By continuing to browse this site you agree to us using cookies as described in About Cookies
Notice: Wiley Online Library will be unavailable on Saturday 7th Oct from 03.00 EDT / 08:00 BST / 12:30 IST / 15.00 SGT to 08.00 EDT / 13.00 BST / 17:30 IST / 20.00 SGT and Sunday 8th Oct from 03.00 EDT / 08:00 BST / 12:30 IST / 15.00 SGT to 06.00 EDT / 11.00 BST / 15:30 IST / 18.00 SGT for essential maintenance. Apologies for the inconvenience.
 The results of some recent experiments employing the EISCAT HF “heating” facility at Ramfjordmoen, near Tromsø, Norway as a HF radar transmitter are described. The motivation for the experiments was the detection of “conjugate echoes” caused by geomagnetic field-aligned ducting of the HF wave in the magnetosphere and reflection from the magnetically conjugate ionosphere. No such echoes were detected during the experiments, which is probably to be expected from consideration of the plasma density gradients required to sustain guidance of the waves at the low HF frequencies involved. However, echoes were obtained at ranges which could be consistent with backscattering from ionospheric irregularities in the equatorial and southern auroral regions; this is similar to spread-Doppler clutter observed by large over-the-horizon radar (OTHR) systems. The experimental technique is described and the results discussed with a view toward future attempts to sound the lower magnetosphere using high-power HF transmitters.
 Radar remote-sensing of the ionosphere is a well-established technique in geospace science. Methods relying on coherent backscatter from plasma density irregularities [Greenwald et al., 1978, 1995] and on “incoherent” backscatter from thermal fluctuations in the plasma [Evans, 1969] are some of the most important diagnostics. Radar techniques can survey large areas of space with high temporal and spatial resolution and can measure several parameters of the plasma. Compared to spacecraft, they are relatively inexpensive to build and operate. If it was possible to apply the same technique to sensing of the magnetosphere, these same benefits might be gained in a region which is otherwise the preserve of satellite missions. The prime reason why it is difficult to apply the radar technique to the magnetosphere is the very low plasma density found there which means that the backscatter cross-section is extremely small, regardless of whether coherent or incoherent scattering is involved. Coupled with the large range to the target, this would necessitate transmitter powers and antenna gains far beyond those of any currently operating incoherent scatter radar. Furthermore, the large Debye length of the plasma would demand a long probing wavelength which might be reflected by the intervening ionosphere.
 Despite these fundamental difficulties, at least two possibilities exist for radar remote-sensing of the magnetosphere. A method which is known to work relies on guidance of the electromagnetic wave by geomagnetic field-aligned plasma density irregularities or “ducts”. This phenomenon is perhaps best-known from satellite topside sounder measurements [Knecht et al., 1961; Muldrew, 1963, 1967, 1980; Sharma and Muldrew, 1973, 1975] but can also be observed on the ground [Bukin, 1978; Gurevich and Tsedilina, 1985; Ellis and Goldstone, 1987, 1988, 1990]. Once the wave enters such a duct, it is guided along it until the plasma density becomes sufficiently high to reflect it back along the duct, or sufficiently weak that it leaves the duct again. In the case of ground-based sensing, a wave launched from the ground may penetrate the ionosphere, enter a duct and be guided to the magnetically conjugate ionosphere where it is either reflected or again penetrates, possibly reflecting from the ground. Thus a “conjugate echo” may be detected at the transmitting site. For satellite-based sounders, the satellite typically launches the wave directly into the duct and may then detect echoes from the reflection of the wave from the ionosphere at either end of the duct. The theory of propagation in field-aligned ducts has been explored by Booker  and Gurevich and Tsedilina .
 If it is known that ducting is taking place, it is possible from the theory to obtain an estimate of the size of the density perturbation supporting the ducting and thus obtain a lower bound on the plasma density in the magnetosphere. Furthermore, Ellis and Goldstone [1987, 1988] observed modulation of the delay times of the conjugate echoes which, in some cases, could indicate drift of the duct caused by hydromagnetic waves in the magnetosphere.
