We revealed the spatial distribution of HF radar echoes in the subauroral region by conducting a statistical analysis of scattering occurrence from the midlatitude Super Dual Auroral Radar Network (SuperDARN) radar in Hokkaido, Japan. Consequently, Dusk Scatter Event (DUSE) was identified as a most prominent backscatter target in these latitudes. Past studies have intended to associate an appearance of DUSE with the density gradient at the poleward or sunward edges of the midlatitude trough. However, exact spatial collocation between the source region of DUSE and the midlatitude trough has not been revealed because the SuperDARN radars in the auroral region could not observe the whole part of DUSE due to a limitation of the field-of-view coverage. Thus, it has been unclear which of the density gradients associated with the midlatitude trough is responsible for generating DUSE. The data from the Hokkaido radar enabled us to estimate the lower-latitude boundary of DUSE as around 59° in AACGM magnetic latitude. In addition, by adding the data from the King Salmon radar in Alaska we derived a complete statistical distribution of DUSE. The latitudinal extent of DUSE is about 9° from 59° to 68°, which is approximately 1000 km. The statistical distribution of DUSE was compared with the model of the midlatitude trough. As a result, the source region of DUSE is closely colocated with the minimum of the trough, which suggests that the electron density gradient at the sunward edge of the trough is responsible for DUSE. This means that the location of the duskside sunward edge (i.e., local time extent) of the midlatitude trough can be monitored by using an appearance of DUSE as a proxy. The current statistical analysis also suggests that we can derive a spatial distribution of HF radar echoes from the midlatitudes to high latitudes by combining the data from the Hokkaido and King Salmon radars located in the Far Eastern area, which would be a very powerful diagnostic tool for investigating the global distribution of plasma irregularities.
If you can't find a tool you're looking for, please click the link at the top of the page to "Go to old article view". Alternatively, view our Knowledge Base articles for additional help. Your feedback is important to us, so please let us know if you have comments or ideas for improvement.
 Ionospheric field-aligned plasma irregularities are plasma density fluctuations which have been amplified through plasma instability processes such as gradient-drift instability or two-stream instability [Fejer and Kelley, 1980; Keskinen and Ossakow, 1983]. Such density irregularities at the F region altitudes, whose scale size is about a few hundred meters, often cause signal power fluctuations in transionospheric satellite transmissions known as ionospheric scintillation. Hence, knowledge on the global distribution of ionospheric plasma irregularities is indispensable for maintaining satellite-ground data transmission.
 Several statistical studies of decameter-scale irregularities in the high-latitude F region ionosphere have been carried out using the data from Super Dual Auroral Radar Network (SuperDARN) [e.g., Ruohoniemi and Greenwald, 1997; Milan et al., 1997]. SuperDARN is an international collaborative project based on the network of coherent HF radars located in the high-latitude zones of the northern and southern hemispheres [Greenwald et al., 1995; Chisham et al., 2007]. The radars of SuperDARN cover a vast area in the high-latitude ionosphere and routinely observe backscatter echoes from ionospheric irregularities; thus it is a very useful diagnostic tool for estimating the global distribution of plasma irregularities in the high-latitude part of the ionosphere. The previous statistics have demonstrated that there are two outstanding sources of decameter-scale irregularities in the high latitudes, which are the dayside cusp and nightside auroral zone. Irregularities in these regions are considered to occur in close association with precipitations of ions and electrons of magnetospheric origin. In the past studies, however, the occurrence distribution of irregularities in the region equatorward of the auroral latitudes has not been examined in detail.
Hosokawa et al.  first presented statistical distribution of HF echoes at the subauroral latitudes. They demonstrated that Dusk Scatter Event (DUSE), which was first identified by Ruohoniemi et al. , is the most prominent backscatter target for the SuperDARN at these latitudes. During geomagnetically quiet conditions, they appear immediately after the local sunset and last for 2–3 h. The source region of DUSE lies in the magnetic latitudes equatorward of the auroral oval, which probably corresponds to the plasma density depleted region known as the midlatitude trough [Moffett and Quegan, 1983; Rodger et al., 1992]. Subsequent study by Hosokawa et al.  proposed a model explaining the generation of DUSE which employs a sunward plasma density gradient at the sunward edge of the midlatitude trough and an ambient poleward electric field as key drivers for amplifying density fluctuations through the gradient-drift instability. At the time, however, the latitudinal profile of DUSE has not been revealed because the magnetic latitude of the SuperDARN radars at the auroral latitudes is too high to observe the subauroral ionosphere completely. Thus, the spatial relationship between DUSE and the midlatitude trough is still unknown for which the latitudinal extent of DUSE could not estimated accurately.
