WEBVTT
Kind: captions
Language: en

00:00:00.640 --> 00:00:02.060
Good morning, everyone.

00:00:02.060 --> 00:00:06.339
Thank you for coming out
two days in a row.

00:00:06.339 --> 00:00:09.110
Our housekeeping note for next week
is we’ll be back to our normal

00:00:09.110 --> 00:00:13.750
seminar day and time -- Wednesday,
October 7th, 10:30 a.m.,

00:00:13.750 --> 00:00:17.410
when our speaker will be
Emily Brodsky from UC-Santa Cruz

00:00:17.410 --> 00:00:19.410
presenting Constraints from
Fault Roughness

00:00:19.410 --> 00:00:21.990
on Scale-Dependent
Strength of Rocks.

00:00:21.990 --> 00:00:24.919
Now today, giving a special
encore of the Joyner Lecture,

00:00:24.919 --> 00:00:26.079
we have Paul Somerville.

00:00:26.079 --> 00:00:29.899
And to introduce Paul,
we have Tom Hanks.

00:00:33.560 --> 00:00:36.059
- Yeah. Today’s speaker
is Paul Somerville,

00:00:36.059 --> 00:00:42.269
and he’ll be giving
the Joyner Lecture.

00:00:42.269 --> 00:00:46.940
Paul is the
12th Joyner Lecture for the --

00:00:46.940 --> 00:00:51.289
for the current --
for the current calendar year.

00:00:51.289 --> 00:00:55.199
And I think -- well, many of
the Joyner Lectures have speaked

00:00:55.199 --> 00:00:58.879
in this room in previous room in
previous years, and Joyner Lectures

00:00:58.879 --> 00:01:05.210
are chosen for their work
at the interface -- their work

00:01:05.210 --> 00:01:07.670
and their contributions
at the interface between

00:01:07.670 --> 00:01:11.250
earthquake science
and earthquake engineering.

00:01:11.250 --> 00:01:17.490
And Paul Somerville is certainly
well-qualified to meet the --

00:01:17.490 --> 00:01:25.060
those criteria and is probably the
best-known person in the world

00:01:25.060 --> 00:01:33.189
for bringing real seismologically based
ground motion models and simulations

00:01:33.189 --> 00:01:38.509
into earthquake
engineering applications.

00:01:38.509 --> 00:01:42.780
Paul did his undergraduate work
at the University of New England

00:01:42.780 --> 00:01:48.789
in Australia and then did a Ph.D.
at University of British Columbia.

00:01:48.789 --> 00:01:55.420
He was a post-doc at Earthquake
Research Institute in Tokyo.

00:01:55.420 --> 00:02:01.840
Prior to -- or at the same time, he was
working at Woodward-Clyde and then --

00:02:01.840 --> 00:02:08.039
which moved on to URS and --
which merged recently with AECOM.

00:02:08.039 --> 00:02:11.419
So here’s Paul.
- Thank you, Tom.

00:02:11.420 --> 00:02:14.760
- I’m going to give that to you.

00:02:23.980 --> 00:02:30.760
- Well, thank you very much
for the invitation to speak here.

00:02:30.760 --> 00:02:33.209
I need to explain
that this talk is given

00:02:33.209 --> 00:02:36.690
at the annual meeting
of the EERI and SSA.

00:02:36.690 --> 00:02:40.019
So it has to be something
that’s of interest

00:02:40.019 --> 00:02:43.760
to both engineers
and geoscientists.

00:02:43.760 --> 00:02:49.650
So a lot of what I’ll say is
something you already know.

00:02:49.650 --> 00:02:53.680
And you’ll have to put up
with me talking about things

00:02:53.680 --> 00:02:59.000
in a simple way
to engineering audiences.

00:02:59.000 --> 00:03:03.120
If you ever want to learn -- really learn
about structural mechanics, though,

00:03:03.120 --> 00:03:08.319
I suggest you go to an EERI meeting
because they have these groups

00:03:08.319 --> 00:03:16.120
of students who, every year, take part
in a competition to design, build,

00:03:16.120 --> 00:03:22.150
and shake, you know, scale models
of mid-rise and high-rise buildings

00:03:22.150 --> 00:03:23.940
out of balsa wood.

00:03:23.940 --> 00:03:27.230
And if you hang around just for an
hour or so, they’ll all -- the students --

00:03:27.230 --> 00:03:32.610
you know, it’s like a poster session,
but they’ve got all of these buildings

00:03:32.610 --> 00:03:36.110
that they’ve built and -- designed,
built, and then they test them.

00:03:36.110 --> 00:03:38.560
And if you ask them
what they did -- what they --

00:03:38.560 --> 00:03:42.260
you know, what is their brilliant
new concept, you will learn so much

00:03:42.260 --> 00:03:46.810
[chuckles] about structural engineering.
It’s really unbelievable.

00:03:46.810 --> 00:03:49.569
But you have to understand,
a lot of this stuff I’m going to say here

00:03:49.569 --> 00:03:53.750
was designed for people
like that to understand.

00:03:53.750 --> 00:04:00.260
Okay, so like me,
you must be wondering,

00:04:00.260 --> 00:04:04.129
what was the Joyner Committee
thinking when they chose me.

00:04:04.129 --> 00:04:08.360
I think maybe they thought about my
association with illustrious people.

00:04:08.360 --> 00:04:12.769
I had bosses like
Lloyd Cluff and Ed Idriss.

00:04:12.769 --> 00:04:14.909
My workmates include
Norm Abrahamson,

00:04:14.909 --> 00:04:18.240
Rob Graves, Arben Pitarka,
Dave Wald.

00:04:18.240 --> 00:04:22.470
And those -- you’ve pinched some
of those people from us, right?

00:04:22.470 --> 00:04:27.970
And I’ve borrowed from these people
a lot in the -- in preparing my talk.

00:04:27.970 --> 00:04:33.930
My mentors were Don Helmberger,
Kojiro Irikura in Japan.

00:04:33.930 --> 00:04:37.950
If not that, then maybe they
were thinking I showed enterprise

00:04:37.950 --> 00:04:40.350
by changing companies twice.

00:04:40.350 --> 00:04:43.170
But actually [chuckles]
that didn’t happen.

00:04:43.170 --> 00:04:49.530
I just stayed where I was [chuckles],
and my companies quit on me.

00:04:49.530 --> 00:04:53.680
So the title is sort of phrased
in the way earthquake engineers talk --

00:04:53.680 --> 00:04:58.310
structural engineers --
capacity and demand, right?

00:04:58.310 --> 00:05:02.200
So it’s Geoscientists’ Capacity
to Meet Engineers’ Demands

00:05:02.200 --> 00:05:04.480
for Seismic Hazard Inputs.

00:05:04.480 --> 00:05:08.150
And this is exactly
what we honor Bill Joyner for.

00:05:08.150 --> 00:05:11.160
And the USGS continues
its fine tradition

00:05:11.160 --> 00:05:12.590
of inclusive and
open collaboration.

00:05:12.590 --> 00:05:14.980
You can see I’m listing
all these people --

00:05:14.980 --> 00:05:17.800
David Boore et al.,
ground motion prediction models.

00:05:17.800 --> 00:05:19.930
OpenSHA -- Ned Field.

00:05:19.930 --> 00:05:23.790
Seismic Hazard Mapping -- Art Frankel,
Mark Petersen -- those people.

00:05:23.790 --> 00:05:26.580
Risk Targeted Seismic Hazard --
Nico Luco.

00:05:26.580 --> 00:05:29.070
PAGER Situational Awareness --
Dave Wald.

00:05:29.070 --> 00:05:31.370
Outreach --
Lucy Jones, Ross Stein.

00:05:31.370 --> 00:05:35.840
I mean, I could close my case
right now without even going to

00:05:35.840 --> 00:05:40.750
university researchers who have
contributed on this interface.

00:05:40.750 --> 00:05:47.210
But I have recently been
in Christchurch, New Zealand.

00:05:47.210 --> 00:05:49.970
And -- well, let me show you
what it looks like now.

00:05:49.970 --> 00:05:53.500
Last time I was there,
this is what it looked like.

00:05:53.500 --> 00:05:57.650
80% of the buildings in the
central business district --

00:05:57.650 --> 00:06:04.040
that means downtown --
have been demolished.

00:06:04.040 --> 00:06:08.900
The earthquake that did most of
that damage was magnitude 6.2.

00:06:08.900 --> 00:06:12.100
It occurred on an
unknown fault beneath the city.

00:06:12.100 --> 00:06:14.660
There was a lot of widespread
damage due to liquefaction,

00:06:14.660 --> 00:06:18.930
and that’s the reason, I think,
for a lot of the demolition.

00:06:18.930 --> 00:06:24.900
But it locally produced ground motions
much stronger than code levels.

00:06:24.900 --> 00:06:28.950
And I’ll go into talking about
why that might have been.

00:06:28.950 --> 00:06:35.020
80% of the CBD buildings are
demolished despite the fact that

00:06:35.020 --> 00:06:38.930
professors Park and Paulay from
Canterbury University in Christchurch

00:06:38.930 --> 00:06:43.280
wrote the book on the design of
concrete-frame structures.

00:06:43.280 --> 00:06:48.330
They literally wrote the book.

00:06:48.330 --> 00:06:54.080
The USGS now -- sorry --
GNS now have greatly increased

00:06:54.080 --> 00:07:00.000
the seismic hazard level based on time-
dependent forecasting, up from about

00:07:00.000 --> 00:07:12.220
0.22g to 0.6g -- a 500-year return period,
which is more than Christchurch at 0.4g.

00:07:12.220 --> 00:07:16.370
So an immediate question is, you know,
where else could this happen, right?

00:07:16.370 --> 00:07:19.970
So our work is not done.

00:07:19.970 --> 00:07:24.370
And so, yeah,
here’s the earthquake.

00:07:24.370 --> 00:07:28.350
This is the USGS at the bottom --
a ShakeMap.

00:07:33.260 --> 00:07:37.580
Yeah, so here’s the GS ShakeMap,
and this is the building density map.

00:07:37.590 --> 00:07:42.930
So you can see this little earthquake
was a direct hit on the city.

00:07:42.930 --> 00:07:48.580
And Brendon Bradley has shown
that within 10 kilometers, the --

00:07:48.580 --> 00:07:54.030
this is 1-second spectral acceleration --
was larger than his model.

00:07:54.030 --> 00:08:01.560
But beyond 10 kilometers, it was
ordinary, or maybe lower than expected.

00:08:01.560 --> 00:08:04.700
It suggests that it wasn’t a matter --
a simple matter of high stress drop

00:08:04.700 --> 00:08:09.170
or some high source characteristics,
that instead, maybe there were

00:08:09.170 --> 00:08:13.470
other things -- local things
in Christchurch.

00:08:13.470 --> 00:08:16.570
You can see -- this dashed line
is the building code spectrum,

00:08:16.570 --> 00:08:20.220
and you can see -- these are
the horizontals here -- they vastly

00:08:20.220 --> 00:08:27.000
exceeded the code spectrum,
as much as a factor of 2 at 1 second.

00:08:27.000 --> 00:08:32.190
So that would be really bad
for mid-rise buildings.

00:08:32.190 --> 00:08:34.720
So let’s talk about
contrasting characteristics

00:08:34.720 --> 00:08:37.570
of earthquake engineers
and scientists.

00:08:37.570 --> 00:08:41.380
For earthquake engineers, the proper
course of action is the one that

00:08:41.380 --> 00:08:46.380
maximizes utility, usually defined
as maximizing total benefit,

00:08:46.380 --> 00:08:53.790
human-centered on a moral foundation,
by balancing demand versus capacity

00:08:53.790 --> 00:09:01.690
or cost versus benefit, using
situational conditional thinking.

00:09:01.690 --> 00:09:03.940
Earthquake scientists have really
quite a different enterprise.

00:09:03.940 --> 00:09:07.000
They seek to understand
the workings of nature,

00:09:07.000 --> 00:09:11.940
which are not situational
or conditional.