 A second possibility of remote sensing is closely allied to incoherent scatter. Gurevich et al.  suggested that it might be possible to detect backscatter from ion acoustic waves excited by plasma instabilities in the auroral acceleration region at altitudes around 4000 km. These waves, being driven by instabilities, would have amplitudes much greater than the amplitudes due to fluctuations in a thermal plasma, making them detectable using high power HF transmitters such as ionospheric heaters. They described an experiment using the Sura HF facility near Nizhny Novgorod, Russia, to detect these waves. Some of the results appeared to show spectra consistent with backscatter from ion acoustic waves. Greenwald  pointed out that the assumption of the alternating code modulation used [Lehtinen and Häggström, 1987], that signals from disjoint scattering volumes are uncorrelated, would be likely to fail in the presence of strong, narrow-spectrum ground scatter returns at similar ranges to the wanted target resulting in spectral artifacts which might look like the wanted ion acoustic spectrum. This was confirmed by Hysell et al.  who concluded that none of the observations made with the Sura facility alone showed convincing evidence of backscatter from ion acoustic waves. On the other hand, the observations made using the UTR-2 radio telescope at Kharkov, Ukraine to receive the Sura transmissions appeared more positive, with spectral features of possible geophysical interest being observed at ranges which might correspond to the auroral acceleration region.
 This article describes some experiments using the EISCAT HF Facility, located at Ramfjordmoen near Tromsø, Norway, as a radar to detect conjugate echoes at high latitudes. The experiments were partly about gaining experience in using a high power HF transmitter of this type in radar service. No evidence of conjugate echoes was found, but echoes in the 6000–17,000 km range interval were observed for which possible causes are discussed. Suggestions are made for future experiments aimed at detecting enhanced plasma wave activity in the auroral zone. The HF Facility has been used in radar service before in conjunction with the technique of artificial periodic irregularities [Rietveld et al., 1996] and some consideration has already been given to adapting the facility for better performance as a radar by converting one of the transmitting arrays for reception Rietveld et al., “An HF magnetospheric radar for EISCAT?”, poster presented at the EISCAT Workshop, Leicester, U.K., June 1997.
2.1. Transmitting System
 The EISCAT HF facility is located at Ramfjordmoen, near Tromsø, Norway (69.6°N, 19.2°E). The facility has been described by Rietveld et al. . Twelve transmitters rated at 100 kW RF output can be switched between three antenna systems consisting of phased arrays of crossed-dipoles. Array 1 covers 5.5–8 MHz with a typical gain of 30 dBi, array 2 covers 4–5.5 MHz with a typical gain of 24 dBi and array 3 covers 5.5–8 MHz also with a typical gain of 24 dBi. The actual gain is frequency dependent and the figures quoted later in the text were determined using commercial electromagnetic modeling software. The interconnection of the transmitters and antennae allows the polarization to be selected between O-mode, X-mode or linear and for the beam to be swept in the north-south direction, but not in the east-west direction. The beam sweeping is performed by programming the transmitter output phases and is therefore almost instantaneous.
 The control system normally used for modulating the transmitters for ionospheric pumping experiments is not suitable for radar operations on account of the available timing precision. Therefore a dedicated radar controller unit was built for this purpose. The controller interfaces to the transmitter through two logic control lines, one of which switches the RF output on and off and the other which inverts the RF phase, permitting 0/180 degree phase modulation of the transmission. The timing reference was the on-site UTC clock. The “off” state obtained using this control method does not reduce the RF output to zero. In fact a reduction of 90–96 dB below full output was obtained, corresponding to a power of 0.25–1 mW for a nominal 1 MW full output. This leakage signal is potentially significant compared to the weak signal which might be obtained from backscatter targets.
2.2. Receiving System
 Since conjugate echoes were the main focus of the work and since the echo power was expected to be reasonably strong (due to the focused nature of the propagation) it was felt that reception from a remote site using a simple antenna would be sufficient. The use of a remote site avoids the need for a transmit/receive switch and gives some measure of isolation from the transmitter leakage signal, although only that part carried by surface wave propagation.
 Over the course of the experiments, carried out in March 2003 and June 2004, three different reception sites were used. The first site, used in 2003, was at Skibotn, Norway (69.35°N, 20.36°E), about 50 km east of the transmitter. At this site, a broadband resistively loaded dipole antenna approximately 2 m above ground was used. The second site, also used in 2003 was at Seljelvnes, Norway (69.25°N, 19.43°E) about 50 km south of the transmitter. At this site, two half-wave dipoles, 10 m above ground, for 4 and 5 MHz were available for use. The third site, used in 2004, was near Breivikeidet, Norway (69.6°N, 19.5°E) about 13 km northeast of the transmitter. Here, a broadband resistively loaded dipole antenna was used again, approximately 2 m above ground.
 The receiver was a conventional single-sideband communications receiver, an Eddystone model 1650. The receiver was tuned to a frequency offset from that of the transmitter, typically by 1.5–2 kHz so that the transmitter signal was converted to an audio frequency signal centered on 1.5–2 kHz. The audio signal was digitized by a computer “sound-card”. The software recorded the raw audio samples to file allowing arbitrary off-line processing after the experiments and provided basic real-time displays of the received signal to monitor the progress of the experiment.