 The second midlatitude SuperDARN radar has been operative in Hokkaido, Japan since December 2006. The radar was found to be a very powerful tool for monitoring storm time high-speed subauroral flow [Ebihara et al., 2008; Kataoka et al., 2007, 2009; Koustov et al., 2008] and various thermospheric waves [Ishida et al., 2008; Ogawa et al., 2009; Suzuki et al., 2009; Koustov et al., 2009]. Since the radar field of view (FOV) well covers the possible source region of DUSE at the subauroral latitudes, the radar is suitable for estimating the latitudinal extent of DUSE. In this paper, we statistically analyze the occurrence distribution of DUSE detected by the Hokkaido radar in April 2007 and estimate the latitudinal distribution of DUSE. In particular, the lower-latitude boundary of DUSE is identified clearly. Based on the statistical results, the spatial collocation between DUSE and the midlatitude trough is investigated. In addition, we discuss possible generation mechanisms of DUSE in terms of ionospheric instability process operating at the edge of the midlatitude trough.
 The second midlatitude SuperDARN radar has been operational in Rikubetsu (43.53°N, 143.61°E), Hokkaido, Japan since December 2006. Operation of the radar is on a 24 h, 365 days a year basis. However, the operation program is separated into three categories, common time, special time and discretionary time. The data used in the current statistical analysis were taken from periods when the radar was running in the common time normal scan mode (see Greenwald et al.  for details). In the current version of this mode, the radar sounds along 16 discrete pointing directions separated by 3.24° in azimuth, with the radar boresite pointing at an azimuth of 30° clockwise from geographic north. It takes ≈3 or 7 s to integrate backscatter returns in one beam direction and about 1 or 2 min is needed to complete a scan of all directions. A pulse length is set to 300 μs, which is equivalent to a gate length of 45 km, and a lag to the first gate is 1200 μs (180 km). Operating frequency was around 10.80 MHz in the nighttime (2100–0900 UT) and around 11.07 MHz in the daytime (0900–2100 UT) during the period of statistics, corresponding to a wavelength of ≈30 m. Thus, the radar observes Bragg scatters from irregularities with a wavelength of ≈15 m. In each radar cell, routine analysis of the obtained autocorrelation functions provides backscatter power, spectral width, and Doppler shift of the echo spectra [Baker et al., 1988].
 For most of the SuperDARN radars at the auroral latitudes, radar backscatter echoes have been gated into 75 range bins. According to this tradition, the number of range gates was set to 75 for the Hokkaido radar at the time of deployment. Subsequently, it has been changed to 110 since June 2007 to get backscatter echoes from the auroral latitudes at far ranges. Figure 1 shows the FOV of the Hokkaido radar mapped in the geographic coordinate system, in which contours of constant AACGM magnetic latitude [Baker and Wing, 1989] are superimposed. Here, the FOV of the radar is plotted with the original 75 range gates mode because we use the data in April 2007 in the current statistical analysis. The FOV covers the geographic latitudes from 45° to 75° and the geomagnetic latitudes from 40° to 65°. In Figure 1, the FOV of the SuperDARN King Salmon radar in Alaska (58.68°N, 156.65°W) is also shown. The FOV of the King Salmon radar covers the highest-latitude part of the Hokkaido FOV, which enables us to estimate the latitudinal distribution of HF radar echoes in the Far Eastern sector. In the current statistical analysis, we also employed the data from King Salmon to supplement the Hokkaido radar observations.