00:09:11.940 --> 00:09:14.690
So what engineers do --
this is what they’re thinking.

00:09:14.690 --> 00:09:18.630
Here’s the venerable Joe’s Bar
created by Ron Hamburger.

00:09:18.630 --> 00:09:23.510
The big, red arrow is the
horizontal motion of the ground.

00:09:23.510 --> 00:09:27.670
And the arrows up the side
of the building indicate

00:09:27.670 --> 00:09:34.190
a representation of the forces that that
effectively impose on the building.

00:09:34.190 --> 00:09:37.750
And so it’s basically --
this is a static pushover, right?

00:09:37.750 --> 00:09:42.240
On the vertical axis, we have the force,
or the base shear demand.

00:09:42.240 --> 00:09:45.430
And then we have the
lateral deformation of the roof

00:09:45.430 --> 00:09:48.090
of the building
on the horizontal axis.

00:09:48.090 --> 00:09:52.990
And you can see,
as we go through stronger forces,

00:09:52.990 --> 00:09:57.720
we have various states
of damage to the building.

00:09:57.720 --> 00:09:59.840
In the linear range,
it’s sort of operational.

00:09:59.840 --> 00:10:02.400
Then we get life safe.

00:10:02.400 --> 00:10:09.520
And then we have structurally stable,
and then we have collapse.

00:10:09.520 --> 00:10:15.570
And these levels of ground motion
or damage are associated with

00:10:15.570 --> 00:10:19.700
different levels of ground motion --
more frequent for operational

00:10:19.700 --> 00:10:22.350
and then becoming less frequent,
more rare.

00:10:22.350 --> 00:10:28.430
Typically, 10% in 50 years -- we design
for life safety in building codes for that.

00:10:28.430 --> 00:10:31.120
And then we design for
structural stability,

00:10:31.120 --> 00:10:38.140
or check for structural stability
at 25/100-year return period.

00:10:38.140 --> 00:10:42.420
The work we do as geoscientists
is the hazard, right?

00:10:42.420 --> 00:10:48.510
So we have this objective
scientific hazard,

00:10:48.510 --> 00:10:51.200
characterized by an intensity
measure like spectral acceleration

00:10:51.200 --> 00:10:54.860
or velocity or duration
or something like that.

00:10:54.860 --> 00:11:00.690
And then, conditional on that
ground-shaking level, we have the

00:11:00.690 --> 00:11:04.910
engineering demand parameter,
which measures the effect of that

00:11:04.910 --> 00:11:08.890
on the building in terms of drift -- that’s
the horizontal displacement of the roof

00:11:08.890 --> 00:11:14.610
with respect to the ground, acceleration
in the floors, and then so forth.

00:11:14.610 --> 00:11:18.500
And conditional on those forces
and deformations in the building,

00:11:18.500 --> 00:11:22.380
we have a damage measure,
which is the condition of the building,

00:11:22.380 --> 00:11:25.010
what repairs are necessary,
and so forth.

00:11:25.010 --> 00:11:29.180
And then, conditioned on that,
there are these decision variables --

00:11:29.180 --> 00:11:32.570
dollar loss, downtime,
life-safety.

00:11:32.570 --> 00:11:38.750
These is sometimes called the three Ds --
dollars, downtime, and deaths.

00:11:38.750 --> 00:11:42.390
So that’s the impact, right?
And so the impact is measured this way.

00:11:42.390 --> 00:11:47.250
This is performance-based design
created by the people at PEER

00:11:47.250 --> 00:11:53.490
and very strongly influenced
by people who were very near here --

00:11:53.490 --> 00:11:56.370
people like Alan Cornell.

00:11:59.380 --> 00:12:05.280
There’s a contrast in mindfulness
between engineers and geoscientists.

00:12:05.290 --> 00:12:09.860
Folklore says that scientists
are absent-minded.

00:12:09.860 --> 00:12:13.730
There are few references
to absent-minded engineers.

00:12:13.730 --> 00:12:17.060
Which brings to mind
the story of Sherlock Holmes

00:12:17.060 --> 00:12:18.970
when he went camping
with Watson.

00:12:18.970 --> 00:12:22.930
They were in a tent telling stories
until late in the night.

00:12:22.930 --> 00:12:25.250
And then they fell asleep.

00:12:25.250 --> 00:12:30.420
And then Holmes woke up
with a bit of a start.

00:12:30.420 --> 00:12:32.540
And then he woke Watson.

00:12:32.540 --> 00:12:39.150
And he said, Watson, look up at
the stars and tell me what you surmise.

00:12:39.150 --> 00:12:44.010
And so Watson said, well, there
are billions of stars in universe.

00:12:44.010 --> 00:12:48.660
Some of those stars might have
planetary systems like ours.

00:12:48.660 --> 00:12:54.250
And on some of those planets,
there might be intelligent life like us.

00:12:54.250 --> 00:13:00.800
To which Holmes replied, “Watson,
you idiot, somebody stole our tent.”

00:13:00.800 --> 00:13:04.600
[laughter]

00:13:04.600 --> 00:13:08.900
So I’m trying to find --
I went looking for the antonym.

00:13:08.900 --> 00:13:13.279
I’ve said engineers think
in a situational, conditional way.

00:13:13.279 --> 00:13:17.770
Oddly enough, the antonym dictionary
didn’t have any antonyms for situational,

00:13:17.770 --> 00:13:23.180
but when I looked up conditional,
I got a torrent of adjectives.

00:13:23.180 --> 00:13:28.730
So if we’re the opposite of engineers,
then the antonym dictionary says

00:13:28.730 --> 00:13:34.000
that we are absolute, arbitrary, arrogant,
authoritative, autocratic, coercive,

00:13:34.000 --> 00:13:38.740
commanding, compulsive, compulsory,
controlling, despotic, dictatorial,

00:13:38.740 --> 00:13:43.480
dogmatic, domineering, exacting,
haughty, imperative, imperious,

00:13:43.480 --> 00:13:46.970
irresponsible, lordly, overbearing,
peremptory, positive --

00:13:46.970 --> 00:13:53.779
we’ll take that one -- supreme, maybe --
tyrannical, and unconditional.

00:13:53.779 --> 00:13:58.779
But among all of those things,
I want to tell the engineers that,

00:13:58.779 --> 00:14:04.750
in the 1960s, geoscientists
were also revolutionary.

00:14:04.750 --> 00:14:10.440
And so, of course, I want to
tell them about plate tectonics.

00:14:10.440 --> 00:14:14.470
And here’s a scheme showing,
you know, plate tectonics at work.

00:14:14.470 --> 00:14:18.940
You can see there’s a ridge here,
and then there’s a transform fault.

00:14:18.940 --> 00:14:21.710
Let me -- oh, yeah.
Let me go into detail here.

00:14:21.710 --> 00:14:29.020
So the illustrator has taken the
common-sense 2D interpretation

00:14:29.020 --> 00:14:30.980
of this transform fault, right?

00:14:30.980 --> 00:14:34.160
Obviously, in his mind,
this is a left-lateral fault

00:14:34.160 --> 00:14:37.480
that’s offset these ridges, right?

00:14:37.480 --> 00:14:45.420
But the thinking of Tuzo Wilson --
J. Tuzo Wilson -- barely 50 years ago,

00:14:45.420 --> 00:14:49.740
a new class of faults and their
bearing on continental drift.

00:14:49.740 --> 00:14:55.240
It’s really right-lateral, right?
So when you have 180 degrees' error,

00:14:55.240 --> 00:15:01.060
then it’s time to say game’s up
[chuckles], and you have a total

00:15:01.060 --> 00:15:06.839
revolution in their understanding
of why earthquakes happen.

00:15:06.839 --> 00:15:10.950
We used to have an ML --
peak amplitude on a

00:15:10.950 --> 00:15:13.990
Wood-Anderson
at 100 kilometers’ distance.

00:15:13.990 --> 00:15:17.150
Now we have Mw, derived from
log moment, where moment

00:15:17.150 --> 00:15:21.710
is the moment of each of
this pair of couples,

00:15:21.710 --> 00:15:25.080
which we use to represent
a shear dislocation.

00:15:25.080 --> 00:15:27.360
It took a while to figure out
that that was

00:15:27.360 --> 00:15:30.690
the right force representation
for an earthquake.

00:15:30.690 --> 00:15:36.210
And so we have fault parallel
displacement like this, and then we have

00:15:36.210 --> 00:15:42.649
this huge directivity pulse on the fault
normal component to keep dynamic

00:15:42.649 --> 00:15:48.770
equilibrium that makes this directivity
pulse that I will talk about later.

00:15:48.770 --> 00:15:53.300
And so we have -- for the geologists,
if we have the seismic moment rate --

00:15:53.300 --> 00:15:56.020
oh, yeah -- seismic moment
is the shear modulus

00:15:56.020 --> 00:16:01.490
times the earthquake fault area
times the slip on the fault.

00:16:01.490 --> 00:16:06.339
Moment rate -- if we have geologists
out there measuring slip rate,

00:16:06.339 --> 00:16:10.050
then we can actually calculate
earthquake recurrence from slip rate

00:16:10.050 --> 00:16:13.160
because that gives us
our earthquake budget.

00:16:13.160 --> 00:16:17.800
All very heady stuff.

00:16:17.800 --> 00:16:21.110
Common geoscience and
engineering role in forecasting.

00:16:21.110 --> 00:16:23.860
Geoscience and structural mechanics --
an explanation is valid

00:16:23.860 --> 00:16:26.500
only if it has
predictive capabilities, right?

00:16:26.500 --> 00:16:30.980
So we have to live
by that rule in science.

00:16:30.980 --> 00:16:36.750
In engineering practice,
it relies a lot on forecast probabilities.

00:16:36.750 --> 00:16:39.990
That performance-based
earthquake engineering I talked about

00:16:39.990 --> 00:16:44.100
is basically a means
of forecasting damage.

00:16:44.100 --> 00:16:47.120
And it’s based on probabilistic
seismic hazard analysis,

00:16:47.120 --> 00:16:52.649
which is obviously a forecast
of ground motion frequency.

00:16:52.649 --> 00:16:55.540
In engineering,
a design procedure is valid

00:16:55.540 --> 00:16:59.649
if it has been shown
to be effective in past practice.

00:16:59.649 --> 00:17:03.490
I use the word “prediction” --
earthquake prediction.

00:17:03.490 --> 00:17:06.390
We can’t do it yet.
We’re not sure it’s possible.

00:17:06.390 --> 00:17:08.740
But earthquake forecasting,
we do that all the time.

00:17:08.740 --> 00:17:12.480
It’s the foundation of probabilistic
seismic hazard analysis.

00:17:12.480 --> 00:17:17.329
But in Christchurch, how do we
properly use it in time-varying forecasts?

00:17:17.329 --> 00:17:20.589
So I’m alluding now back to the
fact that, in the aftermath

00:17:20.589 --> 00:17:26.000
of the Christchurch earthquake,
GNS are saying that the seismic hazard

00:17:26.000 --> 00:17:32.900
is three times -- the ground motion is
three times higher than it used to be.

00:17:32.900 --> 00:17:36.020
Contrast in accountability.

00:17:36.020 --> 00:17:39.590
Geoscientists do not
sign design documents.

00:17:39.590 --> 00:17:43.050
They are not usually subject
to economic and political forces.

00:17:43.050 --> 00:17:46.940
A recent exception was, of course,
the ones who got caught up

00:17:46.940 --> 00:17:49.880
in the L’Aquila issue in Italy.

00:17:49.880 --> 00:17:52.660
They’re not worried about precedents.

00:17:52.660 --> 00:17:54.910
They just focus on today’s best answer.

00:17:54.910 --> 00:17:58.580
That’s kind of a
luxurious situation, I think.

00:17:58.580 --> 00:18:01.350
Engineers, by contrast,
sign design documents

00:18:01.350 --> 00:18:03.370
based on codes
that allow collapse.

00:18:03.370 --> 00:18:05.740
[chuckles]
Why do they do that?

00:18:05.740 --> 00:18:11.480
Because, as I’ll show you,
geoscientists can’t bound the hazard.

