Simultaneous observations of transionospheric and HF ionospheric propagation within the polar cap



[1] The total electron content of the ionosphere over Alert, Canada was measured continuously for over 3 years using a dual frequency GPS receiver. Peaks in the measured total electron content are attributed to the presence of convecting polar patches (patches of enhanced ionization), which increase in number and intensity with increasing sunspot number, peaking in October to November 2011. In addition, HF radio transmissions were made at frequencies between 4 and 14 MHz over a path from Qaanaaq, Greenland to Ny-Ålesund, Svalbard and several examples are presented in which features in the received signals are visible by these two means almost simultaneously. The reception of the higher-frequency transmissions is shown to be dependent on the presence of polar patches.

1 Introduction

[2] Many measurements of HF propagation at high latitudes have been undertaken by the University of Leicester and colleagues in the last two decades [e.g., see Warrington et al., 1997; Zaalov et al., 2003; Rogers et al., 1997, 2003; Siddle et al., 2004a, 2004b]. Their observations within the polar cap have shown that deviations from the great-circle path (GCP) of up to 100° can be caused by convecting patches and sun-aligned arcs of enhanced electron density [Warrington et al., 1997; Zaalov et al., 2003]. Patches originate in the dayside auroral oval and convect antisunward across the polar cap to the nightside auroral oval during the high levels of geomagnetic activity which occur when the interplanetary magnetic field (IMF) points south (Bz < 0) [Buchau et al., 1983; Buchau and Reinisch, 1991]. By contrast, arcs occur in the more quiescent periods when the IMF points north (Bz > 0) and travel duskward [Buchau et al., 1983; Buchau and Reinisch, 1991]. Furthermore, the nonuniform nature of the ionosphere results in signals arriving over a range of directions in both azimuth and elevation. For example, standard deviations of up to 35° have been seen on a path from Isfjord, Svalbard to Alert, Canada at frequencies of 2.8, 4.0, and 4.7 MHz [Warrington, 1998]. Other measurements have shown deviations of similar magnitudes at auroral latitudes [Warrington et al., 2006].

[3] A ray-tracing model has been developed which reproduces many of the features seen in the observations of direction of arrival [Zaalov et al., 2003, 2005]. Given appropriate parameters, the model can simulate propagation over any high-latitude path and give, among other things, an estimate of area coverage taking into account the off-GCP effects that are ignored by current prediction techniques (e.g., Voice of America Coverage Analysis Program, VOACAP) [Sweeney et al., 1993]. In order to further develop the model for HF forecasting and nowcasting applications, it is desirable to have additional ionospheric measurements as inputs. One potential source of ionospheric data is dual frequency GPS receivers that have the capability of measuring ionospheric total electron content along slant paths (sTEC) from the receivers to the various satellites. As such, they have the capability of detecting the presence of regions of enhanced ionospheric electron density, such as polar patches that will influence the HF propagation. The aim of this paper is to establish a relationship between GPS sTEC measurements and HF propagation characteristics on a path contained within the polar cap.

2 Experimental Configuration

[4] The 1707 km HF link operated from Qaanaaq, northern Greenland (69.22°W, 77.46°N) to Ny-Ålesund, Svalbard (11.93°E, 78.92°N) (see Figure 1) commenced operation in March 2009 with transmission frequencies of 4.6, 7.0, 8.0, 10.4, 11.1, and 14.4 MHz. The radiated signals comprised 2 s sequences of 13-bit Barker coded phase-shift keying pulses modulated at 2000 baud with a repetition rate of 66.7 coded pulses per second. Since the transmitter and receiver systems were synchronized to GPS, the absolute time of flight of the signals could be determined. The complex amplitudes of the received signals were sampled simultaneously, and the data were processed to provide a measure of the relative times of flight of the propagating modes and their associated Doppler spectra (see the method employed by the DAMSON system, described by Davies and Cannon [1993]).

Figure 1.