Figure 2 shows an example of DUSE observed by the Hokkaido radar on 30 April 2007. The backscatter power, Doppler velocity and spectral width along beam 5 are plotted in the format of a Range-Time-Intensity (RTI) plot. Dashed line gives the latitude of the equatorward boundary of the auroral oval [Feldstein and Starkov, 1967] as modeled by Holzworth and Meng  for the prevailing geomagnetic conditions (Kp = 2). Backscatter echoes with small Doppler shift (generally below 50 m s−1) and narrow spectral width (below ≈30 m s−1) are automatically identified as backscatter from the sea (sea scatter), and are indicated in gray in Figure 2b. Throughout this interval, the radar observed lots of sea scatters in the lower-latitude part of the FOV from 45° to 60°. Just above the band of sea scatter, DUSE were observed from 0830 to 1200 UT. The echoes had a Doppler velocity toward the radar, whose magnitude was as large as 500 m s−1 at the higher-latitude part, but was mostly less than 200 m s−1. As shown in Figure 1, the radar boresite is pointing at an azimuth of 30° clockwise from north. The observed Doppler velocity toward the radar corresponds to the sunward return flow in this local time sector (around 1900–2100 MLT). The spectral width was relatively broader than that of sea scatter echoes, but was mostly less than 200 m s−1. Ruohoniemi et al.  indicated that DUSE can be readily distinguished from other types of late afternoon/early evening scatter by the low values of associated Doppler velocities (<200 m s−1) and spectral widths (<200 m s−1). The Doppler characteristics of the echoes shown in Figure 2 are quite consistent with those of DUSE proposed by Ruohoniemi et al. , which suggests that the echoes in Figure 2 correspond to DUSE in the midlatitude SuperDARN observations. Another important feature to note is that DUSE appeared in the latitudes well equatorward of the model auroral oval, which implies that the echoes were located in the subauroral region. In this study, we do not show any simultaneous data to identify the location of the main aurora oval (e.g., DMSP particle data). We only compared the source distribution of DUSE with the empirical auroral oval model. Thus, we cannot definitely determine whether the source of DUSE is actually located in the subauroral region. Hence, it may be more appropriate to use “midlatitude” rather than “subauroral” to describe the origin of DUSE echoes because midlatitude is a more generic term that makes no specific reference to the auroral oval. However, there are significant differences in the plasma properties between the subauroral and proper midlatitude region. For example, the plasma convection is known to be very low in the midlatitude, while that in the subauroral region sometimes enhances up to 1 km s−1. Thus, we continue to use the term “subauroral” in the text. The latitudinal extent of the echoes was approximately 7° from 58° to 65°, corresponding to the scale of ≈800 km. The echoes continued for almost 3.5 h. This corresponds to the longitudinal extent of 2700 km at 60°. In the following, we will investigate the spatial distribution of the source region of DUSE in a statistical fashion.
 In the current statistical analysis, we employed the Hokkaido radar data in April 2007 because the signature of DUSE was most prominent in spring. Figure 3 shows a variation of Kp index in April 2007. In the beginning and end of the month, Kp index was increased up to around 3–4. However, the Kp index was mostly less than 2 in the rest of the month. In this paper, we concentrate on the spatial distribution of DUSE during quiet conditions. Thus, the data obtained during quiet geomagnetic conditions (0 ≤ Kp ≤ 2) are employed for the statistics.
 The purpose of this study is to reveal the spatial distribution of decameter-scale irregularities present in the subauroral F region ionosphere by examining the occurrence of HF radar echoes. Hence, we need to deal with echoes surely scattered from the irregularities at the F region altitudes. The SuperDARN radars, however, receive considerable amount of backscatters from E region irregularities as well as those from the ground/sea (see Milan et al.  for details). Thus, as a first step, we must exclude these unwanted echoes from the statistics. In the SuperDARN observations, the radar echoes whose range is less than 600 km are regarded as due to E region scatter, while echoes returned from ranges larger than 900 km are considered to be due to F region scatter [e.g., Ruohoniemi et al., 1988]. In the current study, we follow this definition and data obtained from ranges greater than 900 km are employed for the statistics. This assumption is known to be valid in most cases. However, we must bear in mind that backscatter echoes from the E region irregularities often appear in the ranges further than 900 km, which was first suggested by Milan et al. . Echoes that are considered to be the ground/sea scatter are excluded automatically on the basis of their line-of-sight Doppler velocity magnitude and spectral width (criteria is Vlos < 50 m s−1 and spectral width <30 m s−1). However, there is considerable ambiguity in this identification as these characteristics are also consistent with backscatter from slow-moving ionospheric irregularities. This could often be the case in the subauroral region as the convection electric field is much weaker in these latitudes than in the auroral region. Before performing the statistical analysis, we checked all of the Hokkaido radar data in April 2007 and confirmed that the echoes appearing immediately after sunset in the subauroral latitude had Doppler velocities well larger than the criterion introduced above (50 m s−1).