00:18:11.480 --> 00:18:13.960
They are subject to economic
and political forces.

00:18:13.960 --> 00:18:17.330
For example, building costs.

00:18:17.330 --> 00:18:21.290
Owners of earthquake-damaged buildings
that they designed coming after them.

00:18:21.290 --> 00:18:24.510
They’re worried about precedents,
for example, in building codes

00:18:24.510 --> 00:18:29.020
and in the portfolios of buildings
that they have designed.

00:18:29.020 --> 00:18:33.920
They prefer smooth transitions
from past practice.

00:18:33.920 --> 00:18:37.940
Speaking of smooth transition
from past practice, again,

00:18:37.940 --> 00:18:43.280
Ron Hamburger, I think, is the
coolest artist at this kind of thing.

00:18:43.280 --> 00:18:49.160
The MCE used to be a deterministic,
you know, maximum credible earthquake.

00:18:49.160 --> 00:18:53.540
The CE now became maximum
considered earthquake,

00:18:53.540 --> 00:18:58.130
which means beyond a certain ground
shaking level like 2% in 50 years.

00:18:58.130 --> 00:19:02.570
We don’t -- we don’t -- we just ignore
larger ground motion levels than that.

00:19:02.570 --> 00:19:05.660
And that’s obviously a
probabilistic framework.

00:19:05.660 --> 00:19:11.880
So there is a 10% probability of
collapse in 50 years under the MCE.

00:19:11.880 --> 00:19:16.000
Note the differing definitions
of earthquake.

00:19:16.000 --> 00:19:19.070
The engineer and
Webster’s Dictionary

00:19:19.070 --> 00:19:23.470
say that an earthquake
is a shaking of the ground.

00:19:23.470 --> 00:19:29.570
Geoscientists say an earthquake is a
sudden shear dislocation on a fault.

00:19:29.570 --> 00:19:34.059
And the geoscientist
pedantically says that the MCE

00:19:34.059 --> 00:19:36.750
is the ground motion from
an earthquake, right?

00:19:36.750 --> 00:19:41.230
Because if you say this building is
designed for a magnitude 7.5

00:19:41.230 --> 00:19:50.410
earthquake, that doesn’t help very much
if you don’t know where the site is.

00:19:50.410 --> 00:19:53.070
Accounting for uncertainty
and randomness, which this is

00:19:53.070 --> 00:19:59.250
common to PSHA and PBEE,
epistemic uncertainty is

00:19:59.250 --> 00:20:02.020
uncertainty about the
true state of nature.

00:20:02.020 --> 00:20:04.850
The alternatives
are mutually exclusive.

00:20:04.850 --> 00:20:10.620
For example, a fault can’t both have
a vertical dip and a dip of 45 degrees.

00:20:10.620 --> 00:20:15.320
And they represent -- we represent
epistemic uncertainty using logic trees

00:20:15.320 --> 00:20:22.600
outside that PBEE integral -- that big
four-fold integral I showed you earlier.

00:20:22.600 --> 00:20:24.740
Aleatory variability
is something different.

00:20:24.740 --> 00:20:29.630
It’s just event-to-event or site-to-site
or building-to-building randomness.

00:20:29.630 --> 00:20:32.970
Obviously, these
alternatives co-exist.

00:20:32.970 --> 00:20:36.240
And we represent this
using distributions, for example,

00:20:36.240 --> 00:20:39.870
the standard deviation in
a ground motion prediction equation,

00:20:39.870 --> 00:20:44.410
in the hazard term
inside the PBEE integral.

00:20:44.410 --> 00:20:52.870
Here’s a very well-recorded earthquake
in Japan -- data from Hiroe Miyake.

00:20:52.870 --> 00:20:55.600
This is the
Niigata Chuetsu earthquake,

00:20:55.600 --> 00:20:59.500
and this is showing the wide scatter
in recorded peak acceleration.

00:20:59.500 --> 00:21:03.090
That’s -- the peak acceleration
is on the vertical axis,

00:21:03.090 --> 00:21:08.470
and distance is
on the horizontal axis.

00:21:08.470 --> 00:21:11.020
And so we represent
this random variability

00:21:11.020 --> 00:21:14.100
by an unbounded
lognormal distribution.

00:21:14.100 --> 00:21:17.100
People struggle to truncate it.
We’d like to truncate it some way.

00:21:17.100 --> 00:21:24.590
But up to about three sigmas,
it’s hard to justify truncating it.

00:21:24.590 --> 00:21:28.540
And epsilon is the number
of standard deviations.

00:21:28.540 --> 00:21:32.470
And then there’s a model --
the median goes through,

00:21:32.470 --> 00:21:33.840
more or less,
the middle of the data.

00:21:33.840 --> 00:21:36.890
And then there’s -- this is a plus
or minus 1 standard deviation.

00:21:36.890 --> 00:21:40.770
So this is epistemic uncertainty --
okay, so this is -- this is

00:21:40.770 --> 00:21:44.630
a representation of the
random variability in the model.

00:21:44.630 --> 00:21:49.650
But there’s also epistemic uncertainty
in the true value of the median.

00:21:49.650 --> 00:21:55.370
So there may be -- may be 10 or 15%
uncertainty about what that true value is.

00:21:59.940 --> 00:22:04.740
Oh, you can see some of these -- these
open circles were hanging wall sites.

00:22:04.740 --> 00:22:08.850
So you can see the hanging wall
is a lot bigger than the other sites.

00:22:08.850 --> 00:22:11.130
And so if you want to
model hanging wall, then you

00:22:11.130 --> 00:22:18.220
can take out -- get rid of
some of this random variability.

00:22:18.220 --> 00:22:22.290
And what happens, of course,
is with this unbounded lognormal

00:22:22.290 --> 00:22:25.440
distribution, when you do a
seismic hazard calculation --

00:22:25.440 --> 00:22:32.580
this is now peak acceleration
versus decreasing annual probability

00:22:32.580 --> 00:22:35.900
of exceedance or
increasing return period.

00:22:35.900 --> 00:22:41.980
The hazard curve just keeps on
growing indefinitely, forever, right?

00:22:41.980 --> 00:22:44.530
This is a site in
New Zealand -- Kaikoura.

00:22:44.530 --> 00:22:48.750
This is a site in
Melbourne, Australia.

00:22:48.750 --> 00:22:51.920
You can see earthquake ground motions
are about 100 times more frequent

00:22:51.920 --> 00:22:57.130
in Kaikoura than they are in Melbourne.
But still, in both cases, there’s no bound

00:22:57.130 --> 00:23:00.950
to the hazard unless you
truncate that random variability

00:23:00.950 --> 00:23:04.110
that I showed you
in the previous slide.

00:23:04.110 --> 00:23:08.049
This is just showing
some seismic hazard maps.

00:23:08.049 --> 00:23:11.360
So the probabilistic ground motion
hazard grows with increasing return

00:23:11.360 --> 00:23:16.860
period because larger and closer
earthquakes become more likely.

00:23:16.860 --> 00:23:19.710
And the random variability in
ground motions for a given

00:23:19.710 --> 00:23:24.260
earthquake magnitude and distance
also contributes and causes the hazard

00:23:24.260 --> 00:23:27.740
to grow indefinitely
with increasing return period.

00:23:27.740 --> 00:23:31.790
So even when you’ve had enough time
to have the biggest possible earthquake

00:23:31.790 --> 00:23:35.460
at the closest possible distance,
the hazard still keeps growing

00:23:35.460 --> 00:23:38.480
because there’s random variability
in how big the ground motion

00:23:38.480 --> 00:23:41.540
might be for the next event.

00:23:41.540 --> 00:23:44.240
And so there is no way
to bound the hazard.

00:23:44.240 --> 00:23:48.460
And so somebody
makes a policy decision.

00:23:48.460 --> 00:23:51.820
This is not an engineering
or a scientific decision.

00:23:51.820 --> 00:23:58.470
This is a policy decision.
What is the acceptable level of risk?

00:23:58.470 --> 00:24:02.350
And in the international
dam community, for example,

00:24:02.350 --> 00:24:05.830
typically they go for
1 in 10,000-year return -- well,

00:24:05.830 --> 00:24:11.730
an annual probability of 1 in 10,000
or a return period of 10,000 years.

00:24:11.730 --> 00:24:18.169
That’s a political decision
because we can’t eliminate risk.

00:24:18.169 --> 00:24:22.540
Just to show you what happens if you
want to take the deterministic approach,

00:24:22.540 --> 00:24:25.169
I call it a scenario --
that’s what Alan Cornell called it.

00:24:25.169 --> 00:24:28.580
Let’s talk about
the worst-case scenario.

00:24:28.580 --> 00:24:33.480
We can have basically the same
worst earthquake in San Francisco

00:24:33.480 --> 00:24:38.260
as we -- as we can in Melbourne,
or Des Moines, for that matter.

00:24:38.260 --> 00:24:42.090
When we do a PSHA, we typically
allow big earthquakes to occur,

00:24:42.090 --> 00:24:47.169
and we don’t have a ready way
to limit how close they can be.

00:24:47.169 --> 00:24:52.250
So this is showing --
these are response spectra now.

00:24:52.250 --> 00:24:55.799
Probabilistic ones are colored,
and they go from -- let’s see --

00:24:55.799 --> 00:25:04.280
500 years to 50,000 years, right?
And then the median deterministic

00:25:04.280 --> 00:25:09.100
scenario is about the same
as a 500-year return period in Kaikoura.

00:25:09.100 --> 00:25:14.820
And the 84th percentile --
that’s the dashed black curve here --

00:25:14.820 --> 00:25:20.240
that’s about the same as a 5,000-year
return period in Kaikoura.

00:25:20.250 --> 00:25:22.590
If we do the
same thing in Melbourne,

00:25:22.590 --> 00:25:27.360
we can see the 500-year
return period is way down here.

00:25:27.360 --> 00:25:30.660
This gray one is
the 50,000-year return period.

00:25:30.660 --> 00:25:34.929
The median scenario is
way above all of these, right?

00:25:34.929 --> 00:25:37.350
So -- and the 84th
is even higher, right?

00:25:37.350 --> 00:25:40.000
So compared together,
you can see these

00:25:40.000 --> 00:25:42.950
deterministic scenarios
are about the same.

00:25:42.950 --> 00:25:46.970
The probabilistic hazards
are very, very different.

00:25:49.669 --> 00:25:55.980
If you look at, say, the return period of
this scenario in Kaikoura, it’s about

00:25:55.980 --> 00:26:05.760
1,000 years for the median and about
5,000 years for the 84th percentile.

00:26:05.760 --> 00:26:10.040
I extrapolate for Melbourne
the median has a return period

00:26:10.040 --> 00:26:11.950
of a quarter of a million years
in Melbourne

00:26:11.950 --> 00:26:16.510
and 2-1/2 million years
for the 84th percentile.

00:26:16.510 --> 00:26:21.470
So my question obviously is, why would
you design for something that unlikely?

00:26:23.060 --> 00:26:27.179
Kaikoura is sort of meaningful because you
do have big nearby earthquakes.

00:26:27.179 --> 00:26:30.750
But for Melbourne,
they’re very rare, right?

00:26:30.750 --> 00:26:36.530
And so the hazard is really coming from
more distant, smaller earthquakes.

00:26:39.960 --> 00:26:43.309
What happened was, you know,
I found engineers in Australia saying,

00:26:43.309 --> 00:26:47.870
oh, in the states what they do is they cap
the -- they cap the probabilistic hazard

00:26:47.870 --> 00:26:52.520
by saying we’ll only go up to the median
plus one standard deviation, right?

00:26:52.520 --> 00:26:56.030
So they cap the hazard.
They put an upper bound on it.

00:26:56.030 --> 00:26:57.390
And they said, why don’t
we do that in Australia?

00:26:57.390 --> 00:26:59.929
And I said, well, because
[chuckles] if you do that,

00:26:59.929 --> 00:27:04.510
you’ll just get blown
out of the water, right?

00:27:04.510 --> 00:27:06.850
Think about that.
If we go back here.

