Annals 47, 1, 2004, 01/07def 119 ANNALS OF GEOPHYSICS, VOL. 47, N. 1, February 2004 Key words Ultra-Low Frequency emission – seis- micity – Alfven waves 1. Introduction It is now recognized that analysis of seismic data, even sophisticated, is not sufficient to re- solve two essential problems of geodynamics: what are the mechanisms of earthquakes (EQs) origin and how can large EQs be forecast? In such a situation an importance of nonseismic methods is evident. One of them is the variation of magnetic field in the Ultra-Low Frequency (ULF) range 0.01-10 Hz. This effect was first reported by Fraser- Smith et al. (1990) in connection with Loma- Prieta, 1989 (USA) large EQ (magnitude Ms = 7.1) and by Molchanov et al. (1992), Kopy- tenko et al. (1993) in association with Spitak, 1987 (former Soviet Union) EQ (Ms = 6.9). Fraser-Smith et al. (1990) were lucky to ob- serve at a distance of 7 km from EQ epicenter and found that ULF magnetic intensity in- creased about 14 days before EQ, then it de- pressed several days ahead and once again it increased strongly at 4 h before the main shock and continued at a high level after EQ. They Preseismic ULF effect and possible interpretation Oleg A. Molchanov (1), Alexander Yu. Schekotov (1), Eugeniy Fedorov (1), Gennady G. Belyaev (2), Mary S. Solovieva (1) and Masashi Hayakawa (3) (1) Schmidt United Institute of Physics of the Earth, Russian Academy of Sciences, Moscow, Russia (2) Institute of Terrestrial Magnetism, Ionosphere and Radiowave Propagation (IZMIRAN), Russian Academy of Science, Troitsk (Moscow Region), Russia (3) The University of Electro-Communications, Chofugaoka, Chofu, Tokyo, Japan Abstract We present the results of ULF magnetic field observation at Karimshino station (Kamchatka, Russia). Using a case study we discovered an effect of suppression of ULF intensity about 2-6 days before rather strong and iso- lated seismic shocks (magnitude M = 4.6-6.6). It is revealed for nighttime and the horizontal component of ULF field (G) in the frequency range 0.01-0.1 Hz. Then we prove the reliability of the effect by computed correlation between G (or 1/G) and especially calculated seismic indexes Ks for the rather long period of observation from June 2000 to November 2001. Our recent data confirm the validity of the effect. We show here a similar result during a period of frequent strong seismic activity in April-May 2002. It is highly probable that the effect ob- served is connected with the increase in plasma density perturbations inside the ionosphere, which are induced by preseismic water and gas release at the ground surface and following energy transportation into the iono- sphere by atmospheric gravity waves. Two models are discussed and computed: the first is a decrease of pene- tration coefficient of Alfven waves from the magnetosphere due to a turbulent increase in effective Pedersen conductivity in the ionosphere, and the second is a change in wave number (k) distribution of source ionospher- ic turbulence. One of the mechanisms or both could be responsible for the observed 2-3 times suppression of ULF magnetic field noise at the ground. Mailing address: Dr. Oleg A. Molchanov, Schmidt United Institute of Physics of the Earth, Russian Academy of Sciences, ul. Bolshaya Gruzinskaya 10, D-242, GSP-5, 123995 Moscow, Russia; e-mail: oleg@ifz.ru 120 Oleg A. Molchanov, Alexander Yu. Schekotov, Eugeniy Fedorov, Gennady G. Belyaev, Mary S. Solovieva and Masashi Hayakawa found the clearest effect in the frequency band F = 0.01-0.1 Hz. Molchanov et al. (1992), Ko- pytenko et al. (1993) observed ULF variation at a distance of 130 km from the EQ epicenter and noted only the last stage of the process: an in- crease in ULF intensity in time period from 3 h before to several days after EQ. Subsequent research on this subject was mainly produced in Japan. Hayakawa et al. (1996a) reported results of observation the ULF magnetic field variations before the great EQ at Guam, 1993 (Ms = 8.0) at an epi- center distance 65 km. They suggested ana- lyzing the polarization ratio R = Z/H in fre- quency band 0.01-0.05 Hz and found that this parameter increased about 1 month