By continuing to browse this site you agree to us using cookies as described in About Cookies
Wiley Online Library is migrating to a new platform powered by Atypon, the leading provider of scholarly publishing platforms. The new Wiley Online Library will be migrated over the weekend of March 17 & 18. You should not experience any issues or loss of access during this time. For more information, please visit our migration page: http://www.wileyactual.com/WOLMigration/
 We present polar mesospheric radar observations at three frequency bands (MF/HF/VHF) and eight radar frequencies: 2.43, 3.3, 4.53, 4.9, 7.6, 28, 50, and 139 MHz, in order to better understand the well known but still not fully understood Polar Mesosphere Summer Echoes (PMSE). The echo morphology at the different frequencies is described by means of case studies where PMSE events were observed concurrently using at least two radar systems deployed over the Alaskan central region. The identity of MF and HF radar echoes as PMSE is resolved for the first time by means of simultaneous measurements made with VHF radars, the reference sensors employed traditionally for PMSE studies. On the basis of echo duration and radar reflectivity estimates, we suggest that low-power HF radars would be more appropriate for PMSE monitoring. This is confirmed by a radar target analysis of turbulent scattering mechanisms in the polar summer mesosphere. MF radars show highly organized PMSE layers quite often but are more susceptible to ionospheric absorption and higher-altitude returns associated with geomagnetic activity. Both phenomena produce a blanking effect in MF PMSE, which at times can persist for hours. HF and VHF radars are less affected by absorption events, but the PMSE echoes become weaker as the radar frequency increases.
 Polar Mesosphere Summer Echoes (PMSE) have been studied for over 25 years, beginning with the pioneering works of Czechowsky et al.  and others in northern Europe and Ecklund and Balsley  at Poker Flat, Alaska. Balsley et al.  published two consecutive years of 50-MHz radar data (i.e., 1979–1980) taken at the Poker Flat Research Range. The surprising result was that, with their relatively modest-sized radar, they detected remarkably strong echoes centered near 85 km during the summer. At other times of the year sporadic and weaker mesospheric echoes were detected but only below 80 km. Since that time, numerous radar studies have taken place mostly at frequencies at or above the 50 MHz used by Ecklund and Balsley. Mesospheric echoes have been reported at 49.6 MHz [Röttger et al., 1990a, 1990b], 51.5 MHz, 53.5 MHz, 224 MHz [Röttger et al., 1988], 500 MHz [Hall and Röttger, 2001], 933 MHz [Röttger et al., 1990a, 1990b], and 1290 MHz [Cho et al., 1992].
 Strong mesospheric radar echoes in the VHF and UHF frequency bands are unexpected since the corresponding Bragg wavelengths are deep within the viscous subrange of the neutral turbulence. This turbulence leads to variations in the index of refraction, which in turn are responsible for atmospheric echoes. Below the mesosphere the index of refraction is controlled by the pressure and temperature of the atmosphere. In the mesosphere the atmospheric medium is so tenuous that variations in pressure and temperature are insufficient to lead to radar echoes. However, during the day the increase in electron density in the D region of the ionosphere, which is collocated with the mesosphere, can lead to detectable echoes at all seasons for large VHF radars such as the Jicamarca radar in Peru and the MU (Middle and Upper atmosphere) radar in Japan. These electrons behave as a passive scalar mixed by the neutral turbulence.
 Simultaneous radar and rocket data taken in Peru showed a quantitative agreement between the observed spectrum of electron fluctuations in the mesosphere and the Jicamarca 50-MHz echoes [Røyrvik and Smith, 1984]. Their data clearly show that the electron fluctuation spectrum has an inner scale determined by the Kolmogorov microscale of neutral turbulence and displays a break to a k−7 spectral form at 40 m, as expected for the mesosphere [Hocking, 1987]. Similar radar and rocket concurrent observations in the polar summer showed that the electron fluctuation spectrum extends to much smaller scales than does the neutral turbulence [Ulwick et al., 1988; Kelley and Ulwick, 1988]. This has been explained as a reduction of the electron diffusion coefficient due to massive charged ice particles associated with Noctilucent Clouds (NLC) and a corresponding increase in the Schmidt number [Kelley et al., 1987; Cho and Kelley, 1993]. The latter is defined as the ratio of the atmospheric viscosity coefficient to the electron diffusion coefficient (i.e., Sc = ν/D). For a high Schmidt number, Batchelor  showed that the spectrum of a passive scalar such as the electrons will extend beyond the inner scale of turbulence. It is important to note that a large value of Sc also allows sharp gradients in the charged particle distribution, which may lead to partial reflection or Fresnel scatter at MF, HF, VHF, and possibly even UHF frequencies. A good number of radar/rocket measurements at PMSE altitudes have been carried out over the last two decades [Lübken et al., 2002]. Although the original radar/rocket studies showed a colocation between the echoes and the variability in the electron density profile, yet many questions remain. For example, neutral and electron turbulence are not always collocated indicating fossil turbulence [Lübken et al., 2002; Rapp and Lübken, 2003].