Map indicating the area of study. A GSV4004B GPS receiver was located at Alert (A). The outer circle around Alert indicates the range at which the GPS signal paths cross the ionosphere at a height of 350 km (the so-called pierce point) for an elevation angle of 25° (the minimum value used in this paper), and the inner circle indicates the approximate locus of positions for which the GPS signals reach a maximum elevation angle (the satellites are never overhead at this location). The shaded area indicates where those pierce points are near Qaanaaq (i.e., lying within 56°–82°W and <81°N, while still maintaining elevation >25°). HF transmissions were made from Qaanaaq (Q) and received at Ny-Ålesund (N).

[5] Observations of TEC were made with a GSV4004B GPS receiver located in Alert, northern Canada (62.25°W, 82.48°N) (see Figure 1), which has been collecting data since May 2008. The approximate viewing area of the GPS receiver is indicated in Figure 1—the outer circle around Alert indicates the range at which the GPS signal paths cross the ionosphere at a height of 350 km (the so-called pierce point) for an elevation angle of 25° (the minimum value used in this paper), and the inner circle indicates the approximate locus of positions for which the GPS signals reach a maximum elevation angle (the satellites are never overhead at this location). Each satellite transits twice per day (orbital period: 11 h 58 min). The information output by the GSV4004B includes azimuth and elevation of the path to each satellite in addition to TEC and scintillation values. Measurements are presented in this paper for 2008–2011, covering the rising part of the current solar cycle (Figure 2).

Figure 2.

The variation in sunspot number over the period of study, showing the peak around November 2011. The quantity plotted is the monthly mean international sunspot number downloaded from the Space Weather Prediction Center, US National Oceanic and Atmospheric Administration,

3 GPS Measurements of TEC

[6] Figure 3 (top row) shows examples of slant TEC (sTEC) measurements for the period 0600 to 1200 UT on 11 November 2008 and 11 November 2011. Each satellite is visible for about 3 h, and the curved shape of the sTEC traces is due to the varying path length within the ionosphere (lowest for high-elevation angles and greatest for the low-elevation angles at the extremities). In interpreting these measurements, it is important to remember that sTEC measurements represent a summation of electron density over the satellite-receiver line, rather than at a single geographic point. This problem is more severe at low-elevation angles, so data with elevation angles less than 25° were discarded. Other authors undertaking various investigations have used cutoff angles of 15° [Spogli et al., 2009], 20° [Krankowski et al., 2006; Gwal and Jain, 2011; Warnant and Pottiaux, 2000], 30° [Prikryl et al., 2010], and even 50° [Breed et al., 2002]. For our investigations, the choice of angle was restricted as the maximum elevation angle of the satellites was only around 45° at these latitudes. The November 2008 measurements were made close to sunspot minimum (see Figure 2), and the November 2011 measurements were at a peak in solar activity approaching the expected sunspot maximum (at the time of writing, it appears that this may possibly be the peak in the current solar cycle).

Figure 3.

(top row) Slant TEC and (bottom row) estimated vertical TEC measured by GPS transits from 0600 UT to 1200 UT on (left column) 11 November 2008 and on (right column) 11 November 2011. The colors represent different satellites.

[7] There is a clear difference in character of the traces obtained in 2008 and 2011. First, the sTEC values in November 2008 peak (with one exception) at around 15 TEC units (TECU), whereas those for November 2011 peak at around 35 TECU. This is a clear indication of the increase in electron densities associated with increased solar activity. More interestingly, the 2011 traces are not smooth, but contain peaks typically of about an hour's duration and about 5 TECU height. These peaks are thought to be due to the presence of convecting patches of enhanced electron density (during this period, the IMF was mainly southward and Kp = 1- to 2-, conditions consistent with the presence of patches [MacDougall and Jayachandran, 2007]).

