WEBVTT
Kind: captions
Language: en-US

00:00:00.860 --> 00:00:03.240
[Silence]

00:00:03.240 --> 00:00:05.900
Okay.
Welcome, everybody.

00:00:05.900 --> 00:00:11.420
Good morning, and thank you for
coming to today’s ESC seminar.

00:00:11.420 --> 00:00:14.969
A quick reminder to
turn off your cameras

00:00:14.969 --> 00:00:19.900
and put your audio
on mute during the talk.

00:00:21.020 --> 00:00:26.260
Next week, our speaker will be
Lucille Bruhat from ENS in Paris.

00:00:26.260 --> 00:00:33.020
So next week will be our first European
talk, but it’s at the normal time.

00:00:33.020 --> 00:00:35.420
And that’s it for
quick announcements.

00:00:35.420 --> 00:00:41.000
So, at this point, I’ll turn it over to
Clara to introduce today’s speaker.

00:00:41.000 --> 00:00:45.260
- Okay. Hi, everyone.
I’m very happy to introduce today’s

00:00:45.260 --> 00:00:50.670
speaker, Sunyoung Park from
Caltech, who also goes by Sunny.

00:00:50.670 --> 00:00:56.190
Sunny received two bachelor’s degrees
in economics and energy resources

00:00:56.190 --> 00:01:00.170
engineering, followed by
a master’s degree in geophysics

00:01:00.170 --> 00:01:04.400
from Seoul National
University in South Korea.

00:01:04.400 --> 00:01:08.610
Sunny then earned her Ph.D.
in Earth and planetary sciences

00:01:08.610 --> 00:01:13.440
in 2018 from Harvard
University in Massachusetts.

00:01:13.440 --> 00:01:18.600
Advised by Professor Miaki Ishii,
her thesis examined the Earth’s

00:01:18.600 --> 00:01:24.500
internal processes and structure with
novel seismological approaches.

00:01:24.500 --> 00:01:28.159
Sunny is currently a
Texaco Postdoctoral Scholar

00:01:28.160 --> 00:01:31.880
at the Caltech Seismo Lab
in Pasadena, California.

00:01:31.880 --> 00:01:37.240
In June 2021, she will start
as an assistant professor in the

00:01:37.240 --> 00:01:43.000
Department of Geophysical Sciences
at the University of Chicago.

00:01:43.000 --> 00:01:48.510
I first met Sunny at the 2017
SSA meeting in Denver.

00:01:48.510 --> 00:01:50.940
And then, a year later,
at SSA in Miami, I learned she

00:01:50.940 --> 00:01:56.760
was moving to Caltech across the street
from my office in USGS Pasadena.

00:01:56.760 --> 00:02:00.880
I knew that it was for real when
I randomly encountered her on the

00:02:00.890 --> 00:02:05.480
street as we were each independently
looking for an apartment in Pasadena.

00:02:05.480 --> 00:02:08.729
So that was a very
happy moment for both of us.

00:02:08.729 --> 00:02:12.840
Sunny’s research interests are
incredibly diverse, and she has

00:02:12.840 --> 00:02:17.959
applied an extremely versatile toolkit
of observational seismology techniques

00:02:17.959 --> 00:02:21.519
to understand Earth’s structure
and earthquake processes at

00:02:21.519 --> 00:02:27.010
both shallow and deep depths.
Some examples of her research include

00:02:27.010 --> 00:02:32.230
constraining near-surface wave speeds
with body wave polarization, which she

00:02:32.230 --> 00:02:37.829
discussed at her ESC seminar back
in 2017; understanding variations

00:02:37.829 --> 00:02:41.660
in the mantle transition
zone with triplication data;

00:02:41.660 --> 00:02:46.560
and analyzing earthquake rupture
directivity in three dimensions.

00:02:46.560 --> 00:02:50.660
But Sunny is also very comfortable
working with geodetic data.

00:02:50.660 --> 00:02:55.420
Today, she will tell us how she used
GPS data to understand postseismic

00:02:55.420 --> 00:03:00.180
deformation after deep earthquakes.
Sunny, take it away.

00:03:01.180 --> 00:03:05.799
- Thank you very much,
Clara, for that introduction.

00:03:06.760 --> 00:03:09.660
Let me share the screen.

00:03:12.860 --> 00:03:14.240
Okay.

00:03:16.500 --> 00:03:19.400
[Silence]

00:03:19.400 --> 00:03:21.640
Do you see my screen?
- Yes.

