Research in Geophysics 2016; volume 5:5730

Time dependent seismicity Fuca and North America plates corresponds to the Cascadia subduction zone (CSZ), where Correspondence: Evangelos Christou, Planetary along the western coast the giant M9 earthquake occurred in 1700. Sciences Humboldt Universität zu Berlin, of Canada The present study was motivated by the Jablonskistraße 14, Berlin, Germany. occurrence of the 2012 Haida Gwaii islands Tel.: +49.15258299952. Evangelos V. Christou, George Karakaisis, (formerly the Queen Charlotte islands) earth- E-mail: [email protected] Emmanuel Scordilis quake (M=7.7). This was the first major thrust Key words: Time dependent seismicity; Canada; 1 Department of Geophysics, School of event recorded along the strike-slip QCF. In Haida Gwaii 2012 earthquake. this article we present the results of the appli- Geology, Aristotle University of cation of two time-dependent seismicity mod- Acknowledgements: Evangelos Christou grateful- Thessaloniki, Greece els in an attempt to retrospectively predict the ly acknowledges Dr. G. Papadopoulos (Chair) and 2012 mainshock. Moreover, we apply both mod- the Organizing Committee of the International els searching for future strong earthquakes Workshop Mega Earthquakes and Tsunamis in Subduction Zones for the financial support. Maps along the western coast of Canada. Abstract were made using Generic Mapping Tools GMT 4.5 (Wessel and Smith, 1998). We used data from the Models applied and data National Resources Canada (On-line Bulletin), Decelerating generation of intermediate The first of the two time-dependent seismic- http://earthquakescanada.nrcan.gc.ca/stndon/NE magnitude earthquakes (preshocks) in a nar- ity models is based on the triggering of a main- DB-BNDS/bull-eng.php, Nat. Res. Can., (last row region (seismogenic region) and acceler- shock by its preshocks and is called decelerat- accessed: July 23, 2014). We thank the anony- ating generation of relatively larger such mous reviewer for the helpful comments and sug- ing-accelerating seismic strain (D-AS) model. gestions. earthquakes in a broader region (critical There is reliable evidence that an increase in region) has been proposed as an appropriate the occurrence rate of intermediate-magni- Conference presentation: this work has been pre- model for intermediate-term earthquake pre- tude shocks has been observed in a broad area sented in the International Workshop Mega diction. We examined the seismic activity before strong earthquakes.2-8 Recently, howev- Earthquakes and Tsunamis in Subduction Zones- which preceded the Mw=7.7 (October 28, 2012) er, skepticism has been expressed on the reli- Forecasting Approaches and Implications for thrust event that occurred off the west coast of ability of these observations,9,10 which is dis- Hazard Assessment, held in Rhodes Island, Greece (2014). Haida Gwaii, Canada (formerly the Queen cussed later. Quantification of the accelerating Charlotte islands), by applying the decelerat- pattern of these earthquakes that occur in this Received for publication: 24 November 2014. ing-accelerating seismic strain model. We 11 broad region before a mainshock, showed Revision received: 6 December 2015. found that this mainshock was preceded by a that the cumulative Benioff strain, S(t), can be Accepted for publication: 22 December 2015. pronounced accelerating seismic sequence expressed by the following power law: with the time to the mainshock, as well as by This work is licensed under a Creative Commons an equally easily identifiable decelerating seis- Attribution NonCommercial 4.0 License (CC BY- mic sequence. Both precursory seismic (1) NC 4.0). sequences occurred in different space, time and magnitude windows. The behavior of pre- ©Copyright E.V. Christou et al., 2016 where tc is the origin time of the mainshock Licensee PAGEPress, Italy vious mainshocks that occurred close to the and A, B, m, are parameters calculated by the Research in Geophysics 2016; 5:5730 2012 earthquake was also examined by the available data (with m<1, B<0). The quantity doi:10.4081/rg.2016.5730 time and magnitude predictable regional S(t), which is considered as a measure of the model. preshock seismic deformation at time t, is An attempt was also made to identify such seismic strain patterns, which may also be defined as , where Ei is the (M=6.3-9.0) shallow (h<100 km) mainshocks related to the generation of strong mainshocks worldwide.15 Tests performed on synthetic cat- along the western coast of Canada. seismic energy of the ith preshocks and n(t) is alogues15,16 and retrospective predictions of the number of preshocks occurred up to time t. recent strong mainshocks have been used to On the other hand, it has also been observed evaluate the model whereas forward tests led Non commercialthat in the narrow (focal) region of anuse ensuing only to the successful intermediate-term prediction Introduction mainshock, a seismic excitation is followed by of two strong earthquakes in the Aegean.17,18 a drop of seismicity, i.e., a seismic quiescence During the formulation of the D-AS model it Strong and large earthquakes along the period.12,13 Global data were used14 to show that was observed that each of the investigated western coast of Canada are not uncommon. intermediate magnitude preshocks in the focal During the instrumental period (since 1898, region form a decelerating pattern and that the mainshocks was preceded by a decelerating when the first seismograph of Milne type was time variation of the cumulative Benioff strain preshock sequence, generated in a relatively small region (seismogenic region) where the installed at Victoria) several M≥7.0 events up to the mainshock also follows a power-law occurred there. The majority of the strong (relation 1) but with a power value larger than mainshock is also located, and by an accelerat- earthquakes in this area are associated with one (m>1). That is, this pattern of decelerat- ing preshock sequence generated in a broader the motion between the three major lithos- ing strain in the focal region is formed of a region (critical region) and that both precurso- pheric plates, namely Pacific, North America seismic excitation followed by a decrease of ry seismic sequences have predictive proper- and Juan de Fuca (Figure 1). The boundary seismicity of intermediate magnitude ties, related to the ensuing mainshock. between Pacific and North America plates is a preshocks. Decelerating preshocks occur in different right lateral transform fault (Queen Charlotte The formulation of the D-AS model for inter- time, magnitude and space windows than the Fault, QCF) that extends from Vancouver mediate-term earthquake prediction was accelerating preshocks. The latter start earlier

