Tectonophysics 618 (2014) 134

Contents lists available at ScienceDirect

Tectonophysics

j ourna l homepage: www.e lsev ie r .com/ locate / tecto

Invited Review Article

The global aftershock zone

Tom Parsons a,, Margaret Segou b, Warner Marzocchi c

a U.S. Geological Survey, Menlo Park, CA, USAb Geosciences Azur, Francec Istituto Nazionale di Geofisica e Vulcanologia, Rome, Italy

Corresponding author at: U.S. Geological Survey, MS-E-mail address: tparsons@usgs.gov (T. Parsons).

http://dx.doi.org/10.1016/j.tecto.2014.01.0380040-1951/Published by Elsevier B.V. This is an open acce

a b s t r a c t

a r t i c l e i n f o

Article history:Received 27 September 2013Received in revised form 24 January 2014Accepted 28 January 2014Available online 5 February 2014

Keywords:Earthquake triggeringDynamic triggeringAftershocksSeismic hazard

The aftershock zone of each large (M 7) earthquake extends throughout the shallows of planet Earth. Mostaftershocks cluster near the mainshock rupture, but earthquakes send out shivers in the form of seismicwaves, and these temporary distortions are large enough to trigger other earthquakes at global range. Theaftershocks that happen at great distance from their mainshock are often superposed onto already seismicallyactive regions, making them difficult to detect and understand. From a hazard perspective we are concernedthat this dynamic process might encourage other high magnitude earthquakes, and wonder if a global alarmstate is warranted after every large mainshock. From an earthquake process perspective we are curious aboutthe physics of earthquake triggering across the magnitude spectrum. In this review we build upon past studiesthat examined the combined global response to mainshocks. Such compilations demonstrate significant rateincreases during, and immediately after (~45 min) M N 7.0 mainshocks in all tectonic settings and ranges.However, it is difficult to find strong evidence for M N 5 rate increases during the passage of surface waves incombined global catalogs. On the other hand, recently published studies of individual largemainshocks associateM N 5 triggering at global range that is delayed by hours to days after surface wave arrivals. The longer the delaybetween mainshock and global aftershock, the more difficult it is to establish causation. To address thesequestions, we review the response to 260 M 7.0 shallow (Z 50 km) mainshocks in 21 global regions withlocal seismograph networks. In this way we can examine the detailed temporal and spatial response, or lackthereof, during passing seismic waves, and over the 24 h period after their passing. We see an array of responsesthat can involve immediate and widespread seismicity outbreaks, delayed and localized earthquake clusters, to noresponse at all. About 50% of the catalogs that we studied showed possible (localized delayed) remote triggering,and ~20% showed probable (instantaneous broadly distributed) remote triggering. However, in any given region,at most only about 23% of global mainshocks caused significant local earthquake rate increases. These rateincreases are mostly composed of small magnitude events, and we do not find significant evidence ofdynamically triggered M N 5 earthquakes. If we assume that the few observedM N 5 events are triggered, we findthat they are not directly associated with surface wave passage, with first incidences being 910 h later. We notethat mainshock magnitude, relative proximity, amplitude spectra, peak ground motion, and mainshock focalmechanisms are not reliable determining factors as to whether a mainshock will cause remote triggering. Byelimination, azimuth, and polarization of surface waves with respect to receiver faults may bemore important factors.

Published by Elsevier B.V. This is an open access article under the CC BY-NC-ND license(http://creativecommons.org/licenses/by-nc-nd/3.0/).

Contents

1. Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 22. Methods and data . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 33. Observations . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 8

3.1. California . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 93.2. Greece . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 103.3. New Zealand . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 103.4. Southeast China . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 10

999, 345 Middlefield Rd., Menlo Park, CA, 94025,USA

ss article under the CC BY-NC-ND license (http://creativecommons.org/licenses/by-nc-nd/3.0/).

http://creativecommons.org/licenses/by-nc-nd/3.0/)http://dx.doi.org/10.1016/j.tecto.2014.01.038mailto:tparsons@usgs.govhttp://dx.doi.org/10.1016/j.tecto.2014.01.038http://creativecommons.org/licenses/by-nc-nd/3.0/)http://www.sciencedirect.com/science/journal/00401951http://crossmark.crossref.org/dialog/?doi=10.1016/j.tecto.2014.01.038&domain=pdf

2 T. Parsons et al. / Tectonophysics 618 (2014) 134

3.5. Chile and Argentina . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 103.6. Baja California . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 113.7. Australia . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 123.8. Volcanic and geothermal regions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 123.9. Global subduction zones . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 123.10. All catalogs . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 143.11. Regions with no evidence of dynamic triggering . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 14

4. Interpretation of observations . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 154.1. Insights into remoteM 5 earthquake triggering . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 164.2. Interpretation of possibly delayedM 5 earthquake triggering . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 174.3. Delayed dynamic earthquake triggering and tremor in Greece . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 194.4. Possible causes of delayed dynamic earthquake triggering . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 194.5. Delayed higher magnitude dynamic triggering and speculation about earthquake nucleation models . . . . . . . . . . . . . . . . . . . . 22

5. Mainshock characteristics . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 235.1. Mainshock magnitude and range . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 235.2. Focal mechanisms . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 245.3. Comparative peak ground velocity and amplitude spectra . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 245.4. Azimuth . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 255.5. Summary of mainshock characteristics . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 27

6. Conclusions: What have we learned about remote earthquake triggering? . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 27Acknowledgments . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 28Appendix A. Supplementary data . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 28References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 28

Nu

mb

er

A) Microseismicity r>10GSN microearthquake rate before and after 15 M7.1 mainshocks

400

500

600

700

800

900

1000

1100

+2

Mean

Hours

B) M>5 seismicity r>10Catalog rate before and after 260 M7.0 mainshocks

-5 -4 -3 -2 -1 5

-5 -4 -3 -2 -1

0 1 2 3 4

0 1 2 3 4 5

+2

Mean

2

4

6

8

10

Fig. 1. In (A) remotely triggered earthquakes recorded on GSN stations identified byVelasco et al. (2008) are shown. The significant rate increase persists for slightly lessthan 1 h. Little is known about these events,whichwere not located by regional networks.In (B) a search of the 34-year M N 5 catalog shows no rate increase associated with 260M 7 mainshocks.

1. Introduction

Aftershocks of large (M 7) earthquakes can happen nearlyanywhere on Earth because their surface waves distort fault zones andvolcanic centers as they travel through the crust, triggering seismic fail-ures (Anderson, 1994; rnadttir et al., 2004; Beresnev et al., 1995;Brodsky et al., 2000; Cannata et al., 2010; Chelidze et al., 2011; Danielet al., 2008; Doser et al., 2009; Felzer and Brodsky, 2006; Glowackaet al., 2002; Gomberg, 1996; Gomberg et al., 1997, 2001, 2004;Gomberg and Bodin, 1994; Gomberg and Davis, 1996; Gomberg andFelzer, 2008; Gomberg and Johnson, 2005; Gonzalez-Huizar andVelasco, 2011; Gonzalez-Huizar et al., 2012; Hill, 2008; Hill et al.,1993; Hirosi et al., 2011; Hough, 2001, 2005, 2007; Hough andKanamori, 2002; Husen et al., 2004; Husker and Brodsky, 2004; Jianget al., 2010; Johnston et al., 2004; Jousset and Rohmer, 2012; Kilbet al., 2000; Lei et al., 2011; Lin, 2012; Meltzner and Wald, 2003;Miyazawa, 2011; Miyazawa and Mori, 2006; Mohamad et al., 2000;Moran et al., 2004; Pankow et al., 2004; Papadopoulos, 1998; Penget al., 2011a, 2010, 2012; Pollitz et al., 2012; Prejean et al., 2004;Savage and Marone, 2008; Shanker et al., 2000; Singh et al., 1998;Spudich et al., 1995; Stark and Davis, 1996; Steeples and Sreeples,1996; Sturtevant et al., 1996; Surve and Mohan, 2012; Taira et al.,2009; Tape et al., 2013; Tibi et al., 2003; Tzanis and Makropoulos,2002; Ukawa et al., 2002; Van Der Elst and Brodsky, 2010; Velascoet al., 2008; Wen et al., 1996; West et al., 2005; Wu et al., 2011,Yukutake et al., 2011; Zhao et al., 2010). Example results (Velascoet al., 2008) are reprised in Fig. 1; hundreds of Global SeismographNetwork (GSN) stations that recorded surface waves from 15 M 7.1mainshocks were filtered and analyzed for local events. A nearly two-fold rate increase is evident when the observations are stacked(Fig. 1A). We plot results from a catalog search for M N 5 events on thesame time range scales (Fig. 1B), but noM N 5 rate increase is associatedwith 260 M 7 mainshocks (e.g., Huc and Main, 2003; Parsons andVelasco, 2011).

At near radii (r b 1000 km) there is a very clear (~50-fold) M N 5earthquake rate increase during the first hour after 260 M 7mainshocks that decays rapidly by Omori's law, and is obvious for atleast 10 days (Fig. 2). The same analysis for the rest of the planet outside1000 km radii from mainshocks shows no detectible rate increaseduring any period (Fig. 2B). The 1000 km radius was chosen becauseParsons and Velasco (2011) found that to be the greatest distance thatsignificant M N 5 earthquake rate increases were seen. Elevated rates

within a 300 km radius are observed to persist for ~710 years(Faenza et al., 2003; Parsons, 2002).