 Sample timing accuracy and stability was derived from GPS. In 2003, the receiver frequency was controlled by a high-stability reference oscillator. In 2004 this was replaced with a GPS-disciplined source.
Table 1 summarizes the experimental parameters of each observing run. In what follows, the runs are referred to by the number in the first column of the table. Table 2 gives the geophysical conditions during the runs.
Zenith angle of transmitter beam, or 6-position scan with 10 s dwell on each position, starting on UT minute boundary. S = 6°, 12°, 18°, 24°, 30°, 36°S; S–N = 15°S, 9°S, 3°S, 3°N, 9°N, 15°N.
Effective radiated power, relative to an isotropic radiator.
Pulse length and Inter-pulse period (IPP).
13 ms/1 s
13 ms/1 s
13 ms/1 s
Barker 13 ms/1 s
13 ms pair/1 s
13 ms pair/1 s
13 ms pair/1 s
1 ms/200 ms
1 ms/200 ms
1 ms/200 ms
Table 2. Geophysical Conditions During the Observing Runsa
The critical frequencies were measured by the EISCAT Dynasonde at Ramfjordmoen.
3− (18 UT)
5+ (18 UT)
5, 6, 7
3+ (18 UT)
4 (21 UT)
peak at 21:06
8, 9, 10
3− (21, 00 UT)
3.1. Echoes With Ranges in the 6000–17,000 km Interval
 During runs 1 and 2, in addition to ground scatter returns at short ranges (<4000 km), echoes were observed at ranges of around 6500 and 16,700 km between 19:32 and 19:45 UT. The exact start of the echoes was probably masked by interference. The echoes seemed to disappear almost simultaneously. Conjugate echoes would be expected around ranges of 700–800 ms (105,000–120,000 km) but nothing was observed at those ranges. Indeed, nothing (other than perhaps interference) was observed at those ranges at any time during the experiments described here.
 A RTI (range-time-intensity) plot of the received power from runs 3 and 4 is shown in Figure 1. After 19:42, the 13 ms pulse was phase-coded using the 13-bit Barker code with 1 ms bauds. The transition from uncoded to coded pulses is clearly seen at 19:42; in this plot the entire interval was subjected to Barker-decoding. With the Barker coding, features narrow in range can be discerned in the plot. The Barker code was decoded using the traditional matched-filter approach [Nygrén, 1997] and so exhibits range-sidelobes which are 22 dB lower in level than the main lobe and this must be borne in mind when examining the plot.
 Throughout almost the entire period shown in Figure 1, returns can be seen at a range which declines from around 16,000 km to 15,300 km. The range extent appears to vary from as little as 500 km (for example, around 19:50 UT) to as much as 1250 km (around 19:57 and 20:10 UT). Between about 20:06 and 20:19 UT, returns of narrower range extent can be seen at ranges of 6150 and 8250 km. These returns might also have been briefly present at 19:56 UT. The presence of these additional returns seems to coincide with a period of enhanced strength in the longer-range return. The gap in the plot at 19:51–19:52 UT was due to a transmitter outage.
 At this point a note of caution in the interpretation of the data presented in Figure 1 must be sounded. Since the frequency spectrum of the returns is not known, it is not possible to be sure that the use of Barker-coding is justified. Ideally, for the Barker code to be correctly decoded, the channel and target must not impose any phase distortion on the pulse. Such distortion, which could arise due to Doppler shifting or spreading of the spectrum causes the decoding to deteriorate and this tends to result in apparent range-spreading of the target. Therefore it is not certain that the range extent which is visible particularly in the ∼16,000 km return is real or a consequence of failure of the Barker coding.
 The ground scatter region extends out to around 5000 km. Occasionally, fine structure can be seen in this region, for example around 19:55 UT. The strong, fine structured returns at the nearest ranges are probably a mixture of the surface wave signal and reflections from patches of plasma in the ionosphere which were overdense to the transmitted wave.
