Reduction of electron temperature in low-latitude ionosphere at 600 km before and after large earthquakes

Transcription

1 JOURNAL OF GEOPHYSICAL RESEARCH, VOL. 113,, doi: /2008ja013367, 2008 Reduction of electron temperature in low-latitude ionosphere at 600 km before and after large earthquakes Koh-Ichiro Oyama, 1 Yoshihiro Kakinami, 1 Jann-Yenq Liu, 1 Masashi Kamogawa, 2 and Tetsuya Kodama 3 Received 3 May 2008; revised 8 August 2008; accepted 9 September 2008; published 25 November [1] We examine ionospheric electron temperatures (T e ) observed by HINOTORI satellite during three earthquakes; M6.6 occurred in November 1981, M7.4 and M6.6 in January 1982 over Philippine, respectively. It is found that T e around the epicenters significantly decreases in the afternoon periods within 5 days before and after the three earthquakes. The region of ionosphere disturbance extends to degrees in longitude. A tendency exists that duration of the disturbance becomes longer as the increase of earthquake magnitude. F 2 peak frequency, f o F 2 and virtual height, h F from a chain of 4 ionosonde stations located in the longitude zone of 120 E 130 E are used together with electron density(n e ), that is observed simultaneously onboard HINOTORI satellite to find possible cause mechanisms of the abnormal reduction of electron temperatures. Behavior of HINOTORI T e /N e and ionosonde f o F 2 /h F implies the existence of westward electric field over epicentre. Our finding suggests that simple two plasma instruments might be able to play a fundamental role to study ionosphere disturbance associated with earthquake, if the constellation of small/mini satellites is organized and the orbits are properly chosen. Citation: Oyama, K.-I., Y. Kakinami, J.-Y. Liu, M. Kamogawa, and T. Kodama (2008), Reduction of electron temperature in lowlatitude ionosphere at 600 km before and after large earthquakes, J. Geophys. Res., 113,, doi: /2008ja Introduction 1 Institute of Space Science, National Central University, Jhongli, Taoyuan, Taiwan. 2 Department of Physics, Tokyo Gakugei University, Tokyo, Japan. 3 Earth Observation Research Center, Japan Aerospace Exploration Agency, Tsukuba, Japan. Copyright 2008 by the American Geophysical Union /08/2008JA [2] The possible precursor effects of the earthquake (EQ) on the ionosphere have been reported by many scientists [Pulinets and Boyarchuk, 2004]. Reduction of ionospheric total electron content (TEC) produced prior to earthquake occurrence has been reported by Liu et al. [2004]. Devi et al. [2004] reported the reduction of TEC as well as increase of TEC at the crest region of equator ionization. They also found the bite out phenomena of fof 2, which they claimed appearing prior to a large earthquake. Depueva et al. [2007] studied a strong earthquake which occurred near magnetic equator on 15 August 1963 with magnitude M = 7.75 by using Alouette-1 satellite as well as ionosonde data. They found the reduction of F region density before and after the earthquake more clearly in satellite data. At the same time they stressed that the average effect on fof 2 was very small and was observed only in the daytime. [3] In addition to the earthquake effects on low-latitude ionosphere, there are several reports, which study the middle latitude ionospheric f o F 2 and f b E s [Silina et al., 2001; Ondoh, 2000]. Recently Pulinets et al. [2007] conducted an intensive analysis on Irpinia earthquake which occurred on 23 November 1980 by using seismic data, radon emanation, hydrological anomalies, ground based ionosonde network, thermal infrared irradiance, Intercosmos-19 satellite topside sounding. They concluded that air lionization by radon which was emanated during the earthquake preparation could explain all atmospheric and ionosphere parameters. However most of ionosphere researchers are still not fully convinced with the existence of precursor effects of earthquake [Rishbeth, 2007]. [4] Japanese Sun Observation satellite HINOTORI was put into an equatorial orbit at the height of 600 km in February 1981 with the inclination of 31. Although the satellite was dedicated to study solar physics, two unique plasma probes which were developed in Japan [Hirao and Oyama, 1970; Oyama et al., 1999; Oya and Obayashi, 1966] were accommodated to study ionosphere anomalies triggered by solar events such as solar flare. These two probes are resonance rectification probe for electron temperature T e, and impedance probe for electron density N e. The instruments have been flown in a number of sounding rockets as well as satellites and their performances are well established. Especially the resonance rectification probe provides very accurate T e, which is even now difficult to obtain [Oyama, 1976]. Data that were measured with these two instruments on board HINOTORI were accumulated for 16 months, until the satellite operation was terminated in June 1982 because of the battery problem. A set of the T e data has been delivered to NSSDC, NASA and is open for public users. 1of14