00:03:21.640 --> 00:03:23.660
- Mm-hmm. Great.
Yeah.

00:03:23.660 --> 00:03:26.480
Yeah, so thank you for
inviting me again here.

00:03:26.480 --> 00:03:31.010
I cannot believe it’s been
already three years since I gave

00:03:31.010 --> 00:03:36.709
the other talk in the seminar series.
So today I would like to talk about

00:03:36.709 --> 00:03:41.279
something different, so that’s
about postseismic deformation

00:03:41.280 --> 00:03:43.989
following deep earthquakes.

00:03:45.560 --> 00:03:49.900
So, postseismic deformation is
important in studying earthquake

00:03:49.900 --> 00:03:55.360
physics since it may capture slow
processes following main shock.

00:03:55.360 --> 00:03:58.900
And it’s also important in
studying the rheological structure

00:03:58.909 --> 00:04:03.739
since its time evolution of
the deformation pattern

00:04:03.739 --> 00:04:08.510
essentially provides
ideas about the rheology.

00:04:08.510 --> 00:04:12.309
But the emphasis has been
on shallow earthquakes, and so

00:04:12.309 --> 00:04:15.790
we are mostly familiar with
the postseismic deformation

00:04:15.790 --> 00:04:21.300
following shallow earthquake that
are usually in the crust, like here.

00:04:23.140 --> 00:04:28.270
So usually, the time evolution of
displacement on the Y axis looks like

00:04:28.270 --> 00:04:33.669
this in time, on the horizontal axis.
So, at the time of the earthquake,

00:04:33.669 --> 00:04:38.300
you have this coseismic deformation
that’s instant, more or less,

00:04:38.300 --> 00:04:41.560
and then that’s followed by this
time-dependent deformation

00:04:41.560 --> 00:04:44.980
which we call
postseismic deformation.

00:04:46.100 --> 00:04:51.600
And there are major – two major
mechanisms to explain this,

00:04:51.610 --> 00:04:57.180
and first is afterslip.
That’s a slow-slipping process

00:04:57.180 --> 00:05:00.550
just following the
main shock rupture,

00:05:00.550 --> 00:05:04.620
usually in the vicinity
of the main shock rupture area.

00:05:05.500 --> 00:05:11.300
And, depending on the orientation of
the fault and your GPS station location,

00:05:11.300 --> 00:05:16.990
for example, you’ll see afterslip that’s
in the same direction of the coseismic

00:05:16.990 --> 00:05:24.080
deformation or it’s on the opposite
direction from the coseismic slip.

00:05:25.020 --> 00:05:30.340
Now, the second mechanism
that causes postseismic deformation

00:05:30.340 --> 00:05:34.060
is the viscoelastic relaxation.

00:05:34.060 --> 00:05:39.560
And, again, that would actually depend
upon the rheological structure here.

00:05:39.560 --> 00:05:45.479
And, depending on the structure and,
again, the station geometry, you’ll see

00:05:45.479 --> 00:05:50.620
either the same direction with
postseismic viscoelastic relaxation

00:05:50.620 --> 00:05:56.140
or the one that’s opposite from
the coseismic deformation.

00:05:57.120 --> 00:06:02.800
And today, I would like to say
something similar to this, but for

00:06:02.800 --> 00:06:08.860
deep earthquakes. And by deep,
I mean, like, 600 kilometer deep.

00:06:10.840 --> 00:06:16.800
Of course, this hasn’t been done
because, often, you would think that

00:06:16.800 --> 00:06:22.159
you wouldn’t see anything at the
surface from such a deep earthquake.

00:06:22.159 --> 00:06:27.460
So, actually, zeroth order question
to ask here is, do we actually see

00:06:27.460 --> 00:06:33.380
any surface deformation pattern
from these deep earthquakes?

00:06:33.380 --> 00:06:39.840
And the answer is actually yes.
So we can actually track down –

00:06:39.840 --> 00:06:42.860
there are just a handful
of studies on this.

00:06:42.860 --> 00:06:50.360
And this is actually 1994 Bolivia
earthquake that’s magnitude 8.2

00:06:50.360 --> 00:06:57.280
at about 650 kilometer deep.
And this is basically, as far as I know,

00:06:57.289 --> 00:07:01.360
the first really well-recorded
large earthquake that happened

00:07:01.360 --> 00:07:06.770
at such deep depth. And we,
at the time, had some seismometer

00:07:06.770 --> 00:07:10.460
around there, so that happened
around here. This is the epicenter.