Figure 4a shows the scattering occurrence rate obtained from the Hokkaido radar observations in April 2007, in which the data are mapped onto the geographic latitude and local time (LT) coordinate system. The dashed lines give a line of solar zenith angle (SZA) of 90°, 95°, 100°, 105° on 15 April, respectively. There exist several regions of enhanced scattering occurrence rate. However, most outstanding peak is found on the eastern side of the SZA 90° line in the dusk to premidnight sector. Location of this primary peak is roughly consistent with Ruohoniemi et al.  who reported that DUSE typically appears when the SZA ranges between 95° and 105°. If we take a closer look at the distribution, however, the primary peak is distributed even on the western side of the SZA 90° line. Namely, there seems to exist a small disagreement between the current statistical result and Ruohoniemi et al. . However, Hosokawa et al.  later pointed out that DUSE are observed in earlier local time during disturbed conditions, which allows the statistical distribution of DUSE to extend into the sunlit area across the SZA 90° line. The scattering occurrence rate of this peak is approximately 10–15%, which is smaller than that derived by Hosokawa et al.  by a factor of 2. This difference is probably due to the fact that the source region of DUSE is far from Hokkaido (≈3000 km away from the radar site). Thus, here, we define the statistical location (i.e., source region) of DUSE as a region of scattering occurrence enhancement greater than 10%. The latitudinal extent of the source region of DUSE is about 9° from 64° to 73° in the geographic coordinate system. The local time extent of the peak is approximately 3–4 h, which is fairly consistent with the duration of DUSE shown in Figure 2, which confirms that the primary peak in Figure 4a is mainly due to an appearance of DUSE in the subauroral region.
Figure 4b shows the same data as shown in Figure 4a, but the data are mapped onto the AACGM magnetic latitude and MLT coordinate system. The two dashed circles give the equatorward and poleward edges of the auroral oval [Feldstein and Starkov, 1967] as modeled by Holzworth and Meng  for Q = 1 (quiet conditions), respectively. The value of Q used here (Q = 1) roughly corresponds to Kp values around 0–2. A region of enhanced scattering occurrence rate is again seen in the dusk sector between 1800 MLT and 2200 MLT, which is identical to the one seen in the geographic coordinate system (Figure 4a). The magnetic latitude of this primary peak is slightly lower than the equatorward edges of the auroral oval model; thus is probably associated with the density structure within the midlatitude trough. If we take a closer look at the peak of scattering occurrence rate, however, enhanced scattering occurrence rate can be seen even in the auroral oval. This means that some of the echoes in the vicinity of the peak are not directly associated with the midlatitude trough. Recent study by Koustov et al.  presented a case in which the Hokkaido radar observed some echoes in the postsunset period within the auroral oval. Hence, we must bear in mind that all of the Hokkaido radar echoes in the dusk sector do not always associated with the midlatitude trough. The latitudinal extent of the primary peak is about 9° from 59° to 68° AACGM magnetic latitude, which is almost consistent with the case shown in Figure 2. The local time extent of the peak is somewhat broader (1800–2200 MLT) than that in the geographic coordinate system (Figure 4a). However, it is still consistent with the example shown in Figure 2. The important thing to note is that the higher-latitude boundary of DUSE cannot be identified correctly only from the Hokkaido radar data because the radar did not observe the latitudes higher than 68° during the interval of this statistical analysis.
 Recently, Greenwald et al.  reported another class of HF radar echo appearing equatorward of the main aurora oval, which is possibly generated through temperature-gradient instability (TGI). However, they occur in the nighttime subauroral ionosphere well after the sunset. In addition, the Doppler velocity of TGI echoes is very low (less than 100 m s−1). Thus, the echoes investigated in the current statistics are not associated with TGI echoes.
