In addition to being scattered by the ionospheric field-aligned irregularities, HF radar signals can be reflected by the ionosphere toward the Earth and then scattered back to the radar by the rugged ground surface. These ground scatter (GS) echoes are responsible for a substantial part of the returns observed by HF radars making up the Super Dual Auroral Radar Network (SuperDARN). While a GS component is conventionally used in studying ionosphere dynamics (e.g., traveling ionospheric disturbances, ULF waves), its potential in monitoring the state of the scattering surface remains largely unexploited. To fill this gap, we investigated diurnal and seasonal variation of the ground echo occurrence and location from a poleward-looking SuperDARN radar at Rankin Inlet, Canada. Using colocated ionosonde information, we have shown that seasonal and diurnal changes in the high-latitude ionosphere periodically modulate the overall echo occurrence rate and spatial coverage. In addition, characteristics of GS from a particular geographic location are strongly affected by the state of the underlying ground surface. We have shown that (1) ice sheets rarely produce detectable backscatter, (2) mountain ranges are the major source of GS as they can produce echoes at all seasons of the year, and (3) sea surface becomes a significant source of GS once the Arctic sea ice has melted away. Finally, we discuss how the obtained results can expand SuperDARN abilities in monitoring both the ionosphere and ground surface.
 The Super Dual Auroral Radar Network (SuperDARN) consists of pairs of HF radars with overlapping fields of view (FOV) whose primary aim is to map horizontal plasma convection at high latitudes [Greenwald et al., 1985; Chisham et al., 2007]. The plasma convection is reconstructed based on the Doppler shift of the ionospheric scatter (IS) from decameter irregularities that are strongly aligned with the geomagnetic field [Ruohoniemi and Baker, 1998]. The maximum backscattered power is produced when the radar wave propagates orthogonally to the geomagnetic field (aspect condition). The strong ionospheric refraction of HF signals (frequencies f0 = 10–20 MHz) allows to achieve this orthogonality over much larger areas compared with UHF-VHF waves propagating along straight line trajectories. In addition, the ionospheric refraction causes bending of some of the rays to the ground, which provides a basis for the over-the-horizon ground-to-ground HF communication. Upon reaching the ground, a part of the HF energy is scattered back to the radar by the suitably sized and oriented irregularities of the ground surface. These signals are regularly observed by HF radars and classified as ground scatter (GS) echoes based on their intrinsically low Doppler shift and spectral width values.
 While SuperDARN GS echoes, due to their insensitivity to horizontal drifts, cannot be incorporated in monitoring the ionospheric convection, their regular presence over well-defined range bands has proven to be useful in studying a number of other ionospheric phenomena. In accordance with the HF propagation theory, GS power is expected to maximize in the vicinity of the skip zone boundary, where “focusing” of the upper and lower propagation rays takes place [e.g., Hunsucker, 1991, pp. 98–103]. Characteristic variations in the skip zone location caused by atmospheric gravity waves are routinely used for their monitoring by SuperDARN radars (for an extended reference list see, e.g., He et al. ). Importantly, the GS Doppler shifts regularly show periodicities closely related to vertical plasma displacements generated by ULF waves allowing for GS-based ULF wave monitoring [e.g., Ponomarenko et al., 2003, 2005]. In separate developments, Hughes et al.  demonstrated SuperDARN's ability to map maximum usable frequencies, while André et al.  utilized GS elevation data from the Saskatoon radar to reconstruct critical frequencies for the E and F layers.
 In contrast to the ionospheric diagnostics, application of GS to monitoring the underlying surface that actually produces these echoes has been rarely discussed in the literature. We have identified only one such publication, which is concerned with studying the Greenland glacier tongue dynamics using GS from the Iceland West SuperDARN radar [Shand et al., 1998]. Furthermore, we believe that in the literature the role of surface conditions is generally underestimated while interpreting the experimentally observed GS characteristics.
