- Open Access
A new algorithm for the detection of seismic quiescence: introduction of the RTM algorithm, a modified RTL algorithm
© The Society of Geomagnetism and Earth, Planetary and Space Sciences (SGEPSS); The Seismological Society of Japan; The Volcanological Society of Japan; The Geodetic Society of Japan; The Japanese Society for Planetary Sciences; TERRAPUB. 2011
- Received: 8 July 2010
- Accepted: 19 December 2010
- Published: 4 March 2011
There are a number of reports on seismic quiescence phenomena before large earthquakes. The RTL algorithm is a weighted coefficient statistical method that takes into account the magnitude, occurrence time, and place of earthquake when seismicity pattern changes before large earthquakes are being investigated. However, we consider the original RTL algorithm to be overweighted on distance. In this paper, we introduce a modified RTL algorithm, called the RTM algorithm, and apply it to three large earthquakes in Japan, namely, the Hyogo-ken Nanbu earthquake in 1995 (MJMA7.3), the Noto Hanto earthquake in 2007 (MJMA 6.9), and the Iwate-Miyagi Nairiku earthquake in 2008 (MJMA 7.2), as test cases. Because this algorithm uses several parameters to characterize the weighted coefficients, multiparameter sets have to be prepared for the tests. The results show that the RTM algorithm is more sensitive than the RTL algorithm to seismic quiescence phenomena. This paper represents the first step in a series of future analyses of seismic quiescence phenomena using the RTM algorithm. At this moment, whole surveyed parameters are empirically selected for use in the method. We have to consider the physical meaning of the “best fit” parameter, such as the relation of ACFS, among others, in future analyses.
- seismic quiescence
Several methods are currently available to diagnose future seismic activity, including the Z-value (Wyss and Habermann, 1988) and ETAS (e.g., Ogata, 2006), or predict major earthquakes, such as M8 (Keilis-Borok and Kossobokov, 1990), RTP (Shebalin et al., 2004), and Hotspot (Rundle et al., 2002). In addition, there are many reports of seismic activity decreasing prior to the occurrence of major earthquakes in or around the focal region (e.g., Mogi, 1979; Wyss et al., 1981; Kisslinger, 1988; Wyss and Habermann, 1988; Wiemer and Wyss, 1994). This latter phenomenon, called seismic quiescence, is expected to provide useful information for earthquake prediction. In many of the above reports the authors frequently claim that the seismic quiescence and activation occurred simultaneously. Matsumura (2005) has also modeled these phenomena as occurring simultaneously under redistribution of the tectonic stress, resulting in a net quiescence.
Positive values of RTL indicate seismic activation, while negative values indicate quiescence. Huang et al. (2001) showed that the RTL value was approximately −7 to −10 at the epicenter of the Hyogo-ken Nanbu earthquake in 1995 (MJMA 7.3, the so-called 1995 Kobe earthquake about 6 months before the earthquake, where MJMA is the earthquake magnitude defined by the Japan Meteorological Agency (JMA)). Huang and Nagao (2002) showed a clear spatial and temporal quiescence pattern at around the epicenter of the 2000 Tottori earthquake (MJMA 7.3, Japan). Huang (2008) recently applied the RTL algorithm to the 2008 Wenchuan earthquake (MS 8.0, China). Wyss et al. (2004) applied both the Z value and the RTL methods to two major earthquakes in Sakhalin and concluded that the two methods yield almost identical results, strongly suggesting that the observed precursory quiescence anomalies are robust and real.
As shown in Eqs. (1) and (3), r i is used to calculate both R and L. This dual appearance of r i seems to be in contradiction to the original concept of the RTL algorithm, i.e., the equi-importance of space and time. To remedy this situation, we have modified the RTL algorithm and apply this new algorithm (the RTM algorithm) to three Japanese earthquakes in Section 4.
Both the RTL and RTM algorithms contain a number of adjustable parameters, namely, T0, Tmax, R0, Rmax, and others. In the past, the most suitable set of parameters was retrospectively chosen by a trial and error method. Seismic quiescence phenomena do not always result in a large earthquake. Sometimes nothing happens. Therefore, it is essential that seismologists screen out, as much as possible, false (artificial) quiescence. To this end, we perform a multiparameter survey test (more than ten parameter sets) with the aim of obtaining a good understanding of the nature of seismic quiescence phenomena.
There is a strange shift of L in April 1992, as shown by an arrow in Fig. 1. A careful survey of the used data reveal that an earthquake occurred very close to the epicenter of the main shock at that time. The very small r i in Eq. (3) quite naturally makes l i /r i and L very large. Here, note that the completeness of the earthquake catalog is not considered in Fig. 1 because the aim of this figure is to show the effects on the factor L.