00:27:09.230 --> 00:27:13.870
If you say, oh, we’ll just cap the
ground motion at the median

00:27:13.870 --> 00:27:17.200
plus one standard deviation in
Melbourne, you’re at something

00:27:17.200 --> 00:27:23.590
that has a 2-1/2 million year return
period, and obviously that’s ridiculous.

00:27:23.590 --> 00:27:26.299
So I think all things
should be looked at

00:27:26.299 --> 00:27:30.700
probabilistically or with
some kind of return period in mind.

00:27:30.700 --> 00:27:35.080
Okay, back to this issue
of epistemic uncertainty.

00:27:38.010 --> 00:27:42.419
We use expert judgment
to weight the alternatives.

00:27:42.419 --> 00:27:46.600
And this gives rise to uncertainty
in the true value of the hazard.

00:27:46.600 --> 00:27:49.870
It doesn’t necessarily
make it bigger or smaller.

00:27:49.870 --> 00:27:52.789
It just is -- it makes it
more uncertain what the

00:27:52.789 --> 00:27:57.020
true value of the median --
mean or median hazard is.

00:27:57.020 --> 00:28:02.940
Aleatory random variability
does something quite different.

00:28:02.940 --> 00:28:06.010
We quantify this using data,
typically, and --

00:28:06.010 --> 00:28:09.650
like the standard deviation in
the ground motion prediction equation.

00:28:09.650 --> 00:28:12.179
And this causes the
hazard level to increase

00:28:12.179 --> 00:28:13.850
with increasing
return period, right?

00:28:13.850 --> 00:28:18.260
So this is a really nasty thing
because that’s what makes

00:28:18.260 --> 00:28:23.060
the hazard keep growing indefinitely
with decreasing probability.

00:28:25.510 --> 00:28:27.539
So the way we treat
random variability is

00:28:27.539 --> 00:28:31.710
we integrate over it in that
long hazard integral that I showed you.

00:28:31.710 --> 00:28:35.289
For example, the sigma
in a ground motion model.

00:28:35.289 --> 00:28:37.990
And it’s included in the --
in the hazard curve

00:28:37.990 --> 00:28:42.240
and in the response spectra
that we get out of that.

00:28:42.240 --> 00:28:45.530
Epistemic uncertainty
can be due to all kinds of sources.

00:28:45.530 --> 00:28:48.330
For example, in PSHA,
earthquake -- distributed

00:28:48.330 --> 00:28:51.700
earthquake source models,
we don’t know how best to do it --

00:28:51.700 --> 00:28:57.200
spatial smoothing or seismic zones
or some other scheme.

00:28:57.200 --> 00:28:59.070
Earthquake recurrence models,
characteristic or

00:28:59.070 --> 00:29:02.240
Gutenberg-Richter
or something else.

00:29:02.240 --> 00:29:08.960
How do we include active faults?
What’s the recurrence behavior of them?

00:29:08.960 --> 00:29:10.370
Ground motion
prediction models --

00:29:10.370 --> 00:29:12.720
we have got lots of
different ones these days.

00:29:12.720 --> 00:29:14.710
Site response -- how is that handled?

00:29:14.710 --> 00:29:20.590
All of these things are uncertainties
about what is the true state of nature.

00:29:20.590 --> 00:29:24.559
And so we need to give weight
to all viable alternative models.

00:29:24.559 --> 00:29:29.270
We don’t just use a preferred model.
We have to use all of them.

00:29:29.270 --> 00:29:34.600
And we obtain a hazard curve
for each logic tree branch, right?

00:29:34.600 --> 00:29:40.110
So this can provide a
huge set of different hazard curves

00:29:40.110 --> 00:29:42.809
for different versions of nature.

00:29:42.809 --> 00:29:46.179
And all this occurs
outside the hazard integral.

00:29:46.179 --> 00:29:48.830
And so what we do with all
of these logic hazard curves

00:29:48.830 --> 00:29:53.440
is we obtain the mean
and fractiles of them.

00:29:53.440 --> 00:29:55.020
We can take the arithmetic mean.

00:29:55.020 --> 00:30:01.030
We can take the median, which
separates 50% higher from 50% lower.

00:30:01.030 --> 00:30:06.480
And then we can look at
fractiles such as 85% or 95%.

00:30:06.480 --> 00:30:08.799
And here’s an
example of one.

00:30:08.799 --> 00:30:10.799
So the mean and median
are close together.

00:30:10.799 --> 00:30:15.289
The mean is usually a little higher than
the median -- that’s red versus green.

00:30:15.289 --> 00:30:20.690
And then these are
the 85th and 95th fractiles.

00:30:20.690 --> 00:30:24.820
And the whole idea of this
was again invented by people

00:30:24.820 --> 00:30:30.370
like Alan Cornell at Stanford,
Bob Kennedy -- reliability theory.

00:30:30.370 --> 00:30:32.530
And this is really
what’s behind

00:30:32.530 --> 00:30:34.600
performance-based
earthquake engineering.

00:30:34.600 --> 00:30:39.429
They use the term in nuclear power
plant engineering -- HCLPF.

00:30:39.429 --> 00:30:42.410
High confidence in a
low-probability of failure.

00:30:42.410 --> 00:30:47.820
The low probability for
nuclear plants is 10,000 years.

00:30:47.820 --> 00:30:54.690
And the high confidence is the
confidence that, at that probability level,

00:30:54.690 --> 00:31:00.630
the true value of the hazard, for example,
will not exceed the 95th fractile, right?

00:31:00.630 --> 00:31:04.799
So if you want to have
a high confidence of a low probability,

00:31:04.799 --> 00:31:07.630
then this is a 10,000-year
return period.

00:31:07.630 --> 00:31:10.049
You go -- if you don’t want to --
if you don’t need

00:31:10.049 --> 00:31:12.380
such a high confidence,
you take the median.

00:31:12.380 --> 00:31:18.530
If you want a high confidence,
you take the 95th fractile.

00:31:18.530 --> 00:31:23.900
So you convolve these hazard curves
with fragility curves, right,

00:31:23.900 --> 00:31:27.400
and so the hazard curves
have fractiles.

00:31:27.400 --> 00:31:32.230
The fragility curves, which is the
force on the structure versus

00:31:32.230 --> 00:31:37.929
the likelihood of damage --
you convolve hazard curves and

00:31:37.929 --> 00:31:43.569
fragility curves with their fractiles, and
then you get a probabilistic risk result.

00:31:47.110 --> 00:31:49.230
I think I’ve said all of this.

00:31:49.230 --> 00:31:53.370
But the fractiles show the uncertainty
in the true value of the mean hazard.

00:31:53.370 --> 00:31:55.370
It’s got nothing to do
with random variability.

00:31:55.370 --> 00:32:00.720
It’s got to do with how uncertain
you around about the mean hazard.

00:32:00.720 --> 00:32:06.130
Okay, so to wrap up this part,
I just want to talk a bit

00:32:06.130 --> 00:32:11.000
about design spectra
and time histories.

00:32:11.000 --> 00:32:17.520
It used to be that a lot of criteria
were based on deterministic criteria.

00:32:17.520 --> 00:32:22.280
And even now, dams are
treated this way in California.

00:32:22.280 --> 00:32:25.669
The notion being that you’re designing
for the largest possible ground motion.

00:32:25.669 --> 00:32:28.799
I think I’ve persuaded you
there is no such thing.

00:32:28.799 --> 00:32:35.230
So now we prefer to
use probabilistic spectra.

00:32:35.230 --> 00:32:38.490
They have a known
frequency of occurrence.

00:32:38.490 --> 00:32:41.290
And this is a
uniform hazard spectrum.

00:32:43.220 --> 00:32:46.600
In other words, each period
in the response spectrum

00:32:46.600 --> 00:32:51.140
has the same defined
annual probability of exceedance.

00:32:55.860 --> 00:33:00.659
And so if we need to represent the
response spectrum with time histories

00:33:00.659 --> 00:33:06.650
to do a dynamic analysis of a structure,
we need to de-aggregate the uniform

00:33:06.650 --> 00:33:10.690
hazard spectrum to find the relevant
magnitude, distance, and epsilon.

00:33:13.789 --> 00:33:18.840
But when we do that,
there is conservatism.

00:33:18.840 --> 00:33:24.400
And this was demonstrated by
Jack Baker, who proposed

00:33:24.400 --> 00:33:28.190
the Conditional Means Spectrum
to obtain a realistic scenario spectrum

00:33:28.190 --> 00:33:31.570
for time series scaling
or spectral matching.

00:33:31.570 --> 00:33:37.600
Jack Baker is at Stanford.
And I’m going to go into that next.

00:33:37.600 --> 00:33:43.640
The recent trends in seismic hazard
analysis are using site-specific,

00:33:43.640 --> 00:33:50.700
e.g., single-site sigma, ground
motion prediction equations.

00:33:50.700 --> 00:33:54.570
These are empirical models,
but they account for the

00:33:54.570 --> 00:34:01.289
reduced random variability when you
talk about a given source-to-site path.

00:34:01.289 --> 00:34:06.110
And then there’s also physics-based
simulations like those being done now

00:34:06.110 --> 00:34:11.629
on the SCEC Broadband Strong Ground
Motion Simulation Platform.

00:34:11.629 --> 00:34:14.729
If you really know a lot about the
fault and the kind of earthquake

00:34:14.729 --> 00:34:19.460
that might happen on it
and the geological structure

00:34:19.460 --> 00:34:26.679
on the path to the site,
then you can do simulations.

00:34:26.679 --> 00:34:30.409
But getting back to the
Conditional Means Spectrum,

00:34:30.409 --> 00:34:34.799
these slides are
from Jack Baker.

00:34:34.799 --> 00:34:42.129
Here he’s showing a lot of near-fault
ground motion response spectra.

00:34:42.129 --> 00:34:48.440
And here is a design for a building,
probably in San Francisco somewhere.

00:34:48.440 --> 00:34:54.469
It has a natural period of 1 second,
so it’s probably 10 or so stories tall.

00:34:54.469 --> 00:35:00.200
So the design condition is
actually two standard deviations

00:35:00.200 --> 00:35:03.549
above the median level for
that earthquake scenario, right?

00:35:03.549 --> 00:35:08.440
So typically, when we’re designing
in this situation, we’re designing

00:35:08.440 --> 00:35:13.450
for a response spectrum
that’s a lot larger than

00:35:13.450 --> 00:35:17.739
what we expect from the
median value for that earthquake.

00:35:17.739 --> 00:35:23.440
So Jack said, well, let’s plot all
of these recorded motions' spectra.

00:35:23.440 --> 00:35:28.559
And then he -- those are in green --
and then he picked out this black one,

00:35:28.559 --> 00:35:30.539
which actually goes
through the design condition --

00:35:30.539 --> 00:35:36.539
1 second spectral
acceleration, 1-1/4g.

00:35:36.540 --> 00:35:43.700
And he said, even if that spectrum
is that high at 1 second,

00:35:43.710 --> 00:35:46.329
the likelihood is that it
won’t be high everywhere, right?

00:35:46.329 --> 00:35:49.079
It has to be exceptionally high
to be two standard deviations

00:35:49.079 --> 00:35:51.359
above the mean
for that event, right?

00:35:51.359 --> 00:35:57.710
So true enough, at longer periods and
at shorter periods, the spectrum is less.

00:35:57.710 --> 00:36:03.900
Okay, so he said, well, let’s see.
What is the most likely response

00:36:03.900 --> 00:36:07.759
spectrum for this event,
given that this is the

00:36:07.759 --> 00:36:09.799
design condition
shown by the red dot?

00:36:09.799 --> 00:36:13.339
And so he looked at the correlation
between adjacent periods of

00:36:13.339 --> 00:36:17.249
the response spectrum, and then
they’re not perfectly correlated.

00:36:17.249 --> 00:36:20.509
So he used the calculated
correlation coefficients

00:36:20.509 --> 00:36:25.440
between adjacent periods of the
response spectra of all of these data.

00:36:25.440 --> 00:36:27.259
And this is what he got.