 In the present work, we study the temporal and spatial variations of GS echoes over the polar regions in an attempt to establish both the ionospheric and ground conditions that control GS echoes. To achieve our goals, we use data from the Rankin Inlet SuperDARN radar, which conveniently covers a broad range of landscapes. The experimental technique and data processing algorithms are described in section 2. In section 3 we systematize the spatiotemporal patterns that we uncovered in echo occurrence. In section 4, based on ionosonde and satellite data, we show how changes in ionospheric and ground conditions with location, season and local time can explain the observations. Section 5 offers a synthesis of our results and discusses their implications for HF propagation at high latitudes.
2. Experimental Setup and Data Processing Details
 The Rankin Inlet radar (RI, 62°49′N, 92°05′W) was put in operation early in the second half of 2006. Its 16-beam FOV covers a large part of the Northern polar cap with beam 7 pointing toward the magnetic pole (Figure 1). In a normal sounding mode, the radar consecutively scans through all 16 beams with integration time ≃3 s/beam, i.e., the whole FOV scan takes ≃1 min. The azimuthal beam width is ≃3.5° at the half-power level. The line-of-sight spatial extent of the range gate is 45 km. The radar operating frequency, f0, depends on the propagation conditions and it is periodically switched between “permitted” frequency bands to maximize the amount of the ionospheric scatter echoes. In this study we used the first two full calendar years of Rankin Inlet records, 2007 and 2008. Statistical analysis revealed that during this time, frequencies of ≃12 MHz were used during local daytime, while f0 ≃9–10 MHz were usually utilized at night. We limited our study to echoes with signal-to-noise ratio (SNR) ≥ 6dB. Separation of GS and IS signals was performed using the standard SuperDARN algorithm based on narrow spectral widths and low Doppler velocities [Blanchard et al., 2009]. The echo occurrence, PGS,IS, was calculated as the ratio between the measured number of valid GS (IS) echoes observed at a given beam-range cell divided by the maximum possible echo number for the interval of interest (normally 60 measurements per hour). With respect to this, all necessary precautions were made to account for discretionary time modes, when the sampling rate varied from beam to beam, so that the equal time intervals contributed equally to the calculated occurrence rate.
3. Experimental Results
3.1. Seasonal Variations
 We started with an analysis of the average monthly echo occurrence during 2007–2008. Figure 2 shows the statistics for beams 0, 7 and 15 corresponding to the western, central and eastern parts of the RI FOV (Figure 1). PGS and PIS are shown in the top and bottom panels, respectively. Notice the difference in the upper limits of the color scales for GS (≤0.25) and IS (≤0.4). White filling depicts range-time intervals for which no valid GS echoes (SNR ≥ 6 dB) were recorded. Also note that in order to reflect proportionally the group range of the radar returns, as seen in Figure 2, the range scale is shifted back by four range gates, which corresponds to the group delay between the beginning of the radar emission and range gate 0 (180 km in group range).
 The seasonal statistics reveal several distinct features:
 1. IS occurs more frequently than GS.
 2. PGS and PIS are out of phase: there is a pronounced summer maximum for GS, while the ionospheric component is more frequently observed during the autumn-winter-spring period.
 3. GS is observed most frequently beyond range gates #25–30, while IS is generally confined to the closer ranges.
 4. While PIS variations are uniform from beam to beam, GS returns are mostly coming from the central part of FOV.
 5. Both GS and IS exhibit strips of enhanced occurrence at fixed ranges. These enhancements are uniform across the whole FOV for PIS, but for PGS they show a significant azimuthal variability.
 Ground and ionospheric scatter echoes can also overlap in group range, especially when several ionospheric layers coexist. In this case, strong IS returns can distort the actual GS occurrence pattern. To quantify this effect, we compared median SNR distributions for both components, RISmed and RGSmed. In Figure 2 (top) the white diagonal shading represents the areas where IS dominates GS, i.e., RISmed ≥ RGSmed. According to Figure 2, IS effectively “blankets” GS at gates ≤ 20–30). In contrast, at farther ranges GS becomes dominant, and the measured PGS values should accurately reflect the actual seasonal dynamics of the GS occurrence.
 Finally, there is a distinct population of GS-like echoes at very close ranges (gates 0–5), which exhibit a sharp occurrence maximum during the summer months. These returns cannot be actual GS because their group range is too small for them to be reflected from E region to the ground and back (more on this below).