Key questions then are: Why aren't dynamically triggered M N 5earthquakes correlated with passing surface waves across the global af-tershock zone theway smaller earthquakes are? Is there no comparablehazard in the global aftershock zone to that in the local zone? Perhapswe haven't yet observed this simply because M N 5 earthquakes areorders ofmagnitude less frequent than smaller shocks by theGutenbergand Richter law (log(N) = a-bM; Ishimoto and Iida, 1939; Gutenbergand Richter, 1954). However, extrapolation of the Velasco et al. (2008)

0

24

-10 -8 -6 -4 -2 0 2 8

-180 -120 -60 60 1200 -180 -120 -60 60 120 180

-8 -6 -4 -2 0 24 6 4 6 8 10

-24 -18 -12 -6 0 6 12 18 -18 -12 -6 0 6 12 18

Days Days

Days Days

Hours

Nu

mb

er

Hours

C) r1000 km

A) r1000 kmRate of M>5 catalog earthquakes before and after 260 M7 mainshocks

0

500

1000

1500

2000

2500

3000

0

100

200

300

400

500

600

0

1000

2000

3000

4000

5000

6000E) r1000 km

Fig. 2. In (A) allM N 5 earthquakes that occurred in 24 h periods before and after 260M 7mainshocks, and within a 1000 km radius of themainshocks are stacked. A clear rate increaseandOmori-law decay is evident. In (B) the same process is applied except all globalM N 5 events outside of the 1000 km radius are considered. No rate change is evident. In (C) and (D) thesame process is followed except 10 day intervals are considered. Periods of 180 days are shown in (E) and (F).

3T. Parsons et al. / Tectonophysics 618 (2014) 134

observations, assuming that the maximummagnitudes detected lie be-tweenM=2 andM=3, and a b-value=1, implies that about 70M N 5events should be observed within 15 min of surface wave passage(Parsons and Velasco, 2011).

We can gain some insight by examining a specific location such asthe Basin and Range province, which demonstrated widespread remoteearthquake triggering after the 2002 M = 7.9 Denali earthquake(Gomberg et al., 2004; Husker and Brodsky, 2004; Pankow et al.,2004; Fig. 3). While the seismicity rate is clearly increased significantlyby Denali surface waves, the overall rate of triggered earthquakes is toolow to necessarily expectM N 5 earthquakes during the first 24-h periodfollowing the mainshock. This can be determined by extrapolating theGutenbergRichter distribution based on the number of M N 2 eventsat a b-value = 1, which yields an expected rate of M N 5 events to be~0.6/day.

An intriguing (and concerning from a hazard perspective) explana-tion for the lack ofM N 5 remote triggering directly associatedwith pass-ing surface waves is the possibility that larger magnitude earthquaketriggering occurs, but is delayed relative to surface wave arrivals.Many such cases of delayed (hours to days) larger earthquake occur-rence have been temporally correlatedwithmainshocks at remote glob-al distances (e.g., Gomberg and Bodin, 1994; Gonzalez-Huizar et al.,2012; Pollitz et al., 2012; Tzanis and Makropoulos, 2002). If the re-sponse/nucleation time is longer for a larger earthquake than a smallermagnitude event, then there may be information about the initialphases of the earthquake rupture process being conveyed, and a sugges-tion that this may be magnitude dependent.

In global analyses to date, systematic regional observations of seis-mic response to passing surface waves across the magnitude scale are

lacking. Since M 5 triggering during surface wave arrivals appears tobe rare or absent, we want to look at as broad a magnitude range as ispossible on regional networks where non-detection of M 5 events isnearly impossible. We take the approach that if we can amass as manyunequivocal remote-triggering responses (like that in Fig. 3) as possible,then we can more confidently assess large earthquake triggering bygreatly reducing the possibility of coincidental associations.

In this review we examine 21 local and regional seismic catalogsfrom many parts of the world (Fig. 4) for response to 260 M 7 globalmainshocks. This paper is therefore an earthquake catalog review ratherthan a literature review. We address the following questions: (1) howoften is there a significant increase in seismic activity at a given locationin response to an earthquakemore than 1000 kmaway? (2)What is themagnitude distribution of dynamically triggered earthquakes? (3) Iflarge earthquakes are triggered, are they always preceded by a cascadeof lower magnitude events? (4) Is there any information from magni-tude response that might enable speculation about the earthquakenucleation process? (5) Are there identifying features in commonamong mainshocks that cause remote triggering?

2. Methods and data

Looking at stacked data from many locations simultaneously in-creases the number of events and adds statistical power to a triggeringanalysis, but this also makes it difficult to grasp regional frequencyand variability in triggering response to passing surface waves. Further,the events shown in Fig. 2Awere recorded at single stations rather thanby a regional network,meaning no detailed information about locationsand magnitudes is available. The stackedM N 5 events shown in Fig. 2B

-24 -18 -12 -6 0 6 12 18 240

5

10

15

20

25

Nu

mb

er

Time (hr) relative to Denali EQ origin

-130 -125 -120 -115 -110 -105 -100

35

40

45

50 A

B

Nu

mb

er

Magnitude

1

10

100

300

1.2

1.6

2.0

2.4

2.8

3.2

C

Fig. 3. Remote earthquake triggering in the Basin and Range extensional province of the western United States is shown. In (A) amap of seismicity 24 h prior to (blue) and after (red) the2002M=7.9 Denali earthquake is shown. (B) A histogramof earthquake number per 30min is shown that demonstrates the earthquake rate increase observed byGomberg et al. (2004),Husker and Brodsky (2004), and Pankow et al. (2004). The cumulative magnitude frequency of the post-Denali seismicity is shown in (C); extrapolation of this relation to M N 5 ratessuggests an expected rate of ~0.6 M N 5 events/day.

4 T. Parsons et al. / Tectonophysics 618 (2014) 134

have magnitude and location information, but represent only thesparsest part of the magnitude spectrum, and only tell a partial story.

The backbone of this review is thus a compilation of earthquake cat-alogs that are complete to lower magnitudes. These are secured from avariety of sources including the Advanced National Seismic System(ANSS), which assembles numerous regional USA and internationalnetwork catalogs together, the Japan Meteorological Agency, the WorldData Center for Seismology Beijing, Geoscience Australia, GNS ScienceNew Zealand, Istituto Nazionale di Geofisica e Vulcanologia in Italy, TheKandili Observatory in Turkey, and The National Observatory of Athens

Fig. 4. Map of the regions sampled and discussed in this review of global seismic response toenough seismic station coverage to enable a complete earthquake catalog from M 2. In oththe global subduction interface catalog of Heuret et al. (2011) was included and illustrated by

in Greece. The Global Seismograph Network (GSN) catalog is used tofill inwhere no local network observations are available. Areas are select-ed either because of catalog availability constraints, or as representativesampling. All data are assembled prior to analysis, and in no cases arecatalog bounds or other properties altered after examination. We seekcatalogs from active regions with quality networks as well as samplesfrom all continents and different tectonic environments. We end upwith 21 individual catalogs with a cumulative 1,524,873 unique events.

A keymotivation for using these regional catalogs is that their lowercompletion magnitudes (typically M = 2) means that the question of

teleseismic surface waves. Many of these regions were selected because they have denseer cases the ANSS/GSN catalog was applied to sample the major continents. Additionally,the red lines.

-20

-15

-10

-5

0

5

10

15

20

1979 1984 1989 1994 1999 2004 20094.5

5.5

6.5

7.5

Ch

ang

e:

24

ho

urs

of

M>7

glo

bal

mai

nsh

ock

Mag

nitu

de

Basin and Range

Change 1 day M=7.9Denali, AK

M=7.0Kuriles

M=8.6IndianOcean

M=9.2Sumatra

Year

1

2

34

5

2

1

mean

Fig. 5.Daily rate changes in the Basin andRange Province of thewestern USA (see Fig. 4 forlocation) following 260M 7.0 globalmainshocks. Blue spikes show changes in the num-ber of earthquakes in the catalog for 24 h periods before and after each of 260 globalM 7.0 mainshocks. The mean daily rate increases across the whole catalog are shownby the light blue line. One and two standard deviations (1 and 2 ) on daily rate changesare shown by dashed red lines. We take daily rate changes that exceed two standard de-viations as significant. Important events are labeled, numbered, and analyzed in Fig. 6.Local earthquakes ofM 4.5 are indicated by the stick plot along the base of thehorizontalaxis; those shaded in red occurred within 24 h after the global mainshocks.

5T. Parsons et al. / Tectonophysics 618 (2014) 134

remote higher-magnitude triggering can be directly addressed. The re-sults presented in Fig. 1B show stackedM 5 rates that are unchangedafter surface wave passage. Questions from that analysis remain that in-clude the expectedM 5 rates during these short intervals (hours), andpossible masking of events in the global catalog. However, in a regionalcatalog that is complete to lowmagnitudes, it is virtually impossible thata M 5 earthquake could be missed. Further, we can extrapolateexpected numbers of M 5 shocks based on the lower magnituderates, and by assuming GutenbergRichter magnitudefrequencyrelations, determine if there are absent high magnitude events.

A second catalog of 260 global M 7.0 mainshocks is also assem-bled; the M 7.0 threshold is arbitrary, but this magnitude wasshown to be capable of triggering earthquakes at global distances byVelasco et al. (2008), and we adopted the same threshold for theParsons and Velasco (2011) study. The duration of the mainshock cata-log runs from 1979 through 2012 and includes 41 new potentialM 7triggers over the catalog used by Parsons and Velasco (2011), includingthe February 2010M=8.8Maule,March 2011M=9Tohoku, andApril2012 M = 8.6 Indian Ocean events. All earthquakes used in this studyare shallow, spanning 050 km in depth.

In this reviewwewant to test the broadest magnitude spectrum pos-sible for remote triggering, particularly in light of the disparity illustratedin Fig. 1. We therefore include the lowest magnitudes available in eachregion, but we do not imply that this value represents a magnitude ofcompleteness. As described below, we compare 24 h, local earthquakerate changes associated with remote sources, and therefore assume thatdetection thresholds are unchanged over these 48-h periods. The primaryoccurrence that could affect this assumption would be the periodjust after a large local earthquake, when data losses are expected(e.g., Iwata, 2008; Kagan, 2003). To obviate this, we track the occurrenceof larger earthquakes within regional catalogs very carefully, and any sig-nificant daily rate change that is observed is hand checked for local effects.

