2fce auroral roar emissions. The grayscale displays root ppower 10ffiffiffiffiffiffi 10 9 ffiffiffiffiffiffi spectral densities ranging 9from p V/m Hz (white pixels) to V/m Hz (black pixels), on a logarithmic grayscale; intense auroral roar lies near the saturation level of this grayscale. Numerous narrowband fixed-frequency interference lines, probably originating mostly from distant shortwave transmitters but partly from local interference, occur between UT. However, the most outstanding feature in the spectrogram is the auroral roar emission occurring intermittently during UT in the frequency range MHz. Because the interference is weak and usually consists of fixed frequency lines, a computer program can select all pixels exceeding an adjustable threshold over an adjustable array of pixels in frequency-time space on the spectrogram images. Figure 2b (middle panel) shows the mask produced by such a program,pwhich in this case seeks out events ffiffiffiffiffiffi exceeding 50 nv/m Hz over a 20-kHz frequency interval and 5-minute time interval, while excluding interference lines detected over 2-hour timescales. [7] Clearly, the program identifies the auroral roar, but also selects some interference features. However, from visual inspection these features are easily identifiable as interference, particularly when the original dynamic spec- SIA 17-3 trum (Figure 2a) is available for comparison. As a result it is a simple process to inspect the computer generated mask on the computer screen and to use the mouse to box off the regions of interference, which can then be eliminated from the selected data by a second computer program. In this manner, a modified mask is generated, illustrated in Figure 2c (bottom panel), in which only auroral roar emissions are identified. With these automated programs for displaying the data, generating masks where auroral roar occurs, and allowing a trained scientist to interactively modify those masks, an entire year of South Pole data was analyzed in this fashion in a reasonable amount of time. The resulting data base consists of frequencies, amplitudes, and universal times of each pixel identified as auroral roar during January through December, [8] Figure 3 shows scatterplots of auroral roar frequency and amplitude as functions of time of day expressed in UT. Midnight magnetic local time (MLT) occurs at 3.5 UT at South Pole Station. The shaded region indicates hours ( UT) when the receiver was turned off and no data were collected. There is a clear MLT dependence of auroral roar detected at South Pole Station, with a maximum in the pre-midnight MLT hours (approximately UT). The absence of data from UT makes it difficult to tell whether the events actually go away near noon MLT, but despite the data gap, the data strongly suggest a minimum in occurrence rate during the pre-noon MLT hours (approximately UT). Because the ionosphere is in darkness continuously for <6 months above South Pole, the diurnal variation highlights the importance of magnetic local time in controlling event occurrence rate. This result confirms previous statistical studies indicating maximum auroral roar occurrence rates in the pre-midnight hours at stations near the auroral zone [Weatherwax et al., 1995; LaBelle and Hughes, 2001]. Figure 3a shows that the largest auroral roar amplitudes also occur in the pre-midnight/midnight MLT sector. Studies of auroral roar occurrence at magnetically conjugate locations have not yet been possible. [9] The observed frequencies of the auroral roar emissions reach their widest distribution, from below 2.2 MHz Figure 3. Scatterplots of (a) amplitudes and (b) frequencies of auroral roar emissions observed at South Pole in 1999, as functions of time of day (in UT). Cross-hatched regions indicate times when no data are available.

