Integrated Analysis of Multi-Parameter Precursors to the Fukushima Offshore Earthquake (M_j = 7.3) on 13 February 2021 and Lithosphere–Atmosphere–Ionosphere Coupling Channels

Abstract

The preparation phase of earthquakes (EQs) has been investigated by making full use of multi-parameter and multi-layer observations of EQ precursors, in order to better understand the lithosphere–atmosphere–ionosphere coupling (LAIC) process. For this purpose, we chose a specific target EQ, the huge EQ of Fukushima-ken-oki EQ on 13 February 2021 (magnitude M_j = 7.3). We initially reported on EQ precursors in different physical parameters not only of the lithosphere, but also of the atmosphere and ionosphere (Hayakawa et al. followed by Akhoondzadeh et al. and Draz et al., both based on satellite observations). Our first two papers dealt with seven electromagnetic precursors in the three layers (with emphasis on our own ground-based observations in the atmosphere and lower ionosphere), while the second paper dealt with Swarm satellite observations of magnetic field, electron density, and GPS TEC in the ionosphere, and the third paper dealt only with climatological parameters on and above the Earth’s surface (together with GPS TEC). We have extensively reviewed all of these results, and have coordinated the temporal evolutions of various physical parameters relevant to the LAIC system; we have sought to understand which hypothesis is more plausible in explaining the LAIC process. Then, we came to a conclusion that two possible LAIC channels seem to exist simultaneously for this EQ: a fast channel (nearly simultaneous responses on the ground and ionosphere), and a slow channel (or diffusion-type), with a time delay of a few to several days, in which the agent effects in the lithosphere and lowest atmosphere seem to propagate up to the ionosphere with a definite time delay. Finally, we have suggested some research directions for the future elucidation of LAIC channels, and also made some comments on an early EQ warning system.

1. Introduction

Short-term earthquake (EQ) prediction (with a lead time of about one week) is one of the most important topics left in the field of geoscience, but it is still far from realization [1,2,3,4,5,6,7,8,9,10,11,12,13,14,15]. However, there has been a huge amount of progress made in the study of seismo-electromagnetic phenomena during the last few decades, and it is recently becoming a consensus that electromagnetic (non-seismic) precursory anomalies do take place prior to an EQ [3,4,5,6,7,8,9,10,11,12,13,14,15]. That is, various precursory effects are known to occur not only in the lithosphere and the lowest atmosphere, but also in the upper atmosphere (ionosphere). In the lithosphere, the pressure accumulated during the EQ preparation phase leads to microfracturing, resulting in the generation of electric currents in the fault regions of the lithosphere (e.g., [16,17,18]) and the emanation of radioactive radon, etc., (e.g., [19]), with consequences over the near-Earth’s surface (such as temperature change (e.g., [20,21]), atmospheric meteorological parameter changes (e.g., [22,23]), SLHF (surface latent heat flux) (e.g., [24,25]), OLR (outgoing longwave radiation [26,27,28,29])). The upper atmosphere (ionosphere) is found to be most sensitive to pre-EQ effects, as evidenced by subionospheric VLF/LF propagation anomalies (e.g., [30,31,32,33,34,35,36,37,38,39,40,41,42,43,44]), by ionosondes and GPS TEC (total electron content) observations (e.g., [45,46,47,48,49]), and also by satellite observations (e.g., [50,51,52]). It is surprising that the upper-most layer of the ionosphere is very sensitive to pre-EQ lithospheric activity, and so three layers are highly coupled with each other. This leads to the generation of a new concept of lithosphere–atmosphere–ionosphere coupling (LAIC) (e.g., [3,4,5,8,9,10,11,14,15]). Our current concern is the elucidation of the mechanism (or physics) of the LAIC process. The monograph edited by Ouzounov et al. (2018) [10] has emphasized the importance of multi-parameter observations of different phenomena in different layers of the Earth (lithosphere, atmosphere, and ionosphere). Taking into account the importance of multi-parameter observations, which is recognized by many scientists but is a tough job because of difficulties of interdisciplinary coordination amongst multiple observations, there have been several papers published in this direction regarding various huge EQs [53,54,55,56,57,58,59,60,61,62,63,64,65,66,67,68,69]. Despite these difficulties, a few more papers dedicated to multi-parameter and multi-layer observations have been produced with some other case studies [70,71,72,73,74,75,76].

