The propagation of radio signals through the ionosphere has been an essential area of study for HF radio and radar signals. One network of HF radars, the Super Dual Auroral Radar Network (SuperDARN) Greenwald et al. (1995); Chisham et al. (2007), requires significant refraction to obtain details of plasma convection in the terrestrial ionosphere over a very large field-of-view. The signal is required to be approximately perpendicular to the geomagnetic field to scatter from ionospheric plasma irregularities, or perturbations, aligned with the geomagnetic field. This provides details about the ionospheric plasma characteristics and motion. There are also many studies from SuperDARN investigating ground scattering of the radar signal from over the horizon propagation (e.g., Chisham et al., 2007; Nishitani et al., 2019, and references therein).

Highlighting the need to better understand the propagation of radar signals through the ionosphere is a recent publication by Shepherd et al. (2020). The study investigated bistatic operations of the SuperDARN radars and the different propagation modes present. This included the difficulties in properly isolating and imaging the different signal propagation modes, such as F-region coherent scatter and ground scatter, over a large field-of-view with the multiple possible bistatic propagation paths. Future implementation of an imaging interferometry analysis for SuperDARN radars will be greatly beneficial for further analysis of these bistatic multiple propagation paths and better mapping of the scatter.

One benefit of using VHF frequency radars in comparison to HF radars is the requirement of very large plasma densities to produce sufficient refraction of the signal for ground scatter to occur. This allows more easily attainable propagation mode determination, and therefore a better understanding of the ionospheric characteristics required for the signal characteristics observed, including the angle of arrival and spectra details.

While the occurrence of VHF ground and/or sea scatter is rare, it has been shown to occur in the auroral zone during active geomagnetic events (e.g., Kozlovsky and Lester, 2015; Kozlovsky et al., 2019). These occurrences were noticed due to unexpected meteor radar measurement behaviour, specifically the variation in the radar signal over time. From the results it was determined that the meteor radar signal at 36.9 MHz was refracting off the E-region ionosphere, scattering from the ground/sea, and propagating back along a similar path to be received at the radar. Periodic variations in the radar signal were attributed to changing E-region plasma characteristics, potentially from pulsating aurora generating variations in the E-region plasma density profile.

Multiple occurrences of these groundscatter echoes were shown to occur every year, specifically during periods of increased geomagnetic activity around midnight. This corresponds to a potentially enhanced E-region ionization layer that allows the VHF signal to propagate over the horizon due to refraction. In the Kozlovsky and Lester (2015) and Kozlovsky et al. (2019) studies, the interferometry of the meteor radars were not used to determine the height of these expected groundscatter echoes. This could provide further details on the ionospheric plasma medium. The use of interferometry to determine altitudes for VHF ground scatter occurrences is investigated in the present study.

A longstanding issue in the study of E-region coherent scatter is the determination of the altitude of the scatter with respect to the corresponding radar spectra (e.g., Sahr et al., 1991; Sahr and Fejer, 1996). From previous measurements and ionospheric plasma instability theory it is generally accepted that auroral E-region coherent scatter occurs between 95 and 125 km and is strongest at perpendicular angles to the geomagnetic field (Sahr and Fejer, 1996). Measurements of E-region coherent scatter by VHF radars do not always agree with these altitude and aspect angle expectations, resulting in different explanations for the disagreement, mostly to do with refraction of the VHF signal. Some examples of this are provided in Moorcroft (1989), Hall et al. (1990), Hall and Moorcroft (1992), and Uspensky et al. (1994), where tilted plasma density layers are proposed to provide sufficient refraction of the radar signal to obtain expected altitude and aspect angle characteristics of the coherent scatter. Whether it is tilted layers or larger than expected ionospheric plasma densities, from the measurements it is clear that significant refraction of VHF signals occurs in the auroral zone during active ionospheric events.

A study by St.-Maurice and Chau (2016) provided a theoretical framework to describe the reasons for the different E-region coherent scatter spectra based on the altitude of the measurements. The evidence for the explanations were provided in a companion paper by Chau and St.-Maurice (2016), where measurements from VHF multistatic pulsed meteor radars in Germany were able to detect E-region coherent scatter during a very active geomagnetic period. In relation to this, understanding the refraction occurring at VHF radar frequencies during active geomagnetic conditions will aid in better understanding the altitude dependence of the E-region coherent scatter and potentially further validate the theoretical framework proposed by St.-Maurice and Chau (2016). Other reviews on the origin of the different characteristics of E-region coherent scatter can be found in Fejer and Kelley (1980), Sahr and Fejer (1996), Moorcroft (2002), Makarevich (2009), and Hysell (2015).

