Original Paper (with figures)

The VAN Method of Short-Term Earthquake Prediction

by Seiya UYEDA
Research Leader, RIKEN International Frontier Program on Earthquake Research, Professor of Tokai University

In Greece, short-term earthquake prediction has been in practice for more than a decade. The method, called the "VAN", is based on monitoring of geoelectric potential changes. @Incorporating the VAN with other promising methods of different disciplines, it seems possible not only to improve earthquake prediction but also to open new frontiers of earth science leading to better understanding of fundamental crustal processes.

What is VAN Method?
Earthquake prediction aims to specify three elements, namely when, where, and how large the impending earthquake will be. It is generally recognized that such is hard to achieve with the capability of the present seismological science. It may, therefore, sound unbelievable that earthquake prediction has been in practice in Greece for more than a decade.

The method is called the VAN method, after the initials of three Greek scientists, Professors P. Varotsos, K. Alexopoulos and K.Nomicos, all from the University of Athens. Soon after an earthquake hit the Athens area in early 1981, causing serious damage, they started monitoring the geoelectric potential changes, because, as solid state physicists, they anticipated that some electric current may be generated in the earthquake source region before catastrophe.

The measurement is relatively simple in principle. One measures the geoelectric potential differences between electrodes stack into the ground at desired distances. The geoelectric potential is constantly changing, due to causes such as geomagnetic field variations, rain fall, man-made noise, electro-chemical instabilities of electrodes and so on. One has to distinguish meaningful pre-seismic signals, if they exist at all, from these noise. This is a formidable task, which none of the previous researchers was able to achieve. Strictly following the laws of basic physics and through almost unbelievable efforts, the VAN-group accomplished this task and discovered that indeed precursory seismic electric signals ("SES") did exist, only if one chooses right spots ("sensitive stations"). Even at sensitive stations, only less than 0.1% of observed changes were meaningful signals, i. e., the SES, all others being noise. Based on SES observation, the VAN group specifies the three elements of earthquake prediction, accuracy of which depends on how well the stations are "calibrated" largely by experience. Depending on the types of SES, earthquakes occur within several days to several weeks. The location of the focal area can be predicted by a very peculiar nature of SES, called the "selectivity", i.e., a "sensitive station" is sensitive to SES from specific focal area(s) only. The magnitude of the earthquake is estimated through the amplitude of the SES, taking the distance between the focal area and the station into account.

Is VAN Method really successful; pros and cons?
They score their prediction "successful" when the actual earthquake occurred within the above stated time interval, within ca. 100 km from the predicted epicenter and within ca. 0.7 unit of the predicted magnitude. With these criteria, about 60% of their predictions are successful and about 60% of M>5.3 Greek earthquakes are successfully predicted.

This is truly astonishing so that their success has not readily been recognized by the scientific community, including the Greek seismological community. Some argue that VAN's SES are all noise unrelated earthquakes, and others insist that the apparent success was simply by chance and the VAN method has no physical background.

As stated above, the first step of VAN method was to distinguish SES from multitude of noise. In order to achieve this purpose, VAN group installs, at each station, many pairs of electrodes with different separations of the order of 100 - 200 m in both north-south and east-west directions ("short dipoles") and a few much longer dipoles of the order of several to a few tens of km ("long dipoles"). Noise due to geomagnetic variations can be easily recognized because they appear at all the stations simultaneously. Noise due to instabilities of electrodes and nearby sources are eliminated because such changes do not show signal intensity proportional to the length of short dipoles. Noise from the more distant sources can also be distinguished by the use of long dipoles. Experience has demonstrated that the changes survived this process of noise elimination are almost always SES, namely they are followed by earthquakes. Dr. S. Gruszow and others, based on their simultaneous observation at a station near the Ioannina VAN station, claimed that the SES for the MB(NEIS):6.1 1995 Kozani-Grevena Earthquake could be attributed to an industrial noise the origin of which , however , was not determined. However, their arguments do not seem to hold as the VAN group explained in detail (see the last reference in the list).

As to the statistical significance of their success, most tests concluded in favor of VAN. A few critics, like Dr. Robert Geller of the University of Tokyo, score VANÕs success rate very low and argue that such a low score can be attained by chance. However, careful readers will find that most of such criticisms were based on either too stringent conditions or erroneous interpretations. It seems that for them, an acceptable earthquake prediction system has to be rigidly formulated like a game of cards. Earthquake prediction, like any other science in its making, is in an inherently imperfect state. The vital issue here is whether or not there is some physical reality in VAN's SES as a precursor to be further explored, and not to scrutinize if the present VAN system, admittedly far from perfection, satisfies the framework of formal evaluation. It looks to the present author that these critics miss the point by worrying more about formalism than science. Moreover, sometimes their reasoning seems to be almost pathologically dictated by preconceived unproven dogma that earthquakes are unpredictable. Bad science will not stand the test of time anyway. So why a bud of potentially viable new science has to be nipped off because it is not yet perfect?