Another concern might be the data losses for lowmagnitude eventsduring the passage of surface waves across local networks. From D.Oppenheimer, personal communication (2013) we note with regardto ANSS stations, For short-period stations, the passband is 0.530 Hz,so the surface waves are mostly outside the passband, and the pickerdoes a fair job detecting the local, triggered events. However, the shortperiod stations are typically analog, so the signal clips if the surfacewaves are big. In this case, we can't easily time the local events. Onmore modern digital stations (after 2005) we avoid that problem, asthe dynamic range of the sensors is high enough. Therefore it is possi-ble that we lack coverage during the actual passage of surface waves,particularly at lower magnitudes; this problem is reduced at about theM ~ 5 threshold because GSN stations can observe them remotely atmany locations where the mainshock and triggered event arrivals donot interfere.

We apply the following procedure to every catalog. We begin bycalculating the observed daily change in the number of earthquakes ineach regional catalog, excluding the 260 24-h periods after globalmainshocks occurred (Fig. 5). This is intended to establish the expectedbackground daily variability that is not affected by global mainshocks.We establish the mean daily change and the variance on that changeby examining 2-year windows at 0.5 year intervals (the preceding2 years of observed rate changes are evaluated at each 0.5-year inter-val). We do this because virtually all catalogs grow more completeand record more events with time as new stations are installed, thusthe magnitude and significance of the mean daily rate change willchange with time. Additionally, earthquake rates fluctuate dramaticallywhen larger events occur within the region that trigger manyaftershocks. We calculate the mean rate change and significanceindependently for increases and decreases because aftershocks cancause instant rate increases to a degree that cannot occur as a daily de-crease. We experimented with different durations used to calculate themean daily rate changes, and settled on 2 years as an optimal balancebetween having sufficient numbers to calculate a stable mean, while

still representing catalog time-dependence. We do not decluster thecatalogs, because we are looking for clustering behavior caused byremote mainshock triggers.

We calculate time dependent variance and hence standard devia-tions () on the mean rate changes by fitting daily rate changes over2-year periods to negative binomial distributions, which are found tobetter represent clustered phenomena (e.g., Jackson and Kagan, 1999;Vere-Jones, 1970). An indication that a negative binomial distributionis a more appropriate than a Poisson process occurs when the data aredispersive, with the variance greater than the mean. We apply a maxi-mum likelihood regression technique (Cameron and Trivedi, 1998)that starts with fitting a Poisson model, then a null model (interceptonly model), and finally the negative binomial model. We iterate untilthe change in the log likelihood is vanishingly small. We estimate thedispersion () inherent to each catalog from the maximum likelihoodregressions, calculate variance as var = + 2, and find the 1 and2 variations on rate changes from the variance. We note significantdispersion in every catalog that we analyzed, with ranging from0.19 to 0.55, which means a Poisson process is rejected.

We isolate earthquake rate changes in regional catalogs across 24h periods relative toM 7.0 globalmainshock events that happenmorethan 1000 km away from any of the events in the regional catalog. The1000 km distance was chosen because it was the maximum distancewhere earthquake rates were detected significantly above backgroundlevels by Parsons and Velasco (2011) during the first 24 h following205 post-1979mainshocks. It was thus interpreted to be the maximumextent of static stress triggering. Global distance ranges betweenmainshocks and possibly triggered events are calculated with the in-verse method of Vincenty (1975) using the NAD83 ellipsoid. We high-light local rate changes that exceed a 2 level above the mean. Wetake the 2 threshold to be a guideline because an exact confidence in-terval depends on the degree of smoothing that results fromthe duration of the catalog used to calculate it (2 years in incremental0.5-year steps in this review) and on the statistical distribution used(negative binomial). Therefore, if a rate increase approaches the 2 ,or if a specific mainshock was noted to cause remote triggering byother authors, we investigate it as a possible example of remote

6 T. Parsons et al. / Tectonophysics 618 (2014) 134

triggering. Our results depend on the chosen significance threshold,with an increase or decrease changing the number of triggering casesthat we identify. The detailed analyses that we conduct suggest thatwe admit more cases for consideration than we omit.

The 1000-km exclusion zone removes the possibility of local, staticstress change induced processes from being mistaken for remote trig-gering. Significant local events of M b 7.0 that happen within 24 h of aglobal mainshock tend to be associated with their own aftershocks,which then contribute to a significant earthquake rate increase. Indeedthese sequences could be a cascade that is set off by a globalmainshock,or could instead be a coincidence. We therefore examine everysignificant rate increase in detail to establish its character.

We find an array of responses to remote earthquakes in regionalcatalogs that range from: (1) widespread seismicity rate increases,(2) isolated local mainshocks and associated aftershocks, (3) swarm in-vigoration, to (4) no significant response. We define probable remotetriggering as being a widespread seismicity rate increase without anobvious local cause (Fig. 3). We define possible remote triggering as

0

20

40

60

0 5 10 15 20-5-10-15

Nu

mb

er

0 5 10 15 20-5-10-150

5

10

15

20 2004 M=9.2 (Sumatra)

2010 M=7.0 (Kuriles)

37

38

39

40

-120 -11

-120 -113 subregions

3 subregio

2 subregion

M=5.6

M=4.

M=4.2

37

38

39

40

(1

(9/18/20

(8/4/2

(4/11/2

MonoLake

0 5 10 15 20-5-10-150

5

10

15

20

20 days: Basin and Range Region 2

43

44

45

-112

Jacks

Near BCanyo

2004 M=7.0 (C. America)

2012 M=8.6 (Indian Ocean)

0 5 10 15 20-5-10-15

Days after global mainshock

0

5

10

15

20

37

38

39

-113

1

3

4

5

California

A

B

C

D

Fig. 6.Detailed analysis of remote earthquake triggering in the Basin and Range province of thewjust a decaying aftershock sequence. In (B) aM=3.9 event occurs ~3 h after the 2004M=9.2rate increase associatedwith the 2002M=7.9 Denali earthquake shown in Fig. 3. Similarly in (can be tied to remotemainshocks. Their isolationmeans these events could easily be coincidentchanges in 23 0.5 by 0.5 subregions, whereas we calculate a 2 (2 standard deviations) sigaffected 16 subregions. Locations of insets are shown in (A).

being a localized earthquake and aftershocks that may have occurredby chance, ormay have been triggered.We define swarm invigorationas an already active zone of seismicity that intensifies after surfacewaves pass through the region from a remote mainshock. To add asystematic way of defining these responses, we quantify their spatialnature by dividing our study regions (Fig. 4) into 0.5 by 0.5 boxesand calculating the mean and variance of the number of subregionsthat display 24 h seismicity rate changes in 100 random trialsacross catalog durations. We then calculate how many subregions dis-play rate increases for each significant regional response to globalmainshocks. If this number exceeds a 2 threshold in the number of0.5 by 0.5 boxes from random trials and there is no local mainshock,thenwe identify the response aswidespread, and thus probable remotetriggering. In other words, we want to find out what the normal dailyspatial variability in seismicity rate is, and what constitutes an anoma-lous region-wide change.

Examples are shown in Figs. 5 and 6; in this case the Basin and Rangeprovince catalog is analyzed (see Fig. 4 for location). This catalog

8 -116 -114 -118 -116 -114

8 -116 -114 -118 -116 -114

M=4.6

increased (2 threshold=8.9)

ns increased (2 threshold=8.9)

s increased (2 threshold=8.9)

2: 1.1 hrs delay

: 18.5 hrs delay

0/9/2004)

04)

010)

012)

(12/26/2004)

M=3.92 2h 57 min after M=9.0

4 hours: example Basin and Range sub-regions

-110

on Hole, WY

ryce n NP, UT

-111

Nevada

estern USA. In (A) it is demonstrated that the apparent significant rate decrease is actuallySumatra earthquake. This response is spatially limited as comparedwith themore regionalC) and (D) isolated clusters of events near JacksonHole,Wyoming, and Bryce Canyon, Utahal, and not examples of remote triggering. Each of these responses are associatedwith ratenificance threshold of 8.9. The response to the Denali earthquake in the Basin and Range

0.1

1

10

100200

2 3 4 502468

10

2 3 4 502468

10

5.0 6.0 7.0 8.0 9.0 6.0 7.0 8.0 9.0 6.0 7.0 8.0 9.0

5.0 6.0 7.0 8.0 9.0 6.0 7.0 8.0 9.0 6.0 7.0 8.0 9.0

5.0 6.0 7.0 8.0 9.0 6.0 7.0 8.0 9.0

5.0 6.0 7.0 8.0 9.0 6.0 7.0 8.0 9.0

6.0 7.0 8.0 9.0

2 3 4 502468

10

0.1

1

10

100

2 3 4 5

2468

10

2 3 4 502468

10

2 3 4 502468

10

0.1

1

10

100

2 3 4 502468

10

2 3 4 502468

10

2 3 4 502468

10

0.1

1

10

100

2 3 4 502468

10

2 3 4 502468

10Real data

Randomcatalogsamples

Magnitude-frequency: 24 hours after 260 M7 mainshocks and 260 random times

Nu

mb

er

Magnitude

n: M6

Nu

mb

er CBA

FED

HG

KJ

I

Fig. 7. Comparison of (A) observed and (BK) randomized incremental magnitude-frequency distributions for 260 24-h periods in the globalM N 5 earthquake catalog. The inset histo-grams showhowmany of 260 random24-h periods had between 2 and 5M 6.0 earthquakes in them. In otherwords, howmany days out of 260 are there 2 ormoreM 6.0 earthquakesby random chance? The point of this figure is to show that by pure coincidence, at least 4 of 260 24-h periods have 2 to 5M 6.0 earthquakes in them. Except in (A), these periods are notpreceded by a M 7.0 mainshock.