It is important to check the completeness of earthquake catalogs before evaluating the seismic activity. Hi-net, which is the new dense seismic station network in Japan (e.g. Obara 2003) established after the 1995 Kobe Earthquake, has made the detectability of inland earthquakes in Japan very high. The way of confirmation of the completeness itself has been an interesting study, and various ways have been proposed. One of the most sophisticated methods is a probability-based magnitude of completeness (PMC) method (e.g., Nanjo et al., 2010a). However, compared with Gutenberg-Richter (G-R) based methods (e.g., Woessner and Wiemer, 2005), applying the PMC method is more difficult because it requires detailed knowledge about the network and its setup etc.
The RTL and RTM algorithms have a number of adjustable parameters. The most suitable set of parameters is chosen retrospectively through an extensive parameter survey. In this section, we show how the RTL and RTM algorithms work by using the examples of their application to three major earthquakes in Japan.
The devastating 1995 Kobe earthquake (MJMA 7.3) occurred in southwest Japan on January 17, 1995. As already mentioned, Huang et al. (2001) showed the existence of seismic quiescence before this earthquake by the original RTL algorithm. To verify the effectiveness of the new RTM algorithm, we applied it to this earthquake as an example.
At this stage, it may be helpful to compare the RTL/RTM algorithms with other methods by pointing out their respective advantages and disadvantages.
To the best of the authors’ knowledge, there is no published report on seismic quiescence phenomena before the Noto Hanto Earthquake in 2007. We have now demonstrated that the RTM algorithm is able successfully detect such phenomena.
Using the ΔCFS (Coulomb failure stress) calculation, Kumazawa et al. (2010) reported seismic quiescence phenomena preceding the Iwate-Miyagi Nairiku earthquake in 2008 in the regions of the stress shadow area. However, as this algorithm needs data on the focal mechanism of the main shock, it is impossible to obtain the results prior to the occurrence of the main shock. In comparison, the RTL/RTM algorithms are very simple and do not need information on focal mechanism. We believe that this is an advantage of the RTL/RTM algorithms.
The Z-value is one of the well-known methods to evaluate the seismicity, especially seismic quiescence phenomena. This algorithm treats the deviation from the mean seismicity, therefore the values change smoothly. In contrast, the RTL/RTM algorithms treat the product of three factors that are normalized by the background trend, which means that the RTL/RTM values are zero at the usual time. Therefore, the changes in these values tend to be clearly visible when something unusual happens.
The M8 algorithm treats multiparameters (e.g., the absolute seismicity, its deviation, and the spatial concentration of hypocenters, etc.) and routinely cautions “Times of Increased Probability of Strong Earthquakes (TIP)” when some of the parameters exceed their threshold levels. This algorithm focuses not only seismic quiescence phenomena. In this algorithm, if an expected earthquake occurs within 5 years, this caution is judged to be true. However, the period (5 year) is longer than that in the RTL/RTM algorithms (Tmax).
For the 1995 Kobe earthquake, the seismic quiescence does not appear when Mmin ≤ Mc = 2.0 (Table 1). This absence means that the completeness in the earthquake catalog is really important for this kind of analysis.
The best parameter set is different among the three earthquakes. In the case of the 1995 Kobe earthquake (Table 1), parameter sets with the longer Tmax and the larger Mmin seem to be better. On the other hand, in the case of the Iwate-Miyagi Nairiku earthquake (Table 3), those parameter sets with the shorter Tmax and the smaller Mmin seem to be better. This difference is based on a number of complicated factors, such as the seismic event frequency, the degree of declustering, among others.
It is widely known that not all quiescence is followed by the activation of seismicity. We should evaluate the results obtained through multiparameter sets rather than rely on one parameter set as this would help avoid any acceptance of false seismic quiescence. For the three earthquakes in this study, one distinct quiescence and one major earthquake appeared during the calculation period. Further study may clarify whether or not the RTM algorithm can contribute to the development of this aspect. One possible approach to use when the aim is to clarify the statistical significance of the results would be apply calculations using many synthetic catalogs (e.g., Sobolev et al., 2002). However, we have not made this kind of statistical check since the main aim of this paper is to introduce the RTM algorithm.
Based on the results of the test applications described above, we consider that the RTM algorithm is preferable to the RTL logarithm for identifying seismic quiescence. Although it is essentially important to check the completeness of the seismic catalog (e.g., Nanjo et al., 2010a, b) as mentioned above, the present earthquake catalog in Japan is sufficiently complete to be used for seismicity analysis due to the Hi-net, the new dense seismic station network. However, for a full justification of the present conclusion, further detailed investigations on such issues as the reason-ability of the proposed parameter sets and the declustering process are needed.