before EQ but returned to the regular level after it. Later Hayakawa et al. (1999) considered the data once again and showed that slope of ULF spectrum (fractal number) was also changed before the EQ. Hattori et al. (2002) reported observation of ULF magnetic variation around date of two Kagoshima, 1997 large EQs (M = 6.5 and M = 6.3) at a distance about 60 km from both epicenters. They also ana- lyzed the polarization ratio and found its in- crease about 1 month before EQ date. They could not find this signature at the far-dis- tance stations with the same equipment. Kopytenko et al. (2002) observed ULF mag- netic variations using network of stations sit- uated in the Izu and Chiba areas of Japan. They discussed results related to EQ swarm during June-July, 2000 with the strongest shock Ms = 6.4 in the middle of the swarm. Epicenter distances to the stations changed from 70 to 150 km and the authors focused on the polarization ratio near frequencies F1 = = 0.1± 0.005, F2 = 0.01± 0.005 and F3 = 0.005± ± 0.003 Hz. It was shown that ratio R (F3) / R⋅ ⋅(F1) sharply increases just before the start of strong seismic activity, while amplitudes of Z and G component variations and Z/G ratio de- fined in a frequency range F2 during night time intervals (00-06 LT) begin to increase ∼ 1.5 months before the period of the seismic activity. In this paper, we consider ULF perturba- tions in temporal scale from several hours to a few days and focus on the time correlation of our data with seismicity. Some results of obser- vations have already been published (Mol- chanov et al., 2003). Here we briefly review the previous results (Section 2) and present new ob- servational evidence of the preseismic ULF ef- fect (Section 3). Furthermore we discuss the possible explanation of the effect (Section 4). 2. Previous results During 1999-2000, in addition to the exist- ing seismic and geophysical observations, Rus- sian and Japanese scientists established a spe- cial observatory at Karimshino site (52.94°N, 158.25°E) in Kamchatka (Far-Eastern Russia). Its main purpose was to study a correlation of seismic activity with electromagnetic and oth- er nonseismic phenomena. The main advan- tage of this station is quiet electromagnetic environment that allows us to use rather sensi- tive equipment and to check some theoretical ideas. The regular recordings have been under- way since June 2000 and some information about Karimshino station has already been published (Gladyshev et al., 2002; Uyeda et al., 2002). Our three-component induction magne- tometer measures the geomagnetic field varia- tions in the frequency range 0.003-40 Hz. The sensitivity threshold is better than 20 pT/Hz1/2 at frequency 0.01 Hz. It corresponds to 0.02 pT/Hz1/2 at frequencies above 10 Hz. Here we analyze results in the interval from June 24, 2000 to February 25, 2001 (the first interval du- ration of 7 months) and second interval from February 26, 2001 to September 16, 2001 (dur- ing of about 6 months). So, the whole period of the observation has a duration of about 13 months. As mentioned before in this paper we are presenting results on variation with scale more than several hours which is why we use two hour averaging of the data. First of all we have produced the spectrum of ULF intensity for each magnetic field com- ponent (H, D, Z) in the 7 frequency bands: F = = 0.003-0.01 Hz (channel 1), F = 0.01-0.03 Hz (channel 2), F = 0.03-0.1 Hz (channel 3), F = = 0.1-0.3 Hz (channel 4), F = 0.3-1.0 Hz (chan- nel 5), F = 1.0-3.0 Hz (channel 6) and F= 3.0-5.0 Hz (channel 7). We found a conventional corre- 121 Preseismic ULF effect and possible interpretation Fig. 1a-e. Variation of impedance ratio at channel 2 near the date of the EQs with the following parameters: panel a) – July 6, 2000, M = 6.0, distance D = 762 km, Ks = 1.4; panel b) – July 29, 2000, M = 4.9, distance D = = 195 km, Ks = 0.8; panel c) – August 23, 2000, M = 4.6, distance D = 112 km, Ks = 1.35 ; panel d) – Novem- ber 21, 2000, M = 5.1, distance D = 170 km, Ks = 0:75 ; panel e) – February 7, 2001, M = 5.6, distance D = = 210 km, Ks = 1.15. Vertical grid lines are for local midnightofeachday of observation. The intervals in panels c) and d) are shorter due to lack of data. a b c d e 122 Oleg A. Molchanov, Alexander Yu. Schekotov, Eugeniy Fedorov, Gennady G. Belyaev, Mary S. Solovieva and Masashi Hayakawa F ig . 