Some recent E-region coherent scatter studies have been able to locate E-region coherent scatter much more accurately than previously (e.g., Hysell et al., 2012; Chau and St.-Maurice, 2016; Huyghebaert et al., 2019). Even with the enhanced imaging and location determination capabilities, there are still questions as to how much refraction occurs for the radar signal, especially at frequencies of ∼ 30 MHz in the lower VHF band. The study presented here is a basis to better understand the refraction of VHF signals in the presence of strong geomagnetic activity, providing insights on signal propagation modes during these active periods. This will also help provide additional context to ionospheric measurements made by VHF radars in the past, present, and future.

Beginning with the first published implementation of a CW coded meteor radar system by Vierinen et al. (2016), the concept of CW coded signals for meteor radars has been implemented at multiple new radar sites around the World, specifically by the radar remote sensing group at the Leibniz Institute of Atmospheric Physics based in Kühlungsborn, Germany. Their meteor radar network sites are known as MMARIA (Multistatic, Multifrequency Agile Radar for Investigations of the Atmosphere) (Stober and Chau, 2015), with the sites using the CW coded signals being known as SIMONe (Chau et al., 2019) implementations of MMARIA. Besides CW coded signals (i.e., spread spectrum), SIMONe uses multi-input multi-output (MIMO) and compressed sensing concepts, allowing for more reliable and increased resolution when analyzing the received signal using interferometric analysis (Chau et al., 2019; Urco et al., 2019). The MIMO setup consists of both multiple transmit and receive antennas at the sites, as well as multiple sites included in the network. The multistatic meteor radar networks derive neutral winds from ionized meteor trails that drift with the background neutral atmosphere. A detailed description of operational SIMONe systems can be found in Chau et al. (2021).

Using the SIMONe Norway system and its enhanced interferometric capabilities, multiple propagation paths of the VHF radar signal can be isolated by processing the coded CW signal and using MIMO interferometry. Only a single receiver antenna was present at the SIMONe Norway receiver sites for this study, resulting in multi-input single-output (MISO) analysis. A description of the modified processing algorithm to distinguish the different radar propagation modes during active geomagnetic periods is detailed in Section 3. An overview of the results in provided in Section 4. The description and determination of the different propagation modes is provided in Section 5. The radar signal propagation modes detected include meteor scatter, direct ground-wave from the transmitter, multi-site over-the-horizon transmitter links, ground/sea clutter from ionospheric refraction, E-region scatter, and potentially F-region scatter. The summary and conclusions from this event are provided in Section 6.

E-region coherent scatter velocities of > 1 km/s have been previously recorded by different systems (e.g., Haldoupis et al., 1993; Huyghebaert et al., 2019). To further analyze E-region spectra with the SIMONe Norway system, changes to the raw voltage processing were made to prevent aliasing of the signal. For a description of the typical SIMONe signal processing, the reader is referred to Chau et al. (2021).

Through re-analysis of the raw voltage data using multiple repeated code sequences it is possible to increase the integration time of the system. This is important for obtaining satisfactory Doppler resolution measurements for the analysis of E-region coherent scatter. By integrating the signal over a longer period, well defined spectra can be determined for the radar signal at all ranges simultaneously. The system setup and signal processing allows finer analysis of the signal than is typically possible with uncoded CW radars. The integration performed on the data is similar to that of the ICEBEAR (Ionospheric Continuous-wave E region Bistatic Experimental Auroral Radar) system (Huyghebaert et al., 2019). For the SIMONe Norway system analysis presented here, the signal of the different codes transmitted simultaneously on each locally separated transmit antenna are received on a single bistatic receiving antenna at each of the sites. In the ICEBEAR system a single code is transmitted and received on multiple locally separated receiving antennas. A block diagram of the SIMONe processing algorithm for this study is presented in Figure 2.

The matched filter, low pass filter, and decimation stage can be described by the equation, V filtered r , t n = ∑ d V ant t n d rate + d C * r , t n d rate + d (1) where V _ant is the complex voltage measured at an antenna, C* is the complex conjugate of the code transmitted, d _rate is the decimation rate, t _n is the decimated sample number in time, r is the range gate, d is the sample in the decimation summation, spanning from 0 to d _rate—1 in a 0-indexing system.