 The strong scattering cross section of PMSE provides a mechanism for remote sensing of mesospheric ice particles without the need for special lighting conditions. Cho and Kelley  predicted volume scattering cross section (η) as a function of the radar Bragg wave number (k) for mesospheric conditions and several Schmidt numbers. Such calculation shows that η increases as λ11/3, where λ is the radar operating wavelength, which in turn suggests that radars operating at frequencies less than VHF would have correspondingly stronger echoes. In the VHF it is rather a λ3 dependence (i.e., the scattering is from the viscous-convective subrange introduced by Batchelor ). Attempts to verify the PMSE dependency on wave number (or radar wavelength) by means of observations have been made in the MF and HF bands by several investigators. Bremer et al.  reported mesospheric echoes at three frequencies (53.5, 224, and 2.78 MHz) using radars deployed at various locations over northern Europe. Huaman  presented PMSE-like echoes detected with MF radars located in Mawson and McMurdo, Antarctica. Radar measurements using the SURA (midlatitude) facility showed the existence of HF mesospheric echoes, although no supportive information was available to validate them as PMSE [Karashtin et al., 1997]. Similarly, a second attempt using the HAARP (high-latitude) facility showed HF echoes very similar to PMSE [Kelley et al., 2002] but again, no supporting information was available. The characteristics of these observations included their layered nature, seasonality, geographical location, and altitude (near the mesopause), which strongly suggested but did not prove a relationship with VHF and higher-frequency PMSE.
 In this article we use multiple radar frequencies to show that PMSE is indeed detectable in the MF and HF bands. Radar observations collected simultaneously at three radar facilities located in the Alaskan central region are described. PMSE events are reported at eight different frequencies (from as low as MF/2.43 MHz to as high as VHF/139 MHz). Additional on-site instrumentation for ionospheric measurements are employed to better understand the conditions affecting PMSE. Finally, sensitivity estimates for each radar sensor are provided to asses their capabilities at detecting PMSE.
2. Radar Facilities, System Parameters, and Data Sets
 Since the data sets presented in this work cover several summer seasons we describe each radar system used in chronological order or as the data became available. Figure 1 shows a map containing all the radar locations. Two main geographical areas along the Alaskan central region can be identified: the Chatanika/Two Rivers area where the HIPAS and Poker Flat facilities are located at ∼65°N, and the Gakona area at ∼62°N, where the rest of the radar systems are deployed. Table 1 summarizes characteristics of all the radar systems. In addition to system parameters (i.e., geographical coordinates, transmitted power, frequency, etc.), model data (i.e., sky noise and volume reflectivity) are included for the estimation of radar sensitivities.
Table 1. System Parameters and Model Data for the Alaskan Radar Facilities
 The first set of exploratory experiments took place at the High Frequency Active Auroral Research Program (HAARP) facility near Gakona (62.4°N) during summer 2000. A variety of HF frequencies were used as shown in Figure 2. Layered echoes were detected at 3.3, 4.9, and 7.6 MHz with the optimum detections between 4 and 5 MHz [Kelley et al., 2002]. The HAARP transmitter antenna is a phase array, which gave us a lot flexibility for experiments. At the time of these experiments the transmitted power was 960 kW. We operated the system in radar mode with 10 μs long pulses (Δr = 1.5 km) and a 15 ms interpulse period (IPP). The long IPP was used in order to avoid range aliasing from multiple F region reflections. The phase array antenna cannot be used for reception, so we used one of the two relatively low gain receiving antennas at the site. For summer 2000, we used the digisonde antenna for reception, which had a gain of 11 dB, and afterward a spira-cone antenna with a gain of 7 dB. Ramos  offers additional details about the transmitting and receiving systems used to operate HAARP as an HF bistatic radar.