2. Z /G va lu es ( le ft pa ne ls ) an d 1/ G va lu es ( ri g h t pa ne ls ) af te r 1 da y av er ag in g fo r th e sa m e ca se s as i n fi g. 1 a- e. 123 Preseismic ULF effect and possible interpretation lation with Kp index of magnetic activity and evident daily variation, especially in channels 1, 2, 3. But we did not find any clear correlation with a specially developed index of seismicity Ks, which includes dependence on magnitude and distance to earthquake (EQ) epicenter and which is proportional to seismic energy re- leased near the observation site (see Molchanov et al., 2003). Then we apply the method of polarization ratio, which was discussed in many papers since Hayakawa et al. (1996a) and which is reduced to analysis of Z1/2/G1/2 ratio in our re- search. In contrast with amplitude analysis, some correlation with Ks was found at least for the frequency channels 2 and 3 and near the date of large Ks values. We demonstrate sever- al cases in fig. 1a-e, each case during time in- terval ± 14 days around the EQ date and pres- entation is centered on the corresponding date. For simplicity, we present only channel 2 (F = = 0.01-0.03 Hz). It is obvious that nighttime values of Z1/2/G1/2 show an increase at about 2- 7 days time period before EQ date. An impor- tant question arises immediately: what does increase in Z component or decrease of G component or both mean? First of all, we ex- amined the behavior of Z component and found it reveals mainly seasonal changes and sometimes it is exposed to small man-made perturbations but does not show correlation with seismicity. To clarify this we present Z/G values and 1/G values after 1 day averaging in fig. 2 for the same cases as in fig. 1a-e. As a re- sult, we concluded that the effect observed is a depression of horizontal ULF magnetic field several days before EQ. The effect was not found for all the events being masked by the dominant magnetospheric component of am- plitude variation of the geomagnetic field on the Earth surface. Maybe, for some events the difference of parameters of EQs can be impor- tant. To check it we analyzed the geographical distribution of the casual EQs and discovered that correlated cases are mainly concentrated near the sea shore. Then we checked the reliability of the ef- fect by correlation analysis. At the beginning we construct the set of normalized deviations as the following: G G G G G G G G1 1 1 1 i i i i i i i i = - = - d d _ _ ` i i j where i is number of the frequency channel, i = = 1, 2, 3 and 〈G〉, 〈1/G〉 denotes running mean with 1 month window. Taking into considera- Fig. 3a-c. Cross-correlation of (a) δG∗Kpd, b) δ(1/G)∗ ∗Kpd and (c) δ(1/G)∗Ksd in a range of ±15 days for whole 13 months period of observation and different frequency bands. b c a 3. New observational results Previous results were obtained for a period of rather moderate seismic activity, Ks ≤ 3 (see fig. 1a-e). However, since late autumn of 2001 the activity intensified in the vicinity of our sta- tion. In this situation, statistics of the preseis- mic effect do not change essentially but a new feature appears just near the date of strong EQs. Let us consider it for the period from April 17, 2002 to May 17, 2002, including several strong EQs with Ms > 5. Two of them happened near the sea shore A (April 26, Ms = 5.8, dis- tance D = 180 km, Ks = 7.5) and C (May 8, Ms = = 6, distance D = 195 km, Ks = 10.4), the others occurred further from the shore, including B (May 3, Ms = 5.2, distance D = 190 km, Ks = =1.8) (fig. 4a). Impedance ratio in channels 1, 2, 3, 4 is presented in fig. 4b. Preseismic signature 3-4 days before event A and 3-5 days before event C can be noted. It is similar to previous re- sults and in the same manner as earlier we check that it is due to a decrease of G-component of the ULF magnetic field. In contrast, an increase in the impedance ratio near the date of the EQs is connected here with an increase in Z-compo- nent, which can be considered in conventional terms of ULF radiation from underground seis- mic source. 