2 Table 1. List of the Three Earthquakes, Which are Discussed in This Paper EQ1 EQ2 EQ3 Date Longitude Latitude Magnitude Depth (km) 22 November :05 11 January :10 24 January : E E E 18.8 N 13.8 N 14.1 N [5] To study the effects of earthquake on ionosphere, we adapt three procedures. As the first step, we need to grab general features of Te/Ne, depending on local time, season, solar flux, latitude and longitude [Su et al., 1996, 1997, 1998], as well as magnetic disturbance. Secondly, we need to understand the physics of these various features, such as Te in plasma bubble [Oyama et al., 1988], effect of electric field on the morning overshoot [Oyama et al., 1996], Te around equatorial ionization anomaly [Oyama et al., 1997; Balan et al., 1997], annual behavior of Ne/Te [Bailey et al., 2000], and effect of neutral wind upon Ne/Te regarding tilted magnetic meridian [Watanabe et al., 1995; Watanabe and Oyama, 1996; Oyama and Watanabe, 2004]. As the third step, Te model [Oyama et al., 2004] as well as Ne model [Kakinami et al., 2008], which describes normal and average behaviors of the ionosphere are constructed. To check the useful application of Te/Ne model, nighttime increase of Te during geomagnetic disturbance was studied [Oyama et al., 2005]. [6] On the basis of the high reliability of the models, we tried to find out the earthquake effect on the ionosphere by calculating the deviation of Te from the model value (DTe). DTe during three earthquakes are examined. Table 1 summarizes geographic locations of the epicenter, the depths, and the magnitude for 3 earthquakes (earthquake list is provided by United State Geological Survey). [7] Figure 1 illustrates the epicenters of three earthquakes and locations of 4 ionosonde stations, which provide ground based ionogram to be discussed in the later section. collision with neutral gas is less. Owing to the low local Ne and the resulting small heat capacity, Te at these heights is raised rapidly with reducing solar zenith angle in spite of the fact that the ionization of the atmosphere has changed very little. When neutral density increases after the delayed heating, energy of thermal electrons is dissipated through the collisions between thermal electrons and neutral gas, which results in a reduction of Te. [10] Figure 2 displays that the morning overshoot peak is higher in the winter hemisphere than in the summer. The feature comes from lower Ne in winter than in summer. The difference of Ne between two hemispheres is mainly caused by meridional neutral wind [Lin et al., 2007]. Effect of zonal wind appears on the Ne distribution in Asian as well as American zones where magnetic meridional plane tilts with respect to geographic axis [Watanabe et al., 1995]. [11] Afternoon overshoot, which shows more strong latitude and seasonal variations, can be explained mainly by heat flux conducted from the higher altitude along magnetic line of force, although the heating due to photoelectrons travelling from below contributes. Another heat 2. Local Time Behavior of Te at 600 km [8] To detect anomalies, typical variations of Te in various local times, seasons, latitudes, longitudes, and solar activities should be studied as a first step. Figure 2 illustrates one of the typical local time and geomagnetic variations of Te at 600 km during the northern winter months (November, December, and January). It can be seen that Te reaches the minimum value at 0400 LT, rises steeply and yields a peak at 0900 LT (named as morning overshoot ). After the morning overshoot, Te reduces toward noon, starts increasing at 1500 LT, and shows the second peak at 1700 LT (named as afternoon overshoot ). Finally the peak disappears around 2000 LT. The morning overshoot is not so sensitive to geomagnetic latitude, comparing to the afternoon overshoot. In contrast, the afternoon overshoot shows a clear latitudinal variation; its peak is very low over the geomagnetic equator, but becomes higher in higher latitudes. [ 9 ] The morning overshoot peak at 0900 LT is explained as following [Da Rosa, 1966]. Energy of thermal electrons is dissipated through collision with other thermal electrons and with neutral particles. Heating of the neutral atmosphere is delayed because of the large heat capacity and therefore the neutral density is low in the early morning. Accordingly the energy loss of thermal electrons caused by Figure 1. Epicenters of the three earthquakes and locations of four ionosonde stations. 2 of 14