00:36:27.259 --> 00:36:31.890
So this black line now
is the most likely response spectrum

00:36:31.890 --> 00:36:37.559
given that it goes through
1-1/4g at 1 second, right?

00:36:37.559 --> 00:36:39.239
So that’s what the Conditional
Means Spectrum --

00:36:39.239 --> 00:36:45.219
it is the most likely spectrum
if you have this situation.

00:36:45.219 --> 00:36:49.640
All right, so the idea is that,
if you design something

00:36:49.640 --> 00:36:54.130
to match this red spectrum, it doesn’t
happen in a single earthquake.

00:36:54.130 --> 00:36:56.589
It happens collectively
through a lot of different earthquakes

00:36:56.589 --> 00:36:58.619
on different faults and so on.

00:36:58.619 --> 00:37:04.180
But it doesn’t ever happen
in a given event.

00:37:04.180 --> 00:37:08.319
So here is this thing in practice.
This is a project I did recently.

00:37:08.319 --> 00:37:13.309
Oh, and of course, you need --
you want to do this broadband, right?

00:37:13.309 --> 00:37:17.960
So what I’ve done is I’ve -- this is a --
this is one for peak acceleration.

00:37:17.960 --> 00:37:20.640
This is 1/2 a second period.

00:37:20.640 --> 00:37:23.849
This one is 1 second,
and this is 2 second, right?

00:37:23.849 --> 00:37:29.759
So the green line is the median spectrum
for the scenario being considered.

00:37:29.759 --> 00:37:31.779
The blue line is the
uniform hazard spectrum,

00:37:31.779 --> 00:37:36.140
and the red spectrum
is the uniform hazard spectrum.

00:37:36.140 --> 00:37:40.910
And so I -- then I match ground motion
time histories -- recorded ones --

00:37:40.910 --> 00:37:46.479
to these red spectra to represent
the most realistic representation

00:37:46.479 --> 00:37:50.829
of what an individual
single event would do.

00:37:50.829 --> 00:37:56.259
Okay, so now I want to
change courses a bit and talk about

00:37:56.259 --> 00:38:01.599
earthquake ground motions and the
work that I’ve been doing in that field.

00:38:05.280 --> 00:38:08.740
First, ordinary or
median ground motions are,

00:38:08.749 --> 00:38:13.469
in some cases, fairly well-recorded
and understood and modeled

00:38:13.469 --> 00:38:15.619
by ground motion
prediction equations.

00:38:15.619 --> 00:38:17.549
But sometimes they
don’t matter very much

00:38:17.549 --> 00:38:20.529
in places with good
building codes and practices.

00:38:21.940 --> 00:38:26.120
Extraordinary, or high-epsilon,
ground motions, are, in most cases,

00:38:26.130 --> 00:38:28.690
poorly recorded and understood.

00:38:28.690 --> 00:38:31.430
But they often matter a lot,
even in places with

00:38:31.430 --> 00:38:35.640
good building codes and practices,
like Christchurch.

00:38:35.640 --> 00:38:38.700
The seismologist needs to
be able to explain ground motions

00:38:38.700 --> 00:38:40.700
in order to be able to
predict them, right?

00:38:40.700 --> 00:38:45.380
So I set about trying
to understand explaining

00:38:45.380 --> 00:38:47.740
large high-epsilon
ground motions.

00:38:49.440 --> 00:38:53.499
There are two main ways
of modeling ground motions.

00:38:53.499 --> 00:38:56.880
The frequency domain approach by
Dave Boore and others -- the stochastic

00:38:56.880 --> 00:39:02.569
model -- uses a Fourier amplitude
spectrum and a response spectrum.

00:39:02.569 --> 00:39:06.829
You get large amplitudes from
high stress drop using this model.

00:39:06.829 --> 00:39:09.009
There’s a problem of
non-uniqueness.

00:39:09.009 --> 00:39:11.160
Many waveforms can
fit the same spectrum.

00:39:11.160 --> 00:39:14.219
That’s true in any situation.

00:39:14.219 --> 00:39:16.539
And if you use the
frequency domain approach,

00:39:16.539 --> 00:39:19.559
you may have multiple wave types,
like surface waves as well as

00:39:19.559 --> 00:39:23.710
body waves -- reflected waves
as well as direct body waves.

00:39:23.710 --> 00:39:27.769
So you’re looking at sort of a
composite of a lot of things

00:39:27.769 --> 00:39:31.809
in real data that
may not be in this model.

00:39:31.809 --> 00:39:37.489
The time domain model approach --
this is sort of followed by

00:39:37.489 --> 00:39:42.799
Don Helmberger and
using synthetic seismograms.

00:39:42.799 --> 00:39:46.549
By modeling the recorded waveform,
you can identify different wave types

00:39:46.549 --> 00:39:49.910
by their arrival time,
polarization,

00:39:49.910 --> 00:39:53.489
partition among vertical
and horizontal components.

00:39:53.489 --> 00:39:57.390
You can get amplitudes in
many ways that I’ll talk about next.

00:39:57.390 --> 00:40:01.079
There's a problem of non-uniqueness
in inverted kinematic source models,

00:40:01.079 --> 00:40:04.920
so we don’t really know
very well what happens on the fault,

00:40:04.920 --> 00:40:07.829
but we’ve done
our best to find out.

00:40:07.829 --> 00:40:10.700
And this approach has not been
very accessible until recently,

00:40:10.700 --> 00:40:16.390
and now everyone has access to this
ground motion simulation activity

00:40:16.390 --> 00:40:20.589
through the SCEC Broadband Strong
Ground Motion Simulation Platform,

00:40:20.589 --> 00:40:23.709
where you can log in and
think of what kind of

00:40:23.709 --> 00:40:26.079
earthquake do I
want to simulate today, right?

00:40:26.079 --> 00:40:29.380
So you can -- you can
do this stuff yourself now.

00:40:29.380 --> 00:40:32.820
You can choose various
different approaches to simulation.

00:40:34.829 --> 00:40:36.709
So let’s see.

00:40:39.239 --> 00:40:45.309
Conditions giving rise
to high epsilon values.

00:40:45.309 --> 00:40:48.789
Some are predictable from
fault and station geometry.

00:40:48.789 --> 00:40:53.630
One that I’ll talk about is Moho
reflections at large distances.

00:40:53.630 --> 00:41:00.809
Near-fault rupture directivity effects,
which are effectively a sonic boom.

00:41:00.809 --> 00:41:03.459
And if we know
the underground structure,

00:41:03.459 --> 00:41:09.739
basin resonance effects
due to long-period surface waves.

00:41:12.500 --> 00:41:16.700
Okay, so in ground motion prediction
models, the empirical approach is to

00:41:16.700 --> 00:41:20.839
represent the magnitude --
the earthquake source’s magnitude

00:41:20.839 --> 00:41:24.950
and may some other attributes,
like whether it’s strike-slip or dip-slip

00:41:24.950 --> 00:41:28.900
or so forth -- how deep it is
and stuff like that.

00:41:28.900 --> 00:41:33.369
The seismic propagation path is
represented by a scale or number,

00:41:33.369 --> 00:41:35.940
which is the closest distance.

00:41:35.940 --> 00:41:42.309
And the site is represented by, say,
a site category or, more recently, VS30,

00:41:42.309 --> 00:41:47.599
which is one of the
great things that Bill Joyner did.

00:41:47.599 --> 00:41:50.640
If we’re doing simulation,
then we model the earthquake

00:41:50.640 --> 00:41:57.029
as a shear dislocation.
We regard the crust as a wave guide,

00:41:57.029 --> 00:41:59.979
using -- and we use --
we calculate Green’s functions.

00:41:59.979 --> 00:42:05.670
And we may do some complicated
modeling of basin response.

00:42:05.670 --> 00:42:10.220
Speaking of Green’s functions,
George Green --

00:42:10.220 --> 00:42:15.180
son of a Nottingham baker and miller.
This is the mill.

00:42:15.190 --> 00:42:21.930
His formal education lasted one year,
ending in 1802 at the age of 9.

00:42:21.930 --> 00:42:28.579
I think he exhausted his teacher’s
knowledge of the material world.

00:42:28.579 --> 00:42:33.059
He formulated Green’s theorem,
Which is the basis of Green’s functions

00:42:33.059 --> 00:42:37.640
used throughout physics to describe
actions of forces from distributed sources.

00:42:37.640 --> 00:42:42.410
There’s his paper.
He was a modest fellow.

00:42:42.410 --> 00:42:46.539
Didn’t have a lot of mates
at the Royal Society,

00:42:46.539 --> 00:42:51.660
and so he [chuckles] --
it was a self-published paper.

00:42:51.660 --> 00:42:55.849
Which was bad because people in
Europe spent the next 50 years

00:42:55.849 --> 00:43:03.190
chasing after copies of this thing, up to
and including [inaudible].

00:43:03.190 --> 00:43:09.859
And of course, his -- the Green’s
functions are the basis of elastodynamic

00:43:09.859 --> 00:43:15.059
representation theorem that we use for
ground motion simulation. [coughs]

00:43:15.059 --> 00:43:17.910
And it says ground motion can be
calculated from the convolution

00:43:17.910 --> 00:43:22.799
of the slip velocity time function on the
fault with the Green’s function for the

00:43:22.799 --> 00:43:27.739
appropriate distance and depth,
integrated over the fault rupture surface.

00:43:30.529 --> 00:43:35.049
Green was also the first
to explain total internal reflection,

00:43:35.049 --> 00:43:40.209
which controls Moho reflections and
basin waves, which I will also describe.

00:43:42.880 --> 00:43:45.249
Later on, you can see he --
people found him,

00:43:45.249 --> 00:43:50.700
and so he published this one
in the Cambridge Phil mag.

00:43:50.700 --> 00:43:53.180
Okay, so Moho Bounce.

00:43:53.180 --> 00:43:59.039
This is a rather obvious and easy one,
but here’s the source of the earthquake.

00:43:59.039 --> 00:44:00.410
The direct wave is like this.

00:44:00.410 --> 00:44:04.019
The downgoing waves go down,
and then they’re reflected, and then,

00:44:04.019 --> 00:44:09.190
beyond a critical angle, there’s
total reflection back to the surface.

00:44:09.190 --> 00:44:12.589
It seems like this happened in the
Loma Prieta earthquake, for example.

00:44:12.589 --> 00:44:13.969
This is a ShakeMap.

00:44:13.969 --> 00:44:17.450
And as you know,
there was damage in San Francisco,

00:44:17.450 --> 00:44:20.319
in Oakland,
about 100 kilometers away.

00:44:20.319 --> 00:44:25.579
If you look at the recorded motions,
starting near the source

00:44:25.579 --> 00:44:33.509
and going further away,
these are the expected arrival times

00:44:33.509 --> 00:44:37.999
of direct S, Conrad,
and Moho reflections.

00:44:37.999 --> 00:44:40.369
Direct S isn’t very strong here,
but you can see,

00:44:40.369 --> 00:44:45.779
as you go further away, these -- the
Moho, in particular, looks pretty strong.

00:44:45.779 --> 00:44:49.190
And if you do a synthetic seismogram,
you can see the direct wave is

00:44:49.190 --> 00:44:54.539
strong at first, but then it peters out,
and then pretty soon, at 50 kilometers,

00:44:54.539 --> 00:44:59.069
the strongest wave
is Moho reflection.

00:44:59.069 --> 00:45:02.319
And so if you just plot that --
you know, this is a

00:45:02.319 --> 00:45:05.479
attenuation of the
direct wave in green.

00:45:05.479 --> 00:45:10.930
And the reflected wave, or downgoing
waves, reflected back in blue.

00:45:10.930 --> 00:45:14.920
You can see this might
make a kind of a flattening shape

00:45:14.920 --> 00:45:24.380
to the peak acceleration decay.
And that’s what was observed.

00:45:24.380 --> 00:45:27.200
Next interesting thing --
the rupture directivity effect,

00:45:27.200 --> 00:45:30.059
which is a seismic boom.