3.2. Effect of Transmitter Beam Direction
Figure 2 shows the RTI plot for the period 20:37–21:43 UT covering runs 5, 6, and 7. The modulation for these runs was a pair of pulses each 13 ms long with a gap of 13 ms between them, the pulse pair being repeated every 1 s. The purpose of this double-pulse was to give the returns a more readily identifiable appearance in the RTI plot. Because the receiver filter is also 13 ms in length, an infinitely thin target has an apparent range extent of 26 ms (3900 km). A long-range echo (∼16,000 km), showing the expected double-pulse signature, appears at about 20:52 UT and disappears rather abruptly at 21:33 UT. This disappearance coincides with the onset of an auroral absorption event, as indicated by the IRIS riometer located at Kilpisjärvi, Finland (69.05°N, 20.79°E). Only the surface wave signal from the transmitter remains during this event, except for a brief period of partial recovery around 21:38 UT. The vertical feature in the plot (showing reduced background powers) around 21:14 UT was caused by adjustments to the receiver gain. A change in the behavior of the background power can be seen from about 21:10 UT. Prior to this time, the background power (noise, or noise plus interference) is constant with range, although showing some variation with time. After 21:10 UT the background ceases to be constant with range. It is thought that this is due to “pumping” of the receiver overload protection circuit (an automatic gain control which begins to operate when the input voltage exceeds 0.5 mV) by the strong ground scatter returns at near ranges. The receiver gain then recovers as a function of range.
 Separating the recordings into different scan positions (not shown) gives an indication of the dependence of the power of the long-range echo on zenith-angle in the north-south plane. For the period 20:41–21:10 UT, the signal-to-noise ratio (SNR) maximized at 6°S at about 15 dB, falling to about 10 dB at 36°S. From 21:14–21:40 UT, after the change in beam scanning, it was found that the peak was around 9°S at 15–20 dB and the minimum at 15°N, at 10–15 dB. Overall the echo seemed to be strongest slightly to the south of the zenith. By basing these measurements on SNR rather than power, the possible gain control effects are removed, provided that the receiver gain was not reduced so far that the receiver internal noise dominated over the antenna noise.
 Another example of an echo power dependence on zenith-angle is shown in Figure 3 (runs 8, 9, and 10). In the interval 00:01:30 to 00:03:30, the transmitter was on continuously for “tuning-up” on array 3. The wavy lines in the background are due to electrical power supply interference. From 23:30, an echo appears above the background at a range of 9000 km. The echo then appears to broaden to cover the range interval 8000–9000 km and shows very strong modulation with the scanning. Toward 00:00 UT, it becomes apparent that the echo consists of two peaks, one at 8200 km and the other at 8800 km. The echo appeared to be strongest when the beam was 15°N. The transmitter was switched to array 3 (which has a different radiation pattern) after 00:00 UT and operated with the beam 15°N. The echo remained clear after the switch and persisted until the transmission stopped at 00:20. Note that the echo disappears after the transmitter stopped and is thus not due to interference; the RTI plot continues until 00:21 UT.
 When the scan positions are separated, it is found that the echo is essentially undetectable except at beam directions of 9° and 15°N of the zenith. At 9°N the SNR was around 0 dB and at 15°N around 5 dB.
3.3. Other Characteristics
 During the experiments in 2003, long-range echoes were observed on a variety of frequencies in addition to those presented here: 4.5 MHz (2003-03-03 20:42–20:45 UT, ∼15500 km), on 6.77 MHz (2003-03-01 20:40–20:52 UT, ∼6500 and ∼9500 km) and on 6.96 MHz (2003-03-03 21:28–21:33 UT, ∼15,500 km). The ranges correspond to the ranges observed on 4.9 or 5.4 MHz around the same time. The modulation used in these experiments was not very suitable for determining the target spectrum. By treating the 13-ms pulse as a “long pulse” in the incoherent scatter sense, an autocorrelation function (ACF) can be calculated. This was tested for one case at 21:16 UT on 2003-03-04. It was found that the signal ACF decorrelated rather slowly, the power falling to about 70% of the zero-lag value by a lag of 10 ms. Assuming an exponential decay in power, the time-constant is 28 ms, corresponding to a power spectrum of width (FWHM) 11 Hz. Since the ACF decay time constant is more than twice the pulse length, this figure is likely to be inaccurate. The ACF phase was approximately linear with lag and consistent with a Doppler shift of −7 Hz. The same Doppler shift was observed on the surface wave signal and arises from a frequency offset between transmitter and receiver during the 2003 experiments.
3.4. Ground Backscatter Measurements
 As was noted in the Introduction, part of the purpose of these experiments was a more general exploration of the use of the HF Facility as a radar. With regard to its possible use for auroral sounding as described by Gurevich et al.  and Hysell et al. , it was felt that an investigation of the ground scatter returns, which have been largely ignored so far, would be useful. In 2004, a modulation consisting of a 0.5 ms pulse every 40 ms was employed to measure the spectrum of these returns as a function of range. An example result is shown in Figure 4. The sounding was taken at 7.953 MHz with the beam directed 15°S of vertical. Antenna array 1 was used with 12 transmitters delivering 80 kW each. The polarization was O-mode. fOF2 was 4.7 MHz and fOE was 2.8 MHz.