00:07:10.460 --> 00:07:15.200
And, again – actually,
this is the epicenter here.

00:07:15.940 --> 00:07:21.140
And then these are the
broadband seismometers we had.

00:07:21.980 --> 00:07:28.000
So, unfortunately, at that time, we didn’t
have a GPS station right around there.

00:07:28.000 --> 00:07:33.500
And a group of people, for example,
Göran Ekström calculated normal

00:07:33.500 --> 00:07:39.860
mode synthetics for this, which is
like this P wave here, and S wave here,

00:07:39.860 --> 00:07:44.009
and then followed by
this permanent displacement.

00:07:44.009 --> 00:07:49.340
So that’s something that’s an expected
displacement you can expect.

00:07:49.340 --> 00:07:54.720
And, using this seismic array,
people have argued that they

00:07:54.720 --> 00:07:58.789
actually see that – for example,
this permanent displacement

00:07:58.789 --> 00:08:04.300
following this event that mimics –
that look like the synthetics.

00:08:04.940 --> 00:08:09.220
There’s some robustness question.
Because, in order to get something like

00:08:09.220 --> 00:08:14.980
displacement from seismometers, we
have to do what we call deconvolution.

00:08:14.980 --> 00:08:21.280
And you can see that not all stations
have that kind of big displacement.

00:08:21.280 --> 00:08:25.949
For example, here you see the first
and last part is almost the same.

00:08:25.949 --> 00:08:29.810
Same here. But in
some stations, you see that.

00:08:29.810 --> 00:08:34.740
So we possibly have observed this
postseismic static displacement

00:08:34.740 --> 00:08:40.460
from this deep earthquake,
and that’s the first observations.

00:08:41.180 --> 00:08:45.960
Now, interestingly, after following
this earthquake, some group of people

00:08:45.970 --> 00:08:50.880
actually thought about postseismic
deformation from these deep event.

00:08:50.880 --> 00:08:57.560
Unfortunately, we have no observation
of this, but what they did is, they just

00:08:57.560 --> 00:09:02.520
theoretically calculated what’s maybe
expected deformation – postseismic

00:09:02.520 --> 00:09:08.480
deformation following this deep event.
So that’s what I think may be first

00:09:08.480 --> 00:09:12.780
deep earthquake-related
postseismic deformation paper.

00:09:14.180 --> 00:09:17.320
Now there comes
another really big event.

00:09:17.320 --> 00:09:24.460
That’s 8.3 Sea of Okhotsk event
that happened in 2013 around here.

00:09:24.940 --> 00:09:28.560
Now we have GPS
stations around there.

00:09:28.570 --> 00:09:33.500
So this is actually first GPS
observation of such a deep event.

00:09:33.500 --> 00:09:41.030
Again, this is coseismic deformation
that you see in the arrows that are

00:09:41.030 --> 00:09:46.190
basically showing you the horizontal
coseismic displacement here.

00:09:46.190 --> 00:09:53.030
So that study is done by Steblov et al.,
and there was another study by Xu et al.,

00:09:53.030 --> 00:09:59.390
and they actually also tried to discuss
maybe hint of postseismic deformation

00:09:59.390 --> 00:10:04.300
from this, although the robust – there
is some robustness question on that.

00:10:06.580 --> 00:10:11.850
So now we can ask, can we actually
measure postseismic deformation,

00:10:11.850 --> 00:10:16.780
which is even challenging compared
to the coseismic deformation.

00:10:16.780 --> 00:10:21.920
And I’m also going to show you that
the answer is yes to this in this talk.

00:10:21.920 --> 00:10:27.000
And, given that you all probably
see this postseismic deformation,

00:10:27.010 --> 00:10:31.100
it has really important contribution
that it might be able to make.

00:10:31.100 --> 00:10:36.520
First is about afterslip. We essentially
know nothing about afterslip for

00:10:36.520 --> 00:10:40.600
deep earthquakes. Not only – we don’t
even understand why deep earthquake

00:10:40.600 --> 00:10:44.760
really – how they happen.
But, if we see something like

00:10:44.760 --> 00:10:48.790
afterslip following deep earthquake,
that would be really interesting

00:10:48.790 --> 00:10:52.180
in understanding deep
earthquake mechanisms.