[8] Estimations of the vertical TEC at the latitude and longitude of the pierce point (the point at which the ray path from satellite to ground station crosses 350 km altitude, the approximate height of the F region peak) may be made by multiplying the sTEC by the cosine of the angle of incidence to the ionosphere at the pierce point, taking into account the Earth's curvature. Such estimated vertical TEC (evTEC) values are clearly an approximation, particularly so at high latitudes where the ionospheric electron density profiles are likely to vary within the geographic area covered by the slant ray path from satellite to receiver while in the ionosphere. The concept of an ionospheric pierce point is a significant oversimplification that is often adopted by various workers. From the point of view of this paper, an ionospheric enhancement 100 km above the assumed pierce point height of 350 km would have an associated ground range error of between 100 and 220 km (associated with elevation angles of 45° and 25°, respectively). Figure 3 (bottom row) shows the same data from 2008 and 2011, replotted in terms of evTEC, calculated as indicated above. The difference in evTEC for the various satellites is around 10 TECU, compared with around 20 TECU for the sTEC measurements.

[9] A quantitative indication of the variation in the magnitude (in terms of total electron content) of polar patches is given in Table 1 for the month of October in 2008, 2009, 2010, and 2011. In each 6 h period for each day, the largest short-term (i.e., up to 1 h) change in evTEC was noted. These were then averaged over the whole month to indicate the diurnal variation and how this changes with the solar cycle. Some variation was also evident over longer periods (i.e., significantly greater than 1 h), but this was not included in the analysis as it was probably due to effects other than polar patches. A marked increase in patch intensity associated with higher sunspot activity is clearly evident. The remainder of this paper considers particular case studies and short-term statistics from the period October to November 2011, corresponding to the peak in solar activity (see Figure 2).

Table 1. Average Over 1 Month of Maximum Feature Size (in TECU) in Each of the Periods 0000–0600 UT, 0600–1200 UT, 1200–1800 UT, and 1800–0000 UTa
 0000–0600 UT0600–1200 UT1200–1800 UT1800–0000 UT
  1. a

    Values for October in each of 2008, 2009, 2010, and 2011.


4 Comparison of GPS TEC Measurements With HF Vertical Sounding Measurements

[10] A sequence of ionograms for Qaanaaq at 15 min intervals between 02.15 and 03.30 UT on 1 November 2011 is presented in Figure 4. The ionograms differ markedly from typical midlatitude soundings; in particular, the traces are usually spread and often exhibit spread extensions to what appear to be the “normal” F region traces. Dramatic changes in the F region critical frequency (foF2) occur between consecutive ionograms, as does the maximum frequency reflected (fmax) that exceeds foF2 when a spread extension trace is present. The size of the fluctuations in foF2 and their duration are similar to those shown in MacDougall and Jayachandran, [2007] from the network of vertical sounders forming part of the Canadian High Arctic Ionospheric Network.

Figure 4.

(a–f) Sequence of ionograms from Qaanaaq on 1 November 2011, 0215–0330 UT. Data obtained from the Lowell DIDBase [Reinisch et al., 2004]. The red/yellow points are indicative of o-mode reflections and the green/blue/grey points indicative of x-mode reflections.

[11] Figure 5 (top) indicates the 20 min average of the evTEC values of all pierce points that come within the general area of Qaanaaq (56°–81°W and <82°N, while still maintaining elevation >25°, shown as the shaded area in Figure 1). F region critical frequency (foF2) values taken from the Qaanaaq ionosonde are shown for the whole of that day in Figure 5 (middle), and the maximum reflected frequencies (fmax) in Figure 5 (bottom). Despite the variation in location of GPS measurements, the figure shows that there is a good correspondence in the timing of peaks between the two data sources (see in particular the peaks at 0340, 0640, 0825, 1135, 1240, and 1322 UT).

Figure 5.

(top) Twenty minute average of estimated vertical TEC (evTEC) obtained from all pierce points near Qaanaaq (between 56° and 82° west and south of 82° north), and the (middle) FoF2 values and (bottom) maximum observed frequency from the Qaanaaq Digisonde (data obtained from the Lowell DIDBase [Reinisch et al., 2004]).