 The surface wave appears as an intense dot at zero range and zero frequency. At a range of ∼150 km, a band of power appears. This may be due to scatter from E-region irregularities. The next band at ∼700 km range is probably ground backscatter from a 1-hop E-region path and the subsequent bands are probably ground backscatter from 1-, 2-, 3-, and 4-hop F region paths with the latter at nearly 5000 km range and only just discernible. It is immediately clear that most of the power in the ground scatter returns is concentrated in a relatively narrow band of width 3–4 Hz centered close to zero frequency offset from the radar frequency.
 The February–March 2003 experiment set out to look for evidence of conjugate echoes from ducting of the HF waves along magnetospheric field-aligned plasma density irregularities and reflection from the conjugate ionosphere. No evidence of this phenomenon was found. With hindsight, this is not too surprising since the theories [Booker, 1962; Gurevich and Tsedilina, 1985] suggest that ducting would require the existence of plasma density depletions which would be very large compared to (or even larger than) the ambient plasma density on field-lines of the relevant L-value (L = 6.4 at Tromsø). It is likely that these irregularities could only exist relatively rarely.
 Although conjugate echoes were not observed, the February–March 2003 results provided good evidence of some other form of “long-range echo.” Possible causes of this phenomenon will now be discussed.
4.1. Possible Instrumental Artifacts
Figure 2 indicates that the echo disappeared during auroral absorption although the surface wave signal was still present at a high level. If the echo was due to an instrumental artifact (perhaps extraneous pulses from the transmitter, or some signal processing problem in the receiver) one would still expect to be able to see the echo due to the surface wave signal. In the June 2004 results, Figure 3 indicates that the strength of the echo depended strongly on the direction of the transmitter beam and that it disappeared when the transmission was stopped. These factors are evidence against the echo being a result of interference.
4.2. Estimated Scattering Cross-Section
 Knowing the characteristics of the radar system and the power received, the radar scattering cross-section (RCS) can be estimated from the radar equation. Regrettably, no absolute power calibration of the receiving system was available. However, a crude estimate of the power can be obtained by looking at the noise level observed. It will be assumed that the receiver's internal noise is low enough to be neglected, which is normally a reasonable assumption for modern receivers at low HF frequencies. Transmission line losses are also neglected.
 Estimates of radio noise are given by ITU-R Recommendation P.372-8. According to this document, the median noise level due to atmospheric noise at Tromsø during the 2003 experiments would be −174 dBW for 16–20 LT increasing to −154 dBW for 20–24 LT. These powers are in a 1 Hz bandwidth at 5 MHz. The standard deviations are around 4 dBW. These figures assume the “spring” season. The estimated galactic noise at 5 MHz is −168 dBW. For man-made noise, the “rural” or “quiet rural” environment would apply. In the former case, the power would be −156 dBW, for the latter it would be less than the galactic noise at −170 dBW. Roughly, the total noise power level would be in the −170 to −150 dBW range. Figure 2 suggests that the noise level was rather variable.
 The reception of the 13 ms pulse used a filter matched to the pulse length giving a bandwidth of about 150 Hz. The noise power in this bandwidth would then be in the range −148 to −128 dBW or about 2 × 10−15 to 2 × 10−13 W. It was noted earlier that for the results in Figure 2 the SNR was up to about 15 dB. Hence the signal power was in the range 6 × 10−14 to 6 × 10−12 W.
 Since the range to the target (16,000 km) greatly exceeds the separation of the transmitter and receiver (50 km or less) it is reasonable to assume that transmitter and receiver are colocated. Furthermore, as the ionospheric critical frequency was (mostly) below the radar frequency, most of the transmitted power can be expected to have been escaping into the magnetosphere as intended. Therefore the RCS will be estimated assuming the target is in the main lobe of the transmitter radiation pattern.
 Rearranging the radar equation for a “hard” target [Barton, 1967] for the RCS:
 Now for the 2003 experiment typical values for the radar parameters are Pt = 640 kW (transmitter power), Gtmax = 160 (peak transmitter antenna gain, ∼22 dBi), Grmax ≈ 2 (peak receiver antenna gain) and λ = 60 m (wavelength). Taking R = 16,000 km (target range) gives σ = 1 × 107 m2 to 1 × 109 m2. These are roughly equivalent to the area of a square of side 3–30 km.