3 Figure 2. Local time dependence of T e at the height of 600 km with respect to geomagnetic latitude for northern winter for F10.7 < 200. Data include all longitudes. Red circles mark morning overshoot and afternoon overshoot. source might be provided from downward ExB plasma drift near sunset, leading to the observed enhancement of T e [Watanabe et al., 1995; Bhuyan et al., 2002]. Contrary to the morning overshoot, which is mainly controlled by neutral gas density, T e peak of the afternoon overshoot largely depends on N e. It is noted that the T e peak shows the lowest value in equatorial region, where N e is higher than that in higher latitudes. [12] We find that T e in the afternoon overshoot reduces drastically and systematically before and after earthquake, as we report below. 3. Reduction of T e in the Afternoon Overshoot [13] Three examples of T e reduction in the afternoon overshoot before and after the occurrence of earthquakes are shown in Figures 3, 4, and 5. Figures 3, 4, and 5 consist of two figures (a) and (b). Upper panel and lower panel of figures (a) and (b) in the three figures provide T e and N e respectively. At the bottom of each 6 sub-panels, UT (Universal time), local time (LT), longitude (LON), latitude (LAT), and magnetic latitude (MLAT) are provided. [14] Black and blue circles plot values from model and observation respectively. Two thin black curves provide the root square error of 500 K for T e and first and third quartile (25%) for N e in the model. If observed T e deviates more than 500 K from the model, we take the feature as abnormal. [15] Upper panel of Figure 3a shows T e behaviors near afternoon overshoot two days prior to EQ1. The observed T e do not increase when T e model expects the increase at around 0921 UT at the longitude between 91.7 and Reduction of T e continues during the expected afternoon overshoot. [16] In the upper panel of Figure 3b shows the behavior of T e in the afternoon overshoot one day after EQ1. The model illustrates that T e should start to elevate at about 0936 UT in the longitude 76 and shows the peak at around 0947 UT, and disappears at 1000 UT at the longitude close to On the other hand, T e values observed remains constant during the period when T e is expected to elevate. Upper panel of Figure 4a shows the case for 3 days prior to EQ2. The panel (a) indicates that the deviation of T e starts at 1126 UT at the longitude between 69.4 and Upper panel of Figure 4b shows the T e behavior 1 day after the EQ2 occurred. At 0654 UT, T e should start to elevate toward 0704 UT. The T e observed keeps constant value until 0702 UT, starts to elevate, and finally merges to the model value at 0706 UT. The same features as those of Figures 3a and 3b are noticed. However the disturbed region appears much wider than that for EQ1. [17] For EQ3, panels and panel b of Figure 5 are presented for 6 days before and 2 days after the earthquake respectively. Panel (a) shows the small depression of T e in the afternoon overshoot, which should start at 0856 UT according to the model T e. Panel (b) shows that even before afternoon overshoot starts T e becomes lower than the model value at 0626 UT. [18] In the lower panel of Figure 3a, N e observed is nearly equal to the model value at the beginning of T e reduction. At 0930 UT, N e starts to deviate from the model value. [19] The lower panel of Figure 3b shows the similar features; N e follows the model value until 0948 UT and it suddenly drops. The similar features are recognized in the lower panels of Figures 4a, 4b, 5a, and 5b. [20] The features, which we described above for T e and N e data, are common for all the three earthquakes. [21] Figure 6 are shown to stress the spatial distribution of DT e around earthquake days for three EQ events (EQ1, top panel; EQ2, middle panel; and EQ3, bottom panel). DT e is plotted in geographic longitude latitude coordinate along 3of14

4 Figure 3. Examples of the deviation of T e and N e from the model for EQ1 in the afternoon overshoot for EQ1. 2 days before EQ1 (a) (satellite pass, 4065), and 1 day after (b) (satellite pass 4110). N e is shown for each panel at the lower panels. Black and blue dots show the model and observation respectively. Two thin lines show the upper and lower values of standard deviation. Lower panel shows the same but for N e. Location of the spacecraft is listed at the bottom. From the top, UT, LT, longitude, latitude, and geomagnetic latitude. 4of14

5 Figure 4. Same as Figure 3 but for EQ2. (a) 3 days before EQ2 (satellite orbit 4799). (b) 2 days after EQ2 (satellite orbit 4856). 5of14

6 Figure 5. Same as for Figure 3 but for EQ3. (a) 6 days before EQ3 (satellite orbit 4947). (b) 2 days after EQ3 (5065/5066). 6of14

7 Figure 6. Spatial distributions and daily variations of deviation of Te in the longitude -latitude coordinates for three earthquakes. Satellite orbit data are shown for three periods: before EQ, during EQ, and after EQ. D in each figure indicates the earthquake day. Stars in the middle of each panel show the epicentre locations. Note that Te deviation spreads more than 50 degrees in longitude toward both west and east from epicentre. Bold line in each figure indicates geomagnetic equator. each satellite orbit. In the north of the epicentres, data acquired by HINOTORI satellite was scarce. During the access (about 10 minutes) with HINOTORI satellite, over Kagoshima Space Centre ( E, N), the time was used to retrieve the data stored in a data recorder and not sufficient to retrieve the real time data. Red Stars in the middle of the figures (22 26 November 1981, January 1982, and January 1982) for three earthquakes mark the epicentres. For each earthquake, three panels are provided before earthquake (left), during the earthquake (middle), and after the earthquake. The middle panels of Figure 6 show that Te deviation from the model reaches the maximum around earthquake days. The figures also show that near epicentre the reduction of electron temperature is most intense, although the maximum reduction does not always coincide with the latitude of epicentre such as the cases of EQ2 and EQ3. It is noted that the region of large Te deviation ranges from more than 60 to the west and 40 to the east from the epicenter, respectively. Although the data beyond 160 in longitude was not available for EQ2, the region might extend more than 160 in the East. The spatial distributions are nearly symmetric in longitude with respect to the epicenter. Comparison of the disturbed region of EQ2 with those of other two earthquakes appears to suggest that EQ2 (the largest earthquake among the three) has wider region than other two. It seems that the effect of the Pacific Ocean does not exist as far as we examine the EQ2, which has data up to the longitude of 170 east. [22] As we mentioned that the largest deviation of Te with respect to latitude does not match with the epicenter. This suggests that only Te measurement is not enough to identify the location of earthquake epicenter. However identification of the epicenter might be possible by combining equator and polar orbiting satellites, which provide other physical parameter such as Ne and O+. [23] Figures 7 and 8 are presented to show detail diurnal variation of the deviation of Te (D Te) from the model for three earthquakes respectively. The longitude and latitude ranges are limited between 20 and 220, and between 0 and 32, respectively. Dst and Kp indices, which illustrate 7 of 14

