Plasma density gradients at the edge of polar ionospheric holes: the absence of phase scintillation

. TS2 Polar holes were observed in the high-latitude ionosphere during a series of multi-instrument case studies close to the Northern Hemisphere winter solstice in 2014 and 2015. These holes were observed during geomagneti-cally quiet conditions and under a range of solar activities 5 using the European Incoherent Scatter Scientiﬁc Association (EISCAT) Svalbard Radar (ESR) and measurements from Global Navigation Satellite System (GNSS) receivers. Steep electron density gradients have been associated with phase scintillation in previous studies; however, no enhanced scin- 10 tillation was detected within the electron density gradients at these boundaries. It is suggested that the lack of phase scintillation may be due to low plasma density levels and a lack of intense particle precipitation. It is concluded that both signiﬁcant electron density gradients and plasma density levels 15 above a certain threshold are required for scintillation to occur.

using the European Incoherent Scatter Scientific Association (EISCAT) Svalbard Radar (ESR) and measurements from Global Navigation Satellite System (GNSS) receivers. Steep electron density gradients have been associated with phase scintillation in previous studies; however, no enhanced scin- 10 tillation was detected within the electron density gradients at these boundaries. It is suggested that the lack of phase scintillation may be due to low plasma density levels and a lack of intense particle precipitation. It is concluded that both significant electron density gradients and plasma density levels 15 above a certain threshold are required for scintillation to occur.