[12] Simulations, using techniques developed from Zaalov et al. [2003, 2005], of vertical ionograms with a background ionosphere corresponding to the times of the data in Figure 4 are given in Figure 6. The simulations are not an attempt to reproduce the sequence of measured ionograms since to do so would require a detailed knowledge of patch size, shape, position, and intensity that we do not have. Furthermore, the simulations do not include scattering, so traces are not as spread as in the measurements. Figure 6a is an ionogram (showing o-mode only) for the background ionosphere with a critical frequency of around 4.5 MHz with no significant patches present. Figure 6b is a simulated ionogram with a patch overhead the ionosonde. In this case, the ionogram displays a normal shape, but with a raised critical frequency, and is somewhat reminiscent of the ionogram displayed in Figure 4e. Figures 6c and 6d are simulations with the patch not overhead (by several hundred kilometers) the ionosonde, and in these cases, the traces are somewhat more complex. Of particular note in Figure 6d is the feature extending above what appears to be the normal shaped trace up to a frequency of around 6.0 MHz. Extensions of this nature appear to be a common feature of the measured ionograms.

Figure 6.

Simulated Qaanaaq ionograms for 1 November 2011 for an ionosphere containing a patch, approximate size 700 km, close to the ionosonde viewing area. O-mode traces only are shown. (a) No patch present. (b) Patch overhead. (c and d) Scenarios with the patch not overhead the ionosonde.

5 Comparison and Simulation of GPS TEC Measurements With HF Doppler Measurements

[13] Previous experimental results [e.g., Warrington et al., 1997] have demonstrated that polar patches can affect the propagation of HF signals, for example, reflection from the enhanced electron density within the patch can introduce off great-circle propagation and lead to signals being received at times when they would otherwise not be expected. Since the patches are moving, it is expected that the frequency of the signals will be Doppler shifted with a positive shift as the patch moves toward the path and a negative shift as it moves away.

[14] An example of the correspondence between Doppler shift and GPS evTEC measurements from 21:00 UT on 31 October 2011 to 05:00 UT on 1 November 2011 is shown in Figure 7. In this instance, vertical TEC was estimated from three satellites in succession (PRNs 5, 15, and 9) with ionospheric pierce points close to the HF path. The paths of the pierce points for these satellites as a function of time of day are shown in Figure 8. In the period up to around 00:30 on 1 November 2011, three relatively sharp peaks in evTEC are evident at 22:30, 23:10, and 00:30. These appear to be associated with positive-to-negative swings in the Doppler trace at 14 MHz starting at around 22:00, 22:50, and 00:00, respectively. Although a few tens of minute offsets are apparent between peaks in the GPS data and in the start of a positive-to-negative swing in the HF Doppler frequency, the timing of the peaks suggests that the two types of measurement are responding to the same electron density features in the ionosphere (note that the GPS ionospheric pierce points and the HF path midpoint are offset). For these times, there is little evidence of significant Doppler frequency swings at the lower frequencies of 10 and 8 MHz, although there are some swings at 10 MHz between 00:30 and 01:00 UT. From 01:00 UT, the 14 MHz signal appears to be above the path maximum usable frequency (MUF) as propagation has mainly ceased, with the exception of a positive-to-negative Doppler swing starting at 02:30 UT, apparently associated with a evTEC peak for two satellites. At 10 MHz, propagation, when it occurs after 01:00 UT, is associated with positive-to-negative Doppler swings that correspond with evTEC peaks at 01:20, 02:40, and 02:50 UT. Further swings from around 03:20 UT may be associated with a more prolonged evTEC peak. Note that a 1:1 correspondence is not necessarily expected as a patch may be positioned such as to affect the HF path but not so as to be detected by the GPS receiver. At 8 MHz, there is some evidence of Doppler swings after around 01:50 UT.

Figure 7.

Estimated vertical TEC, time of flight for the 14 MHz signal, and the Doppler shifts at 14, 10, and 8 MHz on 31 October 2011/1 November 2011.

Figure 8.