Once the received signal is decoded and filtered, a Fourier transform is taken for each range, thereby determining the spectra for each range separately. This is shown by the equation, S r , f = F F T t V filtered r , t n (2) where f is the frequency of the signal and FFT _t refers to taking a Fast Fourier Transform in the time domain for each range. The result is a range-Doppler shift array of the radar signal, given by S. An example of this data is provided in the next section.

Once the spectra are calculated for each antenna, the cross-spectra can also be determined. Each range-Doppler bin from the spectra of one antenna is multiplied by the complex conjugate of the range-Doppler bin from another antenna. The equation describing this is given by, S 1,2 r , f = S 1 r , f S 2 * r , f (3) where the subscript number refers to the antenna number. This cross-spectra information can be used to perform interferometry on the scattered radar signal.

The SIMONe 10 ms CW code was repeated 20 times to obtain 200 ms of integration time with a matched filter. This was done for each of the codes, as each transmit antenna for the SIMONe system has a different code. The self-clutter in range-Doppler space generated from the 10 ms CW code was ∼ 18 dB from the peak signal. For strong signals the multiple different codes generated a significant amount of self-clutter, which had to be removed for usable data to be obtained. A 0 Hz clutter removal algorithm was first used on the data, similar to that described in Urco et al. (2019). From there another clutter removal algorithm was implemented to remove frequency shifted clutter. It should be noted that since we repeated the 10 ms code, self-clutter was magnified along the ranges for a given Doppler and repeated every 100 Hz in the Doppler space. The self-clutter removal algorithm is described next.

To further remove clutter from Doppler shifted signals, the previous SIMONe implementation of the self-clutter removal algorithm was expanded upon. Though there was a significant processing time increase, this frequency shifted self-clutter removal was essential for analysis of the radar signal. The points in the range-Doppler spectra were selected by finding all points in the range-Doppler domain above 10 dB SNR. The value of these bins were recorded to be re-populated into the spectra after the self-clutter removal algorithm. The transmitted code was then shifted by the range, frequency shifted by the Doppler, and multiplied by the range-Doppler bin value to provide an amplitude and phase shift before being subtracted from the raw voltage data. The equation describing this process is given by, V new t = V t − ∑ p S r p , f p d rate C r p , t e j 2 π f p t × S N R r p , f p − 1 S N R r p , f p (4) where p refers to the range-Doppler points above the SNR threshold for which the self clutter is removed. The SNR term on the right side of the equation scales the signal level so that the noise is not included in the subtraction. The division by the decimation rate is necessary to normalize the values, as the low pass filter and decimation is performed on the voltages before processing. This method is very similar to the CLEAN algorithm used in radio astronomy (Högbom, 1974).

The process was performed for all the range-Doppler bins above a SNR threshold and for all transmit codes. The process was repeated until there were no bins above 10 dB SNR in the data. An error catch was utilized in case a self-clutter reduction could not be attained, in which case the iteration ended early. After the iterative process, the points that were removed were repopulated into the range-Doppler spectra. A text description of the algorithm to produce the 10 × 200 ms integrated spectra and cross-spectra (2 s resolution) is provided in Table 1, following the block diagram provided in Figure 2.

The range-Doppler values corresponding to a SNR above a set threshold were recorded in HDF5 format for further future processing. In the case of the current processing, a 2.0 SNR (∼ 3 dB) threshold was set for recording the data. This reduced the data footprint compared to recording the values from each full range-Doppler spectra.

An example of a range-Doppler SNR plot is shown in Figure 3. Here the left panel does not perform any clutter removal algorithm, while the right panel includes this step. The improvement in the data is obvious, with less clutter included in the results. It should be noted that this algorithm works well for scattering targets with unique ranges and Doppler shifts. If there are two strong scattering targets within the same Doppler shift and range bin, the self-clutter reduction algorithm has difficulty removing the expected signal. This will be seen in Section 4, where self-clutter is still evident during periods of multiple targets with the same range and Doppler shift. It should also be noted that the narrower the spectra for a single target, the better this self-clutter removal algorithm will work. For wide spectra the amplitude of the decoded voltage will decorrelate in time, thereby changing the amplitude of the signal over the integration period. This results in some leftover signal at that frequency after the self-clutter subtraction. In the future, the self-clutter reduction algorithm can be further improved through the simultaneous use of interferometry during the clutter removal stage to separate multiple targets and properly remove their contributions to the radar signal.