7T. Parsons et al. / Tectonophysics 618 (2014) 134

contains 47,791 M 2.0 events. In addition to the very clear rate in-crease associated with the 2002 M = 7.9 Denali earthquake alreadyshown in Fig. 3, three other rate increases at 2 are observed, associatedwith the 2004 M = 9.2 Sumatra earthquake, a 2010 M = 7.0 Kurilesevent, and the 2012M = 8.6 Indian Ocean shock.

It is common to see significant earthquake rate reductions associatedwith 24-h periods after global mainshocks (for example, the eventlabeled 1 on Fig. 5). In every instance throughout our global analysis,these rate decreases are caused by declining aftershock sequences oflocal earthquakes. What happens in these cases is that a moderate tolarge regional earthquake occurs, usually the day before one of the260 global mainshocks, and we thus measure a strong rate decreasefrom a decaying aftershock sequence that has nothing to do with the

global event. This is illustrated in Fig. 6; the rate decrease labeled 1in Fig. 5 is associated with a M = 4.6 earthquake that happened onthe CaliforniaNevada border the day before a M = 7.0 CentralAmerica mainshock. The M = 4.6 event is likely itself an aftershock ofM = 5.6 earthquake at the same location 21 days previously. Thehistogram of daily earthquakes in the local area shows that an after-shock sequence of the M = 4.6 event was decaying when the CentralAmerica mainshock occurred (Fig. 6).

Also demonstrated by Fig. 6 is the variety of remote triggering re-sponse that can only be established by looking at regional networks.The widespread seismicity rate increase observed after the 2002M = 7.9 Denali earthquake is associated with 16 unique 0.5 by 0.5subregions showing a rate increase compared with a mean of 4.4 and

-40

-30

-20

-10

0

10

20

30

40

4.5

5.5

6.5

7.5

A) Northern CA

B) Southern CA

Ch

ang

e:

24

ho

urs

of

M>7

glo

bal

mai

nsh

ock

Ch

ang

e:

24

ho

urs

of

M>7

glo

bal

mai

nsh

ock

Mag

nitu

de

Mag

nitu

de

4.5

5.5

6.5

7.5

1991 M=7.2Mid-Atlantic Ridge

1989 M=7.0Solomon Is.

2003M=7.2New Zealand 2008

M=7.2China

2012M=7.4CentralAmerica

1

1 22

02

34

-40

-30

-20

-10

0

10

20

30

40

Nu

mb

erTime (hours) after M=7.0 New Britain event

0

0 8-8-16

1979 1984 1989 1994 1999 2004 2009

1979 1984 198 1994 1999 2004 2009

-24 16 24

10

20

30

40

50

Aftershock decay fromlocal 2010 M=7.2 El MayorCucapah EQ

Loma Prietacascade fromM=5.1 aftershock

1992 M=7.2 Nicaragua

2010 M=7.5Indonesia

3 Local swarm near Borrego Springs initialtes before 2001 M=7.0New Britain:

Fig. 8. Daily rate changes in (A) northern and (B) southern California, USA. Blue spikes show changes in the number of earthquakes in the catalog for 24 h periods before and after each of260 globalM 7.0 mainshocks. Primary plot features are the same as in Fig. 5. Important events are labeled, numbered, and analyzed in Fig. 9. We identify 4 incidences of significant re-mote triggering in northern California, and 3 in southern California. The event labled 0 in (A) is a coincidence between a localM=5.1 aftershock (and secondary aftershocks) to the 1989M=7.0 LomaPrieta earthquake and aM=7.0 Solomon Islands event. One apparently significant rate change in southernCalifornia (labeled 2) began before the globalmainshock that itis temporally associated with (see hourly histogram at right).

8 T. Parsons et al. / Tectonophysics 618 (2014) 134

a 2 threshold of 8.9 (Fig. 3). However, the other rate increases that aretemporally associated with global mainshocks identified in Fig. 5 are allsmall clusters of events that result from an initial, moderate (M=3.9 toM = 4.6) earthquake that is followed by smaller local aftershocks(just 23 subregions with rate increases). The timing of these initialmoderate events falls within ~1 to 18 h after global mainshocks,meaning that they could be examples of delayed dynamic triggering,or they could simply be coincidental occurrences.

In the following sectionwe report results of similar analyses across awide variety of global regions and tectonic environments to learn moreabout how faults respond to transient strains imposed by passingsurface waves from distant earthquakes.

3. Observations

In the following discussion we will tour and sample the world'searthquake catalogs (Fig. 4).We describe a variety of regional responsesto global mainshocks that range from no significant response towidespread regional seismicity rate increases. We focus on areas withnotable reactions, but also note those regions that do not appear to beaffected (these non-observations are appended in the supplementarydata section).

Before describing individual regional responses, it is necessary tokeep in mind the distinct possibility of coincidental events; we are de-scribing temporal correlations of earthquakes that occur sometimeson the opposite sides of the earth, often in regions of high seismic activ-ity.We therefore look at sets of 260 24-h periods drawn at random fromthe global 19792012 earthquake catalog to find how many M 6.0earthquakes are expected by chance. The M 6.0 threshold is used inthese synthetic tests to ensure consistent catalog completion back toits earliest period to enable a fair comparison to the actual catalog,because randomized 24-h periods could have a different temporaldistribution than the actual mainshocks. A group of 10 assemblies isshown in Fig. 7. In every case, a minimum of four 24-h periods had atleast two M 6.0 earthquakes that occurred without any globalM 7.0mainshock preceding them. Themagnitude frequency relationsof the random draws of M 6.0 earthquakes are not distinguishablefrom that of the actual 24-h periods that follow global M 7.0mainshocks, which means that we are not able to rule out randomchance in cases where we observe a significant rate change that is de-scribed by single local earthquake and its local aftershocks that followa remote M 7.0 mainshock. In other words, any M 6.0 globalevent linked in time with a M 7.0 mainshock can always be a coinci-dence, and that as many as 5 M 6.0 events can happen on a givenday purely by chance.

36

37

38

39

40

0 10-10

25

50

-20 201 2 3 4 5 6 7

0 10-10-20 201 2 3 4 5 6 7

0 10-10-20 201 2 3 4 5 6 7

0 10-10-20 201 2 3 4 5 6 7

24 hours before global mainshocks

24 hours after global mainshocks

Daily RateMagnitude-Frequency (24 hours)

1991 M=7.2Mid-Atlantic Ridge

36

37

38

39

0

25

50

2003 M=7.2New Zealand

0

25

50

25

50

36

37

38

39

0

25

50

0

25

50

2008 M=7.2China

124 12236

37

38

39

124 122 120

0

25

50

0

25

50

2012 M=7.4CentralAmerica

Magnitude Days after global mainshocks

Nu

mb

er

Nu

mb

er

1

2

3

4

10 subregions increased (2 threshold=6.6)

12 subregions increased (2 threshold=6.6)

7 subregions increased (2 threshold=6.6)

3 subregions increased (2 threshold=6.6)

Fig. 9. Detailed analysis of remote earthquake triggering in northern California, USA. Each of the four significant rate increases numbers 14 in Fig. 8 are detailed in beforeaftermaps, incremental magnitudefrequency histograms, and 20 day events-per-day plots. Before information is shaded blue, and after in red. We call these occurrences probable remotetriggering because no precipitating local event is evident, and the rate-increase is regional in nature.

9T. Parsons et al. / Tectonophysics 618 (2014) 134

3.1. California

We compare two 4 by 4 areas in northern and southern Californiathat are centered over the active San Andreas fault system (Fig. 8). Thenorthern California catalog has 233,570M 1.0 events, and the southernCalifornia cataloghas 358,927M1.0 shocks (our use ofM1.0 does notimply a completeness threshold of M = 1.0, but rather that M = 1.0events are present in the catalog). In northern California we note fivecaseswith significant rate increases. The first (labeled 0 in Fig. 8) is like-ly a coincidence because it is associated with a cascade of aftershocks to alocal M = 5.1 earthquake, itself a local aftershock to the 1989 M = 7.0Loma Prieta earthquake. While it is not impossible that the M = 5.1event was triggered by global mainshock, the least astonishing parent

mainshock would be the nearby Loma Prieta rupture. A trio of M = 7.2mainshocks from themid-Atlantic ridge, NewZealand, andChina are con-sidered probable dynamic triggering examples in that they represent re-gionally distributed seismicity rate increases that are not associatedwith any one local higher magnitude event (Fig. 9). The 2 thresholdfor the number of 0.5 by 0.5 subregions showing a rate increase is 6.6,and the responses to these three events indicate effects in 712 subre-gions. None of the four mainshock triggers in northern California overlapwith any thatmay have affected the Basin and Range Province (Figs. 5, 8).This is a common thread throughout our analysis, with virtually no over-lap amongst mainshocks, which suggests that conditions have to be idealfor remote dynamic triggering to occur (e.g., Gonzalez-Huizar andVelasco, 2011; Hill, 2008; Parsons et al., 2012).