4. Possible interpretation of the preseismic ULF effect It was usually supposed that seismo-associ- ated ULF variations could be either due to direct radiation from EQ origin zone (Fenoglio et al., 1995; Molchanov and Hayakawa, 1995) or due to a change in geoelectric conductivity inside and near the EQ zone, which leads to a change 124 Oleg A. Molchanov, Alexander Yu. Schekotov, Eugeniy Fedorov, Gennady G. Belyaev, Mary S. Solovieva and Masashi Hayakawa Fig. 4a,b. a) Map of EQ’s with Ms > 4 from April 17, 2002 to May 17, 2002, including several strong EQs with Ms > 5. Two of them have happened near the sea shore A (April 26, Ms = 5.8, distance D = 180 km, Ks = 7.5) and C (May 8, Ms = 6, distance D = 195 km, Ks = 10.4) the others occurred further from the shore, including B (May 3, Ms = 5.2, distance D = 190 km, Ks = 1.8). b) Impedance ratio in channels 1, 2, 3, 4 is presented. Pre- seismic signature 3-4 days before event A and 3-5 days before event C can be noted. tion that our expected effect has a temporal scale order of several days, we produce 2 days averaging of δGi and δ (1/Gi), which leads to values δ2Gi, δ2 (1/Gi), and find Ksd = K sΣ per day, Kpd = KpΣ per day. Due to a clear daily variation of ULF spectrum in the selected chan- nels, δ2Gi is mainly related to daytime ULF in- tensity,but δ2 (1 / Gi) is mainly related to night- time ULF intensity. As a next step we comput- ed correlation functions F1i(τ) = δ1Gi∗Ksd, F2i(τ) = δ1(1 / Gi)∗Ksd, F3i(τ) = δ1Gi∗Kpd and F4i(τ) = δ1(1 / Gi)∗Kpd using conventional pro- grams, where τ is determined in the interval ± 14 days. The negative value of τ corresponds to the preseismic period and the positive τ-val- ue is for the postseismic period in our formu- lation. Firstly, we present a correlation of ULF in- tensity with global ionosphere-magnetosphere activity (fig. 3a,b). An obvious correlation is observed both for daytime ULF intensity (fig. 3a) and for the night-time values (fig. 3b). It is evident that both day-time and night-time ULF intensity is proportional and concurrent to Kp index. This correlation is understandable. We call a correlation reliable if it reveals for all the intervals and at least twice out of a reliability margin at the whole interval, which is about ± 0.1. Based on this point a correlation in fig. 3a,b is reliable. Then we show a correlation of ULF intensi- ty with seismic index Ks (fig. 3c). Due to our criteria, night-time suppression of ULF intensi- ty near value τ ∼ − 4 days looks a reliable ef- fect. This conclusion coincides with the result of case study. Note that reliability of seismo-as- sociated ULF suppression effect is comparable with the reliability of the well-known effect of connection between ULF variation and Kp in- dex of global magnetic activity. 125 Preseismic ULF effect and possible interpretation a b 126 Oleg A. Molchanov, Alexander Yu. Schekotov, Eugeniy Fedorov, Gennady G. Belyaev, Mary S. Solovieva and Masashi Hayakawa in ULF waves generated by ionospheric sources (e.g., Merzer and Klemperer, 1997). The first mechanism is not compatible with our observa- tional results on preseismic effect because it predicts a preseismic increase in ULF intensity. Indeed, we observed such an increase near the date of several strong EQ shocks (see fig. 4b). Probably the interval of ULF increase could be extended essentially for the very large EQs up to a few days as reported earlier (see Section 1). An explanation by preceding change in the ground conductivity is also not very attractive for us because long-term magnetotelluric obser- vation in U.S.A. and Japan did not show any correlation with seismicity (see, e.g., Park, 1997). It seems that suppression of ULF mag- netic variation happened not inside the ground but in the lower ionosphere. A hint might be in the results of monitoring of the upper atmos- phere and ionosphere around EQ date by VLF transmitter signals, reported by Hayakawa et al. (1996b) and Molchanov et al. (2001). They found clear perturbations of the atmosphere-ion- osphere boundary several days before