Introduction
The F region ionosphere is a weakly ionised plasma in the Earth's atmosphere extending from an altitude of ∼ 150 20 to ∼ 500 km, above which it merges with Earth's plasmasphere. Large-scale plasma structures with a horizontal extent of tens to hundreds of kilometres are routinely observed in the F region high-latitude ionosphere (Tsunoda, 1988). One type of structure commonly observed are polar cap 25 patches, also referred to as patches, which are enhancements of plasma density with at least twice the background value and a horizontal spatial extent of 100 km or more (Crowley, 1996). Buchau et al. (1983) observed such patches of enhanced ionisation drifting anti-sunward with the background 30 plasma flow in the central region of the polar cap at Thule, Greenland: 77.5 • N, 69.2 • W; 85.4 • MLAT (magnetic latitude), 32.4 • MLON (magnetic longitude)CE1 . The patch densities were larger than could be produced due to the observed flux of precipitating particles, and it was concluded that the 35 patches were not produced locally by precipitation. Weber et al. (1984) suggested that the patches were produced on the dayside at auroral or sub-auroral latitudes and then convected anti-sunward to higher, polar latitudes. A comparison of average maps of the electron density and the high-latitude 40 convection pattern suggested that solar-produced plasma was drawn into the polar cap as a continuous density enhancement known as the "tongue of ionisation" (TOI; Foster et al., 1984 TS3 ). Several mechanisms have been proposed to break a TOI into a series of patches, including variations in 45 the high-latitude convection pattern moving flux tubes in and out of sunlight (Anderson et al., 1988), expansion and contraction of the high-latitude convection pattern in response to transient bursts of reconnection drawing in plasma from different latitudes (Cowley and Lockwood, 1992;Lockwood 50 and Carlson, 1992;Carlson et al., 2002Carlson et al., , 2004Carlson et al., , 2006, variations in the y component of the interplanetary magnetic field (IMF) drawing in plasma from different magnetic local 2 L. A. Jenner et al.: Plasma density gradients at the edge of polar ionospheric holes times (MLT; Sojka et al., 1993), variation in the z component of the IMF altering whether plasma could be drawn in to the polar cap (Valladares et al., 1998), erosion of plasma densities due to enhanced recombination during a flow channel event (Rodger et al., 1994;Valladares et al., 1994), and modification of the density of the photoionised plasma transported into the polar cap by particle precipitation (Walker et al., 1999;Millward et al., 1999). Patches have been observed travelling thousands of kilometres across the polar regions (Weber, 1986TS4;Oksavik et al., 2010;Nishimura 10 et al., 2014), and they are primarily associated with times when the z component of the IMF is negative (Buchau and Reinisch, 1991).
Blobs are also plasma density enhancements; however, unlike patches, they occur outside the polar cap. They are further categorised into boundary blobs, sub-auroral blobs, and auroral blobs (Rino, 1983 TS5 ;Jin et al., 2016). Boundary blobs are found near the equatorward auroral boundary, neighbouring the ionospheric trough's poleward wall. Parkinson et al. (2002) observed patches leaving the polar 20 cap, slowing in the anti-sunward direction and then beginning to move zonally. It was suggested that these patches would form boundary blobs; this was later confirmed by Pryse et al. (2006), who compared the plasma density in a polar cap patch to that within a boundary blob that the patch 25 subsequently formed. Sub-auroral blobs have a similar appearance to boundary blobs; however, they are found in the ionospheric trough. Auroral blobs are found within the auroral oval and seem to be longitudinally restricted. The most likely mechanism for their creation is particle precipitation 30 (Jones et al., 1997).
Not all ionospheric structures are enhancements of the background plasma; polar ionospheric holes are regions of low plasma density. Brinton et al. (1978) observed a depletion of this kind under low solar activity (F10.7 = 71 sfu) 35 and low magnetic activity (Kp = 2) conditions. This depletion was also associated with a minimum of electron temperatures, indicating the absence of local particle precipitation. Polar holes are generally located between 21:00 and 06:00 MLT and between 70 and 80 • MLATCE2 and typically 40 have steep plasma density gradients at their boundaries. They are believed to be produced when plasma in the high-latitude convection pattern circulates in perpetual darkness. Plasma loss by recombination in the absence of a plasma source causes density levels to drop. This idea is supported by the 45 conditions under which polar holes have generally been observed, namely quiet geomagnetic activity (Kp 2 or less) when the contribution to the plasma densities from particle precipitation is low (Brinton et al., 1978). The electron densities inside of the polar holes are seen to reach a minimum in 50 the range of 10 8 -10 11 electrons m −3 (Obara and Oya, 1989;Benson and Grebowsky, 2001) and, while there is variation between holes, inside of a singular polar hole the density level is very consistent.
Smaller-scale structures can arise at steep plasma den-55 sity gradients due to instability processes such as the gradient drift instability (GDI) (Keskinen and Ossakow, 1983) and the velocity-shear-driven instability (Kelvin-Helmholtz instability, KHI). Carlson et al. (2008) proposed that the real process involves both mechanisms acting on different 60 timescales. The smaller-scale (tens of metres to tens of kilometres) plasma density structures that arise cause variations in the refractive index of the ionosphere. As a GNSS signal passes through this region, refraction and/or diffraction of the radio wave causes fluctuations in the phase and am-65 plitude of the signal. Ionospheric scintillation is the rapid fluctuation of the received signal which can disrupt applications using GNSS, as thoroughly reviewed by Hapgood (2017). Since the Second World War, large numbers of studies have shown the effect of ionospheric irregularities on ra-70 dio signals, as reviewed by Aarons (1982). The morphology of these irregularities has been extensively studied at high latitudes (e.g. Kersley, 1972 TS6 ), along with the effects upon the propagation of radio signals in this region (e.g. Kersley et al., 1995). 75 More recently studies have focused on Global Navigation Satellite System (GNSS) frequencies, where scintillation poses a substantial threat to the integrity, availability, and accuracy of GNSS positioning, leading to positioning errors and service outages due to signal tracking problems at the 80 GNSS receiver. A direct connection between gradients in the total electron content (TEC) at the edge of a plasma stream with both phase and amplitude scintillation has been observed (Mitchell et al., 2005), and plasma structuring caused by auroral precipitation has been linked to the loss of sig-85 nal lock by a GNSS receiver (Elmas et al., 2011;Smith et al., 2008;Oksavik et al., 2015). A statistical study has shown an agreement between both phase and amplitude scintillation with the asymmetric distribution of polar cap patches around magnetic midnight (Spogli et al., 2009) and that au-90 roral emissions correlate with GNSS signal phase scintillation (Kinrade et al., 2013;. Phase and amplitude scintillation can be associated with the larger spatial structures related to polar cap patches (Alfonsi et al., 2011). The climatology of ionospheric scintillation at 95 polar latitudes in the Northern Hemisphere was determined over almost two solar cycles, and the dependence upon solar cycle, geomagnetic activity, and solar wind conditions was shown by De FranchesciTS7 et al. (2019). Phase scintillation is usually the dominant process at high latitudes (Spogli et 100 al., 2009;Prikryl et al., 2015), and this is the focus of the present study.
Phase scintillation is commonly quantified by the standard deviation of the signal phase, σ ϕ , which is usually computed over 60 s. The refractive component of the signal is usually 105 assumed to be slowly varying and associated with frequencies of less than 0.1 Hz. Therefore, by only considering frequencies greater that 0.1 Hz, the diffractive effects (usually referred to as scintillation) can be distinguished (Fremouw et L. A. Jenner et al.: Plasma density gradients at the edge of polar ionospheric holes 3 al., 1978). However, the 0.1 Hz cutoff can give spurious observations of phase scintillation as a result of erroneous data detrending (Forte and Radicella, 2002). When a GNSS satellite is observed at low elevation angles, the σ ϕ index cannot distinguish between phase scintillation and background noise 5 for weak to moderate phase scintillation (Forte, 2005). Wang et al. (2018) showed that rapid variations in the phase of a trans-ionospheric signal can arise as a result of plasma structures moving rapidly relative to an observer at ground level and can therefore give the appearance of phase scintillation. 10 Rapid changes in the spatial distribution of electron density can also introduce similar effects, as the GNSS satellite-toreceiver ray path can sweep through these irregularities at high speed, resulting in high-frequency refractive variations (McCaffrey and Jayachandran, 2019). 15 The presence or absence of scintillation effects on transionospheric radio signals have been extensively studied for electron density enhancements in the high-latitude ionosphere; however, the effect of the steep plasma density gradients at the edge of depletions, such as polar holes, has not 20 been as comprehensively considered. The purpose of this paper is to report on the effects of such steep density gradients on GNSS signals, observed in three multi-instrument case studies close to northern winter solstice, and to provide observational evidence that supports the work of Aarons 25 (1982).

