Propagation of large amplitude ionospheric disturbances with velocity dispersion observed by the SuperDARN Hokkaido radar after the 2011 off the Pacific coast of Tohoku Earthquake

Abstract

Ionospheric responses to the 2011 off the Pacific coast of Tohoku Earthquake are studied using the SuperDARN Hokkaido radar, which is located at (43.5°N, 143.6°E) and which monitors the ionosphere over a wide horizontal area. The radar observed an oscillation of the vertical motion of the ionosphere with a period of about 1 to 2 min. The disturbance propagated northward, away from the epicenter with the velocity of about 6.2, 4.5, 3.9 and 3.5 km/s. The latter three values are basically consistent with the propagation of the Earth’s surface waves reported in several previous studies. The propagation velocities decreased with time, which has not been reported in previous studies for this propagation velocity range. The peak-to-peak amplitudes of Doppler velocities of ground/sea scatter echoes observed by the radar were up to 200 m/s, which is considerably larger than previously-reported values using HF Doppler measurements, although they are not extremely large for this historical earthquake (M = 9.0). This is the first time that ionospheric data have been obtained with high temporal (8 s) and spatial (22.5 km) resolutions following a giant earthquake, which enables us to discuss the detailed characteristics of the propagation of coseismic ionospheric disturbances.

1. Introduction

Coseismic ionospheric disturbances have been studied using various kinds of instruments such as a GPS receiver network (e.g., Ducic et al., 2003; Otsuka et al., 2006), an HF Doppler system (e.g., Tanaka et al, 1984), magnetometers (e.g., Iyemori et al., 1996) and a mixture of some of these (e.g., Ogawa et al, 1982). Most papers reported that the disturbances propagated away from the epicenter with a velocity of 2–4 km/s, corresponding to the propagation of the Earth’s surface waves (Rayleigh waves). On the other hand, Otsuka et al. (2006) reported that the initial response of TEC to the earthquake propagated at a velocity of 2 km/s but that the main body of the disturbance propagated with a speed of 600 m/s, and they interpreted their result in terms of the oblique propagation of the acoustic waves.

Deployment of a dense network of observation stations capable of high spatial and temporal resolutions simultaneously will enable the detailed characteristics of these coseismic disturbance propagations to be recorded, such as the propagation velocity change with time. One such candidate would be the Super Dual Auroral Radar Network (SuperDARN), which is a network of HF radars (see reviews by Greenwald et al. (1995) and Chisham et al. (2007)). The SuperDARN Hokkaido radar (Nishitani and Ogawa, 2005) is the only one located in Japan, and has monitored ionospheric disturbances since December 2006. The 2011 Tohoku Earthquake, M = 9.0, started at 0546 UT on 11 March, 2011, with its epicenter at (38.1°N, 142.9°E) according to the Japan Meteorological Agency (http://www.jma.go.jp/jma/en/2011_Earthquake.html). Fortuitously, when the coseismic ionospheric disturbances reached the field of view of the SuperDARN Hokkaido radar, it was operating in a special mode such that the radar acquired the data along one specific beam direction every 8 s. Using data taken in this special operation mode, it is possible to discuss the propagation characteristics of the co-seismic ionospheric disturbance with a very high temporal resolution, with the field of view including the region not covered by the GPS receiver network. In this paper, we report on the initial results of the ionospheric disturbances observed by the SuperDARN Hokkaido radar after the 2011 off the Pacific coast of Tohoku Earthquake (hereafter referred to as the 2011 Tohoku Earthquake).

2. Instrumentation

Figure 1 shows the field of view of the SuperDARN Hokkaido radar (geographic coordinates: 43.5°N, 143.6°E). This figure shows the distribution of the expected reflection points of the ground/sea scatter echoes (referred to as the GS-scatter echoes hereinafter) calculated by assuming a reflection height of 250 km.

Fig. 1.

Field of view of the SuperDARN Hokkaido radar.

The radar monitors 16 beam directions (separated by 3.24 degrees), with beam 0 at the western edge of the radar field of view pointing close to geographic north. During the period of interest, the radar was operating with 2 sampling modes. Until 0559 UT, the radar was operating with the ‘normalscan’ modes, sequentially sampling beams 15, 14, …1 and 0 with a 3-s integration time for each beam, and sampling the whole field of view every 1 min. After 0600:12 UT, the radar was operating with the ‘themisscan’ mode (named after the name of the spacecraft program), which sampled beam 4 and the other beams alternately, that is, sampling beams 15, 4, 14, 4, 13, 4, …, 4, 4, 3, 4, 2, 4, 1 and 4, with a 4-s integration time, sampling beam 4 every 8 s and covering the whole field of view every 2 min. Incidentally, beam 4 points close to the direction away from the epicenter, so that it is possible to determine the approximate propagation speed of the disturbances. Beam 0 was not scanned during this mode.