00:45:30.059 --> 00:45:35.119
This is David Wald’s work
for the Landers earthquake,

00:45:35.119 --> 00:45:39.339
which started down at the south end
of the fault where this star is,

00:45:39.339 --> 00:45:42.499
and then it propagated
unilaterally to the north.

00:45:42.499 --> 00:45:45.469
You can see the waveforms
are all plotted on the same scale.

00:45:45.469 --> 00:45:53.809
This is Lucerne, Yermo, and Goldstone.
You can see these huge pulses of energy

00:45:53.809 --> 00:45:57.739
going north, where not much
happened in other directions.

00:45:57.739 --> 00:46:03.029
This is the slip map that
he derived from these recordings

00:46:03.029 --> 00:46:07.619
by fitting synthetic seismograms
to the recorded motions.

00:46:07.619 --> 00:46:11.089
You can see he can explain the
amplitudes and the waveforms

00:46:11.089 --> 00:46:17.309
of all of these recordings, including
these huge ones up to the north.

00:46:20.749 --> 00:46:26.239
This was a very severe thing
in the Kobe earthquakes

00:46:26.239 --> 00:46:34.849
because rupture propagated from
here into Kobe and adjacent cities.

00:46:34.849 --> 00:46:40.499
And it produced this peak velocity,
about a meter per second peak velocity.

00:46:40.499 --> 00:46:46.349
There were two pulses about a meter per
second on the fault normal component.

00:46:46.349 --> 00:46:49.259
Just ordinary, half that level,
on the fault parallel component.

00:46:49.259 --> 00:46:56.890
You can see how much bigger this fault
normal component is than a standard

00:46:56.890 --> 00:47:03.910
building code spectrum, whereas
the fault parallel is fairly ordinary.

00:47:03.910 --> 00:47:07.809
This is -- yeah, when you --
if you look at ground motions recorded,

00:47:07.809 --> 00:47:10.009
say, on cameras, they typically
look jittery, right?

00:47:10.009 --> 00:47:12.829
They just, you know, sort of
fluttering kind of motion.

00:47:12.829 --> 00:47:14.589
You don’t see
any big displacement.

00:47:14.589 --> 00:47:18.499
Okay, but this is what it
looked like in the NHK building.

00:47:18.499 --> 00:47:22.430
A lot of the furniture
is on coasters and rollers.

00:47:22.430 --> 00:47:24.599
One pulse.

00:47:24.599 --> 00:47:26.219
Second pulse.

00:47:26.219 --> 00:47:27.920
Huge -- you know, you can
imagine a meter or so

00:47:27.920 --> 00:47:32.119
of displacement of the ground.
And lights went out.

00:47:32.119 --> 00:47:40.910
I happened to be in Osaka at the time
because EERI and Japanese colleagues

00:47:40.910 --> 00:47:46.920
were holding the first U.S.-Japan
workshop on urban earthquake hazard

00:47:46.920 --> 00:47:51.099
mitigation on the first anniversary
of the Northridge earthquake.

00:47:51.099 --> 00:47:53.719
[chuckles] And so these
earthquakes occurred

00:47:53.719 --> 00:47:56.349
exactly one day --
one year apart.

00:47:56.349 --> 00:48:00.509
In fact, I was here
for the Loma Prieta earthquake.

00:48:00.509 --> 00:48:02.299
I was up in the
PG&E building

00:48:02.299 --> 00:48:06.079
in San Francisco on --
for the Loma Prieta earthquake.

00:48:06.079 --> 00:48:10.680
And so for a while, I thought I was
getting chased by these things.

00:48:10.680 --> 00:48:15.089
These are pictures I took that --
the next day.

00:48:15.089 --> 00:48:25.609
This is just an overpass transition
from steel to concrete superstructure.

00:48:25.609 --> 00:48:28.140
Buildings with missing floors.

00:48:28.140 --> 00:48:31.559
Buildings fallen over.

00:48:31.559 --> 00:48:33.509
Collapses blocking streets.

00:48:33.509 --> 00:48:36.829
So it was a pretty bad mess.

00:48:36.829 --> 00:48:42.650
And a lot of the damage was confined
to this very narrow band offset about a

00:48:42.650 --> 00:48:49.390
kilometer from the subsurface rupture.
So the fault ran through here.

00:48:49.390 --> 00:48:54.940
And I’ll explain why this
happened in a moment.

00:48:54.940 --> 00:49:01.509
But look at how well the Japanese had
done in improving their building codes.

00:49:01.509 --> 00:49:06.190
This column is up to --
these are the damage statistics

00:49:06.190 --> 00:49:11.109
in the Kobe earthquake
for different ages of buildings.

00:49:11.109 --> 00:49:13.459
These ones are older than ’71.

00:49:13.459 --> 00:49:18.789
These are ’72 to ’81.
And these are younger than ’82.

00:49:18.789 --> 00:49:23.339
And they’re broken up this way
because a building code improvement

00:49:23.339 --> 00:49:28.749
occurred here in ’71
and another one in ’82.

00:49:28.749 --> 00:49:34.519
And so these damage states are --
collapse or severe damage is red.

00:49:34.519 --> 00:49:37.299
Moderate damage is yellow
Minor damage is green.

00:49:37.299 --> 00:49:39.699
And then slight damage is blue.

00:49:39.699 --> 00:49:43.259
You can see how dramatically
the performance of these buildings

00:49:43.259 --> 00:49:45.839
improved, especially for steel.

00:49:45.839 --> 00:49:51.079
You can see steel buildings became very
popular, and they performed really well.

00:49:54.440 --> 00:50:00.280
Okay, next topic I want to talk about
is trapping of waves in basins.

00:50:00.289 --> 00:50:03.349
So this is all stolen
from Rob Graves.

00:50:05.120 --> 00:50:11.519
Here’s the case usually
analyzed by simple programs

00:50:11.519 --> 00:50:13.059
like SHAKE or
something like that.

00:50:13.059 --> 00:50:15.979
There’s a flat layer
of soil over bedrock.

00:50:15.979 --> 00:50:20.819
The wave gets in.
If it can get in, then it can also get out.

00:50:20.819 --> 00:50:24.680
And it can go to China or somewhere,
harmless -- for you, right?

00:50:24.680 --> 00:50:29.039
And so this results in a --
kind of reverberations.

00:50:29.039 --> 00:50:35.039
But there’s no really big
wave following the direct wave.

00:50:35.039 --> 00:50:41.679
If, on the other hand, the wave enters
through the edge of a basin, it can enter --

00:50:41.680 --> 00:50:46.969
get in fine because it’s at nearly normal
incidence, as it might have been here.

00:50:46.969 --> 00:50:52.170
But then you can have a post-critical
angle at the bottom of the basin.

00:50:52.170 --> 00:50:56.259
And so although the wave can get in
perfectly well, it can’t get out, right?

00:50:56.259 --> 00:51:00.699
And so it’s trapped
until it finds the thinning edge

00:51:00.699 --> 00:51:04.459
of the basin,
and then it’ll escape.

00:51:04.460 --> 00:51:08.400
Meanwhile, you can see
the direct wave is this one here.

00:51:08.410 --> 00:51:11.009
This is the basin wave here, right?

00:51:11.009 --> 00:51:15.929
And so it can be as bad as
or worse than the direct wave.

00:51:18.480 --> 00:51:20.099
Here’s sort of an example.

00:51:20.099 --> 00:51:23.729
This would say the Whittier Narrows
earthquake occurred on the Puente Hills

00:51:23.729 --> 00:51:28.780
Blind Thrust, what would happen
if we had a bigger earthquake?

00:51:28.780 --> 00:51:32.599
Don’t want to think about that
because you have rupture directivity.

00:51:32.599 --> 00:51:36.469
This is a basin structure here.

00:51:36.469 --> 00:51:43.390
Downtown L.A. is sitting on
the hanging wall of this thing.

00:51:43.390 --> 00:51:46.709
Not good.
So Rob did some simulations.

00:51:46.709 --> 00:51:55.509
So this is now the buried
top of the Puente Hills Blind Thrust,

00:51:55.509 --> 00:51:57.859
about 5 kilometers’ depth.

00:51:57.859 --> 00:52:00.369
The earthquake -- the fault
dips down to the northeast.

00:52:00.369 --> 00:52:06.029
The little blue star
is the hypocenter.

00:52:06.029 --> 00:52:14.940
And so he’s going to
do a 3D simulation of this event.

00:52:14.940 --> 00:52:21.180
Downtown L.A. is just on top of here --
on top of the top edge of the fault.

00:52:21.180 --> 00:52:24.239
And Long Beach
is down here.

00:52:25.460 --> 00:52:29.339
Okay, and so this is a waveform at
downtown with a directivity pulse,

00:52:29.339 --> 00:52:35.390
and then you get this very long basin
wave down in Long Beach.

00:52:35.390 --> 00:52:39.499
These are snapshots at 6 seconds,
when you can see the directivity pulse.

00:52:39.499 --> 00:52:44.150
It’ll be in the fault normal component,
or north-south.

00:52:44.150 --> 00:52:48.650
And then, after 15 seconds,
these are the basin waves.

00:52:48.650 --> 00:52:50.849
And 24 seconds,
basin waves, right?

00:52:50.849 --> 00:52:55.089
If we use the GMPE
without making any adjustments,

00:52:55.089 --> 00:53:00.539
we should have a bullseye centered
on the top corner of this fault.

00:53:00.539 --> 00:53:04.319
These snapshots don’t look
remotely like bullseyes.

00:53:04.319 --> 00:53:08.569
The pink shaded
areas are bedrock.

00:53:08.569 --> 00:53:10.809
But you can see
the ground motions -- this is at,

00:53:10.809 --> 00:53:17.359
I guess, long period 1 or 2 seconds
concentrated in the basin.

00:53:17.359 --> 00:53:22.539
So these are some movies of
this wave field that Rob generated.

00:53:22.539 --> 00:53:27.829
This is the fault normal
component - north-south.

00:53:27.829 --> 00:53:31.180
That was the directivity pulse.
Now you can see the basin waves.

00:53:31.180 --> 00:53:35.199
They just go on and on forever.

00:53:35.199 --> 00:53:37.440
This is 25 seconds.

00:53:37.440 --> 00:53:39.269
They just keep going.

00:53:39.269 --> 00:53:41.789
And in cross-section,
you can see, at the bottom,

00:53:41.789 --> 00:53:43.910
the north -- this is the
fault normal component.

00:53:43.910 --> 00:53:47.539
This is the directivity pulse
getting trapped in the basin.

00:53:47.539 --> 00:53:50.339
The pink is the rock, right,
so you can see nearly all

00:53:50.339 --> 00:53:55.039
of the shaking in this movie
is within the basin.

00:53:55.039 --> 00:54:02.319
And when it escapes, it just escapes
out the side here and goes away.

00:54:02.319 --> 00:54:06.569
Not content with that,
since -- actually, recently,

00:54:06.569 --> 00:54:12.719
he’s been doing
CyberShake at SCEC.

00:54:12.719 --> 00:54:14.900
So what he does here
is he says, well,

00:54:14.900 --> 00:54:21.279
let’s throw out ground motion prediction
equations, and let’s simulate the ground

00:54:21.279 --> 00:54:25.430
motions from all of the scenario
earthquakes in the -- in the UCERF.

00:54:25.430 --> 00:54:32.170
Okay, so he’s got 12,700 fault sources
in southern California from this --

00:54:32.170 --> 00:54:35.869
this is actually an old one --
2002 earthquake source model.

00:54:35.869 --> 00:54:40.269
He has multiple hypocenters and
slip models for each of these scenarios,

00:54:40.269 --> 00:54:44.680
so he has 100,000
ground motion simulations

00:54:44.680 --> 00:54:46.699
for each of these sites
that he’s studying, right?

00:54:46.699 --> 00:54:51.049
So he’s going to do PSHA
by using the simulated ground motions

00:54:51.049 --> 00:54:56.959
to represent all of the
probabilistic hazard events.

