Transcript Magnitude Limits: Implications of the 2011 Tohoku Earthquake

Estimating and testing
earthquake magnitude
limits
David D. Jackson, UCLA
Yan Kagan, UCLA
Peter Bird, UCLA
Danijel Schorlemmer, U. Potsdam
Jeremy Zechar, ETHZ
Quick Summary
• We don’t know what limits earthquake size.
• Maximum magnitude unknowable. “Maximum credible earthquake”
totally subjective. Magnitude with defined excedance
probability in specified time is a more useful concept.
• Common methods for estimating magnitude limits include (a)
historic records, (b) fault or plate boundary segment size, and
(c) moment balance in fixed area. All require subjective choices.
Methods (a) and (b) have failed many times.
• Comprehensive assessments over large areas, e.g. all subduction
zones, is safest method. Bird and Kagan (2004) applied moment
balance to a handfull of global tectonic regions and have yet to
be embarrassed.
• Methodology developed by the Collaboratory for Study of
Earthquake predictability, (CSEP), can be used to test
hypotheses of excedance probability in specified time.
Ways to estimate size limits
• History of earthquakes within region
– May use proxy such as tsunami runup
• Fault-length or –area scaling
– May use largest fault in a region to characterize the whole
region.
• Copying Mmax of region with similar tectonics
• Tectonic and seismic moment balance, assuming a
magnitude – frequency relation.
Problems for estimating size
limits
• Data are limited, especially temporally.
• Choice of region and its size is arbitrary.
• Definition of segments and rupture
termination are imprecise.
• Objectives conflict
– Small region has inadequate data, large region
may miss important tectonics
– Different communities have biases: academics,
developers, insurers, environmentalists.
Historical Record?
Prior estimates of Tohoku
Magnitude Limits
• Ruff and Kanamori, 1980: 8.2 (historical, age and
plate-rate systematics)
• Nishenko, 1991: 7.6 (characteristic), 8.1 (historical,
1611)
• Minoura et al., 2001: 8.3, 200x85 km, based on 869
Jogan Tsunami
• Bird and Kagan, 2004: Corner magnitude 9.6 for all
subduction zones.
• Koravos et al., 2006: 7< Mmax <8, based on
historical earthquakes since 599 AD.
• Annaka et al., 2007: 8.5 based on historical
earthquakes since 1611.
• Stein and Okal, 2010: 9+ for all subduction zones.
Fault Length Scaling,
Global Earthquakes
1000
Wells & Coppersmith
Pegler & Das
Length, km
100
10
1
5.0
6.0
7.0
Mw
8.0
Fault-length scaling doesn’t work;
many S. Calif. quakes don’t stay on mapped faults.
Comparison of
prior fault length
with rupture
length, from
UCLA PhD thesis
of Natanya Black,
2009.
Fault map before (thin black lines) and
after (thicker grey lines) the Elmore
Ranch earthquake of 1987 (magnitude
6.7). From Natanya Black, UCLA.
(Do we really think the the Pacific-North America plate boundary remains
unfaulted in some places? Or, is this map just incomplete?)
The fault that caused the 2001 Bhuj
earthquake may have been previously
mapped [Malek et al., 2000]; its length of 80
km suggests maximum magnitude 7.3 [Wells
& Coppersmith, 1994]. Actual seismic
moment was
3.5x greater than this.
Roger Bilham, CIRES/U CO
*Denali, AK earthquake
of 2002.01.03
(m = 7.9)
jumped from the
Susitna Glacier fault,
to the
Denali fault,
to the
Totschunda fault.
USGS
Other Anomalously Large Earthquakes Violating
“Segment” rules
•
•
•
•
•
•
Sumatra, 2004, 9.4: Broke through “segment boundaries”, violated Ruff
and Kanamori 1980 “fast and young” systematics (Gutscher and
Westbrook, 2009).
Solomon Islands, 2007, 8.1: Ruptured through plate triple junction
(Furlong et al, 2009)
Balleny Islands, 1998, 8.1. Strike slip earthquake on unknown fault in
oceanic crust, 200 km from spreading center (Hjorleifsdottir et al.,
2009).