Instrumentation
The European Incoherent Scatter Scientific Association (EISCAT) operates the EISCAT Svalbard Radar (ESR) at Longyearbyen (78.2 • N, 16.0 • E; 15.2 • MLAT, 30 112.9 • MLON) on Svalbard (Wannberg et al., 1997). The site consists of two antennas, a 32 m parabolic dish, and a 42 m parabolic dish. The 42 m dish is fixed along the direction of the local geomagnetic field lines (azimuth −179 • ; elevation 81.6 • ), whereas the 32 m dish is steerable with respect to both 35 azimuth and elevation. Observations of the electron density, electron temperature, ion temperature, and ion drift line of sight velocity in the ionosphere from this incoherent scatter radar (ISR) are used in this study.
The Super Dual Auroral Radar Network (SuperDARN) is 40 a network of high-latitude coherent scatter radars (Greenwald et al., 1995;Chisham et al., 2007;Nishitani et al., 2019) that observe line-of-sight plasma velocities in the F region. These measurements are assimilated using the map potential technique (Ruohoniemi and Baker, 1998TS8), which uses an 45 ionospheric convection model to map the electrostatic potential pattern. Electrostatic equipotential lines are streamlines of ionospheric convection flows. As the plasma drift velocity is perpendicular to both the electric and magnetic fields in the F region (E × B TS9 drift), the plasma convection pattern 50 can be directly inferred from the electric potential maps. GNSS signals detected by NovAtel GPStation-6 receivers at the Kjell Henriksen Observatory (KHO; 78.2 • N, 16.0 • E; 15.2 • MLAT, 112.9 • MLON) can be used to infer the effects of the ionosphere on radio waves travelling though this 55 medium. Amplitude scintillation is measured using the S 4 index, which is the square root of the variance of received power divided by the mean value of the received power (Briggs and Parkin, 1963). Phase scintillation is measured using the σ φ index, which is the standard deviation of the 60 detrended carrier phase φ in radians (Fremouw et al., 1978) over 60 s.
The IMF was observed by the Advanced Composition Explorer (ACE), which is a NASA spacecraft orbiting the L1 Lagrangian point of the Earth-Sun system, roughly 1.54 × 65 10 6 km from the Earth (Zwickl et al., 1998)CE3. In addition to the x, y, and z components of the IMF, the clock angle, given by arctan |By| |B z | , is also considered. When the clock angle is greater than 45 • either |B y | > |B z | or B z < 0; in either case a two-cell convection pattern is expected with anti-70 sunward flow drawing plasma from day to night across the polar cap (Thomas and Shepherd, 2018).
Total electron content (TEC) maps are used to put these measurements into context. These were obtained from the Madrigal database at the MIT Haystack Observatory (Rid-75 out TS10 and Coster, 2006; Vierinen et al., 2016). Two other indices are used within this study: the Kp index is used as a proxy for disturbances to the geomagnetic field, and the F10.7 cm solar flux is used as a proxy for solar activity. These indices were both obtained from the UK Solar System Data 80 4 L. A. Jenner et al.: Plasma density gradients at the edge of polar ionospheric holes Centre (UKSSDC) at the Rutherford Appleton Laboratory, UK.