Island up to Alaska and the Fairweather Fault based on the examination of the patterns than the former (tsa>tsd), and their magnitudes (FF), whereas the boundary between Juan de described above, which preceded strong are larger the magnitudes of the decelerating

[Research in Geophysics 2016; 5:5730] [page 1] Article

preshocks. The strain acceleration, qa, as well parameter and Pd is the probability that a than the relatively short span of the earth- as the strain deceleration, qd, (also called qual- decelerating strain release fulfils relations (2) quake catalogues used, it is preferable to con- ity indexes), vary with the time to the main- and (4). Both quality indexes qd and qa are very sider seismogenic sources, i.e., circular shock.17 Both start with low values, attain their useful in searching for decelerating and accel- regions that include, in addition to the main largest values several years before the main- erating seismicity patterns since they attain fault where the largest mainshock occurs, shock occurrence and cease gradually about their largest values at the seismogenic and other smaller faults where smaller mainshocks three years before the mainshock, i.e., the critical region, respectively. Global observa- occur. On the basis of this idea, the time and seismic activity declines in the critical region tions15 resulted in the following cut-off values magnitude predictable regional (TIMAPR) and increases in the seismogenic region. of the parameters that describe the decelerat- model has been proposed,21 which makes use The radii r (km) of the seismogenic region ing and the accelerating precursory seismic of numerous interevent times of strong earth- and R (km) of the critical region, which are sequences: quakes (mainshocks) generated in a region, in assumed circular, are given by the relations:15 order to investigate the time-dependent seis- (8) micity of this region. This model has been developed by considering a large sample of (9) global data,21 which was used to derive the fol- (2) lowing two equations that relate the interevent