Macquarrie Ridge, 1989, 8.2: Strike slip on 120km long unknown fault,
5km oceanic crustal thickness.
Darfield, New Zealand, 2010, 7.1. Unknown fault in well studies area.
El Mayor – Cucupah, Mexico, 2010, 7.1. 75 km rupture on combination of
known and unknown faults
Global Tectonic Zones
Bird, Kagan, and Jackson 2010
Magnitude-frequency relation for all subduction zones combined
[Bird & Kagan, 2004, BSSA]
+Sumatra, +Sendai
5. Even outside of broad orogens, dangerous
intraplate faulting is evident in catalogs:
(c) The corner magnitude of intraplate earthquakes is
>7.6, and unconstrained from above, on the
moment magnitude scale [Bird & Kagan, 2004, BSSA].
How to test Mmax models
prospectively
• Specify Mmax on a global (or very
comprehensive) grid, 0.1 degree.
– Requires definition of Mmax
– Few people bold enough to risk writing it down
– Requires location to be epicenter.
• Global CSEP probability forecasts with
lower thresholds of 7.5, 8.0, 8.5, 9.0, 9.5.
– Fits existing CSEP template
– Allows finite time window
Global long-term potential
based on smoothed
seismicity. from the CMT
catalog since 1977.
Earthquake occurrence is
modeled by a timeindependent process.
Colors show the long-term
probability of earthquake
occurrence.
Conclusions
1.
2.
3.
4.
5.
6.
We don’t know what limits earthquake size. It does not seem to be
segment boundaries, pre-existing fault extent, plate boundary triple
junctions, or crustal thickness.
Efforts to estimate maximum magnitude based on geographically
specific earthquake histories or tectonic situations tend to
underestimate because of limited observations.
Some global or generic estimates (e.g. all subduction zones) haven’t yet
been falsified, but constructing facilities to withstand the implied
upper limits could be impossibly expensive.
Unconditional maximum magnitude is not a scientifically meaningful
concept, as it can’t be tested in a finite time. Time limited
probabilistic hypotheses are more useful and could possibly be
testable.
Simple hypotheses that can be applied over much of the globe can be
tested in a reasonable time; more complicated ones can’t.
A CSEP experiment could fairly test some forecast probabilities of
earthquakes over 7.5, 8.0, 8.5, 9.0 and 9.5 globally for 5 or 10 year
period.
For example, this area of monotonous granodiorite in the southern Sierras
has only a few faults mapped...
“Can diligent and extensive mapping of
faults provide reliable estimates of the
expected maximum earthquakes at
these faults?”
No.
Peter Bird
UCLA
2010 Fall AGU, S23B-02
2. Fault trace “lengths” are unreliable guides
to maximum magnitude.
• Fault networks have multiply-branching, quasifractal shapes, so fault “length” may be
meaningless.
• Naming conventions for main strands are
unclear, and rarely reviewed.
• Gaps due to Quaternary alluvial cover may not
reflect deeper seismogenic structure.
• Mapped kinks and other “segment boundary
asperities” may be only shallow structures.
• Some recent earthquakes have jumped and
linked “separate” faults:
4. A recent attempt [Bird, 2009, JGR] to model neotectonics
of the active fault network in the western United States
found that only 2/3 of Pacific-North America relative motion in California
occurs by slip on faults included in seismic hazard models by the
2007 Working Group on California Earthquake Probabilities.
Simple Scaling Model for Length,
Displacement, and Down-dip
Width
What is Mmax?
• Mmax poorly defined, and not measurable.
• Maximum credible earthquake (MCE) is subjective, and
still not measurable.
• Corner magnitude, where tapered GR (TGR) relationship
tapers, presupposes the TGR distribution but it is
measureable for large areas and long times.
• Assuming TGR with estimated corner magnitude and fixed
time duration, one can estimate a “functional magnitude
limit: (FML) which will be exceeded only at arbitrarily
small probability. This may be more useful than “Mmax”
because
– Structures have finite lifetimes
– FML can be estimated with enough data
– Assumed FML can be tested
Nishenko (1991) Circum Pacific
Characteristic Rupture Zones
but seismicity suggests that there may be at least one more ...