Here a couple of our evaluations on the VAN results are presented: namely the degree of success for large earthquakes (MB†5.5), and the magnitude dependence of their alarm rate. Fig.1 shows most of large events in the Greek territory were "successfully" predicted. Fig. 2 shows the total number of earthquakes, "successfully" predicted earthquakes and their ratios (alarm rate) for the MB threshold values on the abscissa. It is clear that the alarm rate increased dramatically, ruling out the possibility of random prediction.
Admittedly, physics of SES has not been completely clarified yet. But serious investigation is underway. In fact, there have been several possible theoretical models and laboratory studies on the generation of SES. The mechanism to bring about the "selectivity" is ascribed to electrical heterogeneity of the earth's crust, that has never been recognized before. Verification of these possibilities will be a truly challenging task in geophysical research in the years to come.

Emerging Interests in VAN
The situation has been gradually, and now rather rapidly, changing. In 1992, American experts on the subject gathered at Lake Arrowhead, California, for a NSF supported Workshop entitled "Low-Frequency Electrical Precursors to Earthquakes: Fact or Fiction?", inviting the VAN- group. After discussing with them, initially skeptic experts came to conclude that "the seismic electric signals observed in Greece are generated in the Earth, and the apparent correlation with earthquakes is promising". On March 5 and 26, 1993, two strong earthquakes hit Pirgos City, Greece. Despite heavy damage of the City, there was practically no loss of life because citizens were prepared thanks to VAN prediction. Such a dramatic success happened several times since 1988. ICSU's Scientific Committee for IDNDR (Sir James Lighthill: chair), became interested in the VAN-method and held, jointly with the Royal Society of London, a special meeting "A Critical Review on VAN" in 1995. Moreover, American Geophysical UnionÕs Geophysical Research Letters published a special issue "Debate on the VAN" in 1996 (Dr. Robert Geller, ed.). Genuine interest on VAN seems emerging.

In Japan, our group detected geoelectric potential changes highly suspect of precursory signals before large earthquakes, such as M:6.6 Off Noto Peninsula (Feb. 3,1993) and M:7.8 Southwest Off Hokkaido (July 12, 1993) earthquakes although our measuring system was not sophisticated enough to achieve complete noise rejection. Moreover, several research groups in Japan, China, Russia, and USA have been generating many promising results of electro-magnetic precursors in a broad range of frequency and of hydro-geochemical precursors such as notable changes in water level and radon emission. Together with the interesting characteristics of SES stated above, these new findings are presenting really fascinating problems and solving them will open a new horizon in the earth science.

Importance of earthquake prediction has become more apparent by the tragic disaster of 1995 M:7.2 Kobe earthquake. The type of earthquake that hit the City of Kobe can cause enormous devastation for its magnitude especially when its epicenter is close to mega-cities. Whereas it has been recognized that short-term prediction is difficult by the conventional methods, it is hoped that the new approaches like electro-magnetic, hydro-geochemical and GPS geodetic methods may shed new lights in monitoring precursors for this type of inland earthquakes also.

For further reading:
"Measurements and Theoretical Models of the EarthÕs Electric Field Variations related to Earthquakes", P. Varotsos and O. Kulhanek (eds.), Tectonophysics, 224, No.1/3, August 30, 1993.

"Electromagnetic Phenomena Related to Earthquake Prediction", M.Hayakawa and Y. Fujinawa (eds.), 677pp., Terra Sci. Publ. Co., Tokyo, 1994.

"A Critical Review of VAN - Earthquake Prediction From Seismic Electrical Signals -", Sir James Lighthill (ed.), 376pp., World Scientific, Singapore, 1996.

"Debate on VAN", R. Geller (ed.), Geophysical Research Letters, 23, No. 11, May, 1996.

P. Varotsos, K. Eftaxias, M. Lazaridou,K.Nomicos, N. Sarlis, N. Bogris, J. Makris, G. Antonopoulos and J. Kopanas, Recent Earthquake Prediction Results in Greece Based on the Observation of Seismic Electric Signals, Acta Geophysica Polonica, XLIV, 301-327, 1996.