Case study: 17 December 2014
The 3-hourly Kp values observed on 17 December 2014 be-5 tween 12:00 and 23:59 UT ranged between 1− and 1+, indicating quiet conditions. The F10.7 cm solar flux was relatively high; the value of 198.5 sfu is typical of solar maximum. The IMF observed by the ACE spacecraft between 12:00 and 23:59 UT (Fig. 1) was characterised by a positive value for the IMF B y (mean value 3.9 nT). IMF B z was more variable but generally showed smaller values (mean value of 1.7 nT). The clock angle was generally greater than 45 • from 14:00 to 19:00 UT, and the corresponding SuperDARN plots (discussed later in this section) show that a two-cell convec-15 tion pattern dominated until at least 20:00 UT.
Total electron content (TEC) maps (Fig. 2) show the overall plasma density throughout the high-latitude regions. The TEC maps at 12:00 and 15:00 UT display values of ∼ 2 TECu (dark blue) in the polar cap. At 18:00 UT and 20 21:00 UT larger electron densities can be observed crossing the polar cap in a two-cell convection pattern, with values of ∼ 15 TECu (yellow), indicating that plasma produced by photoionisation on the dayside is being drawn into the polar cap. This plasma is being drawn into the polar cap during 25 relatively quiet conditions (Kp ∼ 1) and is consistent with a two-cell convection patternCE4 .
The electron densities and temperatures observed by the field-aligned 42 m dish of the EISCAT Svalbard Radar (ESR) between 12:00 and 23:59 UT are shown in Fig. 3. The scales 30 for this plot have been chosen to enable a clear comparison with other figures presented in this paper. A clear depletion in the electron densities is observed between approximately 16:00 and 18:00 UT at all altitudes. The electron and ion temperatures are not elevated at this time with values of approxi-35 mately 1000 K, suggesting that this depletion is void of particle precipitation and did not arise from enhanced recombination due to Joule heating. The ESR does not show a substantial plasma velocity aligned with the radar beam. This radar observed at an elevation of 81.6 • , which is aligned with the 40 magnetic field line in the F region. There was no substantial component of velocity observed along the magnetic field line. In order to further investigate the electron density de-   pletion, a line plot of the maximum detected electron density from 90 to 400 km is shown (Fig. 4). In addition to the maximum density, two other values are present on the plot: the average value for the whole day and 35 % of the average value. The depletion was defined as the time when the electron den-5 sity dropped below the 35 % line; in this case, the depletion was defined as starting at 16:29 UT and ending at 18:00 UT. Figure 5 shows the high-latitude convection pattern inferred from the SuperDARN radars for three representative times during the period that the electron density depletion was observed by the ESR. These clearly show a two-cell convection pattern, with plasma drawn anti-sunward across the polar cap. The ESR observes at a given location, which rotates under the convection pattern. The depletion, identified in Fig. 4, is indicated by a black line. At midwinter, Svalbard 15 is in perpetual darkness. On 14 December, the ground-level terminator is at a maximum latitude of 68 • N, which corresponds to a maximum magnetic latitude CE7 of 76 • MLAT at 21:00 UT. This depletion is nightward of the terminator, and the SuperDARN convection patterns suggest that this plasma 20 is circulating in perpetual darkness. It is interpreted as a polar hole.
The data collected by the GNSS receiver were from the GPS, Galileo, and GLONASS systems, and the receiver provided the azimuth and elevation of the satellite with respect 25 to the receiver. This was converted into a latitude and longitude using the radio wave path and assuming that the data correspond to 350 km in altitude, in line with previous studies (e.g. Cervera and Thomas, 2006;Forte and Radicella, 2002). At low elevation angles, the GNSS TEC and scintillation data 30 can become unreliable due to multi-path issues; thus, observations at an elevation of less than 30 • were discarded. This cutoff has been used in previous studies, such as Mitchell et al. (2005). Signal lock times below 240 s were also discarded, in line with previous studies (e.g. van der Meeren 35 et al., 2015). The satellite tracks were overlaid onto Super-DARN plots (Fig. 5).
TEC and phase scintillation data from GNSS receivers were taken during times when the polar hole was observed. This hole is observed for 1.5 h, and several satellite paths 40 are present during this time window. The GNSS TEC data clearly show lower TEC levels at and around the area marked by the ESR as a hole, and, on some of the satellite trajectories, sharp changes can be seen with the edge of the hole. A one-to-one correspondence between the GNSS TEC data and the EISCAT data is neither expected nor observed. It is highly likely that the polar hole will evolve during the time for which it is observed; therefore, the plots in Fig. 5 include both spatial and temporal variation. The ESR ob-5 serves the polar hole for 91 min, and the plasma velocity surmised CE8 from the electric potential patterns inferred from the SuperDARN radars (Fig. 5) at this location is of the order of 150 m s −1 ; this indicates that the polar hole has a horizontal extent of some 800 km in a direction parallel to the plasma 10 flow. In summary, the combination of the EISCAT and GNSS TEC measurements indicate that the polar hole is present for an extended period of time (of the order of hours) over a large spatial scale (hundreds of kilometres).
Panels showing the location of phase scintillation on the 15 satellite tracks are also displayed in Fig. 5. A threshold of 0.2 rad was used to identify phase scintillation. Different authors have used different thresholds for phase scintillation, including 0.2 rad (e.g. van der Meeren, 2015), 0.25 rad (e.g. Alfonsi et al., 2011), and 0.3 rad (e.g. Kinrade et al.,20 2013). The purpose of using a low threshold within the present study was to ensure that any possible indication of phase scintillation was included. As TEC and scintillation are collected simultaneously, comparing the two might be expected to show increased scintillation where there are 25 changes in TEC. No scintillation was observed on the edges of the holes.