For this study the same angle of departure (AOD) software as the standard SIMONe analysis was used Chau et al. (2021). This provides the center of the scattering volume for the spread target ionospheric echoes, which is sufficient for the current propagation mode analysis. Future analysis of ionospheric scatter data from the SIMONe Norway system can incorporate imaging techniques to determine the size of the scattering region, as this is important for multi-instrument comparisons and properly calculating the extent of the plasma turbulence region (e.g., Meyer and Sahr, 2004; Huyghebaert et al., 2021).

From the coherence of the signal derived from the measured spectra and cross-spectra, the AOD is determined. The scatter can then be mapped in 3-D space.

The initial results from the AOD determination showed that the side lobes of the array point-spread function were commonly chosen as solutions, likely caused by a combination of the spread target nature of ionospheric scatter and multiple low SNR scattering points (noise effects). To reduce these erroneous AOD occurrences, the spectra for the full FOV were fit for each range, and the spectra and cross-spectra were averaged within the fit bounds based on the central Doppler shift and spectral width determined. This makes the assumption that all points within the fit Doppler spectra for a given range are originating from the same point. It is expected that the points in the spectra are typically originating from a similar region, but the AOD could vary slightly with changing Doppler frequency. Due to the significant and similar amount of refraction that will be experienced at the different Doppler shifts for the same range, we decided that the reduction in fine scale AOD determination was worthwhile to reduce the occurrence of erroneous AOD determination. Future implementation of MIMO capabilities with multiple receive antennas will greatly reduce the occurrence of sidelobe selection in the interferometry analysis.

To further filter the self-clutter occurrences after the initial filtering of the signal described in Section 3.2, Section 3.3 conditions were implemented. These three conditions and the order of implementation were:

1. SNR > 6 dB

The spectra were then fit with a Gaussian distribution to obtain a peak-SNR, Doppler shift, and spectral width value for each range with data. The cross-spectra and spectra were averaged within the bounds based on this fit, and the AOD were determined from these averaged values. The data available for each range for a 2 s scan were then: peak-SNR, Doppler shift, spectral width, total propagation distance, azimuth, and elevation.

The direct path signal is at ∼ the sixth range gate (18 km). This signal is consistently removed with the self-clutter reduction algorithm without issue. The distance matches with what is expected for direct path from the transmitter to receiver for the Andenes-Saura link. This is seen during all periods of operation for the system.

The direct path propagation mode is searched for in the self-clutter reduction algorithm for the Andenes-Tromsø and Andenes-Straumen links but is not present. This is not unexpected due to the much larger distance between the sites compared to the Andenes-Saura link.

Due to the lack of interaction with the ionosphere for this direct TX-RX link, the other propagation modes for the VHF signal are focused on.

Another detected signal that appeared as direct feed through in the data occurred at total propagation distances > 1400 km. It was determined to originate from Kühlungsborn due to the radar site using the same set of transmit codes as the Andenes site. Upon searching for the Juliusruh SIMONe signal with its unique code sequence, it also was detected in the data at the expected range and is included in the summary plot data. For this propagation to occur, significant refraction of the VHF signal is necessary.

Interferometric analyses of the signals were performed to investigate how and where the OTH propagation was occurring. It was found that the signal was likely refracting from the E-region ionosphere, showing that there was significant ionospheric plasma density at geomagnetic latitudes of ∼ 58.5° (geographic latitudes of ∼ 62°). Figure 8 provides the geographic coordinates of the signal and a histogram of the altitude for each OTH link. It is clear that the majority of the data is mapped to an altitude between 90 and 200 km and occurs between the sites, as expected.

The amount of E-region plasma density required for OTH propagation at 32.55 MHz is on the order of ∼ 1× 10¹²/m³. While this is a large plasma density for the evening E-region ionosphere, plasma densities of this amount are not necessarily uncommon during active geomagnetic periods. An example of another VHF meteor radar detecting significant refraction of the signal is provided by Kozlovsky and Lester (2015). This radar operated at ∼ 37 MHz and was located in Sodankylä, Finland, a similar latitude as the SIMONe Norway system. Also, from the ionosonde and Saura PRR data presented in Figure 5, we know that there were occurrences of particle precipitation penetrating to at least E-region altitudes.