In this paper, we use GS-scatter echoes. Figure 2 shows the expected geometry of the radio-wave propagation of GS-scatter echoes (1-hop mode). The HF waves obliquely emitted from the radar are reflected by the bottom of the F-region ionosphere, backscattered by the irregular structure of the ground/sea surface, and go back to the radar via the ionospheric reflection point. By measuring the Doppler shift of the received echoes, we can monitor the upward/downward motion of the ionosphere. The range resolution of the radar throughout the period of interest is 45 km. This corresponds approximately to the 22.5-km spatial resolution when the data are mapped onto the reflection points.

Fig. 2.

Working geometry of the HF radar observing the effect of the upward/downward motion of the ionospheric reflection point (modified from Hayashi et al., 2010).

3. Observation

Figure 3 shows the Range-Time-Parameter plot of the Doppler velocities of the GS-scatter echoes observed by beam 4 of the SuperDARN Hokkaido radar as a function of Universal Time (UT) and geographic latitude of the reflection point. Owing to the change in the radar operation mode, the temporal resolution of beam 4 changed from 1 min to 8 s at 0600:12 UT. The positive value in the GS-scatter echo Doppler velocities observed by the radar corresponds to a downward ionospheric motion (not a southward motion: see Fig. 2). Considering the expected geometry of the radar-wave propagation, the real vertical speed of the ionosphere is v/2sin? (v is line-of-sight velocity and ? is the elevation angle of the received signal). For example, given a Doppler velocity of 20 m/s, a ground radar range of 500 km and a reflection point height of 250 km, we obtain an elevation angle of 26.57° and a vertical motion speed of 22.36 m/s (Hayashi et al., 2010).

Fig. 3.

Range-Time-Parameter plot of the line-of-sight velocities observed by beam 4 of the SuperDARN Hokkaido radar. The horizontal and vertical axes are UT and the geographic latitude of the ionospheric reflection point, respectively. The hatched regions are affected by 2-hop echoes, and are excluded from the analysis.

It should be noted that the echoes beyond 50.5 deg latitude are affected by so-called 2-hop geometry echoes with the propagation geometry of radar-ionosphere-surface-ionosphere-surface target one way. We can see the effect from the fact that echoes having large amplitude oscillations at 50.5-52.5 deg latitude for 0600-0602 UT and 0606-0607 UT have approximately the same Doppler velocities as those at 46.5-47.5 deg latitude simultaneously, with a range half that of the former one. We do not say that all the echoes in these regions are 2-hop echoes, but might include them. Therefore, we excluded the region (hatched areas in Fig. 3) from the analysis, whose Doppler velocities are influenced by ionospheric motions at closer ranges. This limits the maximum detectable propagation distance to be 1400 km away from the epicenter.

In Fig. 3 there are several disturbances propagating northward. Among these, the most notable one is seen at about 0600 UT, where the line-of-sight velocities began showing a periodic oscillation with a period of 1 to 2 min, propagating northward. There are also other disturbances propagating northward. They are probably coseismic disturbances, because traveling ionospheric disturbances in the daytime, generated by lower atmospheric activities or high-latitude geomagnetic disturbances, usually propagate southward, as shown by Kotake et al. (2007) and Ishida et al. (2008) (they interpreted the result in terms of the wind filtering theory of gravity waves, where the favored azimuth of TIDs-propagation counters the background wind, which in the thermosphere, is mainly northward in the dayside region). The maximum peak-to-peak amplitude of the disturbances is about 200 m/s, as clearly illustrated in Fig. 4, showing line plots of the Hokkaido radar line-of-sight velocities of ranges 14, 16, 18, …. 30 and 32 (all for beam 4), with 5-point smoothing applied. It is difficult to identify the initial disturbance because before 0600 UT the radar was operating in the ‘normalscan’ mode with a lower temporal resolution. Judging from the velocity profile of beam 4, it is reasonable to say that the first disturbance reached the beam 4 field-of-view between 0559 and 0600 UT. This disturbance obviously propagated northward throughout the whole GS-scatter region. The northward propagation speed, obtained by calculating the ground ranges from the radar for the pixels along beam 4 and estimating the propagation velocities of the disturbances along beam 4, which is approximately along the propagation direction away from the epicenter, decreased with time; for the positive disturbance between 0600 and 0601 UT, the propagation velocity along the beam 4 direction is estimated to be 6.2 km/s. For the negative, positive and negative disturbance between 0600 and 0602 UT, 0602 and 0604 UT, and 0604 and 0605 UT, the speeds are estimated to be 4.5, 3.9 and 3.5 km/s, respectively (for later disturbances the waveforms are less clear, so that it is difficult to estimate the propagation velocity precisely). In Fig. 3 the black line is shown as one example of the disturbance propagation, with a velocity of 3.5 km/s. The characteristic periods of the disturbances are estimated to be 1 to 2 min, with the earlier ones having shorter time-scales. The wavelength of the disturbance is then about 400 km. It should also be noted that the peak-to-peak amplitude decreases as it propagates to more distant ranges. It is consistent with the fact that the apparent Doppler velocity is proportional to sin θ where θ is the elevation angle, which becomes lower at more distant ranges for a constant ionospheric height.