 For volume scattering, the equation for the received power is [e.g., Evans, 1969]:
where σV is the per-unit-volume RCS, assumed constant throughout the scattering volume, Gt, Gr are the transmitter and receiver antenna gain functions relative to an isotropic radiator and θ, ϕ are the zenith angle and azimuth angle.
 In the experiments described here, the transmitting antenna was very much more directional than the receiving antenna which was a dipole over ground. In this case, Gr is approximately constant over the main lobe of the transmitter radiation pattern and can be taken as the peak value Grmax. The transmitting antenna gain is well-approximated by the function
where k = 2.783/Δθ and Δθ is the −3 dB beam width in radians, although the implied azimuthal symmetry is not strictly correct for a square, filled phased-array.
 With these approximations, and taking Δθ = 14.1° the integral in (2) takes the value 31. Taking T = 13 ms gives σV = 2.2 × 10−13 m−1 to 2.2 × 10−11m−1. It is less obvious whether this is large or small. As a benchmark, it can be compared to the corresponding value for incoherent scattering from a plasma. According to Evans  this is σV = 4πre2Ne where re = 2.82 × 10−15 m is the classical electron radius and Ne is the electron density. This expression holds for the case where the radar wavelength is of the order of the Debye length or less. If the wavelength is much greater, the RCS is less than this value. For an altitude of 16,000 km in the auroral zone, a value of 1 cm−3 for Ne seems plausible and this gives σV ≈ 10−22 m−1. Hence the RCS is about 9–11 orders of magnitude greater than that for incoherent scatter. It should be stressed that, given the uncertainty in the estimate of the noise level, these estimates of the RCS could easily be in error by two orders of magnitude.
4.2.1. Artificial Satellites
 The repeatability of the ∼16,000 km echoes observed in 2003, both in terms of the near-constancy of its range and observation on three almost consecutive nights led to the suggestion that they might be due to artificial satellites. Counting against this hypothesis is the estimated RCS, although some objects might have resonant structures leading to enhanced cross-sections [Weinstock, 1967]. Also, the observations made with the Barker coding suggest that the target might have significant range-extent (several hundred kilometers) but it must be borne in mind that the Barker code might have been affected by Doppler shifting (satellite motion) or spreading (ionospheric scintillation).
 In an attempt to eliminate an artificial satellite as the cause, a search of all objects whose orbital elements are in the public domain was carried out. The search looked for objects penetrating the field-of-view (FOV) defined by the main lobe of the transmitter beam (−3 dB point) and assuming no ionospheric refraction of the beam. Runs 1–10 were checked. Although at least one object passed through the FOV on each day, none of them had range-time characteristics which matched the observed echoes. There remains, of course, a small risk that an object or objects whose orbital elements are classified might have been responsible for the echoes. Consideration of typical orbital characteristics suggests that it is quite unlikely that any object would remain so constant in range and in the FOV (especially considering the narrow transmitter beam width) for so long a period as observed. It appears, therefore, very unlikely that artificial satellites could have been responsible for the observations.
4.2.2. Mode Conversion in the Ionosphere
 An O-mode wave can be mode-converted to the Z-mode at the reflection level. This Z-mode wave can reflect at a higher altitude and as it propagates back toward the O-mode reflection level, it becomes converted into an electrostatic (Langmuir) wave [Mjølhus and Flå, 1984; Gurevich and Tsedilina, 1985]. The electrostatic wave has a smaller group velocity than the incident electromagnetic wave. If the electrostatic wave transforms back through the reverse process into an electromagnetic wave which leaves the ionosphere for the ground, the total delay time could be significant even though the distances involved are only a few hundred kilometers. If an X-mode wave is transmitted instead of an O-mode wave then the coupling to Z-mode cannot take place and an echo due to this mechanism would disappear. On 2004-06-10, an echo was being received at a range of 8500 km in the interval 20:31–20:50 UT (data not shown). Initially, O-mode polarization was used but was temporarily switched to X-mode. No change was detected in the echo strength. This strongly suggests that the mode conversion mechanism was not responsible in this case.
4.2.3. Ionospheric Backscatter
Figure 5 shows how the ranges of the echoes relate to location on the Earth both geographically and geomagnetically. The geomagnetic reference points are the approximate locations of the northern and southern auroral ovals according to the parametrization by Holzworth and Meng  and the region within 20° corrected geomagnetic latitude of the magnetic equator [Gustafsson et al., 1992]. The auroral ovals are shown for 22 UT and moderate geomagnetic activity. Concentrating on the 16,000 km echo, it can be seen that the range rings denoting the limits of this echo run near-parallel to the southern auroral zone over about an eighth of their circumference. Similarly, there is a significant overlap between the shorter-range echoes and the magnetic equatorial region. Both the auroral and equatorial regions are well-known to exhibit high levels of electron density irregularities in the ionosphere [Fejer and Kelley, 1980; Tsunoda, 1988].