8 Figure 7. Deviation of T e observation from model value for November 1981 (EQ1). Top panel shows Dust (blue) and Kp indices (black). Bottom panel shows the T e deviation with respect to date. Depending on the local time when the data were obtained, the data points are colored from dark blue (early morning) to deep red (midnight) evening. Red arrow shows the days when earthquake EQ1, occurred. Please note that points plotted by orange color, which correspond to afternoon overshoot, show the largest deviation of T e. Deviation, which shows minimum around 11 November is considered to be caused by magnetic disturbance. magnetic disturbance at low and equatorial regions, are plotted at the top of the figures. Vertical color scales at the right in three figures show the local time when DT e is observed. At the bottom panel, deviation of T e (DT e )is shown with earthquake day marked by a star. Figure 7 shows the EQ1. Magnetic disturbance occurs on 23 November 1981 with the lowest Dst values of 75 nt. The deviation of T e from the model starts to increase on 18 November. Two days data gap exist, and no measurement is available on earthquake day. It appears that the largest deviation of T e in the afternoon overshoot is found around 23 November Large negative deviations of T e from the model are seen in the morning overshoot from 6 November to 13 November, which might be attributed to magnetic disturbance. It is noted that deviation of T e from the model scatters in both positive and negative values during morning overshoot period for magnetic disturbance.. It is also noted that daytime T e varies in antiphase with N e for magnetic disturbance [Oyama et al., 2005]. While for earthquake disturbance deviation of T e is only biased to negative value, and shows gradual steady reduction. Antiphase relation between T e and N e is not found. [24] In Figure 8, EQ2 and EQ3 cases are shown. Dst value is above 50 nt except 31 January. Kp indices are below 4 most of the days. Two red stars in the horizontal axis at the bottom panel indicate the days when EQ2 and EQ3 occurred. Before 8 January 1982, no data were acquired because of New Year holidays in Japan. Further data were not obtained on 10, 17, and 24 January 1982 because of the non-operation of the satellite (one day off every week). [25] T e deviation shows the two minima around 11, 12, and 13 January, and around 24, 25, and 26 January These days are close to the days when EQ2 and EQ3 occurred. Although we cannot acquire firm conclusion, all three cases indicate that T e deviation appears to reach its maximum about 2 days after the earthquakes. It is noted that the recovery to non-earthquake state is about one day slower for EQ2 than EQ1 and EQ3. [26] Findings obtained from the analysis of HONOTORI data are summarized as followings. [27] Reduction of T e in the afternoon overshoot is found prior to and after earthquake in low/middle latitude region. [28] Deviation of T e in the afternoon overshoot starts about 5 days prior to the earthquake and recovers to the original value after about 5 days. Deviation of T e tends to appear earlier and the recovery is slower as the magnitude of the earthquake increases. [29] The disturbed region is nearly symmetric in longitude with epicentre and ranges from to the west and to the east in longitude from the epicentre. A tendency exists that larger earthquake has larger disturbed area. [30] In latitude wise the range of deviation extends from north to south with steep reduction in high latitude. 8of14