00:54:56.959 --> 00:54:59.359
And when he does that --
this is Whittier Narrows --

00:54:59.359 --> 00:55:06.180
you can see this is what he got --
the red curve, this is a hazard curve --

00:55:06.180 --> 00:55:11.269
3-second spectral acceleration versus
return period or annual recurrence.

00:55:11.269 --> 00:55:14.619
This is what you get
with an Abrahamson-Silva

00:55:14.619 --> 00:55:18.180
ordinary model
without any basin effects.

00:55:18.180 --> 00:55:21.359
And then CyberShake,
when you do the simulations,

00:55:21.359 --> 00:55:23.989
you capture the directivity
and the basin effects, right?

00:55:23.989 --> 00:55:28.689
So you can see that’s why
you get these different results.

00:55:34.219 --> 00:55:38.660
And finally, I want to talk about
the basin edge effect.

00:55:38.660 --> 00:55:41.709
This is fairly easy to understand.

00:55:41.709 --> 00:55:46.999
You have a direct wave coming into
a site through the front door.

00:55:46.999 --> 00:55:50.589
But if there’s a basin edge,
then you can have another wave

00:55:50.589 --> 00:55:53.299
arriving at the same time
through the side door, right?

00:55:53.299 --> 00:55:58.019
So it’s just simply superposition
of waves that are able to arrive

00:55:58.019 --> 00:56:04.400
at the same time because of
this messy geological structure.

00:56:04.400 --> 00:56:11.479
Rob noticed that -- this is the Northridge
earthquake, which occurred way up

00:56:11.479 --> 00:56:16.459
north here in the San Fernando Valley
north of the Santa Monica mountains.

00:56:16.459 --> 00:56:23.130
This is the Santa Monica fault,
which, in cross-section, looks like this.

00:56:23.130 --> 00:56:29.769
So as you go from north to south,
away from the earthquake source,

00:56:29.769 --> 00:56:34.309
you are stepping onto --
well, there’s still soil there,

00:56:34.309 --> 00:56:39.420
but if there wasn’t, you’d be
dropping 3 kilometers to bedrock, right?

00:56:39.420 --> 00:56:45.930
So -- and you can see the red dots
are red tag buildings indicating

00:56:45.930 --> 00:56:49.739
that shaking more severe south of
the fault than to the north, right?

00:56:49.739 --> 00:56:54.029
So over here, you get -- you get
a little bit of basin wave developing --

00:56:54.029 --> 00:56:55.910
you can see in
these recordings here.

00:56:55.910 --> 00:56:58.699
And then when you get
to Santa Monica City Hall,

00:56:58.699 --> 00:57:02.079
which is right on the edge here,
you get this huge pulse.

00:57:02.079 --> 00:57:07.569
And the simulations he did
show the same kind of thing.

00:57:07.569 --> 00:57:11.680
This is kind of bedrock conditions,
and then this is the evolution

00:57:11.680 --> 00:57:17.430
of the wave field as you
go across this basin edge.

00:57:17.430 --> 00:57:20.400
We think the same thing
happened in the Kobe earthquake.

00:57:20.400 --> 00:57:25.959
So this black line is showing the fault --
the subsurface fault that ruptured.

00:57:25.959 --> 00:57:30.430
The purple band is this zone
of very intense damage.

00:57:30.430 --> 00:57:36.410
I guess the Japanese had abandoned
intensity -- JMA intensity 7.

00:57:36.410 --> 00:57:38.689
They thought, we won’t
have that anymore in Japan.

00:57:38.689 --> 00:57:41.199
Our building codes
are too good.

00:57:41.199 --> 00:57:48.289
But after the Kobe earthquake,
they re-instated intensity 7.

00:57:48.289 --> 00:57:51.439
So you can see this band
is offset from the fault.

00:57:51.439 --> 00:57:53.849
And in this case --
it’s a bit hard to see --

00:57:53.849 --> 00:57:58.019
but the fault is creating
the basin edge in this case.

00:57:58.019 --> 00:57:59.749
The earthquake occurred
on this fault.

00:57:59.749 --> 00:58:02.069
Here’s the basin.

00:58:02.069 --> 00:58:06.729
And here’s a wave field on the right
without any basin structure --

00:58:06.729 --> 00:58:08.749
calculated by Arben Pitarka.

00:58:08.749 --> 00:58:13.199
And on the left, we’ve got the
basin structure, and you can see

00:58:13.199 --> 00:58:19.369
the very complicated wave field
generated by this basin edge.

00:58:19.369 --> 00:58:23.239
So I began talking about the
Christchurch earthquake.

00:58:23.239 --> 00:58:27.670
The situation where, within
10 kilometers, there were these

00:58:27.670 --> 00:58:35.719
remarkably large ground motions.
And further away, not so large.

00:58:35.719 --> 00:58:41.949
So people like Brendon Bradley
have looked at things like ruptured

00:58:41.949 --> 00:58:48.529
directivity effects, basin resonance
effects, soil amplification effects.

00:58:48.529 --> 00:58:50.979
There might been some of --
all of these things.

00:58:50.979 --> 00:58:56.420
Certainly directivity was there
in the Darfield earthquake,

00:58:56.420 --> 00:58:59.949
which was the main shock --
not so damaging, but you can see --

00:58:59.949 --> 00:59:06.019
this is a directivity pulse from
that earthquake on the

00:59:06.019 --> 00:59:09.229
fault normal component --
north-south and not anywhere

00:59:09.229 --> 00:59:12.930
on the [chuckles] --
the east-west component.

00:59:12.930 --> 00:59:17.420
This is the Christchurch earthquake now --
the aftershock that did all the damage.

00:59:17.420 --> 00:59:19.819
And this is a recording on rock.

00:59:19.819 --> 00:59:23.449
So this is the Lyttelton volcano
recording on rock.

00:59:23.449 --> 00:59:26.170
This is -- looks like the
directivity pulse here.

00:59:26.170 --> 00:59:30.479
And then, on theses sediments --
this is an estuary --

00:59:30.479 --> 00:59:33.309
this is a recording
on this Quaternary basin.

00:59:33.309 --> 00:59:39.339
You can see the directivity pulse and
then the basin waves that it generated.

00:59:41.960 --> 00:59:44.980
So it’s difficult to rule out
a magnitude 6.2 earthquake

00:59:44.989 --> 00:59:47.420
almost anywhere
in the world.

00:59:47.420 --> 00:59:50.099
For one thing, they typically
don’t break the ground surface.

00:59:50.099 --> 00:59:53.999
So we need to know what conditions
caused such high ground motions in

00:59:53.999 --> 01:00:01.269
Christchurch and then assess
where else such conditions may exist.

01:00:01.269 --> 01:00:08.180
So I spent quite a bit of time in
Christchurch in the last few years.

01:00:08.180 --> 01:00:11.059
Using earthquake forecast models,
Gerstenberger et al. estimated

01:00:11.059 --> 01:00:16.509
the seismicity rate in the region
has increased about 20-fold.

01:00:16.509 --> 01:00:20.799
This is actually the David Rhoades
EEPAS earthquake forecast --

01:00:20.799 --> 01:00:24.789
Every Earthquake a Precursor
According to Scale.

01:00:24.789 --> 01:00:29.569
So their 500-year peak acceleration
for deep soil in Christchurch has

01:00:29.569 --> 01:00:36.619
now gone from 0.22g to 0.6g, whereas --
and that’s higher than Wellington.

01:00:36.619 --> 01:00:40.759
Wellington, if you recall, is sitting on
a subduction zone with the Wellington

01:00:40.759 --> 01:00:48.449
Fault running right through it.
So that’s pretty amazing.

01:00:48.449 --> 01:00:53.429
The building code spectrum
only increased from 0.2 to 0.3g.

01:00:56.519 --> 01:00:59.420
And all of this impacts
assessment of damage

01:00:59.420 --> 01:01:03.680
like residual capacity of buildings
that are still standing.

01:01:03.680 --> 01:01:09.900
To meet requirements for safety tagging,
code-mandated repair,

01:01:09.900 --> 01:01:14.109
insurance loss assessment,
decision to repair or demolish --

01:01:14.109 --> 01:01:21.099
I mean, knowing how to estimate
residual capacity is a huge issue.

01:01:21.099 --> 01:01:25.140
People are even trying to apportion
damage among these different events

01:01:25.140 --> 01:01:27.650
[chuckles] to the same building
in Christchurch.

01:01:27.650 --> 01:01:31.030
That is a -- that is a
grand challenge.

01:01:34.499 --> 01:01:38.539
And to close, if you want to learn more
about the Christchurch earthquake

01:01:38.539 --> 01:01:43.009
and get closer to Christchurch,
then I suggest you

01:01:43.009 --> 01:01:48.099
come to Sydney on
November 6th to the 8th.

01:01:48.099 --> 01:01:52.069
We are holding the 10th Pacific
Conference on Earthquake Engineering,

01:01:52.069 --> 01:01:55.049
sponsored by the Australian Earthquake
Engineering Society, of which

01:01:55.049 --> 01:02:01.789
I am president, and the New Zealand
Society for Earthquake Engineering.

01:02:01.789 --> 01:02:04.640
So that’s at the end of a week.
At the beginning of that week,

01:02:04.640 --> 01:02:06.589
there will be the 6th International

01:02:06.589 --> 01:02:11.099
Conference on Earthquake Geotechnical
Engineering in Christchurch.

01:02:11.099 --> 01:02:14.229
So if you want to really learn
all about this stuff,

01:02:14.229 --> 01:02:17.429
go to Christchurch,
and then come to Sydney.

01:02:17.429 --> 01:02:19.460
Thank you.

01:02:19.460 --> 01:02:23.680
[ Applause ]

01:02:26.620 --> 01:02:28.600
- Questions?

01:02:33.220 --> 01:02:36.680
- Thanks a lot, Paul.
It was a very comprehensive overview.

01:02:36.689 --> 01:02:38.689
I have a very simple question.

01:02:38.689 --> 01:02:43.349
Could you go back to the slide --
the design -- yeah, this one right here.

01:02:43.349 --> 01:02:50.469
It says the 500-year PGA for deep soil
has increased from 0.22g to 0.6g.

01:02:50.469 --> 01:02:53.339
- That’s a GNS estimate, yeah.

01:02:53.339 --> 01:02:58.079
- But you recorded much more -- much stronger
than 0.6g, so why would ...

01:02:58.079 --> 01:03:00.319
- Right.
- ... why would the design

01:03:00.319 --> 01:03:03.930
be so far below what’s recorded?

01:03:03.930 --> 01:03:08.069
- Because the -- that earthquake
was a really diabolical earthquake.

01:03:08.069 --> 01:03:12.499
There were things like directivity --
you know, the basin effects -- all of

01:03:12.499 --> 01:03:18.499
those things combined to make
really high ground motions locally.

01:03:20.820 --> 01:03:22.040
- And?
- And ...

01:03:22.060 --> 01:03:24.740
- You don’t think it’ll ever happen again?
- We don’t -- we don’t design for the

01:03:24.749 --> 01:03:28.080
largest recorded ground motions.
That’s doesn’t happen. [chuckles]

01:03:28.080 --> 01:03:31.479
Right?
Buildings have a lot of capacity

01:03:31.479 --> 01:03:36.339
to handle ground motions higher
than what they’re designed for.

01:03:37.960 --> 01:03:45.320
[ Silence ]

01:03:47.760 --> 01:03:52.680
- Paul, you pointed it out the probabilistic
ground motions with CyberShake go up.

01:03:52.680 --> 01:03:57.260
With GMPEs, they’re probably
already pumped up a little bit.

01:03:57.269 --> 01:04:02.099
What’s an engineer to do who’s
dealing with a distributed system?

01:04:02.099 --> 01:04:08.059
Where are they going to pick the
ground motions and not over-design?

01:04:08.060 --> 01:04:10.820
- Yeah, well, yeah, it becomes
very complicated because then

01:04:10.829 --> 01:04:14.209
you have to think about the
correlation of the ground motions

01:04:14.209 --> 01:04:17.199
at different places that
occur in the same event, right?

