Get documents about "GE5950 Volcano Seismology 30 March 2009
Document Sample
GE5950 Volcano Seismology What is a VLP?
30 March 2009
• Very-long period earthquake
– Used to describe signals that are below the LP band
Today’s topics .5 Hz or 2 s period to about 4 Hz
• VLP seismicity
– If they periods are lower than long period
earthquakes, they must be very long period
Homework
• Read introduction and section on “Flow of low-viscosity magma” – Low end typically limited by the seismometer
in Gilbert, J. S., and S. J. Lane (2008), The consequences of fluid • 30 sec. sensors (30 sec - 50 Hz or so) are commonly
motion in volcanic conduits, Geol. Soc. London Spec. Pub., 307, 1- called “broadband” seismometers
10, doi:10.1144/SP307.1. • 120 s sensors (120 sec - 50 Hz or so) are harder to
• Type: dx.doi.org/10.1144/SP307.1 into a web browser and you install, but much broader band
will find the paper.
– By ~200 s, signals are sometimes called ULP (ultra
LP)
Wednesday
• Interpretation of VLPs – When VLPs are detected, the lower end might be
missed due to limitations of the sensor
Thursday
• Lab on VLP analysis
VLP signals are weak VLP signals found at many types of volcanoes
– Typically one cycle • Mafic
– rarely multiple cycles (Erebus) – Erebus (Rowe et al., 1998)
– Kilauea (Ohminato et al., 1998)
– Decay fast, so can only be seen on close – Stromboli (Chouet et al., 2003)
stations – At Erebus and Stromboli, VLP signals accompany explosions
• Within a few km of the source
• Andesitic
– Often lower amplitude than oceanic – Augustine
microseism peak 0.1-0.2 Hz (5 - 10 s – Merapi (Hidayat et al., 2002)
– Popocatepetl (Chouet et al., 2005)
period)
– So, periods below about 10 s, but • Dacite
– Mount St. Helens (Waite et al., 2008)
• Sometimes seen above the microseismic noise so,
anything below the LP band (2 s) can be
• Hydrothermal
considered VLP – Aso Volcano (Kaneshima et al., 1996)
Mechanisms are thought to involve two phase fluid acceleration
1
VLP signals are difficult to locate VLP signals are difficult to locate
– Emergent - cannot pick onset of first motion – Use ratio of N to E and take the inverse tan
– Typically have all energy in the ray path – 90º-atan(N/E)
• Compressional and dilatational motion – Still have uncertainty in which direction
• P wave – Use vertical
– Can be located using particle motion analysis • If 1st motion is compressional (up) radial points away
• Determine the orientation of horizontal particle motion from the source
Map View Map View
Locating VLPs Locating VLPs
– Once radial direction is known, look at radial versus – What about free-surface affects?
vertical, which points to source • Conversion to SV
• Long wavelengths (100s of km) smooth over
heterogeneities making a homogeneous half space – What about topography?