large EQs at nighttime or during night-to-day transition (so-called terminator time). Molchanov et al. (2001) provide arguments that water and gas eruptions before EQs could origin «mosaic» and «twinkle» spots of atmospheric temperature and density variations leading to the generation of Atmospheric Gravity Waves (AGW) turbulence. There are reports on increased intensity of ionospheric irregularities (so called E-spread and F-spread events) several days before large EQs (see, e.g., review by Meister et al., 2002). As considered by Mareev et al. (2002) AGW produce turbulent variations of density and elec- tric field in the lower ionosphere with horizontal scales order of AGW source size near the ground (∼ 50-150 km). Alperovich et al. (2002) showed that AGW perturbations inside the ion- osphere can modify plasma conductivity, espe- cially Pedersen type (along electric field of the perturbation). So, for explanation of ULF field suppression before EQ we considered two models: first is a decrease of penetration coefficient of Alfven waves from the magnetosphere due to turbulent increase of effective Pedersen conductivity in the ionosphere, and second is a change in wave number (k) distribution of source ionospheric turbulence. We assume that the ULF magnetic field at the ground bg is sum of the fields from two generation regions: the magnetosphere, in which downgoing Alfven waves generate with amplitude bm above ionosphere, and the iono- sphere, in which turbulent electric fields ∆Ei generate with random phase (〈∆Ei〉 = 0). The cartesian coordinates with the z-axes ver- tical upward is used. The main geomagnetic field B0 = B0ẑ. The electromagnetic field is expanded over the harmonics ( )exp i t ik r+ - +~ = where xk kk yx y= += t t . Let the x-axes be directed along the k=, kx = k= ≡ k, ky = 0. The following rela- tionship can be written for the electromagnet- ic field in the ionosphere and on the Earth sur- face: ( ) ( , ) ( , ) ( , ) ( , ) ( , ) ( , ) . b T k b k dk T k C E k dk T k C E k dk xg A m ix A ix iy A iy 2 2 2 2 2 2 = + + + + ~ ~ ~ ~ ~ ~ ~ ∆ ∆ # # # (4.1) Here TA is the penetration coefficient of the Alfven wave and Tix, Tiy are the penetration co- efficients of the normalized electric field of the ionospheric turbulence. Furthermore, we take into consideration that under the above assump- tions and real conductivity of the atmosphere byg / bxg << 1. In the first model, we neglect the input of the ionospheric turbulence and produce full wave computations using IRI-90 ionospheric profiles at midnight under the quiet conditions. The procedure of computation is described in many papers beginning from (Hughes and South- wood, 1976). Assuming ( , ) ( ) ( )b k b k km m 1= -~ ~ d (4.2) where δ is Dirac function and k1 is determined by peculiarities of Alfven wave generation and propagation in the magnetosphere. Then in cor- respondence with results of Alperovich et al. (2002) we suppose that ionospheric turbulence induced by seismicity leads to increased effec- tive Pedersen conductivity. Results are present- 127 Preseismic ULF effect and possible interpretation Fig. 5. Dependence of TA (ω, k1) obtained from the full wave solution using IRI-90 ionospheric profiles at mid- night under the quiet conditions. ΣP and ΣH are integral Pedersen and Hall conductivities. ΣA is the Alvfen wave conductivity. The oscillating regime at high frequencies is the manifestation of the ionospheric Alfven resonance. Fig. 6. The frequency dependence of the ratio of the penetration coefficients in the disturbed ionosphere to one in the undisturbed ionosphere calculated as shown in fig. 5. 128 Oleg A. Molchanov, Alexander Yu. Schekotov, Eugeniy Fedorov, Gennady G. Belyaev, Mary S. Solovieva and Masashi Hayakawa Fig. 7. Supposed distribution of the ionospheric turbulence in k-space. Thin line is for regular distribution ( k2 1- = 1000 km) thick line is for seismo-induced k-distribution ( k2 1- = 200 km). Fig. 8. Integrated penetration coefficients Tix ( f, k2) and Tiy ( f, k2). 