10 T. Parsons et al. / Tectonophysics 618 (2014) 134

The southern California catalog shows three significant earthquakerate changes that are associated temporally within 24 h of 260 globalmainshocks. The first signal in 1992 comes 65 days after the regional1992 M = 7.4 Landers earthquake, and activation is concentrated inthe Landers aftershock zones (Supplementary Figure S1), though theaffected area is broad enough to be classified as probable triggeringwith 16 0.5 by 0.5 subregions having rate increases compared witha 2 threshold of 8.5. This could therefore be a case of aftershock invig-oration induced by remote dynamic stressing, or it could be a processrelated directly to the Landers earthquake. There is another apparentrate increase in 2001 (Fig. 8), but this is a swarm that actually initiated2 h before the global mainshock it is associated with (a M = 7.0 NewBritain event). This appears like a 24 h rate increase because theregionwas very quiet before the swarm such that there aremore cumu-lative events in the post mainshock period. This sort of occurrence dem-onstrates the importance of careful study of each apparent rate change.The last significant southern California rate change is associated with a2010 M = 7.5 Indonesian mainshock (Fig. 8). This again follows aregional mainshock, the 2010 M = 7.2 El Mayor-Cucapah event, andactivity is again almost entirelywithin the aftershock zone of that earth-quake, and affects 8 subregions, and is thus considered possible remotetriggering.We note that all three cases for remote dynamic triggering insouthern California are ambiguous because of their association withprior swarm and/or aftershock sequences

3.2. Greece

We study a large region that encompasses Greece and parts ofTurkey (Figs. 10, 11) following the same procedures as applied toCalifornia and the Basin and Range province. This catalog spans from1983 through 2012 and contains 131,016 M 1.0 events. It is clearfrom examining Fig. 10 that the completeness of this catalog is stronglytime dependent (the initial portion is mostlyM 3.0 events). We noteseven cases that demonstrate a significant rate increase that can be tiedto global mainshocks (Fig. 10). At least four cases in Greece can be clas-sified as probable dynamic triggering because in each instance there is aregionally broad response that is difficult to tie to a local mainshock(Fig. 11), with each case having more than twice as many 0.5 by 0.5subregions showing rate increases than the 2 threshold of 13.4.

Other features of note from Greece include a case where a 2008aftershock sequence from a local M = 5.1 in decline is possibly

-100

-80

-60

-40

-20

0

20

40

60

80

100

1979 1984 1989 1994 1999 2004 20094.5

5.5

6.5

7.5

Ch

ang

e:

24

ho

urs

of

M>7

g

lob

al m

ain

sho

ck

Mag

nitu

de

Year

1995 M=7.2 (Kermedec)

2007 M=7.1 (New Hebrides)

2008 M=7.4 (Sumatra)

2006 M=7.6(Kamchatka)

2009 M=7.0 (Sumatra)

Greece

12

53

46

2009 M=7.4(Kurils)

2010 M=7.1(Solomon Islands)

7

Fig. 10. Daily rate changes in Greece; see Fig. 11 for location. Primary plot features are thesame as in Figs. 5, and 8. Seven significant rate increases can be tied to global mainshocks.

reinvigorated by a M = 7.4 mainshock in China, and becomes the siteof an M = 6.1 event (Fig. 11, event labeled 4). Additionally, a 2007M =7.1mainshock from the NewHebrides region is temporarily associ-atedwith a persistent and regional seismicity rate increase the goes on forat least 20 days (Fig. 11, event labeled 3). Fig. 12 shows a before/aftermapping of this rate increase, and its regional and temporal extent. Thishappened during the period between September 2006 and May 2007that was identified as a seismic crisis by Bourouis and Cornet (2009).We find more cases of possible and probable remote dynamic triggeringin Greece than any other region we study, which is consistent with theconclusions of Brodsky et al. (2000) that the Greek region has a low trig-gering threshold, and is subject to superswarms. We do not includetheir 1999 M = 7.4 Izmit mainshock example in our analysis because itfalls within the 1000 kmexclusion zonewe apply throughout this review.

3.3. New Zealand

We examine a large catalog (329,044 M 1.0 events) that encom-passes the islands of New Zealand and note six significant earthquakerate increases that can be associated with global mainshocks (Figs. 13,14). We interpret four of these rate increases as probable remote trig-gering based on our defined criteria of regional response without aclear local trigger (events labeled 25 on Fig. 14). The response labeled6 in Fig. 14 falls into our category of possible remote triggering be-cause the rate increase is caused by aftershocks of a M = 6.7 localmainshock that occurred 22.9 h after a M = 7.2 Aleutian Islands earth-quake. Another rate increase we note falls into another category we callswarm invigoration, where an ongoing swarm appears to be en-hanced by the occurrence of a remote mainshock. In this case (event3 on Fig. 14) an earthquake swarm just south of Rotorua in theTaupo Volcanic Zone was ongoing at the time of the 2008 M = 8.1Antarctic plate earthquake, and then the rate of events doubled in thefollowing 24 h. We note a few other cases of swarm invigoration inother regions.

In all we find 4 probable cases of remote dynamic triggering in NewZealand with the number of 0.5 by 0.5 subregions showing seismicityrate increases exceeding the 2 threshold of 16.2 (Fig. 14).

3.4. Southeast China

We study a catalog of 6384M 3 events recorded inmoderately ac-tive southeast China that is likely to be complete at that level (Mignanet al., 2013). This region was chosen for study because it is adjacent tothe very active western Pacific subduction zones, and is thus an areathat is frequently traversed by high amplitude surface waves from justoutside our 1000 km exclusion zone. Despite this characteristic, wenote only one possible case of remote triggering that is associatedwith a California mainshock, the 1989M=7.0 Loma Prieta earthquake(Fig. 15A).We consider this a case of possible remote triggering becauseall the activity is isolated within the 1989 Shanxi Datong earthquakeswarm (Zhang et al., 1995), that began ~15 h after the globalmainshock(Fig. 15B). Thus this could be coincidental or related.

3.5. Chile and Argentina

We sample a catalog in southern South America that contains 19,840mostlyM 4.0 events from GSN sources (Fig. 16). We observe two sig-nificant rate increases that are not associatedwith localmainshocks. Thefirst happened in 1994, and is temporally associatedwith a remoteM=7.1New Zealand mainshock. The seismicity in Chile during the 24 h after thisglobal mainshock is interesting because it begins about 2.9 h afterthe New Zealand earthquake, and consists of swarm-like M ~ 3.5 to M~ 4.5 events (and likely many small magnitude events not present inthe GSN catalog). Similar to the case described in China above, we clas-sify this as possible dynamic triggering. The second rate increasewe ob-serve is associated with the 2010 M = 8.8 Maule earthquake and its

34

36

38

40

42

44

2

0-10 10 20-20

0-10 10 20-203 4 5 6 7

2 3 4 5 6 7

2 0-10 10 20-203 4 5 6 7

2 0-10 10 20-20

0-10 10 20-20

3 4 5 6 7

2 3 4 5 6 7

0-10 10 20-20

0-10 10 20-20

2 3 4 5 6 7

2 3 4 5 6 7

0

5

10

15

20

1995 M=7.2 (Kermedec)

Magnitude Days after global mainshock

Nu

mb

er

Nu

mb

er

34

36

38

40

42

44

2007 M=7.1 (New Hebrides)

34

36

38

40

42

44

2008 M=7.4 (Sumatra)

0

10

20

30

40

0

50

100

150

M=5.1

34

36

38

40

42

44

15 18 21 24 27 30 33 15 18 21 24 27 30 33

2009 M=7.0 (Sumatra)0

25

50

75

24 hours before global mainshocks 24 hours after global mainshocksDaily Rate

Magnitude-Frequency (24 hours)

0

20

40

60

80

0

40

80

120

160

0

50

100

150

200

0

25

50

75

0

50

100

150

0

50

100

150

Before

After

1

34

36

38

40

42

44

2006 M=7.6 (Kamchatka)

Nu

mb

er

0

20

40

60

80

2

3

4

6

34

36

38

40

42

44

0

10

20

30

40

0

20

40

60

80

34

36

38

40

42

44

0

10

20

30

40

2009 M=7.4(Kurils5

2010 M=7.1 (Solomon Islands)7

10 subregions increased (2 threshold=13.4)

13 subregions increased (2 threshold=13.4)

8 subregions increased (2 threshold=13.4)

17 subregions increased (2 threshold=13.4)

15 subregions increased (2 threshold=13.4)

27 subregions increased (2 threshold=13.4)

28 subregions increased (2 threshold=13.4)

Fig. 11.Mapping of the significant rate changes associated with global mainshocks in Greece as identified and numbered in Fig. 10. Blue shading in the earthquake epicenters and mag-nitude frequency plots denotes 24-h periods before globalmainshocks, and red shading 24 h after. Histograms show20days around globalmainshocks. There are four cases (events 4,5, 6, and 7) of probable dynamic triggering because they affect a wide region, and are not associated with local mainshocks. Event 3 is associated with a long-term rate increaseacross Greece (see Fig. 12). Event 4 may have reinvigorated an aftershock sequence from a local M = 5.1 event that began about 6 h before theM = 7.4 Sumatran global mainshock.

11T. Parsons et al. / Tectonophysics 618 (2014) 134

aftershocks (Fig. 16A). The Maule shock occurred 10.6 h after aM=7.0Okinawa, Japan event. There currently is no way to know if this was re-mote triggering or if this was a coincidence; we thus classify it as possi-ble dynamic triggering.

3.6. Baja California

A relatively small (777 M 4 events) catalog from Baja California,Mexico (Fig. 17) has one significant rate increase that is associated

34

36

38

40

42

44

15 18 21 24 27 30 33 15 18 21 24 27 30 33

2007 M=7.1 (New Hebrides)20 days after global mainshock

0-10 10 20-200

40

80

120

160

Fig. 12.Mapping of thepersistent seismicity rate increase inGreece associatedwith a 2004M=7.1NewHebridesmainshock (see also Fig. 11). In this figure themaps show20days of pre-and post-global mainshock events, and illustrate how regional the effects are.

12 T. Parsons et al. / Tectonophysics 618 (2014) 134

with the 2012 M = 8.6 Indian Ocean earthquake (also noted byGonzalez-Huizar et al., 2012; Pollitz et al., 2012). The largest possiblytriggered earthquake that happened within 24 h of the globalmainshock is aM= 7.0 event that was preceded by a cluster of severalsmaller earthquakes of M = 3.7 to M = 6.1. All activity is delayed byalmost 20 h, though we do not know if smaller (M 4) events beganhappening before that. It appears that the M = 7.0 earthquake mayhave been triggered locally because M = 4.7, M = 4.9, and M = 6.1foreshocks happened 3.4 km, 6.1 km, and 19.4 km away respectively,meaning that local static or dynamic stress changes could havetriggered the M = 7.0 event.