model appropriate
2
VLP analysis- Particle motions VLP analysis- tilt
• Simplest analysis • Ground tilt can look like VLP on broadband seismometers
• Typically have linear particle motions that approximately – Results from change in acceleration due to gravity
point to the same location – Produces characteristic signal with width of low corner
• Must incorporate topography
– Source at center of cone-
shaped volcano won’t be 120 sec instrument
distorted horizontally, but
vertical will be
30 sec instrument
– Can correct for this
– Or model full waveforms
– Can be difficult to distinguish
– Must model it
• At Merapi up to 22% of displacement signals could be due to tilt
induced by crack source
Neuberg & Pointer, GJI, 2000
Kilauea VLPs Kilauea VLPs
• Long-term tilt events are related to a collection of small events
recorded with tiltmeters and seismometers
Ohminato et al., JGR, 1998 Ohminato et al., JGR, 1998
3
Kilauea VLPs Kilauea VLPs
• Here, the data are filtered between 8-50 seconds (why 8?) • Particle motion analysis of a single VLP
• Horizontal and vertical particle motions are linear
• Particle motions from all stations point to about the same place
Ohminato et al., JGR, 1998 Ohminato et al., JGR, 1998
Kilauea VLPs Kilauea VLPs
• Particle motion analysis of a single VLP • VLPs associated with recent activity at Halemaumau point to a
• Horizontal and vertical particle motions are linear similar location
• Particle motions from all stations point to about the same place
Ohminato et al., JGR, 1998 Phil Dawson, personal communication 2008
4
Stromboli VLPs Stromboli VLPs
• VLPs associated with strombolian explosions • VLPs recorded at station T6
– The best studied VLPs • Highly repetitive
– 10-30 per hour during Sept 1997 seismic experiment • VLPs are clearly evident without filtering
– Two active vents
– Repetitive explosion waveforms
– Both explosions dominated
by VLP energy
– VLPs seen all across network
Chouet et al., JGR, 2003; Chouet et al., Geol. Soc. London, 2008 Chouet et al., JGR, 2003; Chouet et al., Geol. Soc. London, 2008
Stromboli VLPs Erebus VLPs
• VLP signals precede strombolian explosions by 1.5 s
• Particle motions are linear and generally point to the vent • Peaked at 20, 12, and 7 Hz
– In these plots positive radial points toward the source - this is not the • Polarized in radial vertical plane
standard convention for “radial”
• Highly repeatable
• Stations farther down the slope have less vertical velocity
• Oscillate for up to 200 s
• First motions all down - deflationary source
Chouet et al., JGR, 2003; Chouet et al., Geol. Soc. London, 2008 Rowe et al., 1998, 2000
5
Erebus VLPs Erebus VLPs
• Velocity (left) and displacement (right) • Displacement particle motions vary slightly with period and time
• How did they get displacement? – Are later oscillations deeper?
Rowe et al., 1998, 2000 Rowe et al., 1998, 2000
Filtering can make an artificial VLP VLP analysis- Particle motions
• One has to be careful when looking for VLP • Given all the caveats about using particle motions, how can
signals he particle motions be analyzed quantitatively?
• Filtering can make a impulsive signal look like – Covariance matrix
it has a VLP component, when it does not – Eigenvalue-eigenvector decomposition
– An impulsive signal can have a broadband signature
• The signals on the right are filter dependent • Covariance is a measure of how much two variables (or
functions) change together
– If two functions change together in the same direction from
their mean, the covariance will be positive
– The covariance is negative when the functions are
anticorrelated
Frequency (Hz) Frequency (Hz)
a) VLP signal is dominant b) spectrum is filter dependent
6
VLP analysis- Particle motions Eigenvalues and eigenvectors
• Covariance matrix in matlab
– Diagonal elements are variances
– Off-diagonal elements are covariances
– For two n x 1 column vectors (seismograms), x and y:
the covariance of x and y is a symmetric 2 by 2 matrix. The
two off-diagonal elements are equal because the describe the
covariance of x with y and y with x, respectively.
The diagonal elements are the variance of x and y,
respectively.
- The covariance matrix holds the information about the
orientation of the data
- To extract it, use eigenvalue-eigenvector decomposition
From Stein & Wysession
The determinant of a singular matrix is zero
From Stein & Wysession From Stein & Wysession
7
Eigenvalues and eigenvectors in 3 dimensions
From Stein & Wysession From Stein & Wysession
Eigenvalues and eigenvectors
From Stein & Wysession From Stein & Wysession
8
From Stein & Wysession From Stein & Wysession
From Stein & Wysession
From Stein & Wysession
9
Compute the eigenvalues and eigenvectors of Compute the eigenvalues and eigenvectors of the
the covariance matrix covariance matrix
– Vector associated with the maximum eigenvalue should
– Vector associated with the maximum eigenvalue point in the direction of polarization
should point in the direction of polarization
– Works in 3D too, of course
This is your lab assignment for this week.
Azimuth = -63º Azimuth = 76º
10
Related docs