Here, we offer a brief description of the mechanism of the LAIC process. This topic has been investigated extensively for the last 10 years, and a few hypotheses have already been proposed (e.g., [4,5,7,9,10,13,14,77,78,79,80]). As is given in Hayakawa et al. (2004) [81], the first is the so-called chemical hypothesis, in which the emanation of radioactive radon, charged aerosols, and/or gases plays the main role, leading to the modification of atmospheric conductivity and the generation of an electric field, thereby driving the variation in ionospheric plasma density [4,9,11]. Additionally, air ionization in this hypothesis leads to the generation of thermal anomalies near the Earth’s surface as the consequence of different physical/chemical processes (changes in surface and air temperature, surface latent heat flux (SLHF), outgoing longwave radiation (OLR), etc.). The second is the acoustic hypothesis, in which atmospheric oscillations including atmospheric gravity waves (AGWs) and acoustic waves are excited by the precursory deformation of ground motion and/or gas emanation, or thermal irregularities, propagating upwards to the lower and upper ionosphere and leading to perturbations in the ionosphere [5,82,83,84,85,86,87,88,89,90,91]. The third is the electromagnetic hypothesis [81], in which electromagnetic waves generated in any frequency range (either in the lithosphere or in the atmosphere) propagate upwards into the ionosphere and magnetosphere, inducing particle precipitation into the upper atmosphere due to wave–particle interactions in the magnetosphere (e.g., [92,93,94,95]). Finally, a fourth electrostatic channel is proposed based on the laboratory experiments, in which positive holes are generated when the ground of interest is stressed by accumulated pressure [78,96]. These processes have been discussed extensively by different authors (e.g., Ouzounov et al. (Eds), 2018 [10]), but none of the above hypotheses have been evidenced by any definite observational data, necessitating further studies until the process of LAIC is well understood.

The present paper deals with the coordination of multi-parameter analyses for a specific huge EQ, Fukushima-ken-oki EQ, which happened offshore of the Tohoku area of Japan at 23 h 07 m on 13 February 2021 (LT) (or UT = 14 h 07 m on 13 February) with a magnitude of M_j = 7.3 (here, we use the magnitude given by the Japan Meteorological Agency, M_j), and with a depth of 55 km. Another EQ happened at a small distance from this EQ, but with a smaller M_j of 6.9, on 20 March 2021. Hayakawa et al. (2021) [66] performed a multi-parameter analysis for these two EQs, but some other groups paid attention only to the first stronger EQ, so we will restrict ourselves only to the first EQ in this paper.

The purpose of this paper is to integrate the multi-parameter observational results for this particular EQ. Hayakawa et al. (2021) [66] first published a paper on their observations of multi-parameters (a total of six physical parameters) by making full use of our own ground-based observations, especially observations of perturbations in the lower ionosphere, mesosphere, and in the stratosphere. Additionally, Akhoondzadeh et al. (2022) [97] presented additional results for the same EQ with data from Swarm satellite observation of electron density and magnetic field, mainly paying attention to medium-term EQ prediction (with a lead time of a few months). Furthermore, very recently, Draz et al. (2023) [23] presented observational results mainly on atmospheric parameters (sea surface temperature, air temperature, relative humidity, OLR) as well as conventional TEC for the same EQ. So, combining different precursor results for one particular EQ is an important idea, which will allow us to find and discuss many mechanisms, and we can promote a much better understanding of the physics of the LAIC process in the 2021 Fukushima-ken-oki EQ.

2. The EQ in This Paper

We observed a significant enhancement in seismic activity on the Pacific Ocean side of Japan, probably as aftershocks of the disastrous 2011 Tohoku EQ. There were two successive EQs in the Tohoku offshore area in February and March 2021.

The first EQ (called Fukushima-ken-oki) took place at 23 h 07 m (LT) on 13 February (or UT = 14 h 07 m on 13 February), with a magnitude of M_j = 7.3 and with a depth of 55 km. Its epicenter was at the geographic coordinates of 37.7° N, 141.7° E. The second EQ (called Miyagi-ken-oki EQ) happened at 18 h 09 m on 20 March 2021 LT (UT = 9 h 9 m on 20 March), with its epicenter located at 38.5° N, 141.6° E with a magnitude of M_j = 6.9 and a depth of 59 km. The former Fukushima-ken-oki EQ occurred as the result of thrust faulting near the subduction zone of the interface plate boundary between the Pacific and North America plates. This EQ event took place in the vicinity of the rupture area of the 2011 March 11 Tohoku mega-EQ, probably as an aftershock. We here select only the first EQ as our target for analysis, and this EQ is plotted in Figure 1, together with the Kakioka (KAK) ULF (ultra-low-frequency) station and two stations (Nakatsugawa (NAK) and Shinojima (SHI)) of the Chubu University ULF/ELF (extremely-low-frequency) network. However, two other papers, by Akhoondzadeh et al. (2022) [97] and Draz et al. (2023) [23], did not pay any attention to the second EQ, so we focus on the first larger-magnitude EQ in this paper.