129 Preseismic ULF effect and possible interpretation ed in figs. 5 and 6. Dependence of TA (ω, k1) is shown in fig. 5. The frequency dependence of the ratio of the penetration coefficients in the disturbed ionosphere to one in the undisturbed ionosphere is shown in fig. 6. It is clear from the figure that the magnetic field on the Earth surface at frequencies below ∼ 10−1 Hz decreas- es with growth of Pedersen conductivity. Contrarily, the second model considers only the ionospheric source and neglects the input from magnetospheric Alfven waves and the change in ionosphere conductivity. For simplic- ity we also assume that turbulence is isotropic, i.e. E E Eix iy i 2 2 2 = =∆ ∆ ∆ . Hence ( ) ( , ) ( , )b T k E k C dkxg i i A 2 2 2 =~ ~ ~∆# (4.3) where Ti (ω, k) = Tix (ω, k) + Tiy (ω, k). Then we take into account that the turbulence devel- ops as ionospheric eigenmode, i.e. ω = ω (k), ( , ) ( )E k E ki i 2 2 =~∆ ∆ and assume the following k-distribution: 2 ( )E k k k k k1i 2 2 11 6 + +∆ 22 _ _i i9 C (4.4) which is reduced to classic Kolmogorov’s dis- tribution k−5/3 for the k >> k2, where k2 has a meaning of inverse scale (L2) of the turbulence external source. So, we assume k2 = 10−3 km−1 (L2 ∼ 1000 km) in the usual (regular) situation and k2 = 5 ∗ 10−3 km−1 (L2 = 200 km), if the main source is AGW induced by seismicity. Both types of the distribution are demonstrated in fig. 7. Integrated penetration coefficients Tix ⋅ ⋅ (ω, k2) and Tiy (ω, k2) are shown in fig. 8, while change in the common penetration coefficient Ti (ω, k2) due to modification of main scale of the turbulence k2 is presented in fig. 9. 5. Discussion and conclusions We found the effect of suppression of ULF magnetic field variations about 2-6 days before rather strong seismic shocks in a case study. It is revealed for night-time and for horizontal Fig. 9. Change in the common penetration coefficient Ti ( f, k2) due to modification of the main scale of the tur- bulence k2. 130 Oleg A. Molchanov, Alexander Yu. Schekotov, Eugeniy Fedorov, Gennady G. Belyaev, Mary S. Solovieva and Masashi Hayakawa component intensity (G) in the frequency range 0.01-0.1 Hz. We prove a reliability of the effect by computed correlation between G (or 1/G) and specially calculated seismic indexes Ks. Based on the simple criteria, we conclude that reliability of seismo-associated ULF suppres- sion effect is comparable with the well-known effect of connection between ULF variation and Kp index of global magnetic activity. Using Ks in our formulation for analysis of preseismic ef- fect means an assumption that seismic shock is a result of some dynamic process (like instabil- ity) and intensity of preseismic perturbations is proportional to the energy of seismic shock it- self. Some justification of this approach is found in the well-known correlation between magnitudes of foreshocks and main shock (see, e.g., Scholz, 1990). This effect can also be supposed in the pre- vious observations at least in those where a pre- seismic increase in polarization ratio had been found (see Section 1). As is shown here, a de- crease of G could lead to observed increase of ratio Z /G. Note furthermore that depression of G amplitude several days before the main shock was noted in one of the first papers by Fraser- Smith et al. (1990). Like some other non-seismic precursors, our effect looks a sporadic one in a case study and can be recognized only by statistics. We be- lieve that ULF-seismicity connection becomes more clear and regular after integration on space, i.e. using a network of stations, which we plan to do in future. Our interpretation of the effect is not very speculative if seismo-induced AGW influence on the ionospheric turbulence is to be believed. The connection between AGW and the turbulent vari- ation of plasma density and electric fields inside ionosphere is well-known. It is evidenced by both ionospheric sounders and radars from the ground and in direct satellite observations (see review in the latest paper by Molchanov et al., 2002). Bia- gi et al. (2003) based on 24-years hydrogeo- chemical observations at Kamchatka area report- ed a clear correlation of hot-water eruptions and change in content with the occurrence of large EQs in the time intervals from several days be- fore EQ up to 1-2 weeks after it. Mareev