 In the subsequent summers (i.e., 2001–2003) we used transmission frequencies of 4.9 MHz at HAARP and 4.53 MHz at HIPAS (High Power Auroral Stimulation Observatory), the other Alaskan HF facility, which is located about 288 km to the north of Gakona, near Two Rivers (64.9°N). Similarly, HIPAS was operated as an HF bistatic radar. HIPAS has a transmitted power of 400 kW. The transmitter antenna consists of five crossed dipole antennas. A delta loop antenna was used for reception. For additional details about the HIPAS HF radar system, see Collins et al.  and Ramos .
 During the course of the investigation we also secured access to MF radar data from a system that operates continuously at the Poker Flat Research Range near Chatanika (65.1°N) about 40 km from HIPAS. The Poker Flat MF radar consists of five antennas, one used for transmission and four for reception. The radar transmitted power and frequency are 50 kW and 2.43 MHz, respectively. The backscattered signals, collected simultaneously with the four receivers, are analyzed to deduce horizontal winds, using the Full Correlation Analysis (FCA) technique, and electron density measurements, using the Differential Absorption Experiment (DAE) method [Murayama et al., 2000].
 Other radar instrumentation at the HAARP site included a permanent 139 MHz radar system, installed in April 2001 and in operation on a campaign basis since summer 2001. The antenna system used for transmission/reception consists of a Co-Co array with beam steering capability. A variety of geophysical phenomena have been studied since then, including meteors and PMSE (F. Djuth, personal communication). Two portable Air Force Research Laboratory (AFRL) 28 and 50 MHz radars were available during a short period of time during the summer of 2003. Due to the relatively small size of these radars and their modest peak power, the PMSE echoes were relatively weak. Nonetheless, when the signal was strong enough, the indisputable signature of PMSE was clear in the data sets. Absorption data provided by on-site riometers deployed at the HAARP and Poker Flat sites were used as an addition to the radar measurements at multiple frequencies. Ionosonde data from HAARP was available during the majority of the experiments.
 All the HF and VHF radar data sets shown in this work were obtained from the analysis of raw binary files collected during the experiments. Off-line coherent integration was employed to improve the signal-to-noise ratio (SNR) of the measurements. Barker decoding was necessary for the analysis of 139 MHz radar data. The MF radar data files were provided in ASCII format as backscattered power, SNR, and wind profiles, with a spatial resolution of 2 km, after oversampling, and a temporal resolution of 3 min. Although the original data analyses were performed using SNR data, for the purpose of multifrequency radar comparisons all the SNR values were expressed as radar reflectivities using system parameters and the assumption of a “pure” turbulent medium and noise dominated by the sky temperature. In section 5 we describe details about the radar equation for turbulent scatter and deriving reflectivity from SNR.
3. Comparison of HF With Higher-Frequency Observations
 An example of a PMSE event observed with three significantly different radar frequencies (4.9, 28, and 50 MHz) is shown in Figure 3. Figure 3 (top) shows the typical HF signal [Kelley et al., 2002], which we believe to be PMSE. Figure 3 (middle and bottom) shows upper HF and VHF data (i.e., 28 and 50 MHz) collected simultaneously during the HF transmissions. It is clear that the three radar targets are closely related. The median reflectivities were found as: 1.1 × 10−12, 1.3 × 10−13, and 1.3 × 10−13 m−1, for the 4.9, 28, and 50 MHz radars, respectively. An absorption jump of about 5 dB, as indicated by the HAARP riometer, caused that the detected signals for the three systems to disappear at about 1830 UT. The 4.9 MHz echoes lasted approximately 26 min. The 28 and 50 MHz echoes lasted 33 and 28 min, respectively. The PMSE altitude of the peak signal was virtually the same for the three radars (i.e., ∼83 km) being slightly lower for the 28 and 50 MHz systems. The HF signal was significant up to about 88 km. The old Poker Flat VHF radar [Ecklund and Balsley, 1981], which was more sensitive than the VHF radars used here, also found echoes up to such heights. Since VHF radars are the accepted marker for PMSE, this comparison is definitive evidence that PMSE and mesospheric summer echoes, in general, can be detected reliably by relatively narrow-beam HF systems such as those at HAARP, HIPAS, and the Russian site SURA [Karashtin et al., 1997].