The loci of the ionospheric pierce points (assumed to be at 350 km height) for satellites PRN 5, 15, and 9 at times corresponding to the data in Figure 7 in relation to the HF path. Numbers along the track indicate universal time.

[15] A second example for an 8 h period on 9 December 2011 is presented in Figure 9. Vertical TEC was estimated from two satellites in succession (PRNs 4 and 5) with ionospheric pierce points close to the HF path. The paths of the pierce points for these satellites as a function of time of day are shown in Figure 10. It is interesting to note that, in this example, propagation at 14 MHz only occurs at times corresponding to peaks in evTEC, and that there is evidence at these times (albeit not always strong) of positive-to-negative Doppler swings. At 10 MHz, there are few swings, except for one marked example at around 15:30 UT.

Figure 9.

Estimated vertical TEC, time of flight for the 14 MHz signal, and the Doppler shifts at 14, 10, and 8 MHz on 9 December 2011.

Figure 10.

The loci of the ionospheric pierce points (assumed to be at 350 km height) for satellites PRNs 4 and 5 at times corresponding to the data in Figure 9 in relation to the HF path. Numbers along the track indicate universal time.

[16] In order to relate the movement of the patches to the observations presented here, a simple model has been developed. Some of the relevant parameters of the model are given in Figure 11. In the model, the following assumptions are made:

  1. [17] a spherical earth is used;

  2. [18] only one patch exists at a given time;

  3. [19] the enhanced electron density in the patch is sufficient to reflect the signal or that irregularities present in the patch are sufficient to scatter the signal and that if the patch is above the horizon (although see 6), a signal will be received;

  4. [20] each polar patch is a point reflection source with a constant speed v, direction relative to the path (θ), and virtual height, hB;

  5. [21] the signal propagates via a single ionospheric interaction (i.e., a one hop mode), and the effects of refraction (e.g., group retardation) are ignored; and

  6. [22] the antennas at transmitter and receiver are isotropic, but an elevation threshold of 5° has been adopted below which no signal is launched or received (a reasonable assumption for the antennas employed), and this realistically restricts the effective horizon.

Figure 11.

A schematic of the parameters used in the simple model.

[23] In the model, the patch moves on a great circle (GC) that includes the start position of the patch and the point where it crosses the path, Pc, and lies at an angle θ to the path. If the GC distance between the transmitter and receiver is PL, then, Pc is defined as the fraction (in %) of PL measured from the transmitter (i.e., 0% would mean that the patch crossed the path at the transmitter, 100% at the receiver, and 50% at the path midpoint). For patches that cross the GC that connects the transmitter and receiver at points not between the terminals, if the patch crossed at a point beyond the transmitter, then Pc is negative, while if the patch crossed at a point beyond the receiver, then Pc is greater than 100%. If the total distance from the transmitter to the patch and the patch to the receiver is given by D = d1 + d2, then the time of flight is D/c, where c is the speed of light, and the Doppler frequency, fD, is proportional to −dD/dt.

[24] The characteristic curves of the Doppler frequency and time of flight as a single patch moves through the midpoint of the Qaanaaq–Ny-Ålesund path at different values of θ are presented in Figure 12, where the values of D and fD have been calculated every 10 s of simulation time. A patch speed of 600 m s−1 was used in the simulation presented to correspond with measurements made by SuperDARN at 23:00 UT on 31 October 2011. For this time, it should be noted that the angle between the GC path and the convection flow direction is approximately 30°. When the patch travels toward the path (for time <0), the path length shortens and as a consequence the Doppler shift is positive (and the time of flight (TOF) decreases, albeit only very slightly in marked contrast to simulations for a shorter path reported by Stocker et al. [2013]), while when the patch has crossed the path and is moving away, the Doppler frequency decreases and the TOF increases. For this path and patch parameters, θ has a small effect on the TOF, except at times shortly after it comes into view and just before it is no longer visible. While the shape of the Doppler curve is only weakly affected by θ, the rate of change of Doppler does change with the direction the patch is moving in. The length of time the patch is visible also depends on θ, because there are positions where the patch is not visible (because it is below the horizon at either the transmitter or the receiver or both) and this then depends on the trajectory. The simulated data in Figure 12 share many of the characteristics of the measured values in Figure 7; and therefore, despite the limitations of the model, it can provide a useful way of inferring information about the patch speed, direction, etc. from experimental observations.