Case study 2: 10 December 2015
The F10.7 cm solar flux for this case was lower than in the first study, with a value of 108.5 sfu. The Kp index was 30 higher, with a value of 3 from 12:00 to 18:00 UT and a value of 4 at 21:00 and 24:00 UT, indicating an active state, but not storm levels. Once again the IMF was variable, with B z taking positive and negative values. B y was consistently larger than B z and was dominantCE9 . As in the previous case study, 35 a two-cell convection pattern was observed.
The TEC maps at 18:00 and 21:00 UT are shown in Fig. 7. As in the previous case study, these maps indicate higher density plasma produced at lower latitudes being drawn across the polar cap within the high-latitude convection pattern, 40 with this effect maximising at 21:00 UT.
The 42 m ESR observations (Fig. 8) for this day show an electron density depletion that contains all of the previously discussed markers, with no significant velocity in the fieldaligned direction. 45 Using the same method as in the previous case, the hole was identified, with the start and end times given as 15:15 and 16:43 UT respectively CE10 . The 32 m ESR observations (Fig. 15) show a depletion at around 15:00 UT.
The high-latitude convection pattern was inferred from the 50 SuperDARN radars (Fig. 11), with the location of the polar hole observed in the 42 m ESR observations and GNSS TEC and phase scintillation measurements overlaid as in the previous case study. The 32 m ESR observations (Fig. 9) were directed poleward, indicating that this is a polar hole rather than 55 the ionospheric trough, which would be located equatorward of the radar. A substantial plasma velocity of some 300 m s −1 towards the radar was observed at 16:00 UT, indicating crosspolar flow in the equatorward direction. The high-latitude convection pattern inferred from the SuperDARN radars also 60 shows anti-sunward cross-polar flow but with a more asymmetric convection pattern than was observed on 17 December 2014. On 10 December 2015 there was a clear dominant dusk cell, drawing plasma across the polar cap from the prenoon sector. The polar hole observed with the 42 m dish of 65 the ESR was in the sunward return flow in the dusk convection cell. CE11 The phase scintillation plot for 15:16 to 16:14 UT (Fig. 11b) has some satellite trajectories passing through the hole boundary but displays no significant scintillation on any 70 of the paths. The later plot (Fig. 11e) does contain phase scintillation, although none of the elevated scintillation matches up to hole boundaries; instead, the scintillation is seen in regions of high and elevated electron density.