The second of the models applied in the time, Tt (in years), until the next mainshock present work makes use of interevent times of and its expected magnitude, Mf, to the magni- (3) strong earthquakes in an area, which depend tude, Mp, of the previous mainshock in the on the rate of tectonic loading. However, since region examined, the long-term seismicity 1/2 such earthquakes, that occur on a single fault, level, Sd (the seismic strain rate in Joule per with M being the mainshock magnitude and sd usually have recurrence times much larger year) and the minimum mainshock magni- 1/2 4 2 and sa (in J /y × 10 km ) is the Benioff strain in the seismogenic and critical region, respec- tively. The following two relations hold also for the decelerating and accelerating preshock sequences:15

(4)

(5)

where tsd and tsa are the start times (in years) of the decelerating and accelerating preshock sequence, respectively. The curvature parameter, C, has been pro- posed19 as a measure of accelerating strain release and equals to the ratio of the RMS error of the power-law fit (Eq. 1) to the corre- sponding linear fir error. C takes positive val- ues smaller than 1, becomes equal to 1 for lin- ear fit and decreases when the accelerating Benioff strain release becomes more intense. A quality index, qa, has been defined Nonto meas- commercial use only ure the intensity of the accelerating pattern:20

(6)

where ma is the exponent in relation (1), Ca is the curvature parameter and Pa is the probabil- ity that an accelerating pattern fulfills rela- tions (3) and (5). A similar relation quantifies the intensity, qd, of the decelerating strain:

(7) Figure 1. Main tectonic features of the studied area. FF, Fairweather Fault; QCF, Queen Charlotte Fault; CSZ, Cascadia Subduction Zone. Epicenters of all earthquakes with with md being the exponent in relation (1) for M≥6.0 that occurred after 1912 are shown along with the epicenter of the giant M9 of 1700 in Cascadia. the decelerating pattern, Cd is the curvature

[page 2] [Research in Geophysics 2016; 5:5730] Article

tude, Mmin, considered: quake of the remaining catalogue is consid- Canada on-line bulletin (http://earthquakes- ered as a mainshock and its associated shocks, canada.nrcan.gc.ca/stndon/NEDB-BNDS/bull- defined as previously, are also excluded from eng.php, last accessed: July 23, 2014) for the (10) the catalogue. This continues till no event with period 1992-2014. The studied area is bounded magnitude larger than a certain cut-off magni- by the 120°W-140°W meridians and 47°N-57°N (11) tude Mmin remains in the original catalogue. parallels. Earthquake magnitudes are mainly For the identification of a circular seismogenic expressed in the ML, MS and mb scales and Mw The mean values of q and w and their stan- source the examined area is covered by a for recent events. Although earthquake dense grid of geographic points. Each point is detectability studies26 suggest that the earth- dard deviations s/q and sw can be calculated by the available data for a certain seismogenic considered as the center, K, of a circular seis- quake catalogue for the examined area may be 22 assumed to be complete for M 5.0 since 1951 source. It was showed that the ratio s/Tt of mogenic source and circles with varying radius ≥ the observed interevent time, T, to the calculat- (e.g., r=30-200 km) are defined. The earth- and M≥4.0 since 1971, we used the ZMAP soft- 27 ed, Tt (by Eq. 10), follows a lognormal distribu- quake catalogue for each circle is declustered ware as well as the traditional frequency- tion with a mean equal to zero. The application according to the procedure described previous- magnitude distribution28 to check space and of this model requires a declustered catalogue ly and relations (10) and (11) is applied on the time variations of magnitude completeness. of mainshocks, which is derived after adopting resulting mainshocks. The optimum seismo- We found that the data are complete for the an appropriate time window, defined by the genic source is selected for the circle (K, r) for whole area under investigation for M≥4.8 for the time interval 1960-1990 and for M 4.0 for ratio s/T (where T is the mean interevent time ≥ and s its standard deviation), since this ratio is which the optimization factor, , the time interval 1991-2014. To ensure homo- a measure of seismic clustering. For ratio val- geneity of the catalogue in respect to the mag- ues smaller than 0.5 an earthquake catalogue where N is the number of interevent times, nitude, mb and MS magnitudes were converted 23 exhibits quasi-periodic behaviour. Global has the largest value (N≥3). In the present into the moment magnitude scale, Mw, by appropriate relations derived by the use of data show that for a time window Dt≥15 years case the geographic point with the highest OP 24 29,30 this ratio becomes smaller than 0.5 value in the vicinity of the 2012 earthquake is global data, whereas ML magnitudes were