3.2. Diurnal Variations
 To study diurnal dynamics we analyzed data from beam 7, which is approximately aligned with both magnetic and geographic meridians (Figure 1). Figure 3 shows the UT dynamics of GS echo occurrence during different seasons represented by March, June, September and December 2007. As with Figure 2, the white diagonal shading corresponds to the data dominated by IS. The vertical dashed line indicates an approximate time of LT-MLT noon at the radar site (≃18UT).
 At range gates ≥25–30, which are not affected by the overlap with IS, the winter, spring and autumn data show a dominance of daytime echoes. At first sight, this is not as obvious in the winter data (lower right). However, one should notice the detectable change from no data at night (white filling) to at least some data, in spite of the low occurrence (dark blue), during daytime. The range of GS daytime occurrence maximum increases from gate 25 in summer to 40–45 in spring and autumn. By contrast with other seasons, the summer echo occurrence remains relatively high for most of the day except for a 2–3 h gap centered on 02 UT. It shows a clear diurnal trend with the maximum occurrence observed at closer ranges during daytime and farther ranges at local night. The summer enhancement in PGS at the very close ranges mentioned in the previous section is mostly observed across the morning-daytime LT sector (≃10–20 UT). All these features are also present in the 2008 statistics (not shown).
3.3. Relation to Geographic Signatures
 As we mentioned in the Introduction, the spatial distribution of GS depends on the scattering and reflecting properties of the underlying surface. For reference, in Figure 4 we present a relief map of the Arctic Archipelago, where white dashed lines show approximate RI FOV boundaries. To study the surface effects, the monthly PGS values across the RI FOV were overplotted on the geographic map (fan plots). The four plots in Figure 5 represent GS statistics for March, June, September and December 2008 (the respective plots for 2007 are almost identical). To emphasize the spatial structure of PGS rather than its magnitude, in each frame the color scales are normalized to the maximum monthly occurrence.
 The spring, summer and autumn intervals all have similar spatial distributions characterized by an enhanced PGS in the central part of the FOV. The maximum in PGS is observed at closer ranges in June and moves to farther ranges in March and September. The GS almost disappears in December except for a relatively “bright” area in the northern part of Baffin Island.
 In general, the most frequent echoes are coming from the land, while the other surface types produce significantly lower numbers of returns (open sea) or even virtually no echoes at all (Greenland ice sheet). A more detailed analysis reveals that the GS enhancements are colocated with Ellesmere, Devon and Baffin Islands (Figure 4) and the coastal areas of Greenland, while the landmasses in the western part of the Arctic Archipelago produce a much lower number of returns. Another salient feature in Figure 5 is the increase in the sea scatter occurrence from Baffin Bay and the southern part of Arctic Archipelago in September.
4. Discussion and Interpretation
 In determining GS characteristics, we need to analyze two physical systems, (1) the ionosphere as a conduit for HF radiation and (2) the ground surface as a target producing radar returns.
4.1. Ionospheric Propagation Effects
 The ionospheric effects were analyzed using data provided by the Canadian Advanced Digital Ionosondes (CADI) [see e.g., Grant et al., 1995] from RI and Resolute Bay (RB) (geographic coordinates 74°43′N, 94°40′W) (Figure 1). The RB instrument is conveniently located near the center of the FOV at ≃1400 km from the radar so that its data are more relevant for analyzing the F modes. The E modes are affected by the ionosphere at the closer ranges ≤500 km and are better represented by the RI ionosonde. Unfortunately, neither instrument covered the whole 2007–2008 interval so that we had to resort to analyzes of the RI data for 2007 and the RB data for 2008.
 The ionosonde data were processed in the following way: for each month, all echoes with SNR ≥ 20 dB were binned by UT hour and frequency, so that median virtual heights, hv, and occurrence rates were calculated for each bin. Figures 6 and 7 show results for the RB (2008) and RI (2007) ionosondes, respectively. The four plots in Figures 6 and 7 correspond to the same months as those in Figure 3 except the winter frame for RI (Figure 7 (bottom right)), which is represented by January because there were too few days of the ionosonde data available for December 2007. We analyzed separately the data from below and above the 150 km virtual height, which we attributed to E and F region reflections, respectively.