Fig. 4.

Line plots of Hokkaido radar line-of-sight velocities of ranges 14, 16, 18, … and 32 (all for beam 4), with 5-min smoothing applied. Echoes with line-of-sight velocities beyond ±200 m/s have been excluded. The positive velocities are toward the radar.

Figure 5 shows the sequential plots of the two-dimensional distribution of the line-of-sight Doppler velocities of GS-scatter echoes observed by the SuperDARN Hokkaido radar. For the 0600:00 and 0600:12 UT scans, shown in panels (a) and (b), only some of the beams (beams 12 to 15 and beams 1 to 13, respectively) have data because the radar changed operation mode at 0600:12 UT. Within one whole field-of-view scan there are several beam 4 scans; in Fig. 5, we plotted a beam 4 scan taken 8 s after beam 5 and 8 s before beam 3. The hatched regions are affected by 2-hop echoes, and excluded from the analysis. Since the beams other than beam 4 were scanned every 2 min during the themisscan mode, it is not easy to discuss the two-dimensional characteristics of the propagation of disturbances with periods less than 2 min. Nevertheless, the northward propagation of disturbances can be clearly seen, at least in the 0600:12 UT scan. Beam 13 (the easternmost beam between 0600:12 and 0602:00 UT), which was scanned first, shows the disturbance at the nearest ranges, and beam 1 (the westernmost beam), scanned last, shows the disturbances at the farthest ranges. This is consistent with Fig. 3, where the disturbances propagate northward along beam 4. At later times, (c) and (d), the two-dimensional structures are more complicated, probably because of the mixing of disturbances with several modes.

Fig. 5.

Sequential images of a 2-dimensional distribution of the line-of-sight velocities of the GS-scatter echoes observed by the SuperDARN Hokkaido radar. The wavefronts of the traveling disturbances at 0600:12 UT, scan (b) are illustrated by broken lines. The hatched regions are affected by 2-hop echoes, and are excluded from the analysis.

4. Discussion and Summary

In this paper, we present initial results on the properties of ionospheric disturbances after the 2011 Tohoku Earthquake, observed by the SuperDARN Hokkaido radar. The most notable aspect of this event is that it was monitored with a very high temporal (8 s) and spatial (22.5 km) resolution, the results of which are utilized to show the characteristics of the present historical earthquake as compared with previous ones, later in this section. These resolutions are higher than those of the GPS receiver Network in Japan, GEONET, where the temporal and spatial resolutions are usually 30 s and 25 km, respectively. Another advantage over GEONET is that the radar is able to monitor the ocean area which GEONET cannot cover, which enables us to monitor co-seismic disturbances as far as 1400 km from the epicenter.

The most notable point is that the propagation shows a dispersive signature, with speed decreasing with time. We are not certain about what produces this dispersive signature, mainly because there have been no numerical simulations dealing with the coseismic ionospheric disturbances with a temporal scale shorter than 1 min. At least we can say that the faster velocities (occurring first) correspond to shorter periods, and slower velocities (occurring later) are associated with longer periods. This signature might be related to the dispersive characteristics of atmospheric waves, or that the disturbance propagation path changes, or some other unknown factors. It is a subject for future studies.