3.7. Australia

A catalog containing 15,754M 1.0 earthquakes covering the en-tire continent of Australia shows one significant rate increase thatcan be associated with a global mainshock, a 2001 M = 7.5Indonesia event (Fig. 18). The delayed (14.5 h) response is spatiallyisolated compared with the mean variability in Australia, and wethus classify this as possible remote triggering. Remote triggeringin Australia was noted by Velasco et al. (2008) and Gonzalez-Huizar and Velasco (2011) through high-pass filtering of broadband

-200

-150

-100

-50

0

50

100

150

200

1979 1984 1989 1994 1999 2004 2009

1992 M=7.4 (Landers, CA)

1997 M=7.1 (Pakistan)

2001 M=7.1 (Japan)

1998 M=8.1 (Antarctic plate)

1

5

2

4

6

4.5

5.5

6.5

7.5

New Zealand

2007 M=7.2 (Aleutians)

3

2001 M=7.1 (S. Australia)

Fig. 13. Daily rate changes in New Zealand; see Fig. 14 for location. Primary plot featuresare the same as in Figs. 5 and 8. Five significant rate increases can be tied to globalmainshocks.

records; that these events do not emerge in catalog tests suggeststhat they are of low magnitude.

3.8. Volcanic and geothermal regions

It has been pointed out that volcanic and geothermal areas may beespecially susceptible to dynamic strains induced by seismic waves(e.g., Cannata et al., 2010; Hill et al., 1993; Hirosi et al., 2011; Mangaand Brodsky, 2006; Miyazawa, 2011; Moran et al., 2004; Surve andMohan, 2012), though triggering is certainly not confined to these set-tings (e.g., Brodsky et al., 2000; Gomberg et al., 2003). We study threevolcanic centers, the Coso geothermal field of southeast California,USA, the Yellowstone Caldera in Wyoming, USA, and the HawaiianIslands (Fig. 19). All three cases showbetween 4 and 5 possible episodesof remote dynamic triggering, but none stand out as being significantlymore susceptible than other regions that we have examined. The Cosoand Yellowstone sites have examples that we classify as probable re-mote triggering based on our 0.5 by 0.5 subregion criteria, however,these areas are very small (1 by 1) compared with other catalogs westudy, and virtually any seismicity rate increase could affect much ofthe catalog areas in these cases. The larger (1 by 2) Hawaiian Islandssite shows four cases of possible remote triggering (affecting from oneto three 0.5 by 0.5 subregions vs. a 2 threshold of 4.6).

3.9. Global subduction zones

We have examined catalogs from mixtures of every type of tec-tonic setting, and while each may have a dominant strain mecha-nism, all have strong variation. An opportunity exists to isolatemechanisms through an earthquake catalog specific to global sub-duction zone interfaces; the catalog consists of 3281 M 5 eventsthat have been identified as being directly on the interplate contactin global subduction zones by Heuret et al. (2011), and ends in 2007.We find no significant rate increases in the subduction interface catalogthat are associatedwith globalmainshockswhenwe apply themethodsthat we use on regional catalogs (Fig. 20A). If we extend the analysis to10 day rate changes we find three significant increases, all related toindividual large subduction events and their aftershocks (Fig. 20B). Ofcourse the odds of having other large earthquakes occur randomly in-creases with the longer period we consider, which illustrates the con-founding nature of delayed dynamic triggering. As long as the numberof possibly triggered large earthquakes is small, it becomes very difficultto establish any causation. The lower magnitude threshold in the sub-duction zone catalog is M ~ 5, similar to the global catalog used byParsons and Velasco (2011). It is therefore difficult to know if the sub-duction setting is not conducive to triggering, or if it is a magnitudeeffect.

48

46

44

42

40

38

36

0

100

200

300

0-200

50

100

150

200

48

46

44

42

40

38

36

0

100

200

300

0-200

50

100

150

200

1992 M=7.4 (Landers, CA)

1

1997 M=7.1 (Pakistan)

2

48

46

44

42

40

38

36

0

100

200

300

0-200

50

100

150

200

1998 M=8.1 (Antarctic plate)

3

48

46

44

42

40

38

36

0

100

200

300

0-200

50

100

150

200

2001 M=7.1 (Japan)

Rotorua Volcanic Swarm

4

0

100

200

300

0-200

50

100

150

200

164 168 172 176 180

48

46

44

42

40

38

36

0

100

200

300

168 172 176 180

Days after global mainshock

0-20

-10 10 201 2 3 4 5 6 7

-10 10 201 2 3 4 5 6 7

-10 10 201 2 3 4 5 6 7

-10 10 201 2 3 4 5 6 7

-10 10 201 2 3 4 5 6 7

-10 10 201 2 3 4 5 6 70

50

100

150

200

2007 M=7.2 (Aleutians)

6

Magnitude

24 hours before global mainshocks

24 hours after global mainshocks Daily Rate

Magnitude-Frequency (24 hours)

M=6.7 22.9 hrsafter global mainshock

48

46

44

42

40

38

36

2001 M=7.1 (South of Australia)

5

10 subregions increased (2 threshold=16.2)

21 subregions increased (2 threshold=16.2)

17 subregions increased (2 threshold=16.2)

19 subregions increased (2 threshold=16.2)

21 subregions increased (2 threshold=16.2)

17 subregions increased (2 threshold=16.2)

Fig. 14. Mapping of the significant rate changes associated with global mainshocks in New Zealand as identified and numbered in Fig. 13. Blue shading in the earthquake epicenters andmagnitude frequency plots denotes 24-h periods before global mainshocks, and red shading 24 h after. Histograms show20 days around global mainshocks. We consider cases 2 through5 as probable dynamic triggering because they appear to affect awide region, and are not necessarily associatedwith a single localmainshock. Event 6 is a case of possible remote triggeringbecause the rate increase is entirely the result of aftershocks from aM= 6.7 event that occurred ~23 h after aM= 7.2 Aleutian Islands mainshock, which could be coincidental.

13T. Parsons et al. / Tectonophysics 618 (2014) 134

-20

-15

-10

-5

0

5

10

15

20

1979 1984 1989 1994 1999 2004 2009

35

37

39

41

43

45

47

49

111 113 115 117 119 121 123

Ch

ang

e:

24

ho

urs

of

M>7

glo

bal

mai

nsh

ock

Nu

mb

er p

er d

ay

Year

Days after 1989 Loma Prieta EQ

0

0 20-20-40-60-80 100806040-001

5

10

15

201989 Shanxi Datong earthquake swarm

1989 Shanxi Datong earthquake swarm (t=15.0 hours)

1989 M=7.0 Loma Prieta (California)

A

B

C

Fig. 15. (A) One global mainshock, the 1989M= 7.0 Loma Prieta earthquake in northernCalifornia, is temporally correlated with (B) the onset of the Shanxi Datong earthquakeswarm in southeastern China (e.g., Zhang et al., 1995), which began ~15 h later, andoccurred in (C) an otherwise seismically quiet period.

-40

-30

-20

-10

0

10

20

30

40

1979 1984 1989 1994 1999 2004 2009

1994 M=7.1 New Zealand

2010 M=7.0 2010Okinawa: M=8.8 Maule t=10.6 hours

-40

-35

-30

-25

-20

-80 -75 -70 -65 -60

3 3.6 4.2 4.80

1

2

3

4

5

Nu

mb

er

Magnitude

-10 -5 0 5 100

4

8

12

16

Ch

ang

e:

24

ho

urs

of

M>7

glo

bal

mai

nsh

ock

Nu

mb

er p

er d

ay

Days after 1994 M=7.1 New Zealand EQ

A

B

C

Fig. 16. (A) A 2010 M = 7.0 mainshock in Japan can be temporally associated with theM = 8.8 Maule, Chile event 10.6 h later, or it could be coincidental. A 1994 M = 7.1New Zealand event, is temporally correlated with (B) a cluster of earthquakes in coastalChile. The events do not appear to be associated with a local mainshock, as the histogramof magnitudes indicates. The duration (C) of the rate increase is less than 24 h.

14 T. Parsons et al. / Tectonophysics 618 (2014) 134

3.10. All catalogs

We describe one last test using the regional catalogs that combinesthem all together. The idea here is that perhaps global mainshockseach cause subtle rate increases everywhere, but when examining anyone region they are not significant. We could potentially detect this be-havior by stacking all the catalogs together and analyzing them simulta-neously. However when do this, we find no rate increases beyond thosealready found in our region-by-region studies (Fig. 21). This pointsagain to a conclusion that stress, faulting, and surfacewave polarizationconditions may need to be optimal for remote dynamic triggering tooccur (e.g., Gonzalez-Huizar and Velasco, 2011; Hill, 2008; Parsons

et al., 2012). We also point out that coincidences do occur; a M = 7.0shock in Japan happened just 3.2 min after a 2007 M = 7.1 Vanuatuearthquake, too soon for the fastest seismic waves to have traveledthere (Fig. 21).

3.11. Regions with no evidence of dynamic triggering

We focused on describing regions with at least possible remote trig-gering responses in the sections above. These represent about half of thecatalogs studied (12 of 21) (Fig. 22). We briefly comment here on theregions that showed no evidence of remote triggering. These catalogsinclude some continental interior regions like East Africa, and the NewMadrid area of the central United States. Our observations are consistentwith the results of Iwata and Nakanishi (2004) and Harrington andBrodsky (2006) in that Japan does not appear very susceptible to remotetriggering. A similar high strain rate subduction zone setting in south-central Alaska also does not exhibit any significant rate changes thatcan be associated with global mainshocks. The very active Sumatra re-gion has had so many localM 7 earthquakes that it might be very dif-ficult to find rate increases associated with remote mainshocks becausethe local seismicity rates are already so high. Similarly, we could notidentify significant rate increases along the North Anatolian fault zoneof Turkey. A detailed catalog in the Apennines of Italy showed no signif-icant rate increases, nor did a catalog centered on the Philippine Islands.