 In Figures 6 and 7, a monthly average ionogram for each UT hour is represented by a vertical set of diamonds, whose colors correspond to the ionosonde frequencies as specified by the color scale at the bottom. The symbol sizes are proportional to the echo occurrence rates with gray diamonds at the bottom left corners corresponding to a 100% occurrence rate (unit probability, p = 1.0). Finally, the vertical coordinate describes the median virtual height. Note that plotting results for all frequencies on the same graph proved to be ineffective because it made the plots very “busy” and incomprehensible. To avoid this, we show only a representative set of evenly spaced frequencies stepping by 0.5 MHz across a 2–8 MHz range.
 First, we analyze F layer variations using the RB ionosonde data. During summer (Figure 6 (top left)), the F region data are observed for the whole day and they show a pronounced diurnal variation in the reflection height characterized by the minimum at ≃12 LT (hv ≃200–250 km) and maximum near the LT midnight (hv ≃300–350 km), which is consistent with the variation in the solar zenith angle, χ. In December (Figure 6 (bottom right)) a sufficient number of F region echoes is detected during local daytime-evening (∼08–20 LT, 14–04 UT) and from higher altitudes (hv ≃300–350 km). By contrast to the regular behavior of the summer F layer, the winter heights and critical frequencies vary considerably. This is consistent with the fact that winter ionization is mainly produced by a combination of the soft particle precipitations and/or patches of enhanced ionization generated at the lower latitudes on the dayside and propagating antisunward across the polarcap with photoionization playing a secondary role [Buchau et al., 1983; McEwen et al., 1994]. The spring and autumn data (Figure 6 (left)) reflect the transition between winter and summer conditions. While the F layer maximum height varies significantly with the season, the critical frequencies stay around fcF ≃4.5–5 MHz through the whole year. Quite unexpectedly, the daytime occurrence in September is significantly higher than in March. This spring-autumn asymmetry appears to be the result of relatively higher magnetic activity in March (≃28% of time with kp ≥ 3), which led to enhanced ionospheric absorption as compared to September (≃9% of time with kp ≥ 3). The above asymmetry was duly accompanied by the noticeably higher GS occurrence in September (Figure 5).
 The E region echoes are much more frequent at RI than at RB. Here, the returns from the high-density (fcE ≥ 8 MHz) E region often dominate at night (00–10 UT) as a result of the intense particle precipitation in the auroral zone. The simultaneous presence of both auroral E and F region returns is regularly observed at night which argues for a patchy (semi-transparent) nighttime E layer. During daytime, the E layer changes to the conventional Chapman type, which is governed by χ and characterized by much lower critical frequencies fcE ≤ 4–4.5 MHz. This layer is easily observed during summer but it disappears in winter as expected from the variations in the solar zenith angle.
 The general tendency for F mode GS echoes to shift to closer ranges in summer and to move farther away in spring and autumn agrees with the seasonal variations in the ionospheric propagation conditions. The lower-altitude summer F layer refracts radar signals back to the Earth surface at closer ranges. The low occurrence rate of GS beyond range gate 25 in winter is conversely due to the higher reflection heights with a skip zone boundary for the F region shifted to the far ranges that cover the Northern polar ice cap with sea ice representing a relatively weak target [e.g., Shand et al., 1998]. This effect culminates in the nighttime winter conditions, when the F region densities might be too low even to bend the radar ray path to the ground anywhere within the FOV.
 To check if the above interpretation agrees quantitatively with the observation results, we modeled HF propagation at 12 LT for summer and winter conditions using the numerical ray-tracing routine described by Ponomarenko et al. . To estimate the input ionospheric parameters, we adjusted a synthetically generated ionogram, based on a three-layer Chapman model, to the average Rankin Inlet ionograms from 18 to 19 UT (≃12–13 LT) for June and December 2007. The resulting midday electron density profiles are shown in Figure 8 by black and red curves, respectively. The summer profile is characterized by three distinct layers, E, F1 and F2, while the winter ionosphere is only represented by the F2 layer. We ran the ray-tracing routine for both profiles using a typical daytime RI SuperDARN frequency f0 = 12 MHz and obtained the number of rays reaching the ground as a function of range gate number (Figure 9). In agreement with the experimental data, the ray tracing confirmed that the winter ionosphere illuminates the ground surface at relatively large distances beyond the range gate 50 (i.e., F layer skip zone boundary), while in June the radar signals can reach the ground as close as range gates 16–17 (E layer skip zone boundary). By contrast to December, the June dependence exhibits several ray density enhancements reflecting the multilayer nature of the daytime summer ionosphere.