3. Summary of EQ Precursors for the 2021 February Fukushima EQ

There have been four papers published dealing with the electromagnetic precursors of this 2021 Fukushima EQ. The first paper, by Hayakawa et al. (2021) [66], was based on multi-parameter precursor observations (mainly based on our own observations), which included lithospheric, atmospheric and ionospheric parameters for a total of six physical parameters. Our own paper was by Hayakawa et al. (2021) [98], which deals with anomalies in Schumann resonances (SRs) for this EQ, providing us with the information on the mesospheric and lower-ionospheric plasma disturbances. In the following paper by Akhoondzadeh et al. (2022) [97], they applied deep learning to the Swarm satellite data, such as magnetic field, electron density and electron temperature in the upper ionosphere, and TEC anomalies (mainly around the F-region). Finally, the most recent paper was published by Draz et al. (2023) [23], who mainly studied the atmospheric parameters (together with TEC anomalies), again with the use of deep learning, but with the main emphasis on the Earth’s surface physical parameters, such as sea surface temperature (SST), air temperature (AT), relative humidity (RH), and outgoing longwave radiation (OLR). In the following, we will summarize the results obtained from these four papers, but we will start with a discussion of the bottom layer of the lithosphere, then work up to the upper ionosphere.

3.1. Lithospheric Effect

The main source of the LAIC process is definitely located in the bottom layer of the lithosphere, so we will review first the pre-EQ phenomena taking place within the lithosphere. There are two main tools used to investigate the possible changes taking place in the lithosphere: (a) SES (seismic electric signal) (Varotsos, 2015 [99]) and (b) lithospheric ULF radiation (e.g., a review by Hayakawa et al. (2023) [18]). Unfortunately, there are no available observations of SES (or geoelectric currents) in the area of our concern because of the presence of DC railways, so we have to rely on the results from the ULF observation at Kakioka (geographic coordinates: 36.23° N, 140.18° E) (Figure 1) (see Hayakawa et al., 2021 [66]). Due to the self-organization of the lithosphere, we can expect to see the generation of electric currents through different generation mechanisms [66]. Our ULF data based on polarization analyses [18] indicate that we detected no clear signature of seismogenic ULF emissions at Kakioka for this particular EQ (this is indicated as “no emission (KA)” in Figure 2), but this is probably due to the large epicentral distance of ~250 km, if we use the empirical formula for the threshold distance of detectability used in Hayakawa et al. (2023) [18]. So, no definite conclusion can be reached on the presence of seismogenic ULF emissions near the EQ epicenter.

3.2. Atmospheric Effects

We here define atmospheric effects as those covering the Earth’s surface up to the stratosphere and the lowest ionosphere. So, we review the physical parameters by starting from the bottom and working up to the upper region. The intensity of OLR (outgoing longwave radiation) is attributed to ionized air molecules resulting from the emanation of radioactive radon moving up and attracting wave vapor and releasing latent heat, contributing to an anomalous increase in OLR flux and to some consequences of changes in meteorological parameters on the Earth’s surface (temperature, humidity, pressure etc.) [22,23]. This may be closely related to the chemical hypothesis of the LAIC process in the previous section.

3.2.1. Earth’s Surface Parameters (Climatological Parameters)

The following results derived from monitoring multiple parameters on the Earth’s surface and the surrounding lowest atmospheric region are based on the paper by Draz et al. (2023) [23]. Its main emphasis is on different kinds of Earth-surface physical parameters, including sea surface temperature (SST), air temperature (AT), relative humidity (RH), and outgoing longwave radiation (OLR), but with the use of a new approach of deep machine learning (artificial neural network (ANN), nonlinear autoregressive network with exogeneous inputs (NARX), and long short-term memory (LSTM)) in order to find possible precursors within a seismogenic region (35.6–39.5° N, 139.4–143.9° E) on the basis of the data of AIRS/Aqua, NOAA, etc. The results derived with the application of AI (artificial intelligence) are quite similar to those derived by more conventional statistical methods, re-confirming the statistical analyses’ results, so we only show the results derived by machine learning.

(a)

RH

The results on climatological parameters are first presented, including RH, AT and SST, followed by OLR. The RH daytime value showed a negative anomaly (decrease in RH) 6 days before the EQ, and the RH nighttime value showed a negative anomaly 5 days before the EQ. These detected clear “negative” anomalous signatures 5–6 days preceding the EQ. When we observe an anomaly on a particular day, a black triangle is plotted on that day in the summary plot of Figure 2. D and N mean that it was observed at local nighttime and daytime, respectively. Any considerable drop in RH is due to air ionization and thermal energy absorption, which change humidity and air temperature in the lowest atmosphere.