75
A series of polar ionospheric holes have been detected in the high-latitude nightside ionosphere in case studies close to winter solstice, under varying solar intensities and geo- magnetic disturbance levels. The first study on 17 December 2014 was characterised by high levels of solar activity (198.5 sfu) and quiet geomagnetic conditions. The second case study on 10 December 2015 also had lower levels of solar activity of (108.5 sfu) but had more active geo-5 magnetic conditions (Kp = 3) than in the previous study. A third case study conducted under quiet geophysical conditions (Kp ≤ 2) and moderate solar activity (F10.7 cm solar flux = 116.7 sfu) on 12 December 2015 showed similar results (not shown).
Ionospheric polar holes contain much lower electron densities than those detected during the rest of the day. This study used the point when the maximum density at a given time dropped 35 % below the daily average maximum density to identify these holesCE12. The changes in electron 15 density are associated with large electron density gradients. Table 1 shows the electron density gradients and the average hole electron density, based on observations from the ESR 42 m. The average polar hole density observed in this study is comparable to those previously reported of 10 8 -20 10 11 electrons m −3 (Obara and Oya, 1989;Benson and Grebowsky, 2001). Steep electron density gradients are observed at the edges of the holes, and these are expressed in units of N e m −3 h −1 CE13 . Although these gradients are expressed in units of per hour (h −1 ), they were calculated from suc-25 cessive observations by the ESR 42 m (these measurements are typically 1 min apart). The spatial extent of these holes was at least several hundred kilometres, as inferred from the GNSS TEC measurements (all studies) and the ESR 32 m observations (case study from 17 December 2014). Polar 30 holes are usually associated with quiet geomagnetic conditions (Kp < 2). It is notable that, on 10 December 2015, a   The IMF conditions during the time when the polar holes were observed, and for several hours beforehand, were appropriate for anti-sunward cross-polar convection.

5
The ground-level solar terminator for winter is only above 70 • MLAT between 15:00 UT and slightly after 21:00 UT, reaching a maximum latitude of just under 76 • MLAT on the dayside at around 21:00 UT, creating the possibility that plasma within the high-latitude convection pattern could cir-10 culate in perpetual darkness; thus, the plasma undergoes recombination whilst simultaneously being insulated from photoionisation, or precipitation, creating a polar holeCE14.
Phase scintillation has previously been observed to coincide with large plasma gradients such as on the edge of iono-15 spheric enhancements, including polar cap patches (Jin et al., 2017), the tongue of ionisation (van der Meeren et al., 2014), plasma structures associated with the aurora (Kinrade et al., 2013;Oksavik et al., 2015;, and the mid-latitude trough (Pryse et al., 1991).

20
The structures that cause scintillation arise due to the gradient drift instability and/or the Kelvin-Helmholtz instability (Keskinen and Ossakow, 1983;Carlson et al., 2008). In the present study, once the boundaries and the large electron density gradients associated with them were identified, these 25 boundaries were investigated for elevated levels of phase scintillation. A threshold of 0.2 rad was used; the purpose of this low value was to ensure that any possible indication of phase scintillation was included. Across all of the observed GNSS points coinciding with the polar hole boundaries, no 30 such levels of phase scintillation were detected. Phase scintillation usually dominates at high latitudes (e.g. Prikryl et al., 2015), although amplitude scintillation has also been observed (e.g. Mitchell et al., 2005). The present study focuses upon phase scintillation as no amplitude scintillation, defined 35 as when the S 4 index was greater than 0.2, was observed on any of the TEC gradients at the boundaries of the polar holes. This is not the first time a plasma density enhancement has been observed without corresponding phase scintillation. Van der Meeren et al. (2016) observed a Sun-aligned polar 40 cap arc under quiet geomagnetic conditions without corresponding scintillation. In the present study, some phase scintillation was observed; however, these points coincide with increases in TEC and the edges of spikes in electron densities at other locations. In the second case study  ber 2015), phase scintillation was observed at a point associated with elevated TEC (Fig. 11e, fCE15 ), but this was not associated with the assumed boundary of the polar hole.
When phase scintillation was observed, it was always associated with electron density gradients; however, the con-50 verse is not always true. Therefore, it appears that some minimum level of overall electron density is needed for phase scintillation to occur. Given that it is the presence of smallscale structures that cause scintillation, this suggests that these small-scale structures have not arisen. 55 Figure 12 shows phase scintillation as a function of TEC and the TEC rate of change. This figure also includes data from a third study (using data from 12 December 2015) that were consistent with the interpretation presented here but have been omitted in the interest of concision. Low scintil-60 lation can be seen at all TEC levels and for a majority of the range of TEC rates of change. Conversely, elevated scintillation levels are only seen above approximately 6 TECU, suggesting that a minimum electron density is required. This is not a new idea, in a review paper Aarons (1982) commented 65 "if the ionosphere is perturbed on a percentage basis, change in N in the trough will be small since N is low; scintillations will then be low." The current paper provides observational evidence to support this suggestion that a minimum electron density is required. The current paper is also con-70 sistent with suggestions made by Prikryl et al. (2015), where the strongest phase scintillations were found to be highly co-locatedCE16 with regions that are ionospheric signatures of the coupling between the solar wind and magnetosphere. Po-12 L. A. Jenner et al.: Plasma density gradients at the edge of polar ionospheric holes  lar holes appear to be areas of weak coupling, and, hence, less scintillation.
In this study, the phase scintillation index (σ φ ) has been calculated across a 60 s interval, which is in line with common practice within this field. However, if this index was 5 computed across a shorter time interval, it is possible that elevated values of σ φ may be associated with the edge of the polar hole. This would be an interesting topic for a future paper. Further developments upon this work would expand the observations of the polar holes discussed to a larger number 10 of examples under a wider range of geophysical conditions. Polar ionospheric holes could be tracked by making observations with a higher temporal resolution at a large number of regularly spaced locations. The advent of EISCAT-3D (Mc-Crea et al., 2015), which will give unprecedented temporal 15 and spatial coverage, will enable such studies in the European sector of the high-latitude ionosphere. The ability to observe the evolution of polar holes over time will give a new, deeper understanding of these features and how they influence practical radio systems such as GNSS.