9 Figure 8. Same as for Figure 7, but for January 1982 (EQ2, and EQ3). Two red arrows show the days when earthquake EQ2, and EQ 3 occurred. Please note that orange color data points show the largest deviation of electron temperature. [31] N e does not change or shows small increase at the beginning of T e reduction, but later N e reduces drastically. The reduction of N e is not because of the local time variation, but because of the satellite location in higher latitude. 4. Ionosonde Data at Manila and Taipei [32] Figures 9 and 10 show ionogram data obtained at Taipei and Manila for EQ1, and EQ2 and EQ3, respectively. At the top of the panels Dst and Kp indices are added to make it easier to refer geophysical situation. The second and third panels show daily variation of f o F 2 and h F for Taipei. Two black thin lines indicate first and third quartile of 15 days. Blue lines illustrate the 15 days median values. Red lines show observation data for each day. The fourth and fifth panels provide f o F 2 and h F at Manila. Colors for lines are the same as for Taipei. [33] In Figure 9, h F at Taipei does not indicate a clear reduction associated with earthquake. While at Manila, h F in the afternoon shows clear reduction of about 100 km, and small reduction of about 5 10 km in the early morning between 18 and 25 November. These features are also seen in Figure 10 for EQ2 and EQ3 before and after the earthquakes. It is noted that reduction of h F in Manila around 11 January 1982 when EQ2 occurred, is clearer than other two cases. [34] Variations of f o F 2 due to earthquake effect are difficult to be identified from Figure 9 and 10. However at Manila, small increase of f o F 2 is found in the late afternoon on 7, 9, 12, and 13 January Clear reduction of f o F 2 is seen between 7 and 16 January in the early morning when f o F 2 shows the minimum. It is noted again that EQ2 is the largest earthquake among three. Detail feature of f o F 2 at these two stations will be described later regarding the mechanism of the T e reduction. 5. Mechanism of T e Reduction [35] Our main finding from HINOTORI satellite is that the afternoon overshooting of T e totally disappears or partially disappears. N e, which is very close to model value or slightly higher than the model value at the beginning of T e reduction, starts to decrease some time after the Te reduction, which means that N e depletion, occurs in higher latitude. Ionosonde data show reduction of ionosphere F region height of about 100 km at Manila. However f o F2 does not change during this phenomenon. This feature is very puzzling, because it is a well-known fact that during daytime when N e reduces (increases), T e shows increase (reduction). Two components might play the role. One is related to heat input from higher altitude, another from lower altitude, as we described in section 1. For both cases, a region of enhanced N e should exist between h F height and 600 km height. This region might be produced as a combination of eastward dynamo electric filed (which is originally generated by neutral wind) and westward electric field that is generated prior to earthquake, and remains even after. Near equator the westward electric field weakens the eastward electric field, and as a result, crests of the equator anomaly shift to the equator ward, which has been reported before [Liu et al., 2001, 2002]. In addition to this movement, h F reduces due to the weakening of the eastward electric field as Manila ionosonde shows clearly. Above F 9of14

10 Figure 9. Ionogram data from Taipei and Manila for EQ1. From the top, first panel shows Dst and Kp indices. The second and third panels show f o F 2, and h F for Taipei respectively. The fourth and fifth panels provide f o F 2 and h F for Manila respectively. Two black thin, blue, and red lines in the second to fifth panels show first and third quartiles of 15 days, 15 days median value, and observed f o F 2 and h F, respectively. region height, the westward electric field drives ionosphere plasma downward under earth magnetic field, and as a result, a region is formed where electron density becomes denser and this dense electron region remains late in the afternoon. [36] Heating of thermal electron due to the heat conducted from higher altitude reduces when enhanced N e region exists below 600 km and T e both at lower height and at the height of 600 km decreases because of high-heat conductivity along magnetic line of force. [37] Photoelectrons, which are mainly generated in upper E region and F region, and travel along magnetic line of force, might also contribute to the behavior of T e at the height of 600 km. Generally the photoelectrons continuously heat thermal electrons up to the height of 600 km. However, photoelectrons that travel through the density-enhanced region dissipate their energy there, and as a result T e reduces along magnetic line of force. [38] In the higher latitude, a region of depleted N e exists because of the downward motion of plasma. This situation is shown in Figure 11. Left and right panels are provided for non-earthquake and earthquake case respectively. In both panels direction of the natural eastward electric field is shown by circled dot (pointing vertical from the paper). In the right panel, westward electric field (vertical to the paper, from the front toward back) is indicated by a crosssurrounded by a circle and three red vertical arrows show the epicenters of 3 earthquakes. Dotted line shows the altitude of HINOTORI satellite. A slant large arrow shows the direction of plasma drift produced by the westward electric field. [39] Figure 12 shows the latitude distributions of fof 2 around equatorial ionization anomaly with respect to LT for EQ1, EQ2, and EQ3. Every 1 hour f o F 2 data from Manila (121.1 E, 14.7 N), Okinawa (127.8 E, 26.3 N), Taipei (121.2 E, 25.0 N), and Yamagawa (130.6 E, 31.2 N) are used. Data between stations are calculated by linear interpolation. Top, middle, and bottom of the left panel show 15 days median values during 5 9 November 1981 for EQ1, 24 December 1981 to 7 January 1982 for EQ2, and 8 22 January 1982 for EQ3. In the right panel, top panel shows f o F 2 on 20 November 1981 (2 days before earthquake). Middle panel shows f o F 2 on 8 January 1982 (3 days 10 of 14