Figures 4 shows two additional examples of unambiguous HF PMSE echoes occurring simultaneously with VHF PMSE. Figure 4a (14 July 2003; 1800–1840 UT) shows PMSE layers at around 83 km at 4.9 and 50 MHz. There is little absorption on this day and the HF signal lasts longer than the VHF signal. This extension is consistent with the frequency dependence of turbulent scatter discussed by Cho and Kelley . A highly disturbed ionosphere above 90 km is evident from the HF data, which is in agreement with an E layer reported by the HAARP ionosonde for that same time interval. The higher electron density levels above 90 km seem to be related to enhanced scattering from the lower PMSE layer, particularly at HF. The median reflectivities were 1.3 × 10−14 m−1 for the 4.9 MHz radar and 4.1 × 10−14 m−1 for the 50 MHz radar. Figure 4b shows a similar data set collected 2 days later (16 July 2003; 21–22 UT). The median reflectivities were found as 9.4 × 10−14 m−1 at 4.9 MHz and 5.2 × 10−14 m−1 at 50 MHz. Both radars gave good indications of PMSE processes occurring near 88 km of altitude although the VHF echoes showed up during 50% of the time only.
4. Comparison of MF With Higher-Frequency Observations
Figure 5 shows simultaneous observations made with the Poker Flat (MF/2.43 MHz) and HIPAS (HF/4.53 MHz) radar facilities on 18 July 2002; 1800–2200 UT. These two systems are located about 40 km apart with beams nearly overlapping over the HIPAS facility (see Figure 1). HIPAS has a half-power beam width of 17 degrees at a transmission frequency of 4.53 MHz. The MF radar has a beam width of about 40 degrees during both transmission and reception. The color scale shows a well defined layer centered near 86 km at HF (Figure 5, middle) whereas the MF data (Figure 5, top) show echoes coming from a wider range of altitudes, specifically, between 79 and 88 km. In Figure 5 (bottom) we have plotted peak reflectivity versus time for both frequencies. The HF data collection was carried out between 19 and 21 UT whereas the MF data were available for a longer period of time. Both radars observed similar mesospheric dynamics according to the peak reflectivity histories. The MF echoes fluctuated between about 10−13 and 10−11 m−1 during the 4-h data segment whereas the HF echoes fluctuated between 10−14 and 10−13 m−1 approximately during the 2-h data collection. The time offset between reflectivity minima and maxima at both frequencies seems to be consistent with the radar locations and background zonal winds of −20 m/s reported by the MF radar. For an echo structure detected at HIPAS under a purely east-west wind front, the same echo would be detected at Poker Flat at approximately 30 min later. A similar time offset can be observed between the HF and MF reflectivity maxima observed at 86 km/2000 UT and 83 km/2030 UT, respectively, indicating PMSE possibly being advected by the background mesospheric winds. In general, the layered nature of PMSE is clear in both data sets although the MF echoes show a more erratic nature possibly associated to the larger antenna collection area and longer averaging performed by the radar signal processor. The enhancement seen in these polar summer MF data is thus unusual and, we believe, is due to a PMSE effect.
 A much less convincing MF data set is presented in Figure 6. These multifrequency data sets were collected on 13 July 2003; 23–01 UT (15–17 LT). While the VHF echoes, detected at 50 and 139 MHz, respectively, are very clear, the MF signals do not show the obvious layering peculiar of PMSE. The VHF reflectivity values were significantly lower that those seen normally at MF/HF. The median MF reflectivity was 2.6 × 10−12 m−1 and 8.1 × 10−14 and 1.6 × 10−14 m−1, respectively, for the two VHF frequencies. Patchiness associated with PMSE and periodic motions were clearly observed in the VHF echoes during almost the entire data collection period. MF echoes from altitudes above 90 km produced some uncertainty discriminating PMSE at 2.43 MHz. The PMSE signature was pretty obvious at both VHF frequencies during the entire data segment. Notice the slight differences in echo altitude between the three frequencies. The mean MF echo altitude was about 81 km whereas the mean VHF echo altitudes were 83 and 84 km, for 50 and 139 MHz, respectively. Recall that the MF radar site was 300 km north of the other two.