 In order to locate the higher-latitude boundary of DUSE, we combine the data from King Salmon radar in Alaska with those from Hokkaido. Figure 5a again shows the scattering occurrence distribution obtained from the Hokkaido radar observations in April 2007. Figure 5b gives scattering occurrence rate obtained from the SuperDARN King Salmon radar in the same period. In Figure 5b, the color scale for the scattering occurrence rate is different from that for the Hokkaido radar to determine spatial distribution of occurrence peaks more clearly. Since the source region of DUSE is far from Hokkaido (≈3000 km away from the radar site) the scattering occurrence rate is smaller in the Hokkaido data (≈15%) than in the King Salmon (≈25%). Thus, we define the statistical location (i.e., source region) of DUSE in the King Salmon data as a region of scattering occurrence enhancement greater than 15%. In the King Salmon data, the statistical location of DUSE echoes is identified equatorward of the statistical auroral oval near the dusk terminator. This peak is clearly associated with an occurrence of DUSE in the subauroral region. Note that, however, low-latitude boundary of the echo distribution cannot be identified only from the King Salmon data. Previously, Koustov et al.  studied echo occurrence distribution of the King Salmon radar. However, they did not identify an enhancement associated with DUSE. This is probably due to the fact that only the data from westward looking beams (beams 0 to 5) were employed in the statistics of Koustov et al. , while data from all beams are used in the current statistics. Hosokawa et al.  pointed out that signature of DUSE is less prominent in beams looking westward. This could be the main cause of the difference between our statistical results and those of Koustov et al. .
 In order to compare the echo distributions obtained from the Hokkaido and King Salmon radars, the data shown in Figures 5a and 5b are merged and plotted together in Figure 5c. Here, we just cut the data from below 65° of Figure 5a (Hokkaido) and above 65° of Figure 5b (King Salmon) and pasted them together. Hence, it should be bear in mind that the color coding of the scattering occurrence percentage is different between below and above 65°. Interestingly, the major peaks in the two data sets (Figures 5a and 5b) are well overlapping each other, which confirms that the dusk echoes observed with the Hokkaido radar are surely corresponding to DUSE that have been observed by the SuperDARN radars in the auroral region. Not only that, by combining the data from the two radars we can obtain the latitudinal distribution of DUSE. The latitudinal extent of DUSE is estimated to be 9° from 59° to 68° in the AACGM magnetic latitude, which corresponds to about 1000 km.
4. Discussion and Summary
 It is well established that if irregularities are generated through gradient-drift instability (GDI), the fastest growth rate at the linear stage is obtained where the background plasma convection and the electron density gradient are exactly in the same direction [Keskinen and Ossakow, 1982]. At the subauroral latitudes in the dusk sector, the background plasma convection is mostly directed westward, which corresponds to the most equatorward part of the duskside convection cell [e.g., Baker et al., 2007]. In contrast, the configuration of electron density gradient is somewhat complicated due to an existence of the midlatitude trough. Namely, direction of the electron density gradient in this region depends highly on the spatial distribution of the midlatitude trough. The midlatitude trough is a density depleted region having a horseshoe-shaped structure extending from dusk to dawn through midnight. There are two possible density gradients contributing to the generation of DUSE in the vicinity of the midlatitude trough, i.e., sunward wall and poleward wall. If we consider the sunward wall (i.e., westward gradient in the dusk sector), configuration of the density gradient and background convection is favorable for generating irregularities through GDI. In contrast, the poleward density gradient at the poleward edge is almost perpendicular to the background plasma convection; thus fast growth of irregularities cannot be expected. Keskinen and Ossakow , however, demonstrated that irregularities can be generated through GDI in anyplace except where the background plasma convection and the electron density gradient are exactly antiparallel although faster growth rate is obtained where the convection and density gradient are more parallel. In addition, ionization associated with auroral particle precipitations just poleward of the midlatitude trough could steepen the density gradient at the poleward edge. Thus, we cannot reject the possibility that the density gradient at the poleward edge contributes to forming irregularities within DUSE. That is, we cannot determine which of these gradients is responsible for generating irregularities within DUSE only from the theoretical consideration as described above.
 Past statistical studies have intended to associate an appearance of DUSE with the density gradient at the poleward [Ruohoniemi et al., 1988] and sunward [Hosokawa et al., 2001] edges of the midlatitude trough. However, exact spatial collocation between the source region of DUSE and the midlatitude trough has not been revealed because the SuperDARN radars in the auroral region could not observe the whole part of DUSE due to a limitation of the FOV coverage. Later, Hosokawa et al.  employed simultaneous observations of DUSE and background electron density with the SuperDARN Finland radar and EISCAT UHF system in Tromsoe and demonstrated that an appearance of DUSE was actually corresponding to a horizontal density gradient in its vicinity. Their observations suggested that the source region of DUSE is likely to be colocated with the sunward edge of the midlatitude trough. At the time, however, the EISCAT UHF radar observed the electron density only above Tromsoe, which did not enable them to investigate two-dimensional spatial collocation between DUSE and the midlatitude trough. Namely, it is still unknown which of the density gradients (i.e., poleward or sunward) is actually responsible for generating irregularities within DUSE. In this paper, we employ the data from the Hokkaido radar and succeed to derive a complete spatial distribution of DUSE in a statistical manner. Comparison of the statistical distribution with the model of the midlatitude trough could allow us to reveal how spatial collocation between the source region of DUSE and the midlatitude trough is.