At this moment, whole surveyed parameters are empirically selected. Therefore, we have to consider the physical meaning of the “best fit” parameter, e.g., the relation to ΔCFS (e.g., Scholz, 1990), among others, in the future. For instance, the RTL and RTM algorithms use an exponential decay in the factors R and T. Theoretically, seismicity is governed by the re-distribution of tectonic stress. Therefore, the factor R may be as a function of 1/r3. This area definitely needs further investigation.
In retrospective studies, we “know” the hypocenter and magnitude of “impending” earthquakes. Consequently, we can perform multiparameter surveys around the known location and magnitude. What can we do in a preseismic stage when none of this information is known? We would like to propose a study in which RTM calculations are routinely made at every grid-point of a region in order to draw a seismic quiescence map, such as those in Figs. 5 and 8, in a multiparameter survey. Computer capability will allow such a study any time now or in the near future. The experience gained in such a study may lead to a narrowing down of the ranges of parameters and, ultimately, even regional progressive diagnosis of seismic quiescence will be realized in a not so distant future. The authors consider this paper to be the first step in future analyses of seismic quiescence phenomena along the line of the RTL algorithm proposed by Gennady Sobolev in 1997.
The earthquake catalog used in this study is from JMA. We would like to thank H. Takayama of JMA for providing a declustering program. We would like to thank also A. Yoshida of JMA for his advice on decluster parameters. Our thanks are extended to S. Uyeda, Q. Huang, G. A. Sobolev, H. Katao, and K. Nanjo who gave us valuable comments and suggestions to improve the manuscript. This research was partially supported by Observation and Research Program for Prediction of Earthquakes and Volcanic Eruptions, the Ministry of Education, Culture, Sports, Science and Technology (MEXT) of Japan.
- Chen, C.-C., J. B. Rundle, J. R. Holliday, K. Z. Nanjo, D. L. Turcotte, S.-C. Li, and K. F. Tiampo, The 1999 Chi-Chi, Taiwan, earthquake as a typical example of seismic activation and quiescence, Geophys. Res. Lett., 32(22), L22315, doi:10.1029/2005GL023991, 2005.View ArticleGoogle Scholar
- Frohlich, C. and S. D. Davis, Single-link cluster analysis as a method to evaluate spatial and temporal properties of earthquake catalogues, Geophys. J. Int., 100, 19–32, 1990.View ArticleGoogle Scholar
- Huang, Q., Search for reliable precursors: A case study of the seismic quiescence of the 2000 western Tottori prefecture earthquake, J. Geophys. Res., 111, B04301, doi: 10.1029/2005JB003982, 2006.Google Scholar
- Huang, Q., Seismicity changes prior to the Ms8.0 Wenchuan earthquake in Sichan, China, Geophys. Res. Lett., 35, L23308, doi:10.1029/ 2008GL036270, 2008.View ArticleGoogle Scholar
- Huang, Q. and T. Nagao, Seismic quiescence before the 2000 M=7.3 Tottori earthquake, Geophys. Res. Lett., 29(12), doi:10.1029/ 2001GL013835, 2002.Google Scholar
- Huang, Q., G. A. Sobolev, and T. Nagao, Characteristics of the seismic quiescence and activation patterns before the M=7.2 Kobe earthquake, January 17, 1995, Tectonophysics, 337, 99–116, 2001.View ArticleGoogle Scholar
- Kasahara, K., Earthquake Mechanics, Cambridge Univ. Press., Cambridge, 248 pp, 1981.Google Scholar
- Keilis-Borok, V. I. and V. G. Kossobokov, Premonitory activation of earthquake flow: Algorithm M8, Phys. Earth Planet. Inter., 61, 73–83, doi:10.1016/0031-9201(90)90096-G, 1990.View ArticleGoogle Scholar
- Kisslinger, C., An experiment in earthquake prediction and the 7 May 1986 Andreanof Islands earthquake, Bull. Seismol. Soc. Am., 78, 218–229, 1988.Google Scholar