00:10:53.020 --> 00:10:57.680
Secondly, if we see viscoelastic
relaxation pattern from this,

00:10:57.690 --> 00:11:03.240
that’s also really interesting.
Because, for shallow earthquakes,

00:11:03.240 --> 00:11:08.850
when you study viscoelastic relaxation,
you can get sort of shallow rheological

00:11:08.850 --> 00:11:14.450
structure. But imagine you have
earthquake at 600-kilometer depth.

00:11:14.450 --> 00:11:18.880
We can actually get to
understanding viscoelastic structure

00:11:18.880 --> 00:11:23.280
of the Earth that are
much deeper depth.

00:11:24.430 --> 00:11:29.020
So actually, even though this
mantle viscosity is really crucial

00:11:29.020 --> 00:11:32.880
in understanding, for example,
mantle convection and everything,

00:11:32.880 --> 00:11:37.380
it’s actually one of the
poorest-constrained parameter.

00:11:37.380 --> 00:11:43.430
And that has been mainly constrained by
gravity or glacial isostatic adjustment.

00:11:43.430 --> 00:11:46.060
And this will be a
completely new way

00:11:46.060 --> 00:11:51.040
to probe the rheological
structure of the mantle.

00:11:52.200 --> 00:11:56.200
So let’s first think
about afterslip case.

00:11:56.200 --> 00:12:04.140
So assume afterslip happens basically
near the source main shock rupture area.

00:12:05.140 --> 00:12:09.500
However, we are actually observing it
at the surface, which is far away

00:12:09.510 --> 00:12:13.900
from the source area.
So basically, the location of the

00:12:13.900 --> 00:12:19.120
afterslip and the main shock
is basically the same point from

00:12:19.120 --> 00:12:24.050
the surface point of view.
So what that means is that the pattern

00:12:24.050 --> 00:12:31.110
we expect from afterslip is essentially
the same as coseismic pattern.

00:12:31.110 --> 00:12:36.230
So that’s something that’s different from
shallow earthquake cases, so that means

00:12:36.230 --> 00:12:42.580
we will only see afterslip that are
inconsistent with coseismic direction.

00:12:42.580 --> 00:12:47.500
And we wouldn’t see afterslip
that looks something like this.

00:12:48.080 --> 00:12:54.640
So, in turn, that actually means that,
if we see anything – any signal that’s

00:12:54.640 --> 00:13:01.370
opposite sign from the coseismic, that
actually means we are detecting this

00:13:01.370 --> 00:13:07.870
viscoelastic relaxation because that can
only come from viscoelastic relaxation.

00:13:07.870 --> 00:13:12.060
So let me just show you
what surface deformation pattern,

00:13:12.060 --> 00:13:16.120
then, do we expect from
these deep earthquakes?

00:13:16.120 --> 00:13:21.470
In order to do that, I’m going to
use the mechanism of the event

00:13:21.470 --> 00:13:25.120
that I just shown you,
which is magnitude 8.3

00:13:25.120 --> 00:13:29.050
Okhotsk event that’s
sitting at 600 kilometer.

00:13:29.050 --> 00:13:32.040
So the mechanism actually is
normal event, but this is just

00:13:32.040 --> 00:13:36.960
a cross-section, so that’s
why it looks like strike-slip.

00:13:36.960 --> 00:13:40.100
So let’s first think about
the afterslip again,

00:13:40.100 --> 00:13:44.200
which is basically the
same pattern as coseismic.

00:13:44.200 --> 00:13:51.540
So expected pattern of coseismic
of east component here can be –

00:13:51.540 --> 00:13:55.050
the map view would look something
like this, where the epicenter is

00:13:55.050 --> 00:14:03.220
at the middle, and the east is the red
color, and west is the blue color here.

00:14:03.220 --> 00:14:06.840
One thing that’s really interesting to
note here is that we are looking at

00:14:06.840 --> 00:14:12.000
really, really large-scale deformation,
for example, of the radius of

00:14:12.000 --> 00:14:15.540
1,500 kilometer.
So that’s not the scale we are

00:14:15.540 --> 00:14:20.000
often thinking about
for shallow earthquakes.

00:14:21.260 --> 00:14:26.980
Now, for north component,
we have some quadrant feature here.

00:14:26.980 --> 00:14:32.080
Again, with the north being
red and the south being blue.

00:14:33.220 --> 00:14:37.480
Then we have a
vertical component here.

00:14:37.480 --> 00:14:43.860
In this case, the upward motion is in
blue and downward motion is in red.