The procedure followed for creating a main- 51.6°N-131.0°W, with OP=19.35 and r=180 km. converted to Mw using the relations valid for shock catalogue in a seismogenic source The declustered mainshock catalogue within western Canada.31 through declustering starts after considering this seismogenic source, which will be subse- the largest earthquake of the available com- quently used for the retrospective prediction of plete earthquake catalogue of the source as the 2012 event, includes the following shocks: the first mainshock. This mainshock and its 1912 M=6.5, 1929 M=7.0, 1949 M=8.1, 1970 Results associated shocks (earthquakes of the original M=7.4, 1992 M=6.7. catalogue that occurred in the seismogenic Data used in the present work have been Previous studies15,16,24,32 showed that the source within a time window ±15 years from extracted from the Seismic Hazard Earthquake centers F(j, l) and Q(j,l) of the seismo- the mainshock origin time) are excluded from Epicenter File (SHEEF) for the period 1627- genic and critical regions, respectively, are the original catalogue. Then, the largest earth- 199125 and from the National Resources located in areas bounded by the two parallels

A

Non commercialB use only

Figure 2. A) Decelerating (dots) and accelerating (small open cir- cles) preshocks of the 2012 mainshock (denoted by a star) that occurred within the circular seismogenic and critical regions, respectively. The retrospectively predicted epicenter is denoted by the grey circle. B) Plots of the time variation of the decelerating (left) and accelerating Benioff strain release, S(t) (right), are also shown at the lower part of the figure, along with the curves that fit the data.

