Comparative stochastic modeling of the Al Hoceima, Morocco, aftershock sequences in 1994 (Mw 6.0), 2004 (Mw 6.4) and 2016 (Mw 6.3).

Abstract

In this study, a stochastic modeling of the three aftershock sequences occurred at Al Hoceima, Morocco, starting on May 1994 (Mw 6.0), February 2004 (Mw 6.4) and January 2016 (Mw 6.3) has been carried out. A version of the restricted trigger model, named RETAS, has been used to perform the temporal analysis of the three selected sequences. The conditional intensity function of the RETAS model is similar to the Epidemic Type Aftershock Sequence (ETAS) model, with the restriction that only aftershocks with magnitude greater than or equal to a certain threshold magnitude (Mth) can trigger secondary events. Varying the threshold magnitude, some variants of the RETAS model have been examined, which ranges from the modified Omori formula (MOF) to the ETAS model, including such models as a limit cases. The best fit models identified, based on Akaike Information Criteria (AIC) and for each of the three sequences, differ from one to the other. The obtained difference suggests that, although the activated faults are close together (in the range of 10 to 20 km), their stress regimes could be different. Furthermore, a stochastic model is proposed to analyze the Benioff strain release after a strong earthquake. The model is developed following a compound Poisson process, and contours the evolution of strain release during the aftershock sequence following the main shock occurrence. The suggested model has been applied to our data. Initially, the temporal evolution of the aftershock decay rate was modeled by a RETAS model, and after that, the best fit model recognized is integrated into the strain release stochastic analysis. The applied stochastic model of Benioff strain release empowers a more detailed study by detecting possible deviations between observed data and model. The comparison between the real values of the cumulative Benioff strain release and the expected modeled ones shows some deviation at the beginning of all the three sequences. This shows that it happens large aftershock clusters at the beginning of the sequence immediately after the occurrence of the main shock, more than if their occurrence could be completely aleatory. Strain release spatial analysis reveals release patterns, changing during each aftershock sequence.

Full text

European Seismological Commission
36th General Assembly
2-7 September 2018
Valletta - Malta
ESC2018-S24-834
COMPARATIVE STOCHASTIC MODELING OF THE AL HOCEIMA, MOROCCO,
AFTERSHOCK SEQUENCES IN 1994 (MW 6.0), 2004 (MW 6.4) AND 2016 (MW
6.3)
Mohamed Hamdache* (1), Dragomir Gospodinov (2), Jose Pelaez (3), Boyko Rangulov (4), Jesus Henares (5)
(1)CRAAG, (2)Faculty of Physics and Technology.Plovdiv University "Paisii Hilendarski", (3)Department of Physics. University of Jaen,
(4)University of Mining and Geology “St. Ivan Rilski”, 1700 Sofia;, (5)University of Jaen. Spain
* m.hamdache@craag.dz
In this study, a stochastic modeling of the three aftershock sequences occurred at Al Hoceima, Morocco,
starting on May 1994 (Mw 6.0), February 2004 (Mw 6.4) and January 2016 (Mw 6.3) has been carried out. A
version of the restricted trigger model, named RETAS, has been used to perform the temporal analysis of the
three selected sequences. The conditional intensity function of the RETAS model is similar to the Epidemic
Type Aftershock Sequence (ETAS) model, with the restriction that only aftershocks with magnitude greater
than or equal to a certain threshold magnitude (Mth) can trigger secondary events.
Varying the threshold magnitude, some variants of the RETAS model have been examined, which ranges from
the modified Omori formula (MOF) to the ETAS model, including such models as a limit cases. The best fit
models identified, based on Akaike Information Criteria (AIC) and for each of the three sequences, differ from
one to the other. The obtained difference suggests that, although the activated faults are close together (in the
range of 10 to 20 km), their stress regimes could be different.
Furthermore, a stochastic model is proposed to analyze the Benioff strain release after a strong earthquake.
The model is developed following a compound Poisson process, and contours the evolution of strain release
during the aftershock sequence following the main shock occurrence. The suggested model has been applied
to our data. Initially, the temporal evolution of the aftershock decay rate was modeled by a RETAS model, and
after that, the best fit model recognized is integrated into the strain release stochastic analysis. The applied
stochastic model of Benioff strain release empowers a more detailed study by detecting possible deviations
between observed data and model. The comparison between the real values of the cumulative Benioff strain
release and the expected modeled ones shows some deviation at the beginning of all the three sequences.
This shows that it happens large aftershock clusters at the beginning of the sequence immediately after the
occurrence of the main shock, more than if their occurrence could be completely aleatory. Strain release
spatial analysis reveals release patterns, changing during each aftershock sequence.