• (b) AT

Deep machine learning analyses indicate that AT values showed clear anomalies (positive) (i.e., increase in AT) 5 days before the EQ. This result seems to be consistent with previous results [24,25,100] suggesting that surface temperature increases about one week before an EQ. The anomalies are indicated by triangles in Figure 2.

• (c) SST

The data of SST show clear negative anomalies 7 and 6 days before the EQ. This result on the SST is also consistent with those from early works [24,25,100], and the anomalies are also plotted by triangles in Figure 2.

• (d) OLR

OLR is one of the most vital parameters used to define the Earth’s radiation emitted from an area of interest; it is a combination of emissions from clouds, the lower atmosphere, and the ground, and is used to study the Earth’s radiation climate (e.g., [26,27,28]).

The machine learning results indicate the detection of a positive anomaly (or an increase) in OLR observed 5 days before the EQ. Previous works [26] have indicated that the lead time of OLR is extremely variable, ranging from a few days to a month, but the anomalies for this EQ are found 5 days before the EQ (rather small lead time). The anomalies are plotted in the same format as before in Figure 2.

3.2.2. Stratospheric Effect

The region of the stratosphere up to the lowest ionosphere can be monitored via a few methods: (1) Stratospheric AGW (atmospheric gravity waves) activity [87,88] and (2) Schumann resonance (SR) anomalies (see our review paper by Hayakawa et al. (2023) [18]). The stratospheric AGW activity has been studied using ERA5 data (as in [87,88]), and in the paper by Hayakawa et al. (2021) [66], we found that AGW activity was continuously enhanced at the altitude of ~22 km 6 days to 2 days before the EQ in Figure 2 (active days are marked with bigger triangles). This analysis is based on the use of electric potential energy, with temperature height profiles based on ERA5 reanalysis data sets from the European Center for Medium-Range Weather Forecasts. Another useful approach to monitoring the mesosphere and lower ionosphere is the modification of SR data. As is summarized in a monograph by Nickolaenko and Hayakawa (2014) [101] and in Hayakawa et al. (2023) [18], SR is known as the global electromagnetic resonance produced by ELF radiation coming from the three lightning chimneys in the world [102]. Hayakawa et al. (2020, 2023) [18,103] summarized the seismogenic anomalies in this SR phenomenon, and [98] found anomalies only around the EQ date in terms of an enhancement in SR intensity at the fundamental frequency of 8 Hz and its harmonics, as observed at Nakatsugawa near Nagoya in Japan. This anomaly can be interpreted in terms of any changes in the electron density profile in the mesosphere and lower ionosphere [18,103].

3.2.3. Atmospheric Effects

The atmospheric effect (of the middle atmosphere) can be studied with the atmospheric ULF/ELF radio emissions [104,105,106], which are impulsive electromagnetic emissions (just like lightning discharges) in the frequency range of ULF/ELF bands (such as f = 1~20 Hz) in possible association with EQs. Hayakawa et al. (2021) [66] succeeded in detecting atmospheric ULF/ELF radiation at a remote station (from the EQ epicenter) of Nakatsugawa near Nagoya (see Figure 1) 6-1days before the EQ, indicating the generation of seismogenic ULF/ELF radiation as a pre-EQ effect of this EQ. The presentation of this anomaly is exactly the same as before, shown in Figure 2, but larger triangles mean stronger intensity. However, the relevant generation mechanism is the least well understood [18].

3.3. Ionospheric Effects

3.3.1. Lower Ionospheric Effects

With the use of subionospheric VLF (very-low-frequency)/LF (low-frequency) propagation data, Hayakawa et al. (2021) [66] studied the lower ionospheric perturbations in possible association with this EQ. The VLF/LF receiving station is located in Kamchatka, Russia (PTK), and the transmitters are all Japanese—JJY (40 kHz, Fukushima) and JJI (22.2 kHz, Ebino, Miyazaki, Kyushu). Notably, the propagation path of JJY-PTK exhibited propagation anomalies exceeding -2 σ level (σ, standard deviation; amplitude depletion) on 2 (11 days before the EQ), 6–7 (7-6 days before the EQ), 13 (day of EQ), and 18 February (5 days after the EQ). These anomalies are plotted in the same way as before in Figure 2.