Single station analyses of Global Seismograph Network (GSN) sta-tions revealed at least twofold rate increases at some stations in everyregion that we examined (Velasco et al., 2008). That these events are

-10

-8

-6

-4

-2

0

2

4

6

8

10

1979 1984 1989 1994 1999 2004 2009

M=8.6IndianOcean

Ch

ang

e:

24

ho

urs

of

M>7

glo

bal

mai

nsh

ock

Year

Hours after global mainshock

Mag

nit

ud

e

A) Baja California

117 114 111

21

24

27

30

3

4

5

6

7

19 20 21 22 23 24

B)

C)

Fig. 17. One global mainshock, the 2012 M = 8.6 Indian Ocean earthquake, is temporally correlated with a cluster of earthquakes in (B) the Gulf of California (red events). The primarycluster is shown as a time series vs. magnitude in (C); in this case the largest event (M = 7.0) is not the first to occur, but is preceded by smaller shocks. The first event in the series isa M = 3.9, almost 20 h after the Indian Ocean mainshock, though we do not know if smaller events began sooner.

15T. Parsons et al. / Tectonophysics 618 (2014) 134

not picked up in regional catalogs suggests that they have very lowmagnitudes, or were masked and/or interfered with during the passageof surface waves.

4. Interpretation of observations

The first important conclusion we draw about remote earthquaketriggering is how rare it is. In any one region we see at most 7 cases ofpossible or probable remote triggering out of 260 candidatemainshocksthat are more than 1000 km away. These are cases that can be detectedat the threshold magnitudes in our catalogs, which range fromM=1.0toM=4.0. A first quantification of the probability yields at most a ~3%chance of a remote mainshock causing a ~2 local earthquake rateincrease in any of the zones considered in this analysis.

We note four regions where we see at least one case of probable re-mote triggering, defined here as a widespread seismicity rate increase

-40

-30

-20

-10

0

10

20

30

40Australia

Ch

ang

e:

24

ho

urs

of

M>7

glo

bal

mai

nsh

ock

Year 1979 1984 1989 1994 1999 2004 2009

2001 M=7.5Indonesia

1

M=5.2 and aftershocks:delayed 14.5 hr

Fig. 18. One global mainshock, a 2001 M = 7.5 Indonesia earthquake, is temporally correl

(affecting a significantly larger number of 0.5 by 0.5 subregions thannormal variation) that can be associatedwith surface waves of a remoteearthquake. The four regions are: (1) the Basin and Range Province and(2) Northern California of the western United States, (3) Greece, and(4) New Zealand. These four regions represent a full range of tectonicenvironments that include strike-slip, extensional, and subductionzones. One feature they all have in common is the presence of volcaniccenters. However, when we focus just on magmatic provinces such asYellowstone Caldera, the Coso Geothermal center, and Hawaii, we donot find them to be especially responsive to passing seismic waves(Fig. 19).

A slight majority (22 of 40) of seismicity rate increases that can beassociated with global mainshocks are those we classify as possible re-mote triggering. These are isolated clusters of earthquakes that typicallybegin with a moderate magnitude earthquake (M = 4 to M = 6),and are followed by their aftershocks. It is impossible to know if

114 120 126 132 138

39

36

33

30

27

24

21

ated with a M = 5.2 local mainshock that was delayed by 14.5 h after surface waves.

-100

-80

-60

-40

-20

0

20

40

60

80

100

1979 1984 1989 1994 1999 2004 2009

-40

-30

-20

-10

0

10

20

30

40

-40

-30

-20

-10

0

10

20

30

1979 1984 1989 1994 1999 2004 2009

1979 1984 1989 1994 1999 2004 2009

A) Coso

Year

Nu

mb

er

B) Yellowstone

C) Hawaii

Ch

ang

e 2

4 h

ou

rs a

fter

glo

bal

mai

nsh

ock

s

2002 M=7.9Denali, AK

2003 M=7.6Mexico

2003 M=7.6Scotia Sea1997 M=7.0

KermadecIslands

4

2 3

1995 M=7.2Kermedec Islands

1

2 3 4

1

2009M=7.6Tonga

2010M=8.8Chile

2004M=7.0New Guinea

1984M=7.5Solomon Islands

3

24

51

1995 Norris Geyser Basin swarm

00 20-20

100

200

Days after M=7.2 New Zealand

35.5

35.8

36.1

36.4

0

2

4

6

8

Magnitude

Magnitude

Magnitude

Nu

mb

er

Nu

mb

er

44.0

44.4

44.8

0

20

40

60

80

100

19.0

19.4

19.8

20.2

20.6

21.0

-155.2

-118.3 -117.8 -117.30 1 2 3

-111.0 -110.6 -110.20 1 2 3

-156.0 -155.2

0 1 2 3 40

10

20

Nu

mb

er

2009M=7.6Tonga

3

1995 M=7.2Kermedec Islands

1

2003 M=7.6Scotia Sea

3

2010M=7.4New Britain

2006 M=7.0Mozambique

Fig. 19. Three sample volcanic centers are studied, and examplemaps are shown that typify responses. (A) The Coso volcanic center of southeast California, USA yields five significant rateincrease that can be associated with global mainshocks. (B) The Yellowstone caldera region of Wyoming, USA shows four significant rate increases, the first is associated with a 1995Kermedec Islands mainshock and appears to be a swarm invigoration, affecting the Norris Geyser Basin swarm (see inset histogram). Additionally, responses from a 1995M=8.0Mexico(2), and 2000M=7.9 Indonesia (3)mainshocks are noted alongwith the already-discussed regional response to the 2002M=7.9Denali earthquake (4) (Fig. 3). In (C) theHawaiianIslands are shown to have four significant rate increase that can be associated with global mainshocks.

16 T. Parsons et al. / Tectonophysics 618 (2014) 134

these are precipitated by passing seismicwaves, or if they are simply co-incidental; tests with the global catalog using randommainshock timesshow that the expected number of coincidentalmoderate local events isnot surpassed by the observations (Fig. 7).

We show detailed temporal histories of earthquake responses in thefour primary regionswhere we see remote triggering in Figs. 23 and 24.A spectrumof responses is evident that ranges from immediate, swarm-like behavior after seismic waves arrive, to activity that is delayed bymany hours. Delayed responses tend to be the local moderate eventwith aftershocks cases that we call possible remote triggering. There is

no consistent observation that delayed responses are preceded by anysort of gradual build-up of seismicity (Figs. 23, 24), with just one exam-ple in Baja California (Fig. 17).

4.1. Insights into remote M 5 earthquake triggering

One of the key goals of this review is to simultaneously observe abroad magnitude spectrum of remote earthquake triggering by usingregional networks that have catalogs complete to the ~M = 2 level.We already know that M N 5 earthquakes do not occur immediately

-40

-30

-20

-10

0

10

20

30

-40

-30

-20

-10

0

10

20

30

40

1979 1984 1989 1994 1999 2004 2009

A) Change 1 day after global mainshocks

B) Change 10 days after global mainshocks 1996 M=7.0

Mid-Atlantic Ridge

2004 M=8.1 TasmanSea

M=7.0 M=7.9 (t=8.0 days)

M=8.1

M=7.1

1

1

223

3

2007 M=7.1Vanuatu

M=8.1(t=7.8 days)

M=9.0 (t=2.4 days)

Year

Ch

ang

e 2

4 h

ou

rs a

fter

glo

bal

m

ain

sho

cks

Ch

ang

e 1

0 d

ays

afte

r g

lob

al

mai

nsh

ock

s

Fig. 20.We study a global subduction interface catalog assembled by Heuret et al. (2011), and (A) note no significant associations between these events and remote global mainshocksafter 24 h. However it is possible to associate (B) rate increases in three cases if a 10-day period is examined.

17T. Parsons et al. / Tectonophysics 618 (2014) 134

during surface wave arrivals (Fig. 25), but detecting delayedM N 5 trig-gering is difficult because if there is a signal, it cannot be isolated fromgenerally high global activity levels (e.g., Huc and Main, 2003; ParsonsandVelasco, 2011).We obviate the problemof possiblymissing delayedM 5 earthquakes thatmight be overlooked by stacking global catalogsby using regionalM N 2 catalogs. We therefore investigate the questionof whether remote triggering rates are too low for the comparativelyrare M 5 events to be expected, and/or whether higher magnituderemotely triggered earthquakes are always delayed.

We can illustrate the potential problem of searching the stackedglobal catalog for remoteM N 5 triggering by combining all the regionalresults where we observed either possible or probable remote trigger-ing into a single magnitude-frequency distribution (Fig. 26). The distri-bution appears to have a significant deficit of lower (M b 2.5), andhigher (M N 4.5) magnitude events relative to a linear Gutenberg andRichter (1954) (log(N) = a-bM) relation. The likely cause on the low-magnitude end is variation in detection thresholds of different regionalnetworks. The taper on the high-magnitude end could be caused bysmall a-values (activity levels) in each regional response such thatexpected M N 5 rates are low. Alternatively, higher magnitude eventsmay be absent for physical reasons. This sort of taper in magnitudefre-quency distributions is commonly observed (e.g., Kagan, 1993), and canbe simulated with multiple catalogs with different maximum magni-tude thresholds (e.g., Geist and Parsons, 2014; Sornette et al., 1991).

