Sie sind auf Seite 1von 11

Moment tensor inversion of

explosive long period events


recorded on Arenal volcano, Costa
Rica, constrained by synthetic
tests

Presenter:
Maria R.P. Sudibyo
Advisor:
K.I.Konstantinou
Introduction
• Volcanoes are complex and have a range of earthquake types
including high frequency (volcano tectonic) events, explosions, long
period events, and tremor.
• Long period event is normally characterized by strongly peaked
spectra, having energy concentrated between 0.2 and 5 Hz, and can
caused by fluid movements inside the volcanic conduits (Chouet,
2003), or dry failure of the volcano edifice (Bean, 2014).
• This paper investigates how the presence of single force can affect the
moment tensor solution. The synthetic tests to obtain the correct
source time function and source mechanism are performed.
Methodology
• The nth component of the displacement which is recorded at position x and time t, can
be written as
(1)
Mpq = the force couple in the pq direction acting at the source
Fp = the single force acting in the p direction
Gnp and Gnp,q = The Green’s function and their derivatives.
• Green function is calculated using 3D-full wavefield numerical simulations including
topography and the best estimate of the velocity structure obtained using spatial
autocorrelation (SPAC) method. Spatial derevatives are given by the central finite-
difference
(2)

Where Gnp,q(r,s) is spatial derivative of the Green’s function Gnp, and Δq is the spatial grid
spacing
Methodology
• In frequency domain, equation (1) can be written as
(3)
Where un(ω) is the spectra of the displacements, Mpq(ω) contains the moment
tensor components, Fp(ω) contains the single forces, Gnp(ω) and Gnp,q(ω)

• Equation (3) can be written in matrix form as u=Gm,


where u is the data matrix, G is matrix containing the Green’s function and the
derivatives, and m is the moment tensor and single forces components.
Methodology

(4)

Due to the symmetry of the moment tensor, we have

(5)
Methodology
• The equation d=Gm is overdetermined, with the misfit (R) between calculated
and observed data is expressed by
(6)
where W = diagonal weighting matrix of the quality of the waveforms. W can be
expressed in explicit matrix format as

(7)
• The solution given by weighted least square can be written as
(8)
Result of
Synthetic
Test
• Synthetic tests are
generated using a Ricker
wavelet, with source is
simulated as an
explosion without single
forces.
• The inversion was
performed for both of
moment tensor and
single forces (MT+SF),
and moment tensor
only.
Result of
Synthetic
Test
• Synthetic test is performed
using synthetic signals
contaminate by random noise,
with noise level is 50% of the
average rms amplitude.
No single forces included.
Result of
Synthetic
Test
• Synthetic test is performed,
including a strong single force
in the west-east direction (Fx),
with a signal-to-noise ratio of
10.
Result of
Synthetic
Test
• Synthetic test is
performed using
incorrect source
position, and signal-
to-noise ratio of 10.
• No single forces
included
Conclusions
• The moment tensor inversion will easily generate spurious (fake)
single forces, when in its process includes noisy data and or
mislocated source.
• The synthetic tests show that in the inversion using noisy data with
amplitude within 10% of the average rms, will generate solution
insensitive to the inaccurate location of the source.