A New Approach To
Paleoseismic Event
Correlation
Glenn Biasi and Ray Weldon
University of Nevada Reno
Acknowledgments:
Tom Fumal, Kate Scharer, SCEC and the USGS.
Big Question
• How do we estimate seismic hazard
when we can’t prove events correlate
between paleoseismic sites?
– Why it matters (1):
• If they correlate: longer, less frequent ruptures
• If not: shorter, more frequent ruptures.
– Why it matters (2):
• Support for future paleoseismic investigations.
2000
0
Calendar Year
The Data: Paleoseismic
event date pdf’s.
Correlating Events Between
Paleoseismic Sites
• Time correlation is not likely to ever be
entirely convincing.
• Example: two exactly overlapping uniform date
pdf’s six years wide give 1/6 chance that the
events are in the same year.
• Probabilities of correlation based on
displacement fall off with site separation
(ask me later what can be done).
Ways to build a rupture history (1)
N. Bend/
S. Bend
No pattern
Big
events
Example from Weldon et al. (2004) Science
Ways (2) Pearls to Scenarios
• Find all ruptures consistent at some
level with the data
• Build a large suite of rupture scenarios
• Select likely histories using other
constraints (slip rate, dating
consistency, etc.)
• Study the properties of likely histories
for recurrence, segmentation, etc.
Linking involves
some rules for
overlap.
Rules don’t seem
to dominate
results.
Pair-wise joint probability range:
~3e-2 to 3e-3; -> absolute
likelihoods are all small
Example ruptures.
Building Scenarios from Ruptures
• Draw from all possible ruptures until each
paleoquake is included exactly once.
• Scenario is one possible history of rupture on
the fault. Construct 10,000 scenarios.
• Core rupture lengths set by sites in rupture.
• Tails added by drawing around the average
per-event displacement d, then tapering by
9900*d. Use measured d where available.
• Tail truncated if rupture would cross a
neighboring site.
• Tails can extend into creeping zone and
Bombay Beach.
Scoring Scenarios
• Degree of time agreement in ruptures
• Total displacement compared to rupture
prediction in some time.
• Recurrence rate in light of hiatus since
1857 (number of ruptures)
The lack of a southern
SAF earthquake since
1857 constrains the
probable number of
ruptures in viable
scenarios.
E.g., 15 rupture
scenarios are twice as
likely as 22 rupture
scenarios.
Fewest
ruptures
case.
Two wall-to-wall
(W2W) ruptures
since AD 900.
Displacement
scales with
length.
W2W cause
serious overprediction of
total displacement. (14.7 m
mean).
Segment
bounds
(WGCEP) frequency
of singleand multiple
segment
ruptures
fall out
directly.
How much
misfit is too
much?
No wall-to-wall
events;
2 pre-1857
north-bend
events, one
bigger.
WW:
several short
events - uncorrelated.
Best case:
1.43 m. avg.
misfit.
Max age
of
complete
record.
predicted from
slip rate
maximum
displacement
among all
ruptures.
Coachella
misfit
Time
overlap
score
Displ. vs. #
of ruptures;
trend below
21 ruptures
is too few.
Time Score
Best time scores
Fewest ruptures, okay
displ. scores
Best displ. scores
Product
score
vs. # of
ruptures
Displ. Score
Product
time
score vs.
displacement fit.
Displ. score
vs.
rupture
number
No. Ruptures
No. Ruptures
Applications of Rupture Scenarios (2)
segs 1, 2
L1
L2
no segs
seg 2, but L1>L2
Ensembles of scenarios => probabilities of
single and multiple-segment ruptures.
20 scenarios - ~450 ruptures.
Rules for counting segments are required.
Conclusions for Data
Collection
• Complete count and rough event dates are
most valuable
• Don’t need great dating precision (but don’t
stop trying to get it)
• Slip-per event measurements are valuable
• New sites are most valuable in large spatial
gaps.
• Do need geologic or geodetic slip rates
Conclusions for Hazard
Estimation
• Scenarios include earthquake location,
magnitude, and frequency: essentials for
seismic hazard estimation
• Ensembles of likely scenarios support hazard
calculations without having to resolve the perevent correlation issues
• Scenarios are data-based - the paleoseismic
record.
• Can quantify single- and multi-segment rupture
frequency for the whole fault (or use it to
question segmentation!)
Surface Slip versus Rupture Length
Relationship for reverse and normal
may be linear but may NOT for
strike-slip
Rate of increase of strike-slip
decreases with length for strike-slip
without apparently reaching a
plateau.
Having rupture
displacements
helps.
Even with
displacement
measurements,
P(correl) is often
small.
This case: all
magnitudes equally
likely:
Pallett Cr. Wrightwood
Pallett Cr. - Carrizo