01:04:17.199 --> 01:04:22.579
So that’s something that I think is,
you know, drawing increasing attention,

01:04:22.579 --> 01:04:25.339
but not a lot of work
has been done on that.

01:04:25.339 --> 01:04:32.160
I think -- let’s see. I’m trying to think.
People in Denver, I think, may -- did --

01:04:32.160 --> 01:04:34.789
Dave Perkins and -- oh, yeah.

01:04:34.789 --> 01:04:38.509
Yeah, I think some work
has been done on that, yeah.

01:04:38.509 --> 01:04:43.799
But for that, you have to look
scenario by scenario, right?

01:04:43.799 --> 01:04:47.289
It’s difficult to do
a probabilistic study,

01:04:47.289 --> 01:04:50.150
right [chuckles], of a --
of a distributed system.

01:04:50.150 --> 01:04:53.839
You have to look at scenarios
the way earthquake loss

01:04:53.839 --> 01:04:58.239
is estimated by -- you know,
by RMS and those people.

01:04:58.239 --> 01:05:00.949
- I guess what I was looking for
was some guidance on

01:05:00.949 --> 01:05:03.719
how you picked that scenario.
- Oh, how you pick the scenario?

01:05:03.720 --> 01:05:07.460
Yeah, well, then -- yeah, good.
Well, that’s a good question.

01:05:09.620 --> 01:05:14.679
Yeah. That’s a matter of looking --
trying to -- well, I think the answer is,

01:05:14.679 --> 01:05:17.880
you have to do
some samples, right? [chuckles]

01:05:17.880 --> 01:05:20.329
And find out which
events seem to control.

01:05:20.329 --> 01:05:25.069
I don’t know how to de-aggregate
when you’re doing scenario by scenario.

01:05:25.069 --> 01:05:29.189
It seems to me like
a pretty hard slog.

01:05:30.600 --> 01:05:36.640
[ Silence ]

01:05:38.360 --> 01:05:46.300
- Yeah, Paul, I’m trying to remember
about the PSAJ calculations

01:05:46.309 --> 01:05:49.969
for the Christchurch area
prior to the earthquake.

01:05:49.969 --> 01:05:52.759
- Yeah.
- And if I remember correctly,

01:05:52.759 --> 01:06:01.239
it was part of a very large,
very low-seismicity region ...

01:06:01.239 --> 01:06:07.779
- Yeah.
- ... in southeastern or eastern New --

01:06:07.780 --> 01:06:10.959
eastern part of the south island.
- Right, yeah.

01:06:10.959 --> 01:06:13.239
- Well-removed, of course,
from the Alpine Fault.

01:06:13.240 --> 01:06:15.180
- Yeah.

01:06:16.580 --> 01:06:20.380
- And maybe there were a couple
magnitude 5 earthquakes

01:06:20.390 --> 01:06:25.709
in the historical record, if I’m
remembering what Mark sent me.

01:06:25.709 --> 01:06:31.900
But what I’m trying to say is,
okay, so this earthquake

01:06:31.900 --> 01:06:34.640
occurred right
underneath Christchurch.

01:06:34.640 --> 01:06:38.480
If it had occurred 100 kilometers
to the west, no one would have

01:06:38.480 --> 01:06:40.040
given to the damn ...
- Right.

01:06:40.040 --> 01:06:42.680
- ... because it would have been underneath
a sheep farm or something.

01:06:42.680 --> 01:06:46.839
- Right.
- And part of the calculation involves a

01:06:46.839 --> 01:06:52.709
very small area, you know,
of the strong ground motion

01:06:52.709 --> 01:06:57.819
that Christchurch --
that this area actually felt

01:06:57.819 --> 01:07:03.929
divided by the area of the entire
low-seismicity zone, which is huge.

01:07:03.929 --> 01:07:09.410
And that’s why the calculation was very
low for the whole region, because ...

01:07:09.410 --> 01:07:10.400
- Yeah.
- ... who knows where that

01:07:10.400 --> 01:07:13.649
earthquake is going to occur?
- Yeah.

01:07:13.649 --> 01:07:18.929
- So do you know that that calculation is
wrong in view of just one earthquake?

01:07:18.929 --> 01:07:23.880
- Well, yeah, no, the seismic hazard
calculation allowed for magnitudes

01:07:23.880 --> 01:07:29.670
as large as 7.1, I think it was,
or 7.2 to occur anywhere in that area.

01:07:29.670 --> 01:07:33.939
So that was already
accounted for.

01:07:33.939 --> 01:07:37.599
And the Darfield earthquake,
which is a 7.1, it really

01:07:37.599 --> 01:07:39.650
didn’t cause a lot of damage
in Christchurch.

01:07:39.650 --> 01:07:46.109
It was this little 6.2 aftershock
that really did the damage.

01:07:46.109 --> 01:07:47.439
And that can happen anywhere.

01:07:47.439 --> 01:07:50.519
I mean [chuckles] --
with very low probability.

01:07:50.519 --> 01:07:54.599
I don’t know any precedent for this --
what’s happened in Christchurch.

01:07:54.599 --> 01:08:00.859
I can’t think of any precedent like this.
It was just a really, really bad, direct hit.

01:08:02.160 --> 01:08:03.170
- Yes.

01:08:03.180 --> 01:08:06.439
And I don’t think we can
do much about it. [chuckles]

01:08:06.439 --> 01:08:09.249
- Right.
- I mean, we just have to live with that.

01:08:09.249 --> 01:08:10.609
- Yeah.
- Correct?

01:08:10.609 --> 01:08:11.969
- That’s right.

01:08:11.969 --> 01:08:15.579
I think there are things that engineers
can do if they really understand

01:08:15.579 --> 01:08:19.890
how to design buildings and,
you know, put in ductility.

01:08:19.890 --> 01:08:23.690
They can -- doing that
can really go a long way.

01:08:23.690 --> 01:08:27.920
You know, avoid irregularities, really
be conscientious about ductility,

01:08:27.920 --> 01:08:30.850
and stuff like that.
They can -- they can go a long way.

01:08:30.850 --> 01:08:33.320
- Well, if they know where to do it.
- Yeah, well, but ...

01:08:33.320 --> 01:08:35.780
- But in this case, you wouldn’t.
Okay.

01:08:35.780 --> 01:08:39.930
- Well, yeah.
They need to learn to do it everywhere.

01:08:39.930 --> 01:08:42.700
- Okay. But that would
drive up the cost considerably ...

01:08:42.700 --> 01:08:44.440
- Not much.
- Not much? Okay.

01:08:44.440 --> 01:08:47.260
- Actually, yeah. Not much.

01:08:48.500 --> 01:08:53.100
- Paul, very interesting talk.
During your talk, you talked about

01:08:53.109 --> 01:08:55.750
some work you did in
western Australia --

01:08:55.750 --> 01:09:00.660
the tailing stem Uniform Hazard
Spectrum near Kalgoorlie.

01:09:00.660 --> 01:09:05.089
Now, Australia is
tectonically fairly inactive.

01:09:05.089 --> 01:09:10.880
And so I was just curious what your
seismic sources were for

01:09:10.880 --> 01:09:15.160
the calculation of your hazard --
Uniform Hazard Spectrum.

01:09:15.160 --> 01:09:23.620
- Well, there are a few mapped faults
in the neighborhood, but the hazard

01:09:23.620 --> 01:09:27.569
is really dominated by just
a distributed source which --

01:09:27.569 --> 01:09:30.770
for which -- well, we used three
different earthquake source models.

01:09:30.770 --> 01:09:34.299
One is the same approach as
Art Frankel’s approach,

01:09:34.299 --> 01:09:37.830
so spatial smoothing.
That’s my model for Australia.

01:09:37.830 --> 01:09:42.200
And then there’s a person who has a --
earthquake source zones.

01:09:42.200 --> 01:09:44.280
And then GA also --
Geoscience Australia --

01:09:44.280 --> 01:09:46.920
have an earthquake
source zone model.

01:09:46.920 --> 01:09:51.190
But no, it’s -- even though there’s
quite a few mapped faults out there,

01:09:51.190 --> 01:09:57.790
their recurrence intervals are so long
that they don’t have much of an impact.

01:09:57.790 --> 01:10:00.700
- Okay, well, thanks.
- Yeah.

01:10:01.600 --> 01:10:08.600
[ Silence ]

01:10:09.840 --> 01:10:12.400
- Just a comment.
I don’t get the chance to disagree

01:10:12.400 --> 01:10:16.670
with Art McGarr very often,
but I think, Art, that Australia is

01:10:16.670 --> 01:10:20.170
the most seismically active continent --
you know, interplayed seismicity,

01:10:20.170 --> 01:10:22.450
Australia is the highest in the world.

01:10:22.450 --> 01:10:25.740
Wouldn’t you agree that it’s ...
- Yeah, it is.

01:10:25.740 --> 01:10:28.920
And it makes it very interesting.

01:10:28.920 --> 01:10:33.710
In fact, there’s a geologist there
named Dan Clark who has

01:10:33.710 --> 01:10:38.290
actually visited Golden, I think.
He’s a really sharp guy [chuckles],

01:10:38.290 --> 01:10:42.250
and he’s sort of almost single-handedly
doing this really great work.

01:10:42.250 --> 01:10:47.510
But he’s got the advantage of lots of
surface traces and lots of earthquakes.

01:10:47.510 --> 01:10:51.390
There have been about
eight surface-faulting earthquakes

01:10:51.390 --> 01:10:54.250
in Australia in the last hundred years,
which is really amazing.

01:10:54.250 --> 01:10:59.000
The Meckering one was about a
meter high for 37 kilometers --

01:10:59.000 --> 01:11:02.690
I mean, magnitude 6.8 --
really amazing stuff.

01:11:02.690 --> 01:11:05.450
And that happens because, you know,
the craton is so strong that

01:11:05.450 --> 01:11:12.350
it’s brittle all the way up to the top.
So, yeah, really fascinating.

01:11:15.760 --> 01:11:18.820
- There’s only been
one fatality in Australia

01:11:18.830 --> 01:11:21.580
in history from
earthquakes, however.

01:11:21.580 --> 01:11:24.750
- No, I think it’s more like
18 or something in the

01:11:24.750 --> 01:11:29.190
Newcastle earthquake in 1989.
- Okay, well, that was the only

01:11:29.190 --> 01:11:31.150
earthquake to cause fatality ...
- I think ...

01:11:31.150 --> 01:11:36.980
- ... and it was only a 5.6 or so.
- Yeah, that’s right. Yeah.

01:11:36.980 --> 01:11:39.650
- Okay.
- But it could be really bad, you know

01:11:39.650 --> 01:11:41.800
if we -- if we had a bigger one.

01:11:41.800 --> 01:11:49.040
For example, Swiss Re sells
cat bonds for Sydney earthquake

01:11:49.040 --> 01:11:52.960
and Queensland tropical cyclone.
They’re so scared.

01:11:52.960 --> 01:11:54.240
Because everybody’s insured, right?

01:11:54.240 --> 01:11:57.250
Everybody in Sydney
is covered for earthquake.

01:11:57.250 --> 01:12:01.640
If they did have an earthquake, like,
even a 6, I mean, they would really --

01:12:01.640 --> 01:12:07.320
the insurer would really
be in a lot of trouble.

01:12:07.320 --> 01:12:08.540
And even the re-insurer, right?

01:12:08.540 --> 01:12:10.830
So we’re talking
about the third level, right?

01:12:10.830 --> 01:12:16.000
The insurer sells a lot of its debt --
its risk to the re-insurer.

01:12:16.000 --> 01:12:21.730
And then the re-insurer offloads to
Warren Buffett or somebody like that.

01:12:24.220 --> 01:12:26.260
- Any more questions?

01:12:29.360 --> 01:12:33.080
Okay, then let’s meet downstairs
in about five minutes

01:12:33.080 --> 01:12:35.530
and walk over to
the patio for lunch.

01:12:35.530 --> 01:12:38.740
And let’s thank our speaker again.
- Thank you.

01:12:38.740 --> 01:12:42.350
[ Applause ]

01:12:43.120 --> 01:12:59.500
[ Silence ]