et al. (2002) showed that even small temperature and density variations near the ground surface effec- tively generate AGW energy flux into the iono- sphere and the time of the energy transportation is between several hours and 1 day for AGW pe- riods from 10 min to 1-2 h and horizontal wave- lengths 30-100 km. Thus our assumptions on seismo-induced modification of the ionospheric turbulence are reasonable. As a result of our first model computations (figs. 5 and 6), we discovered that ULF signal suppression is approximately proportional to tur- bulent increase in Pedersen conductivityina fre- quency range F < 0.1-0.2 Hz. The result is not surprising because in this frequency range TA H A P 2 + +Σ Σ Σ^ h7 A (Hughes and Southwood, 1976), where ΣH, ΣP are integral conductivities of the ionosphere and ΣA is wave Alfven conductiv- ity. In the higher frequency range, the effect is masked by Ionospheric Alfven resonance struc- ture, which is evident in fig. 5 and created due to the passage of Alfven wave through the upper ionosphere. Although the frequency behavoiur of the computed effect coincides with the observed one (see fig. 6), the existence of the EQ-related intensification of the ionospheric turbulence enough to enhance the ionospheric conductivity at several ten percents is not proved. Looking now at the result of our second model computa- tion, it can be seen that frequency and value de- pendence is about the same, but we are free from assumption on increased turbulence intensity. However a new assumption on the redistribution of turbulence k-spectrum after the arrival of seis- mo-induced AGW seems realistic but demands additional research on the nature of ionospheric turbulence. We believe that one of the models or both are helpful to explain our observed effect. Acknowledgements This research was partially supported by ISTC under Grant 1121 and by Commision of the EU (grant No. INTAS-01-0456). Two au- thors (O.A.M. and M.H.) are thankful for sup- port from the International Space Science In- stitute (ISSI) at Bern, Switzerland within the project «Earthquake influence of the iono- sphere as evident from satellite density-electric field data». 131 Preseismic ULF effect and possible interpretation REFERENCES ALPEROVICH, L., I. CHAIKOVSKY, YU. GURVICH and A. MEL- NIKOV (2002): Laboratory modelling of the disturbed D- and E-layers: DC and AC fields, in Seismo-Electromag- netics: Lithosphere-Atmosphere-Ionosphere Coupling, edited by M. HAYAKAWA and O. MOLCHANOV (Terra Sci- entific Publishing Co., Tokyo), 343-348. BIAGI, P.F., O. MOLCHANOV, R. PICCOLO, A. MINAFRA, A. ER- MINI, Y.M. KHATKEVICH and E.I. GORDEEV (2003): Co- postseismic hydrogeochemical anomalies in a volcanic environment, Natural Hazards, 3, 263-267. FENOGLIO, M.A., M.J.S. JOHNSTON and J.D. BYERLEE (1995): Magnetic and electric fields associated with changes in high pore pressure in fault zone-application to the Loma Prieta ULF emissions, J. Geophys. Res. Solid Earth, 100, 12951-12958. FRASER-SMITH, A.C., A. BERNARDY, P.R. MCGILL, M.E. LADD, R.A. HELLIWEL and O.G. VILLARD JR. (1990): Low frequency magnetic field measurements near the epicenter of the Loma-Prieta earthquake, Geophys. Res. Lett., 17, 1465-1468. GLADYSHEV, V., L. BARANSKY, A. SCHEKOTOV, E. FEDOROV, O. POKHOTELOV, S. ANDREEVSKY, A. ROZHNOI, Y. KHABAZIN, G. BELYAEV, A. GORBATIKOV, E. GORDEEV, V. CHEBROV, V. SINITSIN, A. LUTIKOV, S. YUNGA, G. KOSAREV, V. SU- RUKOV, O. MOLCHANOV, M. HAYAKAWA and S. UYEDA (2002): Some preliminary results of seismo-electromag- netic research at Complex Geophysical Observatory, Kamchatka, in Seismo-Electromagnetics: Lithosphere- Atmosphere-Ionosphere Coupling, edited by M. HAYA- KAWA and O. MOLCHANOV (Terra Scientific Publishing Co., Tokyo), 421-432. HATTORI, K., Y. AKINAGA, M. HAYAKAWA, K. YUMOTO, T. NA- GAO and S. UYEDA (2002): ULF Magnetic Anomaly Pre- ceding the 1997 Kagoshima Earthquakes, in Seismo- Electromagnetics: Lithosphere-Atmosphere-Ionosphere Coupling, edited by M. HAYAKAWA and O. MOLCHANOV (Terra Scientific Publishing Co., Tokyo), 19-28. HAYAKAWA, M., R. KAWATE, O.A. MOLCHANOV and K. YU- MOTO (1996a): Results of ultra-low-frequency