[Research in Geophysics 2016; 5:5730] [page 3] Article j±3° NS and the meridians l±3° EW around and seventh columns. The last three columns mean epicenter of the decelerating preshocks; the mainshock epicenter, E(j, l). For this rea- show the values of the curvature parameter, C, ii) the mean epicenter of the mainshocks son each of these areas is covered by a dense for the decelerating and accelerating which were identified after declustering the grid of points (e.g., ±0.2° NS, ±0.2° EW). In preshocks, the logarithm of the strain rate, sd original complete catalogue of earthquakes 4 2 order to define the seismogenic region, where and sa (in Joules per year and per 10 km ) and that occurred in the seismogenic source. The the decelerating Benioff strain release may be the values of the quality indexes qa and qd, difference between the observed and the calcu- observed, each of these points is considered as given by the relations (6) and (7). lated basic focal parameters, in the present the center of a circular seismogenic region The spatial distribution of the decelerating work, are within the model uncertainties, i.e., with radius r (given by relation 2). For various (dots) and accelerating (small open circles) t*: ±2.5 years, M*±0.3, E*: 80±30 km.15,24 r-values, tsd start times and minimum magni- preshocks of the 2012 mainshock (denoted by a In an attempt to search for precursory decel- tudes, different quality index, qd, values are star) that occurred within the circular seismo- erating and accelerating preshock sequences calculated. The geographic point where the genic and critical regions, respectively, is shown that may be related to the generation of a largest of these values is found, corresponds to in Figure 2. The retrospectively predicted epi- future strong earthquake, the area bounded by the center, F, of the seismogenic region. The center is denoted by the grey circle. Plots of the the 48.0°N-57°N parallels and 122°W-138°W same procedure is applied for the definition of time variation of the decelerating (Figure 2B - meridians was covered by a grid of points the circular critical region with radius R left) and accelerating Benioff strain release, spaced 0.5° apart. The magnitude range of the (given by relation 3), and the geographic point S(t) (Figure 2B - right), are also shown at the probably ensuing earthquake was set between where the largest quality index, qa, is calculat- lower part of the figure, along with the curves 7.0 and 8.2 with step equal to 0.2 magnitude ed, corresponds to the center of the critical that fit the data. Both preshock sequences end units. Table 2 lists the basic parameters con- region where accelerating preshocks occurred three years before the mainshock. cerning decelerating and accelerating prior to the mainshock generation. The finally adopted as predicted origin time, preshock sequences, which may be related to a Table 1 gives information on the properties t*, magnitude, M*, and epicenter, E*(f,l), of probably ensuing mainshock. Figure 3 shows of the decelerating and accelerating preshocks the retrospectively predicted mainshock, the spatial distribution of the decelerating and that preceded the 2012 Haida Gwaii mainshock which are given in the last row of Table 1, have accelerating preshocks along with the time and occurred within the corresponding seis- been estimated in the following way. The ori- variation of the respective cumulative Benioff mogenic and circular regions. The first three gin time, t*, is the mean value of the times cal- strain, S(t). columns list the origin time, magnitude and culated by the relations (4), (5) and (10) (the Following the procedure described previous- epicenter coordinates of the mainshock and recurrence interval, Tt, calculated by the last ly, the basic focal parameters of this earth- the next two columns show the centers, F, Q, relation, is added to the origin time of the pre- quake were estimated: i) expected origin time and radii r, R, of the circular seismogenic and vious mainshock that occurred in the region). tc*=2022.5; ii) expected magnitude M*=7.1; critical regions, respectively. The minimum The magnitude, M*, is the mean of the magni- iii) expected epicenter coordinates preshock magnitudes, Mmin, and the start years tude values calculated by the relations (2), (3), E*=49.4°N, 129.6°W. of the decelerating, tsd, and the accelerating, and (11). Finally, the epicenter, E*(f,l), is tsa, preshock sequences are shown in the sixth the mean of two geographic points: i) the

A

Non commercialB use only

Figure 3. A) Decelerating (dots) and accelerating (small open cir- cles) preshocks associated with the probably expected earthquake (grey circle) in area studied. B) Plots of the time variation of the decelerating (left) and accelerating Benioff strain release, S(t) (right), are also shown at the lower part of the figure, along with the curves that fit the data.

[page 4] [Research in Geophysics 2016; 5:5730] Article

Table 1. Properties of the decelerating and accelerating preshocks and their respective circular regions. Date, magnitude and epicenter coordinates of the 2012 mainshock are listed in the first three columns. The centers, F, and Q of the seismogenic and critical regions, along with their corresponding radii, r and R, are then shown. Information follows on the minimum magnitude, Mmin, the start times, tsd and tsa, and the values of the curvature parameter of the decelerating and accelerating preshocks. The last two columns show the log- 4 2 arithm of the strain rate, sd and sa (in Joules per year and per 10 km ) and the values of the quality indexes qd and qa. In the last row the origin time, t*, magnitude, M*, and epicenter, E*(j, l), that were calculated in the present work, for this mainshock, are given. Date M E (j, l) F (j, l) r (km) Mmin tsd C logsd qd 2012.10.28 7.7 52.8-132.9 52.2-134.1 239 4.6 1998 0.53 5.19 3.6

Q (f, l) R (km) Mmin tsa C logsa qa 49.8-133.0 680 5.3 1968 0.38 5.18 7.3 t* M* E* (j, l) 2014.1 7.4 51.9-131.6

Table 2. Properties of the decelerating and accelerating preshocks and their respective circular regions, which may be related to a prob- ably ensuing strong earthquake in the area examined. In the last row the basic focal parameters E*(j, l), t*, and M*, of this earthquake are given. F(j, l) r (km) Mmin tsd C logsd qd 52.2-134.1 239 4.6 1998 0.53 5.19 3.6