00:14:44.700 --> 00:14:49.880
And if you, for example, compare
this prediction to what we saw from

00:14:49.890 --> 00:14:54.500
the 8.3 Sea of Okhotsk earthquake
for horizontal components,

00:14:54.500 --> 00:14:59.860
they actually matches the prediction.
Basically, you’re going toward the

00:14:59.860 --> 00:15:02.570
epicenter and going
outward from the epicenter,

00:15:02.570 --> 00:15:05.920
and that’s the motion
we exactly expect.

00:15:07.100 --> 00:15:11.120
Now let’s think about
the viscoelastic relaxation.

00:15:11.120 --> 00:15:15.440
And, in order to do that,
we need to put in some viscosity

00:15:15.440 --> 00:15:19.170
structure of the mantle.
For example, you can think about

00:15:19.170 --> 00:15:25.640
this four-layer model with elastic crust
and asthenosphere, upper mantle, and

00:15:25.640 --> 00:15:31.240
lower mantle, which happens to be
right below the earthquake, basically.

00:15:32.700 --> 00:15:39.800
And we can use this viscosity value –
these are the viscosity values I used just

00:15:39.800 --> 00:15:46.820
for forward modeling. And these are
the Maxwell time in the parentheses.

00:15:46.820 --> 00:15:52.030
For forward modeling,
I’m using the RELAX software,

00:15:52.030 --> 00:15:55.570
which is basically [inaudible]
correspondence principle based

00:15:55.570 --> 00:15:58.700
calculation and
frequency domain.

00:15:58.700 --> 00:16:05.440
And, as you can expect –
this, again, is east component,

00:16:05.440 --> 00:16:11.000
and this is time-dependent process.
So I’m just going to play movie of this.

00:16:11.000 --> 00:16:15.020
And here I am plotting
not the displacement,

00:16:15.020 --> 00:16:17.970
but actually
the velocity here.

00:16:18.940 --> 00:16:24.860
So, for the first four years, that’s
basically the pattern you expect.

00:16:24.870 --> 00:16:30.100
If you – if I look at the snapshot
of zero and four-year snapshot,

00:16:30.100 --> 00:16:36.180
you see also the places like here,
where basically, the sign of the

00:16:36.180 --> 00:16:40.030
velocity has changed,
which is really interesting.

00:16:41.420 --> 00:16:44.880
Now, the north component
also does something like that,

00:16:44.880 --> 00:16:48.340
where the sign
changes at some point.

00:16:49.600 --> 00:16:53.540
So, if you look at the snapshot,
you almost – you’re looking at

00:16:53.550 --> 00:17:00.000
a lot of opposite polarity compared
to when you first started off.

00:17:00.000 --> 00:17:03.810
For the vertical, we don’t necessarily
see that, but actually the pattern you

00:17:03.810 --> 00:17:10.620
start with is already quite different
from what you expect from afterslip.

00:17:10.620 --> 00:17:14.290
So, as time goes,
the polarity doesn’t change much.

00:17:15.440 --> 00:17:21.020
But, again, yes, the pattern is quite
different from what you see here.

00:17:21.020 --> 00:17:26.420
So, if I just look at one point –
if you had a GPS station at somewhere

00:17:26.420 --> 00:17:31.380
like this, and looked at the time
evolution of the deformation,

00:17:31.380 --> 00:17:36.000
then it looks something like this.
So this is east component, north,

00:17:36.000 --> 00:17:40.640
and up component,
with the years on the right axis.

00:17:40.640 --> 00:17:44.840
And you see this
reversal in velocity.

00:17:44.840 --> 00:17:49.720
So that’s not something that we often
see also in shallow earthquake cases.

00:17:49.720 --> 00:17:55.580
And we realize that this actually
happens because of the presence of

00:17:55.580 --> 00:18:01.660
asthenosphere, like weak structure –
weak viscosity structure.

00:18:01.660 --> 00:18:04.120
And interestingly,
as you might expect,

00:18:04.120 --> 00:18:12.450
this timing of reversal is closely related
to the viscosity of asthenosphere.

00:18:12.450 --> 00:18:15.160
So that’s something we
can keep in mind and look for

00:18:15.160 --> 00:18:18.220
when we look
at observations.

00:18:18.220 --> 00:18:22.180
Now, all of us know that these
deep earthquakes actually happen

00:18:22.180 --> 00:18:28.240
in the slab, not just in
the ambient upper mantle.