Figure 7a presents another example of MF data from Poker Flat along with HAARP 4.9/139 MHz data. Figure 7b presents peak reflectivity histories for each radar frequency. Before 1845 UT, the unambiguous HF and VHF patches are preceded by weaker echoes at HF and a strong patch in the MF data. At 1915 absorption (not shown) increased and the MF signal almost completely disappeared. The absorption step was only 1 dB and seems not to have affected the HF signal. It is interesting that the time delay between the MF signal patch detected at Poker Flat (83 km/1750–1825 UT) and the HF echoes detected at HAARP (85 km/1855–1930 UT) is consistent with the magnitude and direction of the mesospheric winds reported by the Poker Flat MF radar. The VHF echoes, although significantly weaker, showed the same characteristics of the HF echoes detected locally at HAARP.
 Overall in all the multifrequency data sets analyzed, and based on signal detection, duration of mesospheric echoes, and reflectivity estimates, the HF radars proved to be more sensitive to PMSE than the MF or VHF radars. Table 2 summarizes the PMSE events discussed in sections 3 and 4 (6 days total and 15 radar data sets at 6 different radar frequencies) and their corresponding radar reflectivity estimates. The MF radar showed reflectivity values between 1 and 2 orders of magnitude larger than those given by the HF radars and between 2 and 3 orders of magnitudes larger than the VHF radars. Absorption imposed a major barrier for a reliable detection of MF PMSE although its signature, whenever the MF echoes showed up, was indisputable due to the similarity of the radar echoes with measurements at higher frequencies.
Table 2. Summary of Radar Observations and Reflectivity Estimates
Median Reflectivity (m−1)
29 Jul 2003
1.1 × 10−12
1.3 × 10−13
1.3 × 10−13
14 Jul 2003
1.3 × 10−14
4.1 × 10−14
16 Jul 2003
9.4 × 10−14
5.2 × 10−14
18 June 2002
Poker Flat MF
5.7 × 10−12
4.4 × 10−14
13 Jul 2003
Poker Flat MF
2.6 × 10−12
8.1 × 10−14
1.6 × 10−14
3 Jul 2003
Poker Flat MF
1.8 × 10−11
1.2 × 10−13
1.7 × 10−14
 In the next section we intend to quantify advantages and disadvantages of the different sensors and frequencies based on system parameters and a turbulent PMSE scattering model. As a final note on radar data, in all the MF data sets analyzed in this work, we noticed a discrepancy in echo altitude between the MF radar and the radars operating at HF/VHF frequencies. In general, the MF echoes showed up several kilometers lower in altitude with respect to their HF/VHF counterparts. Signal retardation due to ionization is discarded since that would produce an inverse effect (i.e., MF echoes would appear at larger ranges or longer time delays). A possible explanation for this disagreement could be calibration errors introduced by changing ionospheric conditions. As mentioned before, the MF radar is designed for continuous operation. Changing seasonal and/or diurnal ionospheric conditions do not affect the HF/VHF radars since they were operated during short periods (on a daily basis) and we had full control of the experiments and the data processing scheme.
5. PMSE Scattering and Radar Sensitivities
 We now turn to the question, “Why is the PMSE signal so much clearer for a narrow-beam, high-power HF system than for an MF or VHF radar?” The two types of scattering models proposed for the PMSE environment are isotropic turbulent volume-filled scatter and partial reflection scatter. In the present work we consider turbulent scatter only. For this case the radar equation can be written as
where Pt is the transmitted power, Gt is the transmitter gain, η is the cross section per unit volume, V is the volume illuminated, Ar is the effective area of the receive antenna, and r is the target range. Since the area illuminated (Ai) is equal to 4πr2/Gt and the volume (V) is equal to Ai × Δr where Δr is the size of the range gate, this expression can also be written as
 For the purpose of radar sensitivity calculations and interfrequency comparisons we express equation (2) as SNR by dividing the received power (Pr) by the expected noise (N) at the frequencies of interest. Since the SNR was known for all the data sets, we could invert equation (2) to find the cross section per unit volume or reflectivity (η) for the PMSE events shown in sections 2–4. The sky or cosmic noise was assumed to be dominant over other external noise sources and the system internal noise for all the radar systems. This assumption could be questionable for the 28 and 50 MHz radars although absorption data collected during the majority of the PMSE events did not indicate the presence of anomalous events in the D region that could introduce additional sources of background noise at upper HF and VHF frequencies. The average noise temperature of the 139 MHz receiver system was found negligible compared to the sky noise contribution (F. Djuth, personal communication). Twenty-four hours radar observations conducted at Poker Flat, HIPAS, and HAARP showed the expected diurnal variation of sky noise at MF and HF frequencies [Ramos, 2006].