There is another effective tool used to study lower ionospheric seismogenic perturbations: the depression of horizontal components of ULF waves observed on the ground. Its physical background is that the horizontal components of ULF pulsations observed on the ground at night (e.g., Pi2, Pc 4, etc., See Saito (1969) [107]) can be reduced in intensity when passing through the disturbed lower ionosphere (i.e., enhanced absorption). As has already been discussed before, Kakioka station is located at a far distance of ~250 km (see Figure 1), so it struggles to detect seismogenic ULF radiation. However, as is discussed in Molchanov et al. (2003, 2004) [108,109], Schekotov et al. (2013) [110] and Hayakawa et al. (2023) [23], the coverage area for this ULF depression effect is much larger than that for lithospheric radiation. So, a clear ULF depression effect was observed only on a particular day, two days before the EQ, as indicated in Figure 2.

3.3.2. Upper Ionosphere (F-Region and Above)

Hayakawa et al. (2021) [66] did not analyze the foF2 values at Kokubunji, Tokyo, because this ionosonde station seems to be located far away from the EQ epicenter, as indicated by an empirical formula of detectability (the foF2 anomaly set out by Liu (2009)) [111].

The TEC data retrieved from the two stations of USUD (Usuda) in Nagano and MTKA (Mitaka) in Tokyo, within the breeding zone of the EQ, have shown clear seismo-ionospheric anomalies [23]. Prominent positive deviations were observed 6 and 5 days before the EQ (which is summarized in Figure 2) and 7~9 days after the EQ. The EQ precursors claimed by Draz et al. (2023) [23] were also detected in the TEC study in [66], but we considered that these might be related to a small geomagnetic storm on 7 February.

Then, we move on to the observational results given by Akhoondzadeh et al. (2022) [97] based on the analyses of in situ ionospheric physical parameters, such as anomalies in magnetic field, electron density, electron temperature, and TEC above the satellite height, as observed by Swarm satellites at the height of ~500 km. They made use of deep learning as a powerful tool to extract nonlinear patterns from 52 time series of ionospheric precursors. They were interested in not only short-term, but also medium-term precursors, so their analysis period was rather long, ranging from a few months before the EQ to a few weeks after the EQ. However, we are interested only in short-term EQ precursors, so we limit Figure 2 to the period of 2 weeks before to 1 week after the EQ. Comparing the results given by the two implemented methods of median and LSTM (long short-term memory), they found by a voting classification method that clear anomalies could be observed 1, 6, 8, 13, 31 and 32 days before the EQ. These results are also summarized in Figure 2, though days 31 and 32 are outside of our period of interest.

4. Summary and Discussions

4.1. Summary of Different EQ Precursors

Figure 2 shows a summary of the integration of observational results derived from the four papers mentioned in the previous section on the February 2021 Tohoku EQ. At the top of the figure, we plot again the geomagnetic activity (Dst index) (as an important indicator of space weather parameters). Then, the bottom abscissa indicates the date (together with lead time), with a vertical red line representing the EQ date (13 February), and we pay attention only to about two weeks before and one week after the EQ, because we are interested in short-term EQ prediction in order to study the physical parameters arising during the EQ preparation phase. From the bottom and working up to above the abscissa, we refer to lithospheric ULF radiation, various meteorological parameters (RH, AT, SST) on the Earth’s surface, OLR as the lowest atmospheric parameter, stratospheric AGW wave activity, and then ULF/ELF atmospheric radiation and SR in the stratosphere. Further, we move on to the lowest ionospheric disturbances as detected by two independent methods (subionospheric VLF/LF propagation anomaly and ULF depression effect), and the upper ionosphere (around F-region) as VTEC. Finally, we move on to the satellite results for the upper ionosphere, taken at the height of 500 km.

In the paper by Akhoondzadez et al. (2022) [97], they have noted anomalies in the ionospheric parameters 31 and 32 days before the EQ, which seems to be the longest lead time of short-term EQ prediction, of about one month. They have mentioned that around this time, an Mw5.9 EQ happened in the same area as studied by Doborovsky et al. (1978) [112], but its depth was more than 200 km, and its not-so-high EQ magnitude makes them confident in excluding a co-seismic ionospheric disturbance as the fact causing this EQ event, considering instead that the anomalies recorded in this time are more likely to be a possible precursor to our EQ from February 13, 2021. Furthermore, this anticipation time is very close to the ones relating to an increase in electron density before the Mw7.5 Indonesia 2018 (40 days) and the Mw7.1 California Ridgecrest 2019 (33 days) EQs [54,65]. It is interesting to notice that these events have EQ magnitudes comparable with the Japan EQ of interest here, although the Ridgecrest EQ occurred in a different tectonic region. Therefore, it would be compelling in future studies to verify further whether the ionosphere could respond 30–40 days before an EQ with this range of magnitude. This observation would confirm the validity of Rikitake’s law (Rikitake, 1987 [113]), which was also recently established for satellites by De Santis et al. (2019) [51].