00:18:29.230 --> 00:18:33.180
So we might have to think
about slab effect as well.

00:18:33.190 --> 00:18:38.770
So let’s say this example I just
shown you was no slab case.

00:18:38.770 --> 00:18:45.050
But if I put in slab – in this case,
just by a simplistic elastic slab,

00:18:45.050 --> 00:18:51.030
and I can do the forward modeling
again. And I want you to pay attention –

00:18:51.030 --> 00:18:56.700
so the epicenter is here, and the
trench location is about here.

00:18:57.870 --> 00:19:01.500
So I want you to look at what’s
happening on the other side

00:19:01.500 --> 00:19:06.240
of the trench as opposed
to this side of the trench.

00:19:07.070 --> 00:19:12.140
Let me play these east component
together. Then you see that actually,

00:19:12.140 --> 00:19:18.340
on the other side of the trench,
you are very slow in terms of

00:19:18.340 --> 00:19:22.120
varying the velocity field.

00:19:23.400 --> 00:19:28.860
It’s persistent, basically, in the –
the initial value is really persistent

00:19:28.860 --> 00:19:33.270
over time, whereas,
it’s changing abruptly here.

00:19:33.270 --> 00:19:36.960
That happens in all the
components – north again.

00:19:36.960 --> 00:19:39.760
So you have, in the end,
more complex pattern here

00:19:39.760 --> 00:19:43.330
because you want to
keep the initial pattern.

00:19:44.200 --> 00:19:50.780
For the north – for the vertical, sorry,
you actually have already different

00:19:50.780 --> 00:19:55.550
pattern to start off on the other side of
the trench, which is really interesting.

00:19:55.550 --> 00:19:59.880
So that can be something, again,
for us to look for if we wanted to

00:19:59.880 --> 00:20:03.510
see if there’s big
slab effect or not.

00:20:03.510 --> 00:20:09.920
It’s actually quite amazing to think
that slab can affect this because,

00:20:09.920 --> 00:20:13.800
when I was first thinking about
this problem, I thought the deformation

00:20:13.800 --> 00:20:19.340
would be governed by all the weak
structures, especially in early times.

00:20:19.340 --> 00:20:24.930
And slab is actually strong structure.
But it turns out the slab can have such

00:20:24.930 --> 00:20:29.280
a big effect on the surface
deformation pattern as well.

00:20:30.290 --> 00:20:35.720
So I have just shown you this four-layer
model, and I’ve explored all different

00:20:35.720 --> 00:20:41.530
types of models that are actually more
simplistic – say, without asthenosphere.

00:20:41.530 --> 00:20:44.200
And then it looks actually
something like this.

00:20:44.200 --> 00:20:49.710
And, if I also remove lower mantle
and just think about the two-layer

00:20:49.710 --> 00:20:53.140
model case, it looks
something like this.

00:20:53.140 --> 00:20:58.660
So, by comparing these two, they’re
actually dramatically different.

00:20:58.660 --> 00:21:05.120
So, again, it’s telling us that
we are sensitive to such deep structure

00:21:05.120 --> 00:21:08.480
by looking at these
deep earthquakes.

00:21:09.380 --> 00:21:13.940
Now, some people have also argued
that mantle transition zone might have

00:21:13.940 --> 00:21:19.150
some water in it, which would
actually make the viscosity lower.

00:21:19.150 --> 00:21:22.000
So we can also test the
kind of model with the

00:21:22.000 --> 00:21:26.380
weak viscosity at the
mantle transition zone.

00:21:26.380 --> 00:21:32.620
And also – we can also think about
models such as a weak viscosity layer

00:21:32.620 --> 00:21:37.740
in between the upper and lower mantle,
which some people have argued for.

00:21:37.740 --> 00:21:44.540
It turns out these two cases aren’t differ
very much, so they are very similar.

00:21:44.540 --> 00:21:49.430
And if I put the slab structure on them,
they, again, look kind of similar,

00:21:49.430 --> 00:21:52.760
and these are the east
and north component.

00:21:53.820 --> 00:21:57.620
Now, another interesting
feature to think about is

00:21:57.630 --> 00:22:00.590
something like
weak mantle wedge.

00:22:00.590 --> 00:22:04.180
For example, Billen et al.
has argued that Fiji-Tonga

00:22:04.180 --> 00:22:08.240
subduction zone has really
weak mantle wedge structure.