Conclusions
Polar ionospheric holes are regions of electron density depletions containing large electron density gradients at their boundaries. This paper reports case study observations of polar ionospheric holes conducted using the ESR and GNSS 25 receivers. These holes were observed during both quiet and moderately disturbed geomagnetic conditions, under a range of solar activities. Steep electron density gradients have been associated with phase scintillation at GNSS frequencies in previous studies; however, no enhanced scintillation was de-30 tected upon the electron density gradients at these boundaries. Phase scintillation was only observed when electron density levels were elevated above 6 TECU. Aarons (1982) suggested that a minimum density level may be required for scintillation to occur, and the present study provides support-35 ing observational evidence. We conclude that both a minimum electron density level and a sharp gradient in the election density must be present for instability mechanisms to produce scintillation structures.
Data availability. The data used in this paper are publicly avail-40 able at https://www.eiscat.se (last access: TS13 ). Data for TEC processing are provided from the following organisations: UN-AVCO; Scripps Orbit and Permanent Array Center; Institut Geographique National, France; the International GNSS Service; the Crustal Dynamics Data Information System (CDDIS); the National 45 Geodetic Survey; the Instituto Brasileiro de Geografia e Estatística, RAMSAC CORS of the Instituto Geográfico Nacional de la República Argentina; the Arecibo Observatory; the Low-Latitude Ionospheric Sensor Network (LISN); Topcon Positioning Systems, Inc.; the Canadian High Arctic Ionospheric Network; the Centro di 50 Ricerche Sismologiche; the Système d'Observation du Niveau des Eaux Littorales (SONEL); RENAG: REseau NAtional GPS permanent; GeoNet -the official source of geological hazard information for New Zealand; the GNSS Reference Networks; the Finnish Meteorological Institute; and SWEPOS -Sweden. Access to these data 55 is provided by the Madrigal network: http://cedar.openmadrigal.org/ (last access: TS14 ). The Kp index and F10.7 cm solar flux were obtained from the UK Solar System Data Centre at Rutherford Apple-ton Laboratory. These can be accessed at https://www.ukssdc.ac.uk/ (last access: TS15 ). The IMF data were provided by N. TS16 Ness and obtained from the CDAWeb at https://cdaweb.gsfc.nasa.gov/ (last access: TS17 ). CE17 TS18 Author contributions. This work was led by LAJ, under the guidance of AGW. KO provided the GNSS TEC and scintillation data as well as guidance regarding their interpretation. TKY and ARF provided the SuperDARN electric potential maps as well as guidance regarding their interpretation. AJC provided the TEC maps and guidance regarding their interpretation. All authors contributed 10 to the discussion. The paper was prepared by LAJ and AGW.
Competing interests. The authors declare that they have no conflict of interest.
Sweden, and the UK. Assistance from Ingemar Häggström and colleagues at the EISCAT Scientific Association with respect to running the experiments is gratefully acknowledged. Assistance from Steve Crothers and Matthew Wild at Rutherford Appleton Laboratory regarding the data processing is gratefully acknowledged.