Here, we comment on Rikitake’s paper, which is based on seismological measurements, so the lead time of the medium-term precursors used here is generally large, and his formula on the relationship between lead time and EQ magnitude must be useful for medium-term EQ prediction, as recently used by De Santis et al. (2019) [51] and his colleagues. However, Rikitake also mentioned that the lead time of some electromagnetic precursors as studied in this paper is substantially smaller than that for medium-term precursors.

4.2. Coordination of Different Precursors and LAIC Channels

Now we are ready to coordinate the temporal evolutions of various short-term disturbances in different layers. First, we mention that the altitude (500–600 km) of satellites like DEMETER and Swarm satellites is too high to identify the important consequences of this LAIC process, and it seems that on many occasions, unexpected changes in geophysical quantities have been detected on satellites, but these could not be linked with certainty to a given EQ, because essential processes are taking place below the F-region, thereby the information on the lower ionosphere and upper atmosphere (stratosphere) is of greatest value in the study of LAIC process. However, there are only very few papers containing information in those regions [62,66,67], which pay a lot of attention to the intermediate region between the F-region and lithosphere, such as the lower ionosphere and the stratosphere, as an essential bridge between the F-region (and above) and the bottom layer of the lithosphere or the Earth’s surface parameters. Most papers are based on open data that are publicly available, and the data from Swarm satellites and satellite observations of the Earth’s surface parameters (such as meteorological/climatological parameters) are used for LAIC studies (e.g., Haider et al., 2024 [114]). As such, it is needless to say that it is extremely difficult for them to think about or infer any definite causality relationship between the lithospheric or Earth’s surface atmospheric anomalies and anomalies at the satellite altitude, due to an inevitable dearth of information on the intermediate region. The GPS TEC has also been utilized as an indicator of ionospheric F-region perturbations instead of Swarm satellite data, but with the same difficulty.

Here, we look at the summary plot of Figure 2. To derive a total, we accumulated 13 physical parameters. When using multiparameter observations, the most important finding is that all of the electromagnetic anomalies in different layers, such as atmospheric, stratospheric, and lower and upper ionospheric regions, are found to be concentrated in a time window from only about a week before the EQ to the day of the EQ (or even after the day of the EQ (probably an after-effect)). Actually, there were no anomalies at all in January or during the period between our February EQ and the subsequent March 13 EQ (except a few after-effects showing in Figure 2). Hence, these anomalies observed in Figure 2 are almost certainly closely associated with the February Tohoku EQ, mainly as pre-EQ activity. The most serious point when discussing the precursor studies is to pay the greatest attention to the reliability of each anomaly. Most notably, the effects of geomagnetic disturbances and meteorological perturbations can easily interfere with the detection of seismogenic anomalies, even though the general geomagnetic condition during our study period is rather geomagnetically quiet, with the maximum Dst = −50 nT on 7 February. However, recently, some authors [115,116] proposed an innovative (but not well recognized) hypothesis that ULF radiation, thought so far to be an EQ precursor, is triggered by solar activity, just like geomagnetic storms, in which the two phenomena of EQ ULF radiation in the lithosphere and geomagnetic storms are regarded as brothers. This will constitute some of our future work. During the period of interest, there are clear and definite precursors to the EQ. Based on our previous extensive and statistical studies (e.g., [82]), we think that the most convincing precursors are atmospheric ULF/ELF emissions, which are identified to start about 1 week before the EQ and continue up to the day of the EQ. This is considered to be a very regular precursor, and a promising candidate for short-term EQ prediction (see [105]); it was recently utilized in Kamchatka to develop a prospective EQ forecast (Schekotov et al. [101,102,103]). The next most reliable precursor is lower ionospheric perturbations, as detected by ULF depression [18,117] on 11 February (2 days before the EQ), which means that the lower ionosphere must be perturbed around this date. The last convincing precursor is the VLF/LF perturbation (lower ionospheric perturbation) [30,31,32,33,34,35,36,37,38,39,40] that occurred on the day of the first EQ as an imminent precursor, and on some other days. Unfortunately, the propagation path we used is too long, from Japan to Kamchatka, so the VLF anomaly might be contaminated by the geomagnetic storm that occurred on 7 February, resulting in a perturbation in the high-latitude part of the propagation path.

Even though we do not have any uncertainty in the reliability of other anomalies used by other researchers, shown in Figure 2, we can note two likely chains (or cause-and-effect relationships) in the phenomena of the LAIC process during the February EQ: The first is a “fast” channel, acting in such a way that the F-region anomalies (up to the satellite altitude) seem to be coincident in time with the climatological anomalies at 6 and 5 days before the EQ. The other is a slow, “diffusion-type” channel, whereby the effects of the agent occurring 6–5 days before the EQ in the lower altitude seem to propagate progressively upwards to the lower ionosphere as a chain, with some delay of a few days.