11 Figure 10. Same as for Figure 9 but for EQ2, and EQ3. It is noted that observed f o F 2 shows clearly the value lower than the lower quartile, starting from 7 January to 16 January. before EQ2). Bottom panel shows fof2 on 23 January 1982 (one day before EQ3). [40] The panel shows three features on the day of the earthquake: (1) N e start increasing late in the morning, (2) equatorial anomaly crest moves toward equator, and (3) dense electron density region extends to even 20 LT at the latitude of 20 to 25 N. The panel strongly supports the mechanism that we proposed. Figure 11. Cartoon, which shows the energy dissipation of photoelectrons, which travel through high plasma density region below 600 km. Three arrows indicate epicenters of EQ1, EQ2, and EQ3, respectively. Left panel and right panel are supposed to be for no-earthquake and earthquake days, respectively. Direction of original dynamo electric field and westward electric field, which might be associated with earthquakes are drawn by a circled cross and circled dot, respectively. 11 of 14

12 Figure 12. Latitude-local time map of f o F 2 (top panel, EQ1; middle panel, EQ2, and bottom panel; EQ3). Left figures in each panes show f o F 2 averaged over 2 weeks. Right figures in each panel provide f o F 2 2 days, 3 days, and 1 day before EQ1, EQ2, and EQ3 respectively. Common features for three cases are that (1) high plasma density region stays late in the afternoon and (2) EIA cleft moves toward lower latitude. [41] In the above we tried to explain our findings by electric field. The reduction of T e in the afternoon overshoot sometimes appears at magnetic conjugate point (for example, EQ on 9 October 1981, Lat/Long: 10.0/162.1, M = 6.5). Ruzhin et al. [1998] reported similar conjugate point effects for NmF2. We found that atomic ion density measured by DE-2 shows the conjugate point effect. Although reduction of T e in the afternoon overshoot can also be explained by neutral wind at topside ionosphere [Watanabe et al., 1995], relation between N e and T e is in antiphase. It is also noted that the depletion of T e, which we think as due to earthquake effect, is not detected beyond 30 degrees in geomagnetic latitude. This feature is also favorable to electric field. [42] Therefore our next question is on the origin of electric field. Although we do not want to speculate the generation mechanism of electric field at this moment, when information is scarce to get conclusion, we close this section by only showing two plausible mechanisms below. [43] The strong electric field might possibly be generated at the heights where cloud is formed prior to and after earthquakes. The westward electric field which shows a symmetric feature both side of the epicentre suggests the existence of a narrow region of very intense electric field near epicentre along north south direction; positive at eastside and negative at west side of the epicentre. The east/west width of the strong electric field region might be less than 100 km, so that, this small-scale electric field cannot be seen at the ionosphere height. This mechanism assumes the direct penetration of electric filed into topside ionosphere from cloud height. [44] Recently Immel et al. [2006] showed that a newly discovered 1000-km scale longitudinal variation in ionospheric density might be explained by consideration of the dynamo interaction of the tides with the lower ionosphere (E layer) in daytime. England et al. [2006] published a paper that seems to support above idea. [45] Hagan and Forbes [2002] investigated mesospheric and lower thermospheric migration and nonmigrating tidal components that propagate upward from the troposphere, where they are excited by latent heat release associated with deep tropical convection. Pulinets proposes the ionization produced by the radon emanating from the Earth s crust. As a result, hydrated ions are produced, which finally release the latent heat of evaporation [Pulinets et al., 2007]. The second mechanism to modulate lower E region by atmospheric waves seems to be most plausible. 6. Concluding Remarks [46] On the basis of our firm confidence that the data of T e and N e which were measured by Japanese scientific satellite HINOTORI are reliable, we have studied the behavior of T e in the afternoon overshoot for three earthquakes larger than 6.5, which occurred around equator ionization anomaly. 3 events show common features. Continuous systematic reduction of Te in the afternoon overshoot can be found prior to and after the big earthquakes. We presume that a weak electric field, that is generated associated with earthquake, is playing a fundamental role in the region. The electric field is the order of 1 mv/m or less and it should have very slow time variation of the order of 10 days starting from about 5 days before earthquake. 12 of 14