Future work will include the analysis of the modulation of this OTH signal to detect variations in the E-region plasma - similar to what was performed with the Sodankylä radar. The benefit of performing this analysis with the SIMONe Norway system is that a source of ambiguity is removed from the analysis, specifically ground variations affecting the measured signal. Unsurprisingly, the SIMONe Norway system was also able to detect groundscatter measurements during this period of active ionospheric activity, which are explored next.

The occurrence of groundscatter is common in HF signals (e.g., Basu et al., 1973; Greenwald et al., 1995), and has been shown to be present in VHF radar signals (e.g., Kozlovsky and Lester, 2015; Kozlovsky et al., 2019). The capabilities of the SIMONe system are highlighted by its ability to detect groundscatter and perform interferometry on the signal simultaneously while measuring many different propagation modes and determining the details of the spectra for each link. We clearly saw OTH propagation of the signals from SIMONe Germany to SIMONe Norway, so there was obviously sufficient electron density at E-region altitudes for refraction of the VHF signal back to the surface of the Earth.

The altitude of SIMONe Norway measurements with a Doppler shift within ±10 Hz were investigated to determine if groundscatter measurements were detected in the data. The histogram distributions are provided in Figure 9. A peak in altitude at ∼ 200 km is evident, the altitude at which ground/sea-scatter would be mapped to if the signal was refracting from the E-region (∼ 100 km) back to the Earth. This was seen in all three links, though less so in the Andenes to Straumen data. Of course, with large distances between the receiver and transmitter, the geometry for ground and sea scatter is more complicated. With bistatic measurements, refraction of the radar signal from the E-region back to the Earth will occur in two different locations for groundscatter to be measured. VHF signals require large plasma densities for refraction to occur and therefore it is expected that there will be less instances of ground and sea scatter measured for the bistatic radar links with large distances between the transmitter and receiver. The surface also has to be conducive to scattering or reflecting the signal towards the receiver, which will be more likely with certain geometries.

The ground/sea scatter does create difficulties in using the self-clutter removal algorithm described in Section 3.2. There are many instances where signals will be coming from different enough directions at the same range and Doppler shift to cause issues in determining the correct phase of the signal during the self-clutter removal. In Figure 6 there are clear occurrences where the clutter removal for the data fails, commonly due to this issue with removing self-clutter from the ground/sea scatter. The ground/sea scatter typically occurs between 700 and 1500 km, depending on the bistatic link, and is relatively faint in Figure 6.

Due to the geometry of the SIMONe Norway system multistatic links with respect to the geomagnetic field, instances of E-region coherent scatter were unexpected in the data. From Figure 6, Figure 7, there are undeniably occurrences of E-region coherent scatter. With the evidence of significant amounts of refraction of the 32.55 MHz signal, it is therefore expected that the radar signal refracts enough to be perpendicular to the geomagnetic field. In regions where the signal is perpendicular to the geomagnetic field and sufficient plasma turbulence is occurring, we are therefore able to detect E-region coherent scatter with the SIMONe Norway system.

Figure 10 presents a zoomed in version of Figure 6, highlighting the ionospheric scatter measured by the SIMONe Norway system. This figure includes all the filters mentioned in Section 3.3. The times when the self-clutter removal algorithm failed and the data had to removed are shown by the vertical stripes of missing data. While the different links measure similar directions of the scatter at similar times, there is clearly many differences in the measurements between the different bistatic links. Only a preliminary analysis of the ionospheric scatter is provided in this section to highlight the capability of the system, with further analysis of the spectral characteristics and structure of the scatter to be investigated in the future.

Most of the scatter is expected to be from the E-region ionosphere, though the interferometry analysis (Figure 7) shows there are also potential instances of F-region coherent scatter (altitudes of 150–500 km). While F-region coherent scatter is potentially possible with the amount of refraction occurring, due to the current ambiguities in the AOD determination the E-region coherent scatter is focused on.

Figure 11 provides details of the location of the scattered signals and the aspect angle conditions with respect to the geomagnetic field assuming straight line propagation. The ideal aspect conditions occur to the north of the radar links, where the signal is < 5° off perpendicular. Only the data between altitudes of 50–150 km are shown, removing the data expected to be due to ground/sea scatter. The OTH links from the SIMONe Germany system are also excluded.