Q(j, l) R (km) Mmin tsa C logsa qa 49.8-133.0 680 5.3 1968 0.38 5.18 7.3 E*(j, l) t* M* 49.4-129.6 2022.5 7.1

ating shocks that occurred in circular areas Francisco bay region. Calif Div Mines Spec Discussion and Conclusions centered at the mainshock epicenter, which Rep 1959;57:39-48. was based solely on minimizing the curvature 5. Papadopoulos GA. Long-term accelerating We found that decelerating strain, released parameter C. Recent tests on synthetic cata- foreshock activity may indicate the occur- by intermediate magnitude preshocks, which logues of earthquakes with spatiotemporal rence time of a strong shock in the occurred within a narrow region, along with clustering based on the ETAS model in Aegean Western Hellenic Arc. Tectonophysics accelerating strain released by larger shocks in and California16 showed that decelerating and 1988;152:179-92. a broader region, preceded the 2012 Haida accelerating sequences are hardly recogniza- 6. Jaumé SC, Sykes LR. Evolving towards a Gwaii earthquake. We also found that the ble in circular regions centered at the main- critical point: a review of accelerating mainshocks, which occurred in the seismo- shock epicenter (low qd and qa values), in seismic moment/energy release rate prior genic source, where the 2012 epicenter is accordance with the negative results men- to large and great earthquakes. Pure appl located, follow a quasi-periodic pattern which tioned above,9 whereas such sequences are Geophys 1999;155:279-306. allows the estimation of occurrenceNon time and commercialeasily identifiable in circular areas withuse cen- only7. Papazachos CB, Karakaisis GF, Scordilis magnitude of the next mainshock by the use of ters close to the mainshock epicenter. EM, et al. Global observational properties the magnitude and the occurrence time of the of the critical earthquake model. B previous mainshock. Moreover, we searched Seismol Soc Am 2005;95:1841-55. the studied area for currently decelerating- 8. De Santis A, Cianchini G, Di Giovabatista accelerating strain release, which may be References R. Accelerating moment release revisited: related with a probably ensuing mainshock. Examples of application to Italian seismic We cannot ignore the skepticism expressed 1. Cassidy JF, Rogers GC, Hyndman RD. An sequences. Tectonophysics 2015;639:82- during the last several years about the hypoth- overview of the 28 October 2012 Mw7.7 98. esis of accelerating seismicity that this pattern earthquake in Haida Gwaii, Canada: a 9. Hardebeck JL, Felzer KR, Michael AJ. may arise from a combination of data fitting tsunamigenic thrust event along a pre- Improved tests reveal that the accelerating and from the generation of normal foreshocks dominantly strike-slip margin. Pure Appl moment release hypothesis is statistically and aftershocks9 or that it may be due to the Geophys 2014 [Epub ahead of print]. insignificant. J Geophys Res 2008;113: formulation of the accelerating preshock gen- 2. Imamura A. Theoretical and applied seis- 808310. eration as a power-law fit to a cumulative seis- mology. Tokyo: Maruzen; 1937. 10. Mignan A. Retrospective on accelerating micity series.10 The negative results reported 3. Gutenberg B, Richter CF. Seismicity of the seismic release (ASR) hypothesis: as regards the statistical significance of the Earth and associated phenomena. New Controversy and new horizons. AMR hypothesis,9 may be due to the optimiza- York: Hafner; 1954. Tectonophysics 2011;505:1-16. tion procedure applied, in examining acceler- 4. Tocher D. Seismic history of the San 11. Bufe CG, Varnes DJ. Predictive modeling of