4 SIA 17-4 LABELLE AND WEATHERWAX: AURORAL RADIO EMISSIONS AT SOUTH POLE STATION Figure 4. (a) Number of pixels of auroral roar detected on each day in 1999; scatterplots of (b) amplitudes, (c) frequencies, and (d) times of day (in UT) of auroral roar emissions observed at South Pole in 1999, as functions of day of year. The UT times are plotted twice to better reveal trends. The bottom panel (e) shows the daily Ap for reference. to above 2.8 MHz, during the same UT period when the maximum amplitudes occur. The range of observed frequencies narrows as the time of day shifts toward noon MLT from either side. Primarily, this occurs because the lower bound of the frequency range increases as noon is approached, from about 2.2 MHz at 0400 or 2030 UT to about 2.5 MHz at 1000 or 1600 UT. The highest frequency auroral roars are recorded in the immediate post-noon period, where some events exceed 3.0 MHz. Possibly, these observations result from the more variable electron density profiles occurring during the nighttime hours, when auroral bombardment is the primary source of ionization. Figure 1 shows that for large electron densities at the F-peak, the bottomside 2f ce matching point occurs at relatively low altitude and hence higher frequency, and waves generated at the topside 2f ce matching point do not have access to the ground. For lower electron densities, the bottomside matching point would shift to higher altitudes, and waves generated at the topside matching point might gain access to the ground either directly or through holes in the F-region which occur when the ionization is patchy. During nighttime, both high and low densities occur and the F-region can be patchy, and under these conditions the model illustrated in Figure 1 predicts a broad distribution of auroral roar frequencies as observed. For a more slab-like, relatively high-density F-region, the model illustrated in Figure 1 predicts that only the relatively high frequency events should occur, those generated at the bottomside matching point at relatively low altitudes. The resulting frequency distribution would look like that observed at local times far away from local midnight (Figure 3b). Of course, ground level observations of auroral roar are significantly affected by propagation effects, such as absorption and refraction, and considering these effects may lead to alternative explanations for the features seen in Figure 3. [10] Figure 4 shows scatterplots of auroral roar frequencies, amplitudes, and times-of-day in UT as functions of day of year for (Panels b d show include all identified auroral roar pixels, not just daily averages.) For reference, the bottom panel shows smoothed daily Ap values plotted with a seven-day running average. The top panel shows the total number of pixels of auroral roar observed on each day. This plot emphasizes that auroral roar is detected at ground level only during the months of darkness, with the first event occurring on day 86 (March 27) and the last event occurring on day 262 (September 19). The former threshold occurs slightly after the spring equinox, whereas the latter threshold almost coincides with the Fall equinox. These observations confirm the dominant role played by solar illumination in controlling ground level detectability of auroral roar emissions. This control has generally been attributed to photo-ionization enhanced electron densities in the lower ionosphere, which serve to screen the LO-mode auroral roar radiation from the ground. The observation of events before Spring equinox can occur because sunlight on the lower ionosphere, rather than at ground level, controls the E- and D-region densities and hence the screening. The first and last days of occurrence lie considerably closer to the equinoxes than would be predicted by sunrise/sunset at E-region altitudes; for example, E-region sunrise (sunset), defined as 90 solar-zenith-angle at 100 km altitude, should occur 26 days before (after) the spring (fall) equinox, whereas the observed event thresholds are 0 6 days away from the equinoxes. A quantitative explanation of this effect may be more complex than simply calculating onset of darkness at ionospheric altitudes. It appears as though the emissions are prevented from reaching ground level well before (after) E-region sunrise (sunset), at least those defined by 90 solar-zenith-angle at 100 km. [11] Figure 4 shows further that the number of auroral roar pixels, their frequency range, and their amplitude range

5 LABELLE AND WEATHERWAX: AURORAL RADIO EMISSIONS AT SOUTH POLE STATION SIA 17-5 Figure 5. Scatterplot of auroral roar emission frequencies versus their amplitudes, using all 1999 data. For large amplitudes, the distribution of frequencies becomes bimodal, with peaks near 2350 khz and 2650 khz. increase in the last part of the darkness period after about day 200, corresponding to more intervals of relatively high Ap values occurring then. These observations extend previously published evidence of an increase of auroral roar occurrence rates during geomagnetically active periods [e.g., Weatherwax et al., 1995; Hughes and LaBelle, 1998]. Interestingly, lower amplitude auroral roars are detected during approximately the interval than on days before and after. A check of calibration signals during these days shows no change in