The fast channel is characterized by the simultaneous occurrence of the ionospheric perturbation and the anomalies in the Earth’s surface parameters (or climatological parameters, and/or SLHF, OLR, etc.) on the same days, which have often been observed for many EQs, including the 2023 Mw7.8 Turkey EQ [72,114] (Mw: moment magnitude), the 2020 Mw6.5 Monte Cristo range EQ [118], the 2019 Kermadec EQ [68], and the 2022 Luding (China) Ms6.8 EQ [70] (Ms: surface wave moment). Also, some other EQ events such as the 2015 Mw7.8 and Mw7.3 Gorkha-Nepal EQs [60], the 2021 Mw7.2 Haiti EQ [119], and the 2021 November Fukushima EQ (M_j = 7.2) [120] exhibited a nearly fast channel, because different anomalies between the ionosphere, atmosphere and lithosphere took place nearly simultaneously within the one or two days difference (these are very different from other previous case studies because we use the lithospheric ULF radiation and ionospheric perturbation, as detected by the ULF depression effect). On the other hand, a second slow channel was found for the 2017 Mw7.3 Iran EQ by Akhoondzadeh et al. (2019) [59] using Swarm satellite data, GPS TEC and climatological data, and by Yang et al. (2020) [87] for the 2016 Kumamoto EQ based on surface deformation data, AGW data and lower ionospheric data. Further, Sasmal et al. (2021) [62] performed extensive studies on the coupling process for the 2022 Samos (Greece) EQ (Mw = 6.9), and they found a diffusion-type channel. The co-existence of a few possible coupling channels as found in this paper had already been found for the Lushan (China) EQ, but unfortunately only for medium-term EQ precursors (Y. Zhang et al., 2023 [71]).

4.3. Fast and Slow LAIC Channels

This fast coupling would simply imply a pure electromagnetic phenomenon or pre-EQ acoustic waves, such as was proposed by Kuo et al. (2014) [121], though this has been seriously criticized by Prokhorov and Zolotov (2017) [122] due to the unrealistic assumptions. This criticism was also supported by Sorokin et al. (2020) [83] and Surkov et al. (2023) [123], so future studies and investigations will be necessary to better understand this phenomenon even for the fast channel. Even for the fast channel, it seems difficult for us to specify which hypothesized mode of LAIC is in operation; chemical, acoustic, electromagnetic, or electrostatic. The generation of an electric field in the lower atmosphere (either by radon [58] or positive holes [4,78]) may be possible in the fast channel, though there are many papers arguing against the penetration of electric fields into the ionosphere. Pre-EQ acoustic waves are another possibility in the acoustic channel. However, there is no evidence in support of any hypothesis, because we have a limited number of observables; in particular, no information is available from the middle atmosphere and lower ionosphere.

We here turn to observations of the slow, or diffusion-type, channel. The observed time delay of a few days between the Earth’s surface and the ionosphere may be indicative of the propagating nature of the AGW hypothesis, because we can expect some degree of propagation delay under this hypothesis. However, the expected time delay is not as large as a few days (e.g., Lizunov et al. (2020) [86]). Further, the mode of excitation of AGWs is still unclear, as there are a few possible modes of AGW modulation, such as the direct deformation of the Earth’s surface [87], the presence of an AGW resonator in the lithosphere (Chen et al., 2020 [124,125]), or local disturbances in the lithospheric conductivity caused by modulations due to the emanation of charged gases in the EQ preparation phase. Hence, the diffusion-type channel is considered a kind of complicated evolution of different chemical/physical processes, comprising a combination of a few LAIC mechanisms between the lithosphere, atmosphere, and ionosphere.

4.4. Importance of Physical Parameters in the Intermediate Region Bridging between the Earth’s Surface and Ionosphere

Since various kinds of open data are available publicly, more papers will be published in the future on multi-parameter and multi-layer observations based on a combination of Swarm satellite data and climatological data, or a combination of ground-based GPS TEC and climatological data, since there will be many EQs around the globe. However, we think that there will be an essential and inevitable “limitation” in these studies, because the causality relationship between the lithospheric effects and TEC perturbations remains unclear due to a gap in the data from the lithosphere and the ionospheric F-region. So, in order to address this issue and to make a definite step forward, we propose that the coordination of such works with specific ground-based geophysical stations in seismic-prone regions. During the Japanese frontier project (1996–2001), we established such a geophysical station at Kamchatka, Russia, as part of a collaborative work with Russian scientists (Uyeda et al., 2002 [126]), in which different kinds of measurements were installed (such as ULF magnetic field, subionospheric VLF/LF observation, etc.). Recently, Chinese colleagues have established an effective complex station for monitoring ground vibrations and related perturbations in the lithosphere, atmosphere, and ionosphere (called MVP-LAI) in the countryside of the seismically active Sichuan province of China, and it is now in operation (Chen et al., 2021 [127]). We think that they will coordinate their ground-based observations of 14 geophysical parameters with measurements taken by the Chinese satellite (CSES). At least a few such stations are hopefully soon to be established around the world.