[Research in Geophysics 2016; 5:5730] [page 5] Article

seismic cycle of the Great San Francisco Bay Papazachos CB, et al. A forward test of the generation seismic hazard maps of Region. J Geophys Res 1993;98:9871-83. decelerating-accelerating seismic strain Canada. Geological Survey of Canada, 12. Evison FF, Rhoades DA. The precursory model in the Mediterranean. Boll Geofis Open File 6208 2009; pp 16. earthquake swarm in New Zealand. N Z J Teor Appl 2009;50:235-54. 26. Milne WG, Rogers GC, Riddihough RP, et Geol Geophys 1997;40:537-47. 19. Bowman DD, Quillon G, Sammis CG, et al. al. Seismicity of western Canada. Can J 13. Evison FF. Long-range synoptic earthquake An observational test of the critical earth- Earth Sci 1978;15:1170-93. forecasting: an aim for the millennium. quake concept. J Geophys Res 1998; 27. Wiemer S. A software package to analyze Tectonophysics 2001;333:207-15. 103:24359-72. seismicity: ZMAP. Seis Res Letts 2001;72: 14. Papazachos CB, Scordilis EM, Karakaisis 20. Papazachos CB, Karakaisis GF, Savvaidis 374-83. GF, et al. Decelerating preshock seismic AS, et al. Accelerating seismic crustal 28. Gutenberg B, Richter CF. Frequency of deformation in fault regions during criti- deformation in the southern Aegean area. earthquakes in California. B Seismol Soc cal periods. B Geol Soc Greece B Seismol Soc Am 2002;92:570-80. Am 1944;34:185-8. 2005;36:1490-8. 21. Papazachos BC, Papadimitriou EE, 29. Scordilis EM. Globally valid relations con- 15. Papazachos CB, Karakaisis GF, Scordilis Karakaisis GF, et al. Long-term earthquake verting Ms, mb and MJMA to Mw, at EM, et al. New observational information prediction in the Circum-Pacific conver- Meeting on earthquake monitoring and on the precursory accelerating and decel- gent belt. Pure Appl Geophys 1997;149:173- seismic hazard mitigation in Balkan coun- erating strain energy release. 217. tries. NATO ARW, 11-17 September 2005, Tectonophysics 2006;423:83-96. 22. Papazachos BC, Papaioannou ChA. Long Borovetz, Bulgaria, pp 158-161. 16. Karakaisis GF, Papazachos CB, Scordilis term earthquake prediction in the Aegean EM. Recent reliable observations and area based on the time and magnitude 30. Scordilis EM. Empirical global relations improved tests on synthetic catalogs with predictable model. Pure Appl Geophys converting Ms and mb to moment magni- spatiotemporal clustering verify precurso- 1993;140:593-612. tude. J Seismology 2006;10:225-36. ry decelerating-accelerating seismicity. J 23. Kagan YY, Jackson DD. Long-term earth- 31. Ristau JP. Seismotectonics of western Seismol 2013;17:1063-72. quake clustering. Geophys J Int Canada from regional moment tensor 17. Papazachos BC, Karakaisis GF, 1991;104:117-33. analysis. PhD Thesis, University of Papazachos CB, et al. Evaluation of the 24. Papazachos BC, Karakaisis GF, Scordilis Victoria, Victoria, Canada; 2004, 209 pp. results for an intermediate term predic- EM. Time dependent seismicity in the con- 32. Papazachos BC, Papaioannou ChA, tion of the 8 January 2006 Mw=6.9 tinental fracture system. Boll. Geofis Teor Scordilis EM, et al. A forward test of the Cythera earthquake in southwestern Applic 2014;55:617-39. decelerating-accelerating seismic strain Aegean. B Seismol Soc Am 2007;97:347-52. 25. Halchuk S. Seismic Hazard Earthquake model to western south and Central 18. Papazachos BC, Karakaisis GF, Epicenter File (SHEEF) used in the fourth America. Tectonophysics 2008;454:36-43.

Non commercial use only

[page 6] [Research in Geophysics 2016; 5:5730]