instrument sensitivity or frequency response. At the same time, a wider range of frequencies, including lower frequencies, are detected, but the lowest frequency pixels do not coincide with the lowest amplitude pixels any more than would be expected by chance. As shown below, in general low-amplitude events near the threshold are more common than larger amplitude events, but there is no obvious explanation for the prevalence of the lowest amplitude events during the days [12] Figure 5 shows a scatterplot of auroral roar frequency versus amplitude, using all data. Not surprisingly, as mentioned above, there are greater numbers of lowamplitude events, and events are less common, the larger their amplitude. There is a gap between the smallest amplitude events and the absolute amplitude threshold used in the analysis, and the low-amplitude boundaries of the events in this scatterplot and others (e.g., Figures 3a and 4b) are uneven. This arises because the event identification depends not only on amplitude exceeding a threshold, but also on this condition occurring for a range of contiguous pixels, as described above. Because event selection is based not strictly on amplitude, the lowamplitude boundary in the scatterplots is not expected to be a straight line. [13] The most significant feature of Figure p 5 appears at large amplitudes, above V/m ffiffiffiffiffiffi Hz. The frequency distribution of these large amplitude auroral roars is bimodal. Most of the events are distributed around 2650 khz, but a smaller number are distributed around 2350 khz. There are also a handful of high-amplitude auroral roars observed near 2500 khz. Assuming generation at twice the cyclotron frequency above South Pole Station, the frequencies 2650 khz and 2350 khz correspond to source altitudes of 275 km and 610 km, respectively. The source of the bulk of the roars at 275 km, which would normally correspond to the bottomside of the F-region, agrees nearly exactly with previous studies [Kellogg and Monson, 1984; Weatherwax et al., 1995; Hughes and LaBelle, 1998]. However, these South Pole data suggest that a substantial minority of the auroral roar events come from a high altitude source in the topside ionosphere. This observation is qualitatively consistent with the model illustrated in Figure 1, which predicts sources on the topside as well as on the bottomside. (However, Figure 1 has been sketched roughly based on typical ionospheric parameters and should not be compared quantitatively with the data.) The interpretation of the bimodal distribution in Figure 5 as evidence of topside auroral roar sources also agrees with data obtained from an interferometer in Sondrestrom, Greenland. Interferometric measurement of the direction of arrival of auroral roars combined with the assumption that they are generated at local 2f ce makes possible the determination of apparent source altitude of the auroral roar emissions, based on the latitude and longitude of the source as well as the altitude [Hughes et al., 2001]. A histogram of these apparent source altitudes has two peaks, one large peak corresponding to a bottomside source near 275 km and a smaller peak corresponding to a topside source near 410 km [Hughes et al., 2001]. Hughes and LaBelle [2001b] back up this statistical conclusion with case studies comparing electron densities obtained with the Sondrestrom incoherent scatter radar to auroral roar measurements, several of which imply that the emissions originate in the topside of the F-region and reach the ground through a hole or perforation in the F-region, or through an extended region in which the F-region is essentially absent. The South Pole data in Figure 5 provide further evidence that emissions originating in the topside of the F-region often penetrate to ground level. [14] In both the Greenland and South Pole data, the bimodal distribution of source altitudes occurs only for large amplitude roars and is not observed when the much larger number of weak auroral roars are included in the statistics. Two possible reasons have been put forward to explain this. First, possibly the largest amplitude roars consist more often of those which propagate directly in a relatively straight line, suffering less multi-pathing or other propagation effects, while the weaker roars consist of a larger fraction of those whose propagation is severely affected by refraction and multipathing. For these weaker emissions, the mapping between the emission frequency and the source altitude is confused because of variations in source latitude and longitude, and the expected bimodal