4.5. Possible Future Direction of LAIC Studies

Here, we want to make a few more comments on future directions. For the full use of a combination of measurements of surface parameters taken by satellites (NOAA data etc.) and ionospheric parameters taken either by Swarm satellite or GPS TEC, we would like to emphasize the importance of physical parameters bridging between the F-region or above and the Earth’s surface, such as the VLF/LF subionospheric probing of the lower ionospheric propagation perturbations, stratospheric AGW activity in the stratosphere, and atmospheric ULF/ELF electromagnetic radiation, as studied in the papers by Hayakawa et al. [66,67]. Another important direction of LAIC studies will involve an investigation dedicated to any particular hypothesis of the LAIC process. For example, Muhammad et al. (2023) [128] attempted to prove the chemical hypothesis by comparing the ground-based radon observation with the GPS TEC data. Their results suggest that a correlation between radon–TEC changes and EQs does not exist, but radon anomalies are followed by TEC anomalies relative to an EQ. Further, Yang et al. (2020) [87] have paid particular attention to the acoustic channel of the LAIC process by combining AGW wave activity with information on surface deformation collected by GPS. Further, there has been a lot of indirect evidence available on the presence of AGW modulation in VLF ionospheric perturbation (Rozhnoi et al. [33,34,129]), and further support has been provided by recent studies on stratospheric AGW activity as correlated with VLF seismological perturbation [33,34], and by the direct evidence given by VLF/LF Doppler shift observations (Hayakawa et al., 2012 [130]).

Together with further studies of the mechanisms or channels of LAIC, it is highly necessary to work on an early EQ warning system based on various geophysical parameters related to the different layers. Satellite data are a very promising and useful tool that can be used to study various types of possible pre-EQ anomalies due to their wide global coverage and timeliness; further, the possible application of AI (artificial intelligence) to those satellite data (preferably together with extensive ground-based observational data) is of essential interest to the realization of such a short-term EQ warning system, because of the recent developments in machine learning (e.g., [131,132,133,134,135,136,137,138,139,140,141]) and fuzzy inference systems (Akhoondzadeh and Marchetti, 2024 [72]), and so on.

5. Conclusions and Suggestions

In this paper, we have focused on the coordination of multiple observations for a particular EQ (the 2021 February Tohoku EQ) with M_j = 7.3 in order to better study the LAIC channels. We list the important findings and suggestions as follows:

• By paying the greatest attention to previous papers [66,98] on ground-based observations, including atmospheric ULF/ELF radiation, subionospheric VLF/LF propagation anomalies, the ULF depression effect, etc., we have tried to coordinate our results with those given in two recent papers by other researchers (one contains Swarm satellite observational results [97] and the other observational results on the Earth’s surface parameters, as well as those form the lower atmosphere (temperature etc. and OLR) [23]). As a result, all of these 13 physical parameters have been coordinated in this paper for the study of LAIC channels. First, we have found that anomalies in many parameters are predominantly concentrated in a time window from about one week before the EQ to the day of the EQ (and even after the EQ), which are highly likely to be short-term EQ precursors that can be used during the EQ preparation phase;
• Based on the comparison of temporal evolutions, there seem to exist two possible LAIC channels. One is a “fast” channel, in which anomalies in the Earth’s surface parameters (RH, AT, SST) and ionospheric perturbations happen on the same day. The second is a slow, or “diffusion-like” channel, in which the effects in the lithosphere (and Earth’s surface parameters), including any modulations in ground deformation, radioactive radon emanation, etc., tend to propagate upwards into the ionosphere with a definite time delay of a few days;
• We have emphasized the importance of observations of perturbations in the middle atmosphere and lower ionosphere (as observed by subionospheric VLF/LF propagation data), because the information for these regions was notably absent in most previous papers comparing ionospheric perturbations at the satellite altitude and/or the F-region with climatological anomalies on the Earth’s surface;
• A few possible research directions have been suggested. Studies dedicated to a specific LAIC hypothesis are highly required, and also, we propose the application of AI to satellite data for the realization of early EQ prediction.