13 [47] Although the cases, which we have shown here, are for low-latitude earthquake of large earthquake, the detectability of smaller earthquake or high latitude earthquake depends on the accuracy of the model, which can be made with repeated accurate measurements by satellites. [48] Our result suggests that even a small satellite, which carries two simple reliable plasma probes, can play unexpectedly significant role for the study of precursor phenomena associated with earthquake. Height of the satellite is one of the crucial factors for the findings. International collaboration among countries that are suffering from earthquake disasters should be established urgently to launch small satellites from these countries. [49] The reduction of T e in morning overshoot, which appears to be associated with earthquake, is also found in HINOTORI satellite data, but not so clear. It is also stressed that the cases exist where T e in the afternoon overshoot shows the deviation from model when no earthquake is reported in USGS list. The effects of large meteorite shower seem to be not negligible. Cases also exist where clear T e reduction is not found even in the existence of large earthquake. This case seems to happen in high latitude earthquake as well as in ocean earthquake. Although the road to understand the precursor effects is still steep, these issues will be summarized in the separate paper, as the purpose of the present paper is to inform the high possibility of detecting precursor effects of big earthquakes and encourage scientists of various fields to work together as early as possible. [50] Acknowledgments. Prof. S. Watanabe, Hokkaido University, originally constructed T e and N e models. The authors are grateful to Prof. I. Kutiev, Bulgarian Geophysical Institute, Bulgaria for his contribution at the initial stage of this work. Suggestions by Dr. S. Abe of Japan Aerospace Exploration Agency and Dr. T. Maruyama of National Institute of Information and Communications Technology regarding meteor shower events were informative. The manuscript was completed while one of the authors (K.-I. Oyama) was staying at National Central University as a visiting Professor. We also express our gratitude to Mr. T. Izumida of Toshiba Electronics Engineering Corp. for drawing Figures 9 and 10 and Mr. H. K. Jhung of Institute of Space Science, National Central University, Taiwan for drawing Figure 12. This work was partially supported by Japan Aerospace Exploration Agency. [51] Amitava Bhattacharjee thanks the reviewers for their assistance in evaluating this paper. References Bailey, G. J., Y. Z. Su, and K.-I. Oyama (2000), Yearly variations in the low-latitude topside ionosphere, Ann. Geophys., 18, Balan, N., K.-I. Oyama, G. J. Bailey, S. Fukao, S. Watanabe, and M. A. Abdu (1997), A plasma temperature anomaly in the equatorial ionosphere, J. Geophys. Res., 102(A4), Bhuyan, P. K., M. Chamaua, P. Subrahmanyam, and S. C. Garg (2002), Diurnal, seasonal and latitudinal variations of electron temperature measured by the SROSS C2 satellite at 500 km altitude and comparison with the IRI, Ann. Geophys., 20, Da Rosa, A. V. (1966), The theoretical time dependent thermal behavior of the ionosphere electron gas, J. Geophys. Res., 71(17), Depueva, A. Kh., A. V. Mikahilov, M. Devi, and A. K. Barbara (2007), Spatial and time variations in critical frequencies of the ionospheric F region above the zone of equatorial earthquake preparation, Geomagn. Aeron., 47(1), Devi, M., A. K. Barbara, and A. Depueva (2004), Association of total electron content (TEC) and fof2 variations with earthquake events at the anomaly crest region, Ann. Geophys., 47(1), England, S. L., S. Maus, T. J. Immel, and S. B. Mende (2006), Longitudinal variation of the E-region electric fields caused by atmospheric tides, Geophys. Res. Lett., 33, L22205, doi: /2006gl Hagan, M. E., and J. M. Forbes (2002), Migrating and nonmigrating diurnal tides in the middle and upper atmosphere excited by tropical latent heat release, J. Geophys. Res., 107(D24), 4754, doi: /2001jd Hirao, K., and K.-I. Oyama (1970), An improved type of electron temperature probe, J. Geomagn. Geoelectr., 22, Immel, T. J., E. Sagawa, S. L. England, S. B. Henderson, M. E. Hagan, S. B. Mende, H. U. Frey, C. M. Swenson, and L. J. Paxton (2006), Control of equatorial ionospheric morphology by atmospheric tides, Geophys. Res. Lett., 33, L15108, doi: /2006gl Kakinami, Y., S. Wtanabe, and K.