Tanaka et al. (1984) discussed the characteristics of co-seismic ionospheric disturbances with a period of a few minutes using four HF Doppler receivers in Japan. They determined a horizontal propagation velocity of 3.5 to 3.9 km/s. Owing to the limitation of data points, they were not able to confirm that this horizontal propagation velocity was constant. With the present set of data, we can clearly see that at least the initial disturbances propagated through the field-of-view of beam 4 essentially with a constant speed although with slight changes. With the present set of data, it is also possible to discuss the dispersive signature of the disturbance propagation velocities.

The propagation velocities of 4.5, 3.9 and 3.5 km/s for the current event are similar to that of Rayleigh surface waves propagating horizontally on the ground, whereas the earliest one (6.2 km/s) is a little high, which, in future studies, should be investigated in detail by comparing with several other data. The typical horizontal propagation velocity of a Rayleigh wave is 2 to 3 km/s. Ducic et al. (2003) suggested that the Rayleigh waves generate acoustic waves that propagate upward. Since the velocity of the Rayleigh wave is larger than the sound velocity, at a region far from the epicenter the acoustic waves excited by the Rayleigh wave reaches ionospheric altitudes earlier than the acoustic waves excited at the epicenter. Consequently, the ionospheric disturbances would propagate away from the epicenter with the same velocity as the Rayleigh wave.

Although the dispersive characteristics of propagation velocities have been reported in the results of a numerical simulation by Otsuka et al. (2006) and Shinagawa et al. (2007), the main propagation modes discussed by them are the oblique propagation of acoustic waves (770 m/s) and the horizontal propagation of gravity waves (about 800 m/s), respectively. There have been no numerical simulation studies so far which have discussed the detailed propagation characteristics of horizontal Rayleigh waves accompanied by vertical acoustic waves. To the best of our knowledge, this is the first paper to report on traveling ionospheric disturbances with a propagation velocity range of a few km/s. Tanaka et al. (1984) and Ducic et al. (2003) did not consider the velocity dispersive signatures of the traveling ionospheric disturbances. This might be because their events were much smaller, compared with the present case, to enable the investigation of later disturbances, or because the sensitivity of their instruments was not high enough.

Of course, our observations might have been affected by oblique acoustic waves or gravity wave modes. We are not certain whether, or not, the radar data contain these propagation modes, because we have not fully examined the data. This is also a subject for future studies.

The peak-to-peak amplitudes of the Doppler velocities of the ionospheric disturbances are up to 200 m/s. The peak-to-peak HF Doppler frequency changes reported by Tanaka et al. (1984) were 2 Hz (filter applied, as in the present study), which can be converted into a HF radar Doppler velocity value as (HF Doppler frequency change)/(HF Doppler frequency)*(speed of light)/2*2 = 2 [Hz] / 8 [MHz] * (3*10⁸ [m/s]) / 2 * 2 = 75 m/s.

We are not certain about the expected values of Doppler velocities for a M = 9.0 earthquake, but if we assume, as a rough estimate, that the disturbance wave energy is proportional to the energy of the earthquake, then the peak-to-peak amplitude of the wave should be proportional to the square root of the earthquake energy. Accordingly, the expected amplitude for a M = 9.0 earthquake should be times as large as the amplitude for the M = 7.1 earthquake reported by Tanaka et al. (1984). However, the present value (200 m/s) is not extremely large as compared with the above value (75 m/s) for the M = 7.1 earthquake. Both observations were a few hundred km away from the epicenter, so that the condition of distance is similar. One possibility is that the epicenter of the latter earthquake had a 10-km depth of focus (less than that of the former earthquake, viz, 24 km), presumably leading to a more effective generation of the Earth’s surface waves, but we are not sure whether this factor can explain the relative disturbance amplitudes.

In summary, the new findings of the present paper are as follows.

• 1. The disturbance propagation velocities decrease with time, ranging from 6.2 km/s between 0600 and 0601 UT to 3.5 km/s between 0604 and 0605 UT, corresponding to a decrease of 2.7 km/s in about 5 min. This is the first report of the propagation velocity dispersive signatures with a velocity range of a few km/s using a new instrument having high temporal and spatial resolutions.

• 2. The present earthquake causes an ionospheric upward-downward motion with a peak-to-peak amplitude up to 200 m/s, which is considerably larger than the previous event but not as large as expected comparing the size of the present earthquake (M = 9.0) with that of the one reported by Tanaka et al. (1984) (M = 7.1). This provides new information about this historical earthquake.