6 SIA 17-6 LABELLE AND WEATHERWAX: AURORAL RADIO EMISSIONS AT SOUTH POLE STATION Figure 6. (a) The number of auroral roar pixels observed as a function of delay time relative to occurrence of enhanced khz waves of duration 2 minutes. Negative time on the x-axis corresponds to auroral roar emissions preceding the LF bursts. (b) An expanded view of ±100 minutes around zero lag, highlighting the asymmetry whereby auroral roar favors times preceding LF bursts by 1 hour. (c) The number of roar pixels as a function of delay time relative to times when the H- component of the geomagnetic field decreases by 100 g within 10 minutes; (d) an expanded view of ±100 minutes around zero lag. distribution of source heights may not translate into an observed bimodal frequency distribution. Another possibility, suggested by C. Kletzing (personal communication, 2001), is that large amplitude roars correspond to more intense auroral precipitation when the F-region electron density is enhanced by electron impact ionization, in which case the two f uh =2f ce matching points, shown in Figure 1, are separated by a larger altitude difference. This would make the bimodal frequency/altitude distribution easier to discern in such cases. However, there is no indication in the data that the separation of the two peaks increases as the amplitude increases. [15] Finally, the database of South Pole auroral roar observations can effectively be compared with South Pole magnetometer and VLF data, both of which are readily available as 1-minute averages continuous throughout Many previous papers report anecdotal and case study evidence of the relationship between auroral roar occurrences and substorm onsets: Auroral roars sometimes occur for tens of minutes preceding substorm onsets, but then become intermittent or cease to be observed at ground level following substorm onsets [e.g., LaBelle et al., 1994; Weatherwax et al., 1995; Shepherd et al., 1998a; LaBelle and Hughes, 2001]. To further investigate this relationship, Figure 6 shows the results of superposed epoch analyses using South Pole magnetometer and VLF/LF data to identify possible substorm onsets. Figure 6a shows the number of pixels of auroral roar within ±10 hours (600 minutes) of enhanced khz signals lasting at least two minutes, as a function of delay time before (negative values) or after (positive values) the occurrence of the LF burst. Bursts at khz lasting a few minutes usually indicate impulsive auroral hiss which is an effective indicator of a substorm onset [see, e.g., Makita, 1979; LaBelle et al., 1994, 1998]. Auroral roar generally becomes less probable as the delay time relative to the LF bursts increases, but there is an asymmetry in the expected direction: the large peak in the number of auroral roar pixels occurs for times within about one hour before the LF bursts. Figure 6b shows an expanded view of ±100 minutes delay time, showing the asymmetry more clearly. In Figures 6a 6b, LF bursts occurring less than one hour after a previous LF burst are removed from the analysis in order to avoid confusion due to complex or rapidly recurring substorms, although only a few examples are eliminated by this removal, and the results are not qualitatively altered by it. [16] A similar effect, though less clear, is observed when magnetometer data rather than VLF data are used to identify possible substorm onsets. In this case, the reference times for the superposed epoch analysis are determined by instances in which the H-component of the local South Pole decreases by >100 g within 10 minutes; again instances in which such decreases follow one another by less than one hour are eliminated from the survey. Figures 6c 6d shows the results of the superposed epoch analysis, in the same format as Figures 6a 6b. The data are considerably noisier, and it is harder to attribute significance to the peak at delay times of 0 1 hour, corresponding again to auroral roars occurring favorably at times preceding substorm onsets by 1 hour. However, the results, though too noisy to be more than suggestive, agree qualitatively with the cleaner results obtained using VLF data to determine the substorm onsets. They provide further statistical evidence in support of oft-cited anecdotal and case study evidence that auroral roar emissions tend to be observed at ground level for tens of minutes preceding substorm onsets.

7 LABELLE AND WEATHERWAX: AURORAL RADIO EMISSIONS AT SOUTH POLE STATION SIA Summary [17] Extraordinary interference-free MHz spectra have been obtained at South Pole Station during 1999 by moving the receiving antenna to >1.5 km away from the station. This allows semiautomated data analysis for the first time, resulting in a database of the frequency, amplitude, and universal time of auroral roar emissions observed at South Pole Station during The principal results of this paper are discussed above. They include: confirmation that auroral roar is observed at ground level only during times when the ionosphere is not illuminated, and that during such times, the magnetic local time rather than solar zenith angle plays a significant role in controlling auroral roar occurrence and characteristics as observed at ground level; confirmation that auroral roar occurrence rates increase during times of high geomagnetic activity as monitored by the Ap index, and evidence that the distribution of auroral roar amplitudes and frequencies are larger during times of high Ap; further evidence that a substantial fraction of the auroral roar emissions observed at ground level may originate at sources in the topside ionosphere; and the first statistical evidence to support anecdotal evidence that auroral roar emissions tend to precede substorm onsets by tens of minutes, and tend to become intermittent or extinguished following substorm onsets. Undoubtedly the 1999 data base of auroral roars at South Pole may find further uses, and the semiautomated auroral roar identification described above may prove useful analyzing future data from South Pole or other stations, adding