 In order to compare the distribution of DUSE and that of the midlatitude trough, we employ the model of the trough minimum by Kohnlein and Raitt . Kohnlein and Raitt  demonstrated that in April and May the midlatitude trough is observed from around 1900 LT until 0500 or 0600 LT. As local time progresses through the midnight, the midlatitude trough moves to lower latitudes. The invariant latitude of the trough depends also on the geomagnetic activity. Based on these observations, they expressed the magnetic latitude of the trough minimum as functions of Kp and local time as follows:
where t is the time in hours from midnight (positive after midnight and negative before midnight); the range of data is such that −5 h ≤ t ≤ 5 h (i.e., valid only in the dark hemisphere). We overplot the model outputs in Figure 5c as white curves with dots. Here, we employ Kp = 1, 3, 5 as inputs for the model. In the statistics, we only employed the radar data obtained during quiet period (0 ≤ Kp ≤ 2). Thus, we compare the statistical location of DUSE and the trough model for Kp = 1 (curve at the highest latitude). Now, it is clearly shown that the source region of DUSE is closely colocated with the trough minimum. More importantly, the enhancement of scattering occurrence rate is maximized at the trough minimum and distributed on both sides of the minimum. These results mean that generation of DUSE is not associated with the poleward wall of the trough but with the sunward wall. Another interesting feature to note is that the source region of DUSE is located at slightly higher latitudes in earlier local time (for example, 1900–2000 MLT). This tendency is also consistent with the diurnal variation of the trough minimum [Kohnlein and Raitt, 1977]. This again confirms that the sunward wall of the trough is the source region of DUSE.
 In summary, we revealed the spatial distribution of HF radar echoes in the subauroral region by conducting a statistical analysis of scattering occurrence from the midlatitude SuperDARN radar in Hokkaido, Japan. Consequently, Dusk Scatter Event (DUSE [Ruohoniemi et al., 1988]) was identified as a most prominent backscatter target in these latitudes. The data from the Hokkaido radar enabled us to estimate the lower-latitude boundary of DUSE as around 59° in AACGM magnetic latitude. In addition, by adding the data from the King Salmon radar in Alaska we derived a complete statistical distribution of DUSE. The latitudinal extent of DUSE is about 9° from 59° to 68°, which is approximately 1000 km. The statistical distribution of DUSE was compared with the model of the midlatitude trough. As a result, the source region of DUSE is closely colocated with the minimum of the trough, which suggests that the electron density gradient at the sunward edge of the trough is responsible for DUSE. This means that the local time extent of the midlatitude trough can be monitored by using an appearance of DUSE as a proxy. We checked the daily summary plots of the Hokkaido radar data (available at http://center.stelab.nagoya-u.ac.jp/cgi-bin/superdarn/hokkaido.cgi) and found that DUSE occurred almost every day in April 2007. Thus, we can estimate a possible day-to-day variability in the local time extent of the trough by using the data from the Hokkaido radar. The current statistical analysis also suggests that we can estimate a spatial distribution of irregularities from the midlatitudes to high latitudes by combining the data from the Hokkaido and King Salmon radars located in Far Eastern area, which would be very powerful diagnostic tool for investigating the global distribution of plasma irregularities.
 This research was partially supported by Special Funds for Education and Research (Energy Transport Processes in Geospace, Solar-Terrestrial Environment Laboratory, Nagoya University) of the Ministry of Education, Culture, Sports, Science, and Technology of Japan. This work was supported by a Grant-in-Aid for Scientific Research (19340141) from the Ministry of Education, Culture, Sports, Science, and Technology of Japan, and by Project 2 of the Geospace Research Center, STEL. We are also thankful for the data of the King Salmon radar operated by the National Institute of Information and Communications Technology (NICT).