 Finally, in the above context, the winter occurrence maximum coming from the north of Baffin Island (Figure 5 (bottom right)) is located much closer than the predicted value for the daytime F layer skip zone boundary. However, a comparison of Figure 3 (right bottom) and Figure 7 (right bottom) suggests that these echoes are provided by the reflection from the high-density nighttime E layer generated by the intensive particle precipitations in the auroral zone near the radar location.
 While the experimental data are in a general agreement with the seasonal-diurnal variations in the propagation conditions, there is an apparent problem related to the existence of the occurrence enhancements/depletions fixed in range (“preferred” ranges). These observations do not fit the ionospheric variability paradigm and require further analysis. One of the factors that affects the echo occurrence at fixed ranges is the transmitter pulse-overlap interference (POI) caused by the receiver being blocked during the emission of the transmitted pulse. This leads to a decrease in the number of valid echoes from the affected ranges. A detailed analysis (see Appendix A) reveals that for the pulse sequence used in our measurements, POI should cause periodic depletions in the number of valid lags at range gates 5i-4 (i = 1, 2, 3 …), i.e., gates 1, 6, 11, 16, etc., are expected to produce a smaller number of valid ACFs. In Figures 2 and 3 these gates are indicated by the dashed horizontal lines. In Figure 2, while the IS echo occurrence (Figure 2 (bottom)) is clearly affected by POI, the GS echoes (Figure 2 (top)) show no such effect. Instead, PGS exhibits sharp peaks at range gates 26 and 31, where POI-related depletions would be expected. To illustrate this effect more clearly, in Figure 10 we plotted range variations of IS and GS occurrence and SNR in beam 7 for June 2008. The apparent insensitivity of GS to POI might be attributed to its much larger decorrelation time constant, which exceeds the maximum ACF lag ≃40 ms and thus provides a larger quantity of the valid ACF lags for further processing.
4.2. Ground Surface Effects
 The observed lack of correspondence between PGS range variations and POI implies that there should be another factor modulating GS occurrence at fixed ranges. The obvious candidate is the spatial variability in the ground scattering/reflection properties. While the HF radiation from a ground-based transmitter illuminates the surface outside the skip zone, the maximum GS occurrence/power is conventionally expected to come from the skip zone boundary, where the focusing of the upper and lower rays takes place. However, this assumption is accurate only if the backscattering properties of the underlying surface are uniform along the beam direction. This is definitely not the case for the Rankin Inlet FOV. A detailed matching of the occurrence pattern with the geographic map of the Arctic Archipelago (Figure 4) reveals that the most frequent echoes come from the mountainous areas of Baffin, Devon and Ellesmere Islands. This implies that the echoes result from a direct reflection by suitably oriented mountain slopes, thereby providing higher SNR values (notice the coinciding peaks in GS occurrence and SNR at gates 26 and 31 in Figure 10), in contrast to the relatively “flat” landmasses in the western part of the FOV. The nature of the enhancements in PGS observed along the Greenland coast has to be similar due to the presence of the substantial mountains in these regions.
 Therefore, we conclude that the observed range distribution in PGS from Rankin Inlet radar represents a superposition of (1) ionospheric propagation effects causing diurnal and seasonal range variations in accordance with solar zenith angle and (2) ability of the Earth surface to scatter/reflect the HF signals back to the radar, which produce occurrence features at fixed (“preferred”) ranges colocated with mountainous areas. By contrast to the pulse overlap effect always depleting the occurrence rate at certain ranges, the surface variability can lead to both increase and decrease in PGS with respect to the “background” values.