further to knowledge of radio emissions of natural auroral origin. [18] Acknowledgments. The authors acknowledge significant contributions from M.L. Trimpi, who designed and constructed the LF/MF/HF radio receiver deployed at South Pole Station, and Dartmouth undergraduate Julie Sleison who performed much of the semiautomated data analysis leading to the database used in the paper. The work at Dartmouth College was supported by NSF grant OPP The University of Maryland receives support under grants OPP and OPP We thank Umran Inan of Stanford University and Louis Lanzerotti of Lucent Technologies for use of South Pole VLF and magnetometer data, respectively. We further thank the winter-over science technicians at South Pole Station that operated and maintained the instruments during [19] Janet G. Luhmann thanks Paul J. Kellogg and another referee for their assistance in evaluating this paper. References Bell, T. F., and H. D. Ngo, Electrostatic waves stimulated by coherent vlf signals propagating in and near the inner radiation belts, J. Geophys. Res., 93, 2599, Bell, T. F., and H. D. Ngo, Electrostatic lower hybrid waves excited by electromagnetic whistler mode waves scattering from planar magnetic field aligned plasma density irregularities, J. Geophys. Res., 95, 149, Gough, M. P., and A. Urban, Auroral beam/plasma interaction observed directly, Planet. Space Sci., 31, 875, Hughes, J. M., and J. LaBelle, The latitude dependence of auroral roar, J. Geophys. Res., 103, 14,911, Hughes, J. M., and J. LaBelle, First observations of flickering auroral roar, Geophys. Res. Lett., 28, 123, 2001a. Hughes, J. M., and J. LaBelle, Plasma conditions in auroral roar source regions inferred from radio and radar observations, J. Geophys. Res., 106, 21,157, 2001b. Hughes, J. M., J. LaBelle, and J. Watermann, Statistical and case studies of 2f ce auroral roar observed with a medium frequency interferometer, J. Geophys. Res., 106, 21,147, James, H. G., E. L. Hagg, and L. P. Strange, Narrowband radio noise in the topside ionosphere, AGARD Conf. Proc., AGARD-CP-138, , Kaufmann, R. L., Electrostatic wave growth: Secondary peaks in measured auroral electron distribution functions, J. Geophys. Res., 85, 1713, Kellogg, P. J., and S. J. Monson, Radio emissions from the aurora, Geophys. Res. Lett., 6, 297, Kellogg, P. J., and S. J. Monson, Further studies of auroral roar, Radio Sci., 19, 551, LaBelle, J., and J. M. Hughes, Observations of auroral roar emissions at polar cap latitudes: Results from the Early Polar Cap Observatory, Radio Sci., 36, 1859, LaBelle, J., and R. A. Treumann, Auroral radio emissions, 1, Hisses, roars, and bursts, Space Sci. Rev., 99, in press, LaBelle, J., A. T. Weatherwax, M. L. Trimpi, R. Brittain, R. D. Hunsucker, and J. V. Olson, The spectrum of LF/MF/HF radio noise at ground level during substorms, Geophys. Res. Lett., 21, 2749, LaBelle, J., M. L. Trimpi, R. Brittain, and A. T. Weatherwax, Fine structure of auroral roar emissions, J. Geophys. Res., 100, 21,953, LaBelle, J., A. T. Weatherwax, J. Perring, E. Walsh, M. L. Trimpi, and U. Inan, Low-frequency impulsive auroral hiss observations at high geomagnetic latitudes, J. Geophys. Res., 103, 20,459, Lotko, W., and J. E. Maggs, Amplification of electrostatic noise in cyclotron resonance with an adiabatic auroral beam, J. Geophys. Res., 86, 3449, Makita, K., Vlf/lf hiss emissions associated with aurora, Mem. Natl. Inst. Polar Res., 16, 1, Shepherd, S. G., J. LaBelle, and M. L. Trimpi, The polarization of auroral radio emissions, Geophys. Res. Lett., 24, 3161, Shepherd, S. G., J. LaBelle, R. A. Doe, M. McCready, and A. T. Weatherwax, Ionospheric structure and the generation of auroral roar, J. Geophys. Res., 103, 29,253, 1998a. Shepherd, S. G., J. LaBelle, and M. L. Trimpi, Further investigation of auroral roar fine structure, J. Geophys. Res., 103, 2219, 1998b. Shepherd, S. G., J. LaBelle, C. W. Carlson, and G. Rostoker, Latitudinal dynamics of auroral roar emissions, J. Geophys. Res., 104, 17,217, Weatherwax, A. T., J. LaBelle, M. L. Trimpi, and R. Brittain, Ground-based observations of radio emissions near 2f ce and 3f ce in the auroral zone, Geophys. Res. Lett., 20, 1447, Weatherwax, A. T., J. LaBelle, M. L. Trimpi, R. A. Treumann, J. Minow, and C. Deehr, Statistical and case studies of radio emissions observed near 2f ce and 3f ce in the auroral zone, J. Geophys. Res., 100, 7745, Yoon, P. H., A. T. Weatherwax, and T. J. Rosenberg, On the generation of auroral radio emissions at harmonics of the lower ionospheric electron cyclotron frequency: X, O and Z mode maser calculations, J. Geophys. Res., 103, 4071, 1998a. Yoon, P. H., A. T. Weatherwax, T. J. Rosenberg, J. LaBelle, and S. G. Shepherd, Propagation of medium frequency (1 4 Mhz) auroral radio waves to the ground via the z-mode radio window, J. Geophys. Res., 103, 29,267, 1998b. Yoon, P. H., A. T. Weatherwax, and J. LaBelle, Discrete electrostatic eigenmodes associated with ionospheric density structure: Generation of auroral roar fine frequency structure, J. Geophys. Res., 105, 27,589, J. LaBelle, Department of Physics and Astronomy, Dartmouth College, Hanover, NH , USA. (jlabelle@einstein.dartmouth.edu) A. T. Weatherwax, Institute for Physical Science and Technology, University of Maryland, College Park, MD 20771, USA. (allanw@ipst. umd.edu)