- Kumazawa, T., Y. Ogawa, and S. Toda, Precursory seismic anomalies and transient crustal deformation prior to the 2008 Mw = 6.9 Iwate-Miyagi Nairiku, Japan, earthquake, J. Geophys. Res., 115, B10312, doi:10.1029/2010JB007567, 2010.View ArticleGoogle Scholar
- Matsumura, S., Why does the precursory change of seismicity rate tend to be quiescence?, Zisin, 57, 441–444, 2005 (in Japanese with English Abstract).Google Scholar
- Mogi, K., Two kinds of seismic gap, Pure Appl. Geophys, 117, 1172–1186, 1979.View ArticleGoogle Scholar
- Nanjo, K. Z., D. Schorlemmer, J. Woessner, S. Wiemer, and D. Giardini, Earthquake detection capability of the Swiss Seismic Network, Geophys. J. Int., 181(3), 1713–1724, doi:10.1111/j.1365-246X.2010.04593.x, 2010a.Google Scholar
- Nanjo, K. Z., T. Ishibe, H. Tsuruoka, D. Schorlemmer, Y. Ishigaki, and N. Hirata, Analysis of the completeness magnitude and seismic network coverage of Japan, Bull. Seismol. Soc. Am., 100(6), 3261–3268, doi:10.1785/0120100077, 2010b.View ArticleGoogle Scholar
- Obara, K., Hi-net: High sensitivity seismograph network, Japan, Lect. Notes Earth Sci., 93, 79–88, doi:10.1007/BFb0117693, 2003.View ArticleGoogle Scholar
- Ogata, Y., Monitoring of anomaly in the aftershock sequence of the 2005 earthquake of M7.0 off coast of the western Fukuoka, Japan, by the ETAS model, Geophys. Res. Lett., 33(1), L01303, doi:10. 1029/2005GL024405, 2006.View ArticleGoogle Scholar
- Rundle, J. B., K. F. Tiampo, W. Klein, and J. S. Sa Martins. Self-organization in leaky threshold systems: The influence of near-mean field dynamics and its implications for earthquakes, neurobiology, and forecasting, Proc. Natl. Acad. Sci. USA., 99, suppl. 1, 2514–2521, 2002.View ArticleGoogle Scholar
- Scholz, C. H., The Mechanisms of Earthquakes and Faulting, Cambridge University Press, 439 pp., 1990.Google Scholar
- Shebalin, P., V. Keilis-Borok, I. Zaliapin, S. Uyeda, T. Nagao, and N. Tsybin, Advance short-term prediction of the large Tokachi-oki earthquake, September 25, 2003, M=8.1 A case history, Earth Planets Space, 56, 715–724, 2004.View ArticleGoogle Scholar
- Sobolev, G. A. and Y. S. Tyupkin, Low-seismicity precursors of large earthquakes in Kamchatka, Volcanol. Seismol., 18, 433–446, 1997.Google Scholar
- Sobolev, G. A. and Y. S. Tyupkin, Precursory phases, seismicity precursors, and earthquake prediction in Kamchatka, Volcanol. Seismol., 20, 615–627, 1999.Google Scholar
- Sobolev, G., Q. Huang, and T. Nagao, Phases of earthquake’s preparation and by chance test of seismic quiescence anomaly, J. Geodyn., 33, 425–432, 2002.View ArticleGoogle Scholar
- Tsuji, Y., J. Nakajima, and A. Hasegawa, Tomographic evidence for hydrated oceanic crust of the Pacific slab beneath northeastern Japan: Implications for water transportation in subduction zones, Geophys. Res. Lett., 35, L14308, doi:10.1029/2008GL034461, 2008.View ArticleGoogle Scholar
- Wiemer, S. and M. Wyss, Seismic quiescence before the Landers (M=7.5) and Big Bear (M=6.5) 1992 earthquakes, Bull. Seismol. Soc. Am., 84, 900–916, 1994.Google Scholar
- Woessner, J. and S. Wiemer, Assessing the quality of earthquake catalogues: Estimating the magnitude of completeness and its uncertainty, Bull. Seismol. Soc. Am., 95(2), 684–698, doi:10.1785/0120040007, 2005.View ArticleGoogle Scholar
- Wyss, M. and R. E. Habermann, Precursory Seismic quiescence, Pure Appl. Geophys., 126, 319–332, 1988.View ArticleGoogle Scholar
- Wyss, M., F. W. Klein, and A. C. Johnston, Precursor ot the Kalapana M=7.2 earthquake, J. Geophys. Res., 86, 3881–3900, 1981.View ArticleGoogle Scholar
- Wyss, M., G. Sobolev, and J. D. Clippard, Seismic quiescence precursors to two M7 earthquakes on Sakhalin Island, measured by two methods, Earth Planets Space, 56, 725–740, 2004.View ArticleGoogle Scholar
- Yoshii, T., A detailed cross-section of the deep seismic zone beneath northeastern Honshu, Japan, Tectonophysics, 55, 349–360, 1979.View ArticleGoogle Scholar