-I. Oyama (2008), An empirical model of electron density in low latitude at 600 km obtained by Hinotori satellite, Adv. Space Res., 41, , doi: /j.asr Liu, J. Y., Y. I. Chen, Y. J. Chuo, and H. F. Tsai (2001), Variation of ionospheric total electron content during the Chi-Chi earthquake, Geophys. Res. Lett., 28(7), Liu, J. Y., Y. J. Chuo, S. A. Pulinets, H. F. Tsai, and X. Zeng (2002), A study on the TEC perturbations prior to the Rei-Li, Chi-Chi and China- Yi earthquakes, in Seismo- Electromagnetics: Lithosphere-Atmosphere- Ionosphere Coupling, edited by M. Hayakawa and O. A. Molchanov, pp , Terra Sci., Tokyo. Liu, J. Y., Y. J. Chuo, S. J. Shan, Y. B. Tsai, Y. I. Chen, S. A. Pulinets, and S. B. Yu (2004), Pre-earthquake ionospheric anomalies registered by continuous GPS TEC measurements, Ann. Geophys., 22, Lin, C. H., C. C. Hsiao, J. Y. Liu, and C. H. Liu (2007), Longitudinal structure of the equatorial ionosphere: Time evolution of the four-peaked EIA structure, J. Geophys. Res., 112, A12305, doi: / 2007JA Ondoh, T. (2000), Seismo-ionospheric phenomena, Adv. Space Res., 26(8), Oya, H., and T. Obayashi (1966), Measurement of ionospheric electron density by a gyro-plasma probe: A rocket experiment by a new impedance probe, Rep. Ionos. Space Res. Jpn., 20, Oyama, K. (1976), A systematic investigation of several phenomena associated with contaminated Langmuir probes, Planet. Space Sci., 24, Oyama, K.-I., and S. Watanabe (2004), Effects of zonal and meridional neutral winds on the electron density and temperature at the height of 600 km, JAXA RR , Japan Space Exploration Agency, Kanagawa, Japan. Oyama, K.-I., K. Schlegel, and S. Watanabe (1988), Temperature structure of plasma bubbles in the low latitude ionosphere around 600 km altitude, Planet. Space Sci., 36(6), Oyama, K.-I., N. Balan, S. Watanabe, T. Takahashi, F. Isoda, G. J. Bailey, and H. Oya (1996), Morning overshoot of Te enhanced by downward plasma drift in the equatorial topside ionosphere, J. Geomagn. Geoelectr., 48, Oyama, K.-I., M. A. Abdu, N. Balan, G. J. Bailey, S. Watanabe, T. Takahashi, E. R. de Paula, I. S. Batista, H. Oya, and F. Isoda (1997), High electron temperature associated with the prereversal enhancement and the equatorial ionosphere, J. Geophys. Res., 102(A1), Oyama, K.-I., T. Abe, K. Schlegel, A. Nagy, J. Kim, and K. Marubashi (1999), Electron temperature measurement in Martian ionosphere, Earth Planet. Sci., 51, Oyama, K.-I., P. Marinov, I. Kutiev, and S. Watanabe (2004), Low latitude Te model at 600 km based on HINOTORI satellite data, Adv. Space Res., 34, Oyama, K.-I., D. R. Lakshmi, I. Kutiev, and M. A. Abdu (2005), Low latitude Ne and Te variations at 600 km during 1 March 1982 storm from HINOTORI satellite, Earth Planets Space, 57, Pulinets, S., and K. Boyarchuk (2004), Ionospheric Precursors of Earthquakes, Springer, New York. Pulinets, S. A., P. Biagi, V. Tramutoli, A. D. Legen ka, and V. Kh. Depuev (2007), Irpinia earthquake 23 November 1980 Lessons from nature reviled by joint data analysis, Ann. Geophys., 50, Rishbeth, H. (2007), Do earthquake precursors really exist?, Eos Trans. AGU, 88(29), 296, doi: /2007eo Ruzhin, Yu. Ya., V. A. Larkina, and A. Kh Depueva (1998), Earthquake precursors in magnetically conjugated ionosphere regions, Adv. Space, 21, Silina, A. A., E. V. Liperovskaya, V. A. Liperrovsky, and C.-V. Meister (2001), Ionospheric phenomena before strong earthquakes, Nat. Hazards Earth Syst. Sci., 1, Su, Y. Z., K.-I. Oyama, G. J. Bailey, S. Fukao, T. Takahashi, and H. Oya (1996), Longitudinal variation of the topside ionosphere at low latitude: Satellite measurements and mathematical modelling, J. Geophys. Res., 101(A8), 17,191 17,205. Su, Y. Z., G. J. Bailey, K.-I. Oyama, and N. Balan (1997), A modelling study of the longitudinal variations in the north south asymmetries of the ionospheric equatorial anomaly, J. Atmos. Terr. Phys., 59, of 14

14 Su, Y. Z., G. J. Bailey, and K.-I. Oyama (1998), Annual and seasonal variations in the low latitude e topside ionosphere, Ann. Geophys., 16, Watanabe, S., and K.-I. Oyama (1996), Effects of neutral wind on the electron temperature at a height of 600 km in the low latitude region, Ann. Geophys., 14, Watanabe, S., K.-I. Oyama, and M. A. Abdu (1995), Computer simulation of electron and ion densities and temperatures in the equatorial F region and comparison with Hinotori results, J. Geophys. Res., 100(A8), 14,581 14,590. Y. Kakinami, J.-Y. Liu, and K.-I. Oyama, Institute of Space Science, National Central University, Jhongli, Taoyuan, Taiwan. (kaki@jupiter.ss. ncu.edu.tw) M. Kamogawa, Department of Physics, Tokyo Gakugei University, Nakuikitamachi, Koganei City, Tokyo, Japan. T. Kodama, Earth Observation Research Center, Japan Aerospace Exploration Agency, Sengen, Tsukuba, Ibaraki, Japan. 14 of 14