 With respect to the ground surface effects it is also important to analyze the influence of the sea ice cover, which can effectively suppress generation of waves providing conditions for Bragg scatter of HF radiation from the “rough” sea surface. For this purpose, in Figure 5 the gray dotted pattern shows sea regions covered by sea ice for 100% of the time during the given month (the sea ice data used in this study are a 24 km resolution records from “IMS daily Northern Hemisphere snow and ice analysis at 4 km and 24 km resolution. Boulder, CO: National Snow and Ice Data Center. Digital media.” provided by NOAA/NESDIS/OSDPD/SSD at http://nsidc.org/data/g02156.html).
Figure 5 (bottom left) representing September 2008 shows increased occurrence of echoes from Baffin Bay. Figure 11 compares annual variations in average occurrence for range gates 30–35 in beams 7 and 15, which correspond to scatter from land (Ellesmere Island) and sea (Baffin Bay), respectively. By choosing the same range gates for both beams we made sure that the ionospheric propagation effects are essentially the same for both locations. The number of echoes from the landmass (black) maximizes during summer, while the sea scatter occurrence (red) exhibits an additional enhancement in August–October. This would be expected due to the fact that by this time Baffin Bay becomes clear from the sea ice and starts to produce Bragg scatter from the waved sea surface, i.e., indicating open water. Another interesting feature of the September 2008 map is the patchy character of PGS at close ranges (gates 10–25), which apparently resembles the system of straits in the southern part of the Arctic Archipelago. However, to validate this hypothesis it would be necessary to carry out a very careful analysis of the echoes bearing in mind that they are strongly affected by the overlap with the IS returns. Finally, the virtual absence of GS from inland Greenland is in agreement with previous studies [Leonard, 1991] and results from the strong absorption of HF radio waves by the ice sheet.
4.3. Very Close Range Summer Echoes
 The sharp summer enhancement in occurrence of the very close range GS-like echoes is rather enigmatic and cannot be readily explained within the conventional HF propagation paradigm. These echoes were marked as GS due to their relatively low Doppler velocities and spectral width values, but the respective group ranges are too short to accommodate any realistic propagation scenario involving scattering from the ground surface. Also, these echoes cannot result from the conventional Bragg scatter by the field-aligned E region irregularities because at the close ranges the aspect conditions are far from being satisfied under realistic E region densities. There are three previously reported types of the nonaspect radar returns: (1) backscatter from meteor trails [e.g., Hall et al., 1997; André et al., 1998; Hussey et al., 2000], (2) so-called HAIR (high-aspect angle irregularity region) echoes [Milan et al., 2004], and (3) polar mesosphere summer echoes (PMSE) [Ogawa et al., 2004]. While the meteor echoes are most frequently observed between 00 and 06 LT [Hall et al., 1997], the very close range echoes from RI show the summer occurrence maximum close to 12 LT (Figure 3 (top right)) therefore arguing against their meteor origin. A further clarification of the nature of these echoes lies outside the scope of the present paper and will be addressed in a future publication.
5. Summary and Conclusions
 The statistical studies of the ground (surface) scatter occurrence observed by the Rankin Inlet SuperDARN radar over 2007–2008 allowed us to establish the following:
 1. The diurnal-annual variations in PGS range are in general agreement with regular variations in the ionospheric propagation conditions. General occurrence maximizes in summer owing to the increase in illumination of the ground by the HF radiation within the FOV due to the higher ionospheric densities and lower maximum heights. It explains the observed shift in the maximum occurrence to farther ranges in spring and autumn. This effect culminates under the nighttime winter conditions, when the virtual disappearance of the F mode echoes is caused by the inability of the low-density ionosphere to reflect the HF radiation back to the ground within the FOV boundaries. E layer propagation modes are responsible for some of the observed GS features, e.g., nighttime occurrence enhancement in winter.
 2. The spatial pattern in GS occurrence is strongly affected by the type of the underlying surface. The nonuniformity of the scattering/reflecting properties is responsible for the persistent enhancements in PGS at fixed locations, coinciding with mountains and coastal areas. At the same time there are no returns from the Greenland ice sheet. There are only minor contributions to the overall GS statistics from the “flat” land and permanent sea ice. The sea scatter occurrence from Baffin Bay is enhanced in autumn, when there is a minimum sea ice cover.
