There have been found different kinds of electromagnetic precursors of earthquakes (EQs), and the presence of those seismo-electromagnetic precursors is recently considered to be an irrefutable fact, though the physics of these phenomena remains poorly understood. (e.g., Molchanov and Hayakawa (2008) [1], Hayakawa (Ed.) (2009, 2012) [2,3], Uyeda et al. (2009) [4], and Hayakawa and Hobara (2010) [5]). Such electromagnetic EQ precursors can be customarily classified into the two types. The first type is the direct effects such as electromagnetic radiation from the lithosphere in a wide frequency range, while the second is the indirect effect such as seismoatmospheric and ionospheric perturbations as detected by means of propagation anomalies of transmitter signals in different frequency ranges. When utilizing any one of those precursors for the practical short-term EQ prediction, the most important issue is whether the precursor is statistically correlated with EQs or not, which can be only possible by analyzing a huge number of events based on the long-term observation.

Among many EQ precursors [1-5], the most promising one from the statistical point of view is the ionospheric perturbation, because the perturbations both in the lower ionosphere and upper ionosphere (F region) are found to be statistically significantly correlated with EQs. Hayakawa et al. (2010) [6,7] have obtained the significant statistical correlation between subionospheric VLF/LF propagation anomalies and EQs as for the lower ionospheric perturbation. On the other hand, Liu (2009) [8] has presented the result on the statistical correlation of the foF2 in the upper ionosphere with EQs. This paper deals with the lower ionospheric perturbations detected by subionospheric VLF/LF propagation anomalies. Hayakawa et al. (2010) [6] have found that a significant correlation is established for an EQ with significant magnitude (M) ˃ 5.5 [9,10] and shallow depth (D) (D less than 40 - 50 km). These results are based on the observation in and around Japan (that is, mainly inland (or fault-type) EQs). So that, the influence of D is rather evident in generating seismo-ionospheric perturbations for EQs in Japan.

We would like to explore whether this conclusion is still valid for EQs in other geological areas or at different latitudes. By using the VLF/LF subionospheric propagation data from the JJY (Fukushima, Japan) to Petropavlovsk-Kamchatsky (PTK), Russia, we will study, in this paper, the relative importance of M and D of an EQ in generating seismo-ionospheric disturbances over this midlatitude propagation path in order to explore whether there exists a significant difference in the characteristics of seismo-ionospheric perturbations of Japanese EQs and sea EQs. Finally, we discuss this difference in the context of the lithosphere-ionosphere coupling mechanism.

In order to study the relative importance of M and D of an EQ in generating the seismo-ionospheric perturbation at middle latitudes, we will analyze the dependence of amplitude of the LF signal. For the analysis, we use the LF signal recorded at Petropavlovsk-Kamchatsky (PTK) station (geographic coordinates; 53˚09'N, 158˚55'E) in Russia from a Japanese transmitter with call sign of JJY (40 kHz) located at Fukushima [11]. The positions of the transmitter and receiving stations are illustrated in Figure 1 and the distance between the transmitter and receiver is 2300 km. The data from three years (2005, 2006 and 2007) are used.

As a main characteristic of the LF signal, we estimate the difference between the current nighttime amplitude

and the monthly averaged amplitude calculated over night (this is called the conventional nighttime fluctuation method [11,10,6,7]).

The epicenters of EQs with M ≥ 4 from the catalogue of USGS (United States Geological Survey) are shown in Figure 1. We select the EQs with D less than 100 km and with epicenters close to the great-circle path between the transmitter and the receiver. One circle indicates an EQ, with its size reflecting the EQ M. As seen from this plot, many EQs are found to be located in the offsea of Hokkaido, Kurile islands and Kamchatka. At first we select the area around the wave path of JJY-PTK (like wave sensitive Fresnel zone), which is indicated by a black line in addition to the great-circle path of JJY-PTK in Figure 1. Then for every EQs in this area we calculate the radius of a zone for which the ionospheric precursor of an EQ may be found. According to Dobrovsky et al. (1979) [12], the preparation radius is given by R = 10^{0.43 M}. We also compute the distance (L) from the EQ epicenter to the great-circle path. In the following analysis we have included only EQs satisfying tentatively the ratio R/L ≥ 0.7. When we have several EQs on one particular day, we have selected the largest M EQ.

Figure 2 illustrates the distributions of EQs as a function of EQ M. The upper panel represents the total number of EQs on the basis of selection depending on D and distance from the great-circle path (R/L ≥ 0.7). The bottom panel illustrates the number of EQs after the further selection of the largest M on one day. In the latter case (bottom panel in Figure 2), the number of EQs with 4.5 ≤ M ≤ 5.5 is, on the contrary, more than that of EQs with 4.0 ≤ M ≤ 4.5. After this selection the total number of EQs is 543.