Short-term foreshocks and their predictive value G A

Short-term foreshocks and their predictive value G. A. Papadopoulos (1) M. Avlonitis (2), B. Di Fiore (1) & G. Minadakis (1) 1. Institute of Geodynamics National Observatory of Athens, Greece papadop@noa. gr 2. Dept. of Informatics, Ionian University, Greece EARTHWARN

Definitions of short-term foreshocks • No standard definitions…. but • Literature Consensus foreshocks: Spatio-temporal seismicity clusters that exhibit a power-law rise in seismic moment release in the area where a larger mainshock is under preparation, and occurring up to a few months before the mainshock occurrence. • Swarms (Yamashita, 1998): Spatio-temporal seismicity clusters that exhibit a gradual rise and fall in seismic moment release, lacking a mainshock-aftershocks pattern.

First evidence • Power-law increase, b-value decrease - Laboratory experiments (Mogi, 1962, Scholtz, 1968) - Seismic sequences (e. g. Jones & Molnar, 1979) • However, only very few examples were available

Characteristic patterns of short-term foreshocks • • • Time: mode of power-law increase Space: move towards mainshock epicenter Magnitude: b-value drops Foreshock rate? Why some mainshocks have foreshocks and others do not?

Method of analysis • Seismicity is a 3 D process: space-time-size domains • Basic method: in-house FORMA algorithm for the detection of significant seismicity changes - space: select target area, repeat tests by changing - perform completeness analysis - time: seismicity rate changes (z-test, t-test) - Size: b-value changes (Utsu-test)

Good examples of foreshocks: L’Aquila, 6 Apr. 2009, M 6. 3

Chile, 1 Apr. 2014, M 8. 1

Tohoku, 11 March 2011, M 9. 0

S. California, 4 Apr 2010, M 7. 20

S. California 26 Apr 1981, M 5. 75

South Greece, 14. 8. 2011, M 4. 5

South Greece, 14. 8. 2011, M 4. 5

Basilicata (Italy), M 5. 0, 25. 10. 2012

Predictive value: time • Time: power-law mode • Short-term: up to about 6 months at maximum however, 80% in the last 10 days P (t) =A – B (log t)

Alternative: Poisson Hidden Markov Models Orfanogiannaki et al. PAGEOPH (2011) Research in Geophys. (2014) Recognizing changes in the states of seismicity, e. g. Sumatra 2004

Predictive value: space • Space: move towards mainshock epicenter • Topological metrics based on Network Theory : e. g. Betweeness Centrality e. g. Daskalaki et al. , J. of Seismology (2013)

Application in L’ Aquila, 2009

Evolution of Betweeness Centrality L’ Aquila, 2009

Predictive value: magnitude • Mo ≠ Mf ; Mo ≠ duration (f) • However, Mo may depend on foreshock area! Mo ranges from 4. 5 to 9. 0

Foreshock rate? • Current statistics indicates Fr around 40 -50% • In well monitored areas no foreshocks were recognized, e. g. in Parkfield, 2004, M 6. 0 • Earlier statistics indicated Fr around 10 -20% Catalog Problems Foreshock recognition strongly depends on recording capabilities No catalog problems Source properties determines the no foreshock incidence

Conclusions Foreshocks have characteristic 3 D patterns In time: power-law mode In size: b-value drops In space: move towards mainshock epicenter There is evidence that the foreshock area depends on Mo • The predictive value of foreshocks now becomes evident, which is promising for the mainshock prediction • • •