Most of the scatter is observed south of the radars, which could be a result of the propagation of the signals with respect to the geometry of the geomagnetic field associated with a significant amount of refraction. Interestingly, the locations of the scatter for the three links are slightly different. This could correspond to refraction of the radar signals for each link. Significant refraction for a 32.55 MHz signal requires large plasma densities on the order of ∼ 1× 10¹²/m³, so there is likely energetic charged particle precipitation occurring in the region creating an enhanced plasma density layer. It is expected that plasma densities of this magnitude are relatively localized in spatial extent and do not persist without an ionization source due to the relatively high recombination rate of the E-region ionosphere. This ionization source can be realized through energetic charged particle precipitation or also through large electric fields creating enhanced plasma density layers at E-region altitudes (e.g., Nygrén et al., 1984).

The location of the majority of the radar measurements corresponds to relatively large aspect angles ( > 10°) if we assume straight line propagation. With reference to previous studies and reviews, e.g., Moorcroft (2002) and references therein, it is expected that these measurements are not actually at large aspect angles. Ground scatter occurs in the data set with significant refraction from the ionosphere, so it is clear that sufficient plasma density exists for the VHF signal to be refracted to become orientated perpendicular to the geomagnetic field.

Histograms of the Doppler shift and the spectral width of the measurements have also been provided in Figure 12. The Doppler shift of the data has been converted to velocity based on the location of the signal scattering region. Something that should be noted is that a significant portion of the coherent scatter has a Doppler shift that is lower than the expected ion-acoustic speed of the ionospheric E-region plasma. For plasma turbulence at these phase speeds with small spectral widths there have been a few different hypotheses used to explain their occurrence, such as plasma gradients (St.-Maurice et al., 1994) and “Modulated Electron Ohmic Heating by Waves” (MEOHW) (St.-Maurice and Chau, 2016).

The scatter with larger spectral widths and Doppler shifts less than the expected ion-acoustic speed have been described as being due to the radar observing plasma turbulence at angles off parallel with the electron velocity (e.g., Hysell et al., 2012). Empirical formulas were used by Hysell et al. (2012) to derive plasma convection from the mean Doppler shift and spectral width. These results compared favorably with the plasma flows derived from the Poker Flat Incoherent Scatter Radar (PFISR). The empirical electric field determination using E-region coherent scatter will not be investigated in this manuscript, but could be a future direction of study with the SIMONe Norway system measurements.

Due to the amount of refraction occurring for the VHF signal, it is difficult to use the altitude of the scatter to investigate the different hypotheses for the narrow and relatively slow spectra, where the MEOHW mechanism requires altitudes of < 100 km and electric fields on the order of 50 mV/m (St.-Maurice and Chau, 2016). We know there is energetic particle precipitation occurring at E-region altitudes for the signal propagation modes measured, but we cannot claim to have fine-scale altitude resolution due to the significant amount of refraction. This means that there are likely both large electric fields in the region and large density gradients on small spatial scales. This does not help with rejecting either hypothesis for the source of these slow narrow spectra.

Though the radar signal experienced refraction, it should be noted that Figure 7 in Section 4 does show some interesting trends with respect to the altitude and Doppler shift of the scatter. Specifically, there is an increase in altitude with increasing Doppler shift in the E-region, though the number of measurements is small for large Doppler shifts. The E-region coherent scatter with relatively large Doppler velocities and narrow spectral widths is known as “Type IV” E-region coherent scatter (e.g., Providakes et al., 1988). The higher altitude corresponding to these measurements agrees with the notion that the E-region spectra with large Doppler shifts are in the reference frame of the ions, presented by St.-Maurice and Chau (2016). As the altitude increases the component of the ion motion in the E×B direction also increases, resulting in an increase in the required electron velocity for plasma turbulence to occur from the Farley-Buneman, or modified two-stream, plasma instability mechanism (Buneman, 1963; Farley, 1963).

The other interesting trend in Figure 7 is that scatter in the range of ±25–75 Hz is sometimes mapped to very low altitudes. It is highly unlikely that any measurements of coherent scatter would be made at 25 km altitude with the SIMONe radar system, but this trend of low altitude determination is not persistent across all Doppler shift values. The fact that this occurs only for these Doppler shifts will be further investigated in the future. It is possible that this scatter does occur in the lower E-region and potentially even D-region ionosphere, albeit not at some of the altitudes that the figures show.