We examine individualmagnitudefrequency distributions from re-gional probable and possible remote triggering episodes, and extrapo-late them assuming a cumulative GutenbergRichter distribution tofind the expected 24-hM 5 triggered earthquakenumbers for each re-sponse (Table 1). The expected number ofM 5 events is extrapolatedusing a b-value (slope) of 1.0 from event rates at the thresholds

given in Table 1. Of the 28 responses we examine, 8 (29%) have highenough activity rates such that at least one M 5 earthquake mighthave been expected. Of those, 3 (11%) are associated with at least oneM 5 shock. In 5 other instances (18%), noM 5 events were observeddespite high rates at lower magnitudes. This result implies that in mostcases, remoteM 5 triggering is not observed because the overall trig-gered rates are very low. When we restrict the analysis to just probablecases, there are only 3 responses where M 5 seismicity would be ex-pected during the first 24 h, and of those, onewhereM 5 earthquakeswere actually observed (Table 1). Thus one explanation for the absenceof remote M 5 triggering is that the numbers of remotely triggeredearthquakes are too small for high magnitudes to be observed inmost cases, and that the delayed higher magnitude events we do ob-serve are primarily coincidental. If however the possible cases that in-volved possibly delayed higher magnitude triggering are accepted(e.g., Gomberg and Bodin, 1994; Gonzalez-Huizar et al., 2012; Pollitzet al., 2012; Tzanis and Makropoulos, 2002), then more interpretationis necessary.

4.2. Interpretation of possibly delayed M 5 earthquake triggering

Remotely triggeredM N 5 earthquakes are not observed during sur-face wave passage (Fig. 25), though we do note a persistent minimumdelay time of t 9 h for possibly triggered M N 5 earthquakes(Fig. 27). It is unclear how important the 9 h mark is; the compilationshown in Fig. 27A has the potential to be misleading because there aretwo M N 7 possibly triggered aftershocks that happened ~910 h aftersurface wave arrivals (the 2010 M = 8.8 Maule, Chile earthquake andaM=7.4 aftershock), and that are associatedwith their own numerousM N 5 local aftershocks. These subsequent aftershocks give extra weight

-500

-400

-300

-200

-100

0

100

200

300

400

500

1979 1984 1989 1994 1999 2004 2009

1

8 M=7.0 2001New Britain (Coincidental)

9 M=7.1 2001South of Australia (New Zealand)

7

83

412

13M=7.1 2007Vanuatu (Coincidental)

11

M=7.4 2008Sumatra (Greece)

12

M=7.4 2009Kurils (Greece)

13

M=7.0 2009Sumatra (Greece)

14

10 M=7.9 2002Denali, USA (Basin and Range)

15

15 M=7.0 2010Okinawa (Chile)

6 M=8.1 1998Antarctic Plate

11 M=7.1 2007Vanuatu

Year

Ch

ang

e 2

4 h

ou

rs a

fter

glo

bal

mai

nsh

ock

s

10 M=7.9 2002Denali, USA

5001 M=7.3 1992

Landers, USA

M=8.8 2010Maule, ChileT=10.6 hrs

15 M=7.0 2010Okinawa

Hours after global mainshock

Hours after global mainshock

Nu

mb

er

Nu

mb

er

-24 -16 -8 0 8 16 240

10203040506070

Hours after global mainshock

Nu

mb

er

-24 -16 -8 0 8 16 240

10

20

30

40

50

-24 -16 -8 0 8 16 240

10

20

30

40

-24 -16 -8 0 8 16 240

10

20

30

40

50

Nu

mb

er

0

10

20

Hours after global mainshock

Nu

mb

er

-24 -16 -8 0 8 16 24 -24 -16 -8 0 8 16 240

20

40

60

80

100Coincidental M=7.0 Japan & aftershocks removed

25 9 14

11

610

1 M=7.3 1992Landers, USA (New Zealand)

6 M=8.1 1998Antarctic Plate (New Zealand)

7 M=7.1 2001Japan (New Zealand)

3 M=7.2 1995Tonga (Greece)

4 M=7.1 1997Pakistan (New Zealand)

5 M=7.0 1997Kermadec (Hawaii)

2 M=7.2 1992Nicaragua (Southern California)

Fig. 21. Analysis of 21 combined catalogs. When catalogs are stacked together, none of the 260 global mainshocks shows any evidence for significant rate changes beyond those alreadyidentified as affecting a single region. The exception is theM=7.1 Vanuatu event labeled 11, which is associated with a coincidentalM=7.0 shock in Japan that happened only 3 minlater, before its seismic waves arrived in Japan.

18 T. Parsons et al. / Tectonophysics 618 (2014) 134

to the 9-h threshold. We therefore make the same plot in Fig. 27B withall post-Maule aftershocks removed from the catalog. This removesmost of the M N 5 events and makes the 9-h transition less distinct.

As there is uncertainty whether cases of possible remote triggeringare in fact coincidental, we plot only the incidences of probable trigger-ing in Fig. 27C. In this case there is only oneM N 5 event, aM=6.0 event

triggered in Greece ~9.5 h after a 2008M= 7.4 mainshock in Sumatra.This again highlights a repeated result that we find; it is difficult to un-equivocally associate M N 5 earthquakes with passing seismic waveseven if a delay of up to 24 h is allowed. Finally, we plot just the highestmagnitude earthquakes from each possible and probable triggeringresponse vs. time in Fig. 27D, but the delay for larger magnitudes is

Alaska

California (USA)Hawaii

(USA)Baja California

Subductionzone interfaces

Greece

ZealandNew

Japan:Central Honshu

Turkey: Marmara

Basin and Range

New Madrid Coso

Yellowstone

Chile/Argentina

Australia

Philippines

East Africa

China

Italy: Apennines

Sumatra region

No significant reate increases

Possible remote triggering

Probable remote triggering

Fig. 22.Mapping of regional catalogs shaded based on our interpretation of remote triggering response. About 50% of the catalogswe studied showedpossible remote triggering, and ~20%showed probable remote triggering.

19T. Parsons et al. / Tectonophysics 618 (2014) 134

still evident. By contrast, immediate increased rates of lowermagnitudeearthquakes can be clearly associated with surface wave arrivals(Figs. 23, 24, 27).

We gain some insight into the possibility that the apparent delayedM N 5 triggering response is a chance occurrence by conducting thefollowing test. We assemble the magnitude distributions from thepossible and probable triggered events plotted in each panel of Fig. 27(also including those with smaller magnitudes not shown). We assem-ble the occurrence times of these events in separate distributions. Were-associate magnitudes and times at random across 100 trials, andthen track the first occurrences of M 5.0 and M 6.0 earthquakes.These tallies are shown as histograms in Fig. 28 along with the inputdistributions. From these histograms we can show the frequency ofoutcomes when the earliest remotely triggered M 5.0 and M 6.0earthquakes could be expected to occur in the absence of anydelaying physics. From these tests, we note that it would be unlikely forthe N9 h delay to occur given observed magnitude distributions, butpossible, with 96% to 100% of the simulations havingM 5.0 earthquakeshappening before 9 h pass. The exception to this is the probable-triggeredcatalog from Fig. 27C, because it contains only oneM 5.0 event.

4.3. Delayed dynamic earthquake triggering and tremor in Greece

We identified an intriguing, apparently long-lived (at least 20 days)seismicity rate increase that swept across most of Greece after a 2007M = 7.1 New Hebrides mainshock (Figs. 11, 12). The period betweenSeptember 2006 and May 2007, encompassing the occurrence of theM = 7.1 New Hebrides event, was identified as a seismic crisis withswarm characteristics by Bourouis and Cornet (2009). While all theseevents are temporally correlated, it is of course difficult to know ifthere is causation. To learn more, we apply a band-pass filter (cornerfrequencies 28 Hz) to regional broadband records to remove surfacewaves and identify local events. In Fig. 29 we show broadband record-ings after the implementation of a low-pass (0.010.1 Hz in Fig. 29A,traces 13), and a high-pass filter (28 Hz in Fig. 29A, traces 45) thatreveals local events triggered by the global mainshock (Fig. 29B) and

triggered tremor (Fig. 29C). Tremor can be seen at frequencies of up to8 Hz, meaning that the observation is locally sourced and not remnantteleseismic energy (e.g., Peng et al., 2011b). Triggered tremor has beenidentified in different geotectonic environments worldwide (e.g., Penget al., 2009; Rubinstein et al., 2009) and can be initiated by the passingof seismic waves from distant sources.

In Fig. 29A we show triggered regional seismicity that correspondswith the S-wave arrival in the high-passed traces of the radial and trans-verse horizontal components (traces 4 and 5), and triggered tremorwith small amplitudes (105 cm/s) that initiates approximately withthe P-wave arrival. We have identified at least 5 more tremor episodesin the first few hours following the mainshock. Shelly et al. (2011)report that triggered tremor may be a possible mechanism for delayeddynamic triggering, which in this case may explain the persistentseismicity rate increase associatedwith the 2007M=7.1NewHebridesmainshock.We note that this is thefirst identification of triggered trem-or in Greece, and that the Corinth Gulf offers a favorable location sincethe active deformation is related with low-angle normal faulting atseismogenic depths (e.g. Chao et al., 2012) in the back-arc extensionalprovince of the Hellenic subduction zone (Vassilakis et al., 2011). Therelationship between the ambient and triggered tremor could providea physical mechanism to explain the apparent low triggering thresholdin central Greece suggested by Brodsky et al. (2000).

4.4. Possible causes of delayed dynamic earthquake triggering

As long as the possibility exists that dynamic triggering of hazardousearthquakes by seismic waves can be delayed, then there is a need toquantify the probability of this, and to understand the physics behindit. In contrast to possible remote triggering observations, the timing ofnear-sourceM N 5 earthquakes (presumed to be caused by static stresschanges) has no apparent magnitude dependence (Fig. 30). At nearrange, M N 5 aftershocks begin immediately, and follow an Omori-lawtemporal decay that is similar from hourly to yearly time scales. There-fore if we follow the alternative interpretation of our observations, thatthe lack of immediateM N 5 earthquake triggering from remote sources

00-6-12-18 2418126-24

2

4

0

2