 3. We observed a distinct population of very close range GS-like returns, which show a sharp daytime occurrence maximum in summer months. These echoes seem to result from a high-aspect angle E region backscatter or mesospheric echoes (PMSE's), but their exact nature requires a separate study.
 4. Finally, we identified two major artifacts affecting the observed occurrence of SuperDARN GS echoes. These are (1) the masking of the ground scatter returns at close ranges by the more powerful ionospheric scatter component and (2) the effect of transmitter pulse overlap on the quality of the data at fixed ranges.
 In conclusion, we have provided a detailed insight into the nature of the HF ground scatter returns from the polar cap areas. This study can have important application for both theory and practice of HF propagation at high latitudes. SuperDARN data could be used for near-real-time forecasting of HF propagation conditions over the polar cap. These forecasts become particularly important as the number of cross-polar passenger flights increases, and HF is frequently the only available communication option to air traffic control operators in this region. The well-defined correspondence between GS occurrence and geographical features (mountains, coastal lines) provides an independent way to calibrate the ground range of the radar echoes. This information is also very important for a correct interpretation of the propagation conditions, especially the location of the skip zone boundary. Finally, sea scatter occurrence exhibits high sensitivity to the presence of sea ice so that it could be utilized for monitoring the ice cover extent under the SuperDARN FOV.
Appendix A:: Pulse Overlap Interference
 SuperDARN operation is based on the use of a sequence of 7 or 8 nonevenly separated pulses of the same duration, Δt = 300 μs (spatial resolution 45 km) [e.g., Ponomarenko and Waters, 2006]. A complex autocorrelation function (ACF) for each range gate is calculated using different pairs of these pulses to produce different ACF lags (Figure A1). The duration of the shortest pulse separation interval (basic ACF lag), τ, is an integer multiplier of Δt, and all the other lags are integer multipliers of τ. The pulse sequence is emitted every ≃100 ms, while the sampling of the echoes continues with the rate equal to Δt. During the pulse emission the receiver is blocked so that the respective receiver samples are invalid and removed from the analysis. As a result, the ACF lags utilizing the “blanketed” pulses are discarded from the analysis, thereby decreasing the quality of the respective ACFs (pulse overlap interference, POI). To determine the receiver samples for each range gate, the pulse sequence “mask” should be shifted along the receiver time series stepping one range gate at a time. After spanning the first Δn = Δt/τ gates, one of the required samples coincides with the transmitter pulse emission time and has to be discarded, eliminating the respective ACF lags at this particular range. After shifting by further Δn range gates this effect is repeated for a different pulse in the sequence, etc. As a result, the periodic depletion in the quality of the ACF occurs at range gates separated by Δn. In our measurements Δt = 300 μs and τ = 1500 μs, so that POI happens every 5 range gates. In addition, SuperDARN data sampling normally starts at 180 km, which is equivalent to four range gates. Therefore, the POI effect should occur in range gates iΔ n − 4 (i = 1, 2, 3…), so that the first occurrence of POI should be observed in range gate 1 and then repeated with a five-gate step (gates 1, 6, 11, 16, etc.).
 This work was supported in part by the funding from the province of Saskatchewan and Government of Canada for a Canada Research Chair (J.P.S.M.) and discovery grants to J.P.S.M. and A.V.K. The Saskatoon SuperDARN radar operations are funded by a MRS grant from Natural Sciences and Engineering Research Council of Canada (NSERC) and a Canadian Space Agency (CSA) contract. We also acknowledge CSA and NSERC for providing CADI ionosonde data as well as National Snow and Ice Data Center, NOAA, for providing the on-line access to sea ice data. Finally, we thank Chris Meek for the fruitful discussions on close-range echoes and Raj Choudhary for early discussions on ground scatter statistics. Resolute Bay CADI ionosonde as part of the Canadian High Arctic Ionospheric Network (CHAIN) infrastructure is funded by the Canada Foundation for Innovation and the New Brunswick Innovation Foundation. CHAIN operation is conducted in collaboration with the Canadian Space Agency. RB ionosonde data were obtained through the Canadian Space Science Data Portal (CSSDP) at http://18.104.22.168:8080/ssdp/app/home.