 HF over-the-horizon radar (OTHR) systems often observe “spread-Doppler clutter”, which is frequently attributed to the equatorial zone [Buchau et al., 1994; Dandekar et al., 1998]. In the OTHR case, such clutter is often observed at great ranges (10,000 km or more) and is range-aliased into the working range interval. The clutter also appears rather constant in range with frequency. These properties are similar to the echoes reported here. A typical range-extent for the clutter is about 2800 km (1500 nautical miles) which is greater than in the present case. The Doppler-spreading associated with the clutter signal is generally of the order of a few hertz. This is consistent with the crude estimate of the target spectrum described earlier.
 OTHR clutter is also observed from the auroral zones [e.g., Choi et al., 1991]. Coleman  reported that the Jindalee backscatter sounder in Australia observed echoes from the northern auroral regions (12,000 km range) and provided a ray-tracing simulation of the propagation. The observations show that returns are received at the same range over the 5–11 MHz range (the sounder lower-limit apparently being 5 MHz). This seems to support the idea that the longest-range echoes described in the present work originate from the southern auroral zone.
 In runs 1 and 2 (not shown) and runs 3 and 4 (Figure 1) it appears that the strength of the medium- and long-range echoes varies together. In Figure 1, the medium-range echoes are only evident when the long-range echo is at its strongest. This might seem to suggest that the cause of the two echoes is linked, which would go against the hypothesis that they are due to scatter from two very different regions of ionospheric irregularities. However, it is quite probable that the variation in echo power is due to changes in ionospheric absorption local to the transmitter and receiver. An extreme example of this is the auroral absorption event in Figure 2.
4.3. Clutter Cancellation
 In the experiments described by Gurevich et al.  and Hysell et al. , the ground backscatter was a serious radar clutter problem. It has long been known, e.g., [Headrick and Skolnik, 1974] that the ground backscatter returns have a narrow spectrum and this is confirmed by the results presented earlier. The clutter signal has most of its energy concentrated within about ±2 Hz of zero frequency. In contrast, the backscatter from ion acoustic waves sought by Gurevich et al.  exhibits spectral peaks at several hundred hertz away from zero frequency.
 This dichotomy suggests a method by which the clutter signal can be eliminated. Considering the sample vectors obtained from a pair of consecutive pulses, the clutter signal will have changed very little between pulses because the bandwidth is small compared to the pulse repetition rate, whereas the wanted signal will be completely different. Subtracting the pair of sample vectors range-for-range will thus nearly cancel out the clutter contribution, leaving the wanted signal. This technique has been employed in the EISCAT Svalbard Radar to eliminate strong ground clutter returns [Turunen et al., 2000]. Once the clutter signal has been removed or attenuated in this way its impact on the decoding of the alternating code for recovering the wanted target spectrum is eliminated or reduced. It is suggested that this type of technique could be of help in any future experiments to investigate the use of high-power HF transmitters for auroral magnetospheric sounding. The method would also cancel out any leakage signal from the transmitter when in the “off” state. A more directional receiving antenna would also help to suppress the unwanted ground scatter returns, but due to their intensity this alone is unlikely to be sufficient to eliminate them.
4.4. Detectability of Magnetospheric Backscatter
 Finally it is worth briefly considering how detectable scatter from the plasma in the lower auroral magnetosphere might be. Concentrating on altitudes around 4000 km, a typical value for the electron density Ne is 50 cm−3 and for the electron temperature Te is 2 eV, equivalent to about 20,000 K [Kletzing et al., 1998]. For these values, the plasma Debye length D is approximately 1 m. The ratio α = 4πD/λ [Evans, 1969] is then about 0.4 for a radar wavelength λ of 40 m. This means that scatter would be received mainly from the ionic component of the spectrum, as in typical incoherent scatter radars such as EISCAT. In this case, the RCS per electron is no longer the value 4πre2 valid for true incoherent scatter, but is less by a factor which depends on the electron-to-ion temperature ratio Te/Ti. According to Evans  for Te = Ti the RCS is reduced by half and for higher values of Te/Ti ≈ 5 the RCS is scaled by a factor of 0.2 but then rises for higher Te/Ti. Assuming the worst case and taking a factor of 0.2, the RCS per unit volume σV = 0.2 × 4πre2Ne ≈ 10−21 m−1.
