 |
| Titel |
Self-correcting models driven by seismic strain, moment or energy. Applications to the Italian seismicity |
| VerfasserIn |
Elisa Varini, Renata Rotondi, Roberto Basili, Salvatore Barba, Bruno Betrò |
| Konferenz |
EGU General Assembly 2013
|
| Medientyp |
Artikel
|
| Sprache |
Englisch
|
| Digitales Dokument |
PDF |
| Erschienen |
In: GRA - Volume 15 (2013) |
| Datensatznummer |
250076040
|
|
|
|
|
|
| Zusammenfassung |
| The stress release model (Vere-Jones, 1978) provides a stochastic version of the Reid’s
elastic-rebound theory, which is commonly accepted as the most feasible physical
description of the long-term evolution of the earthquake process. It assumes that
the stress X, which governs the state of the system in a region, increases linearly
with time at a constant loading rate Ï imposed by external tectonic forces until it
exceeds the strength of the medium and decreases abruptly generating an earthquake.
This hypothesis is formalised by a self-correcting point process with conditional
intensity function λ(t | Ht) = exp{α + β[Ït - S(t)]}, where Ht is the seismic
history up to time t, S(t) is the cumulative stress release due to all the earthquakes
up to t and α, β and Ï are model parameters. X can be any physical parameter
that constitutes a proxy measure of the strength of an earthquake, therefore we
propose four possible definitions of X and consequently four versions of the stress
release model. Let mw denote the moment magnitude of an earthquake and A be
its rupture area. We consider two classical versions of the model: in the former
X = 100.75(mw-5.3) is the Benioff strain, in the latter X = 101.5(mw-5.3) is the seismic
moment. Then we propose two new versions: the third model is based on seismic energy
X = 102.25(mw-5.3)-A, the fourth one on the scaled energy X = 100.75(mw-5.3)-A, as
defined by Senatorski (2005, 2012). The rupture area A is evaluated by the Wells
and Coppersmith regression with parameters depending on the faulting type of the
earthquake.
We analyse the Italian historical seismicity on a regional basis by subdividing the
Italian territory into eight tectonically-coherent large regions. For each model and
region, a fully Bayesian analysis is carried out in order to estimate the posterior
distributions of the model parameters. We also deal with the forecast problem by
evaluating, for each region, the probability distribution F(t | Hs) of the time t
to the next event conditioned on the past history Hs. The time to the next event
turns out to have an extreme value distribution, known as Gompertz distribution,
with a shape parameter that depends on time through the value of the conditional
intensity. Hence retrospective and prospective forecasts are obtained and validated by
comparing them with the earthquakes of the data set and those recorded in the time
period from the end of the catalog to the present time (2003-2012). In this last
period there have been four earthquakes exceeding the magnitude threshold Mw 5.3,
excluding aftershocks (among these, the L’Aquila earthquake in 2009, Mw 6.3,
and the Finale Emilia earthquake in 2012, Mw 5.9 ). The forecast for these events
falls between the estimated median and mean of the corresponding time to the next
event.
Senatorski, P. (2005) A macroscopic approach towards earthquake physics: the meaning
of the apparent stress, Physica A, 358, 551-574.
Senatorski, P. (2012) Effect of seismic moment-area scaling on apparent stress-seismic
moment relationship, Physics of the Earth and Planetary Interiors, 196-197, 14-22.
Vere-Jones, D. (1978), Earthquake prediction - A statistician’s view, J. Physics Earth, 26,
129-146. |
|
|
|
|
|
|