magnetic field measurements during the Guam earthquake of 8 August, 1993, Geophys. Res. Lett., 23, 241-244. HAYAKAWA, M., O. A. MOLCHANOV, T. ONDOH and E. KAWAI (1996b): Anomalies in the sub-ionospheric VLF signals for the 1995 Hyogo-ken Nanbu earthquake, J. Phys. Earth, 44, 413-418. HAYAKAWA, M., T. ITO and N. SMIRNOVA (1999): Fractal analysis of ULF geomagnetic data associated with the Guam earthquake on August 8, 1993, Geophys. Res. Lett., 26, 2797-2800. HUGHES, W.J. and D.J. SOUTHWOOD (1976): The screening of micropulsation signals by the atmosphere and iono- sphere, J. Geophys. Res., 81, 3234-3240. KOPYTENKO, YU.A., T.G. MATIASHVILY, P.M. VORONOV, E.A. KOPYTENKO and O.A. MOLCHANOV (1993): Detection of ULF emission connected with the Spitak earthquake and its aftershock activity based on geomagnetic pulsations data at Dusheti and Vardziyaobservatories, Phys. Earth Planet. Inter., 77, 85-95. KOPYTENKO, YU.A., V.S. ISMAGUILOV, K. HATTORI, P. VORONOV, M. HAYAKAWA, O. MOLCHANOV, E. KOPY- TENKO and D. ZAITSEV (2002): Monitoring of the ULF Electromagnetic Disturbances at the Station Network before EQ in Seismic zones of Izu and Chibo Peninsulas (Japan), in Seismo-Electromagnetics: Lithosphere-At- mosphere-Ionosphere Coupling, edited by M. HAYA- KAWA and O. MOLCHANOV (Terra Scientific Publishing Co., Tokyo), 11-18. MAREEV, E.A., D.I. IUDIN and O.A. MOLCHANOV (2002): Mo- saic source of internal gravitywaves associated with seis- mic activity, in Seismo-Electromagnetics: Lithosphere- Atmosphere-Ionosphere Coupling, edited by M. HAYA- KAWA and O. MOLCHANOV (Terra Scientific Publishing Co., Tokyo), 335-342. MEISTER, C.V., E.V. LIPEROVSKAYA, O.A. MOLCHANOV, O.A. POKHOTELOV, S.A. SENCHENKOV and O.A. ALIMOV (2002): To the question of spatial scales of seismo- ionospheric effects, in Seismo-Electromagnetics: Lithos- phere-Atmosphere-Ionosphere Coupling, edited by M. HAYAKAWA and O. MOLCHANOV (Terra Scientific Pub- lishing Co., Tokyo), 329-333. MERZER, M., and S.L. KLEMPERER (1997): Modeling low-fre- quency magnetic field precursors to the Loma Prieta earthquake with a precursory increase in fault-zone con- ductivity, Pure Appl. Geophys., 150, 217-248. MOLCHANOV, O.A and M. HAYAKAWA (1995): Generation of ULF electromagnetic emissions by microfracturing, Geophys. Res. Lett., 22, 3091-3094. MOLCHANOV, O.A., YU.A. KOPYTENKO, P.M. VORONOV, E.A. KOPYTENKO, T.G. MATIASHVILI, A.C. FRASER-SMITH and A. BERNARDY (1992): Results of ULF magnetic field measurements near the epicenters of the Spitak (Ms = 6.9) and Loma Prieta (Ms = 7.1) earthquakes: compara- tive analysis, Geophys. Res. Lett., 19, 1495-1498. MOLCHANOV, O.A., M. HAYAKAWA and K. MIYAKI (2001): VLF/LF sounding of the lower ionosphere to study the role of atmospheric oscillations in the lithosphere-ion- osphere coupling, Adv. Polar Upper Atmos. Res., 15, 146-158. MOLCHANOV, O.A., M. HAYAKAWA, V.V. AFONIN, O.A. AKEN- TIEVA and E.A. MAREEV (2002): Possible influence of seismicityby gravitywaves on ionospheric equatorial anomaly from data of IK-24 satellite, in Seismo-Electro- magnetics: Lithosphere-Atmosphere-Ionosphere Cou- pling, edited by M. HAYAKAWA and O. MOLCHANOV (Ter- ra Scientific Publishing Co., Tokyo), 275-296. MOLCHANOV, O.A., A.YU. SCHEKOTOV, E.N. FEDOROV, G.G. BELYAEV and E.E. GORDEEV (2003): Preseismic ULF electromagnetic effect from observation at Kamchatka, Natural Hazards, 3, 1-7. PARK, S.K. (1997): Monitoring resistivitychange in Parkfield, California: 1988-1995, J. Geophys. Res., 102, 545-559. SCHOLZ, C.H. (1990): The Mechanics of Earthquakes and Faulting (Cambridge University Press), pp. 439. UYEDA, S., T. NAGAO, K. HATTORI, Y. NODA, M. HAYAKAWA, K. MIYAKI, O. MOLCHANOV, V. GLADYCHEV, L. BARAN- SKY, A. SCHEKOTOV, G. BELYAEV, E. FEDOROV, O. POKHOTELOV, S. ANDREEVSKY, A. ROZHNOI, Y. KHABAZIN, A. GORBATIKOV, E. GORDEEV, V. CHEBROV and A. LU- TIKOV (2002): Japanese-Russian Complex Geophysical Observatory in Kamchatka region for monitoring of phe- nomena connected with seismic activity, in Seismo-Elec- tromagnetics: Lithosphere-Atmosphere-Ionosphere Cou- pling, edited by M. HAYAKAWA and O. MOLCHANOV (Ter- ra Scientific Publishing Co., Tokyo), 413-420.