 For the radar parameters, λ = 40 m, corresponding to 7.5 MHz has already been chosen. It will be assumed that the EISCAT HF facility is being used with the full output power Pt = 1 MW, that array 1 is used for transmission and that array 3 (covering the same frequency range) is used for reception. The antenna gains can be modeled using the expression (3). For array 1, take Gtmax = 31 dBi and Δθ = 5°. For array 3, take Grmax = 25 dBi and Δθ = 10°. With these parameters, the integral in the radar equation (2) becomes 2.67 × 103. A reasonable pulse length for the altitudes involved would be T = 1 ms. Putting these into the radar equation a received power Pr of 2 × 10−20 W is obtained. The half-width of the spectrum is given approximately by Δ fi = (1/λ)(8kTi/mi)1/2 Hz [Evans, 1969] which for an estimated ion temperature of 4000 K and assuming the ion-mass mi is the proton mass gives Δ fi ≈ 400 Hz. So the spectrum occupies a bandwidth of about 800 Hz. The modulation will occupy a slightly wider bandwidth of 1 kHz. Taking the total bandwidth as 2 kHz and the noise power per unit bandwidth as −170 dBW as earlier, the total noise power is 2 × 10−14 W. Therefore the SNR is of the order of 1 × 10−6, which does not look very promising.
 This figure, of course, is based on scattering from the thermal fluctuations in the plasma. If plasma waves are strongly excited, the RCS can be greatly increased. There have been many observations by incoherent scatter radar of naturally enhanced ion-acoustic waves in the ionosphere, some as high as ∼1000 km altitude [Sedgemore-Schulthess and St.-Maurice, 2001] and these enhancements can be of 1 or 2 orders of magnitude. If the calculation above is repeated for conditions more typical of 1000 km altitude (Ne = 104 cm−3, Te = 5000 K, Ti = 2000 K, O+ ions) the RCS becomes σV ≈ 3 × 10−19 m−1 and the SNR ≈ 0.01. Again, this is for thermal fluctuations in the plasma. It is now clear that with enhancements of the level observed at UHF and VHF, the SNR could reach unity. Therefore the prospect of detecting enhanced ion-acoustic waves at 1000 km altitude is more hopeful. Such observations might contribute useful information on the wave number spectrum of the waves excited in these events, giving an additional clue to the mechanism responsible for them.
 The accuracy of the above calculations depends to a large extent on the reliability of the noise level estimate. A lower limit was used in the calculations above, therefore the estimate may be optimistic. This estimate was determined for 5 MHz. At 7.5 MHz, the galactic and man-made noise is likely to be several decibels less than at 5 MHz but on the other hand, the atmospheric noise may exceed that at 5 MHz by several decibels, according to ITU-R P.372-8.
 Experiments using the EISCAT HF facility as a radar transmitter with the original aim of detecting evidence of conjugate echoes at high latitudes have been described. No evidence of these echoes was found, which is probably to be expected from theoretical consideration of the plasma density required to support suitable magnetospheric ducts [Booker, 1962; Gurevich and Tsedilina, 1985]. However, this does not rule out the occasional existence of ducts in this dynamic region of the magnetosphere.
 Echoes with ranges up to about 16,000 km were observed which seem most likely to be due to ionospheric propagation and backscattering of the wave from large areas of irregular plasma, such as may be found in the auroral and equatorial zones. This is similar to the spread-Doppler clutter observed in over-the-horizon radar systems.
 The characteristics of the ground scatter returns at near ranges and their significance as clutter targets when attempting to detect possible scatter from plasma waves in the magnetosphere (4000 km) have been discussed. It has been suggested that a simple clutter cancellation technique might prove useful in mitigating this problem. It seems that a rather considerable enhancement (4 or more orders of magnitude) of plasma wave amplitude above the thermal level would be required to create detectable backscatter using the EISCAT HF facility for transmission and reception. On the other hand, at lower altitudes (1000 km) the possibility of detecting such backscatter seems much higher, if enhancements are similar to those observed at UHF and VHF.
 EISCAT is an International Association supported (at the time of these experiments) by Finland (SA), France (CNRS), Germany (MPG), Japan (NIPR), Norway (NFR), Sweden (VR), and the United Kingdom (PPARC). IRIS is funded by PPARC. We are grateful to the Radio and Space Plasma Physics Group, University of Leicester for allowing us to use their Seljelvnes field site; to the Dept. of Physics, University of Tromsø for the use of their Skibotn field station; to P. Martinez for assistance with testing the receiving software; and to B. Isham and M. Pinnock for helpful comments. During most of this work, A. S. was supported by a PPARC research studentship and is now supported by PPARC (now STFC) grant PP/C000218/1.