 The sky noise is defined as
where k is the Boltzman constant, Tsky is the sky temperature in K, and B is the receiver bandwidth in Hz. The International Radio Consultative Committee (CCIR)  provides sky temperature data as a function of geographical location (i.e., latitude and longitude) and radar frequency. The receiver bandwidths and the sky noise estimates obtained using the CCIR-322 model are shown in Table 1 for all the radar systems. Finally, turbulent reflectivity data (η) as a function of wave number for Sc = 100 were obtained from Driscoll and Kennedy  and Cho and Kelley .
Figure 8 shows radar sensitivity estimates, as a function of wave number, for the MF/HF/VHF systems used in this work and PMSE modeled as a “pure” turbulent medium. We choose to use the AFRL 50-MHz radar as the reference for all the sensitivity calculations (i.e., all the SNR estimates were normalized with respect to the expected SNR at 50 MHz) given the historical importance of VHF radars for PMSE research and the large amount of published VHF PMSE data. Additionally, the median reflectivity values for all the radar measurements presented in sections 3 and 4 are overplotted for reference purposes.
 For all the radars, the range gates are about the same length so the dominant terms of equation (2) are Pt, η, and Ar. The cross section for turbulent scatter (η) goes as λ11/3, which counteracts somewhat the power advantage at HF versus MF. The antenna receiving areas (or beam width) were nearly similar for the MF/2.43 and HF/4.53 MHz systems although for sensitivity calculations we used the HF/4.9 MHz radar as the “representative” HF system. The 4.9 MHz radar had the biggest collection area of all the receiving antennas (i.e., 2.5 steradians). The expected noise was found about the same for the MF/2.43 MHz, HF/4.53 MHz, and VHF/50 MHz systems, between 5 and 8 dB lower for the 4.9 and 28 MHz radars, and 5 dB higher for VHF/139 MHz system. Based on sensitivity calculations or radar equation predictions we found that the MF/2.43 MHz radar should be about one order of magnitude more sensitive to “turbulent” PMSE than the HF/4.9 MHz and between 5 and 7 orders of magnitude better than the VHF radars. Based on radar measurements, we found that the MF and HF systems should be between 2 and 3 orders better at detecting PMSE than the VHF systems. The PMSE dependency on wave number was found having a very similar trend in both, the radar data sets (median reflectivities) and the scattering predictions. The discrepancy in their magnitudes can be attributed to system and wave propagation losses not modeled in the radar equation. Additionally, having an ideal and/or pure turbulent PMSE medium seems to be very unlikely.
 We conclude that for turbulent type of scatter, the MF and HF systems showed consistently to be more reliable for PMSE detection than the radars operating at higher frequencies (i.e., VHF frequencies). Given the high power consumption of existing HF systems at high latitudes, MF or smaller HF radar systems should be considered seriously as sensors appropriate for PMSE investigation.
 We have presented what we believe to be definitive proof that narrow-beam high-power HF transmitting antennas can be used with low-gain receiving antennas to reliably detect Polar Mesosphere Summer Echoes (PMSE). This conclusion is based on several inherent features of the data, including the altitude, the layered nature, the seasonal behavior, and the signal strength. Even more conclusive, given the historical importance of VHF radar observations of PMSE, is the simultaneous multifrequency data we present to support the HF echo interpretation.
 We are confident as well that we have shown that MF “wind profiler” systems at high latitudes receive enhanced and layered scatter from the region associated with PMSE, although the signature is not as clear as with the HF systems. Since a number of MF stations are located at polar or near polar latitudes, including Antarctica, it may be possible to use the PMSE signature studied here to investigate its long-term variability as well as its low-latitude boundary.
 Research at Cornell was sponsored by the Office of Naval Research under grant N00014-00-1-0658. F.T.D. acknowledges support from the Office of Naval Research under grants N00014-00-1-0658 and N00014-03-C-0482. The authors are grateful to the HIPAS and HAARP operating crews for their considerable assistance in the field. Special mentions to: John Elder, Eric Nichols, Mike McCarrick, David Seafolk-Kopp, Jake Quinn, and Troy Lawlor. We thank the National Institute of Information and Communications Technology in Japan and the University of Alaska Fairbanks for giving access to data from the Poker Flat MF radar.