20
The GNSS TEC and scintillation data were provided by Kjellmar Oksavik at the University of Bergen, which CE18 is supported by the Norwegian Research Council (contract nos. 212014 and 223252). The authors acknowledge the use of SuperDARN data, which are available at https://vt.superdarn.org (last access: TS19 ). 25 SuperDARN is a collection of radars funded by national scientific funding agencies of Australia, Canada, China, France, Italy, Japan, Norway, South Africa, the UK, and the USA. Alexandra R. Fogg is supported by a studentship from the Science and Technology Facilities Council (UK). Assistance from Nathan Brown with re-30 spect to the production of Figs. 5 and 11 is gratefully acknowledged. GPS TEC data products and access through the Madrigal distributed data system are provided to the community (http: //www.openmadrigal.org, last access: TS20 ) by the Massachusetts Institute of Technology (MIT) with support from the US National Financial support. This research has been supported by the NAME OF FUNDER (grant no. GRANT AGREEMENT NO). TS21 Review statement. This paper was edited by Keisuke Hosokawa and reviewed by two anonymous referees.

CE2
Please confirm the change.

CE3
Please note the edits.

CE4
Is this information "is consistent with a two-cell convection pattern" given twice in this paragraph? Please check.

CE5
Are the ion and electron temperatures in this figure (and in Fig. 8) referring to Kelvin? If so, "k" will be changed to "K" during proofreading. Please advise.

CE6
Please check that the panel references are correct throughout this caption and that the meaning of the caption is intact.

CE7
Can "magnetic latitude " be removed here as "MLAT" is used? Please check.

CE8
A word has been changed. Please check that the meaning of your sentence is intact.

CE9
Please check that the meaning of your sentence is intact.

CE10
Please confirm the change.

CE11
Please check all figure references in the following paragraph.

CE12
Please check that the meaning of your sentence is intact.

CE13
Our house standards require that units not preceded by a number are written out in full. Please provide this unit in full written form.

CE14
Please check that the meaning of your sentence is intact.

CE15
Please confirm the figure reference.

CE16
Please note the edit.

CE17
Please review the content of the following sections carefully, as edits are not displayed in the track-changes PDF: "Data availability" and "Author contributions".

CE18
Is the university or Kjellmar Oksavik supported by the contracts? Also, should "contract nos." be "grant nos."? Please advise.

Remarks from the typesetter TS1
Please check this email address. Should it be alan.wood@ntu.ac.uk?

TS2
The composition of all figures and key figure has been adjusted to our standards. This also includes language adjustments to Figs. 1-12. TS3 Foster et al. (1984) is not in the reference list. Do you mean Foster (1984)?

TS4
Weber (1986) is not in the reference list. (1983) is not in the reference list. Do you mean Rino et al. (1983)? TS6 Kersley (1972) is not in the reference list. Do you mean Kersley et al. (1972) Ruohoniemi and Baker (1998) is not in the reference list.

TS9
Please check throughout the text that all vectors are denoted by bold italics and matrices by bold roman.

TS10
Please check. Do you mean Ridout or Rideout? TS11 Figure 6 is not mentioned in the text. Figure 10 is not mentioned in the text.

TS13
Please provide date of last access.

TS14
Please provide date of last access.

TS15
Please provide date of last access.

TS16
Please provide the full first name.

TS17
Please provide date of last access.

TS18
This paragraph has been moved here from the Acknowledgements. Please provide direct links to the data set, if possible. In any case, please provide reference list entries including creators, title, and date of last access.

TS19
Please provide date of last access. Should this also appear in the "Data availability" section? TS20 Please provide date of last access.

TS21
Please note that there is funding information given in the acknowledgements but you have not indicated any funding upon manuscript registration. Therefore, we were not able to complete the financial support statement. Please provide the missing information and double-check your acknowledgements to see whether repeated information can be removed from the acknowledgements. Thanks.

TS22
Please provide the page range if there is more than one page.