Figure 12.

Simulated Doppler frequency and time of flight for a 14.4 MHz signal propagating on the Qaanaaq to Ny-Ålesund path for a patch with hB = 250 km, v = 600 ms−1, that crosses the path midpoint for several values of θ (0°, 20°, and 40°).

6 Variations in MUF Due to Patches

[25] An important effect arising from the presence of the patches of enhanced ionization is that signals may be reflected (often on paths away from the great circle) at higher frequencies than would be the case with the background ionosphere alone, significantly altering the geographical and frequency coverage of HF transmitters operating within the polar regions [Warrington et al., 2012]. There is also evidence of this effect in Figures 7 and 9.

[26] To illustrate the dependence of propagation at the higher frequencies on the presence of patches, the number of days on which signals were observed (out of a maximum of 32 days on which observations were made) between 12 October 2011 and 18 November 2011 as a function of maximum patch intensity (determined as the magnitude of evTEC fluctuations) for the period 0000 to 0600 UT on each day is presented in Table 2. The 0000 to 0600 UT time period was chosen since the background MUF was below the signal frequency; and therefore, propagation, when it occurred, was from enhanced electron density structures. Within each 6 h period, 180 soundings were made on each frequency.

Table 2. Occurrence Frequency for Various Combinations of Maximum Feature Size in EvTEC and Number of HF Data Received During the Period 0000–0600 UTa
 Max Feature Size (TECU)
  1. a

    12 October 2011 to 18 November 2011.

Number of data at 14.4 MHz0–9367
Number of data at 11.1 MHz0–9121
Number of data at 10.4 MHz0–9101
Number of data at 8.0 MHz0–9001
Number of data at 6.95 MHz0–9000
Number of data at 4.6 MHz0–9215

[27] For this analysis period, 3 days had maximum patch intensities (measured by evTEC fluctuation) of 0–1 TECU, 10 days with 2–3 TECU, and 19 days with ≥4 TECU. At 6.95 and 8.0 MHz, good propagation occurred on most days, as expected as the MUF exceeded the signal frequency. With increasing frequency, fewer HF signals are received, as expected, but the number of signals received also becomes more dependent on the patch activity. At 14 MHz, very few signals are observed in this period unless patch activity reaches 4 TECU. However, this relation is not reversible: good propagation does not imply high patch activity.

7 Concluding Remarks

[28] In order to develop HF forecasting and nowcasting applications to support, e.g., communications with the increasing number of transpolar airline flights, it is desirable to have a range of ionospheric measurements as inputs to propagation models in addition to predictions and forecasts of ionospheric and geomagnetic parameters. One source of near real-time ionospheric data is networks of GPS receivers that have the capability of measuring ionospheric total electron content along the paths from ground-based receivers to the satellites. With this need in mind, we have compared measurements from a dual frequency GPS receiver located at Alert with ionograms from the digisonde at Qaanaaq and with oblique HF propagation measurements on a range of frequencies over the path from Qaanaaq to Ny-Ålesund. Good correspondence has been shown between the GPS measurements and the HF propagation characteristics, indicating that GPS measurements are likely to be a useful source of ionospheric data for new nowcasting and forecasting techniques.


[29] The authors are grateful to ARCFAC for supporting the measurements at Ny-Ålesund, to Natural Resources Canada for hosting the GPS receiver at Alert, and to Svend Erik Ascanius of the Qaanaaq Geophysical Observatory for hosting the HF transmitter. The authors also gratefully acknowledge NOAA for the sunspot number data and the University of Lowell Centre for Atmospheric Research for the Digisonde data.