00:22:08.240 --> 00:22:13.850
So this is viscosity in color.
So all this yellow thing is

00:22:13.850 --> 00:22:18.560
really weak compared to
all other surrounding viscosity.

00:22:19.809 --> 00:22:22.900
So we can also think
about those structures.

00:22:22.900 --> 00:22:27.940
For example, I put really simple
mantle wedge that are triangle like this.

00:22:27.940 --> 00:22:34.780
And I can also do a forward modeling
of that and see how different they are.

00:22:38.060 --> 00:22:41.460
They’ve also argued for
this much thicker and

00:22:41.460 --> 00:22:46.580
strong viscosity here compared to
the other side of the trench.

00:22:46.580 --> 00:22:51.720
So that’s something I haven’t tested yet.
I just put in uniform crust here.

00:22:51.720 --> 00:22:56.180
So that’s, again, something
I can test in the future.

00:22:58.010 --> 00:23:03.640
So all these exercises really tell us
that we are sensitive to such slab

00:23:03.650 --> 00:23:07.940
and mantle rheology at relatively
deeper depths than what we have

00:23:07.940 --> 00:23:11.940
for shallow earthquakes,
which is really exciting.

00:23:11.940 --> 00:23:17.270
And now I’m going to move on and
show you some real data observation.

00:23:17.270 --> 00:23:21.050
And actually, I’m going to
focus on the region where they

00:23:21.050 --> 00:23:26.700
argued for this mantle wedge,
which is Fiji-Tonga region.

00:23:26.700 --> 00:23:32.940
So actually, third-largest –
third deep earthquake that

00:23:32.940 --> 00:23:38.730
are as large happened in 2018.
So this is very recent event.

00:23:38.730 --> 00:23:41.820
So that happened in the
middle of ocean in the Fiji area.

00:23:41.820 --> 00:23:46.020
It’s very active
area in seismicity.

00:23:46.020 --> 00:23:51.180
And surprisingly, the coverage
is pretty good, even though

00:23:51.190 --> 00:23:55.070
you think that you’re just sitting
in the middle of the ocean.

00:23:55.070 --> 00:24:00.600
And that’s because we have these GPS
stations sitting at different small islands.

00:24:02.180 --> 00:24:05.700
And this earthquake actually has,
more or less, the same mechanism

00:24:05.710 --> 00:24:09.830
as the Sea of Okhotsk earthquake.
So we can use the forward modeling

00:24:09.830 --> 00:24:14.960
for each of them, and they can
be common to each other.

00:24:16.980 --> 00:24:21.940
So let me just show you some data
example focusing on one station.

00:24:21.940 --> 00:24:27.920
We can first look at, if we at all see
coseismic from this earthquake.

00:24:27.920 --> 00:24:34.560
So here is the time series from this
one station – the east component.

00:24:34.560 --> 00:24:40.670
So the vertical axis is in millimeter,
and here is the horizontal in years.

00:24:40.670 --> 00:24:45.191
So basically, I have data plotted
from the time of the earthquake

00:24:45.191 --> 00:24:50.580
to about just before 2000.
In later analysis, I included more data,

00:24:50.580 --> 00:24:55.800
so it’s almost up
to current time.

00:24:55.800 --> 00:25:01.950
So you can clearly see this step that’s
happening at the time of the earthquake.

00:25:01.950 --> 00:25:05.150
And that’s on
the west direction.

00:25:05.150 --> 00:25:10.270
And that’s at the order of 1 centimeter,
so that’s pretty big, actually.

00:25:10.270 --> 00:25:17.000
So I can just color in the west as
a blue color for the coseismic here.

00:25:17.000 --> 00:25:22.660
And I can look at all these different
stations here and do something similar.

00:25:22.670 --> 00:25:27.970
Here you don’t see actually much
coseismic deformation here.

00:25:27.970 --> 00:25:32.770
But, in other cases, you see
most of this west motion –

00:25:32.770 --> 00:25:37.680
west step here on
each GPS station data.

00:25:39.440 --> 00:25:43.500
So here, I didn’t do any
processing on these data sets.

00:25:43.500 --> 00:25:49.770
Basically, what I did is just take
the raw data and basically flattened

00:25:49.770 --> 00:25:54.170
the part before the
earthquake and just plotted.

00:25:54.170 --> 00:26:01.120
So these patterns – this coseismic
deformation pattern is actually

00:26:01.120 --> 00:26:05.410
basically consistent with what I just
showed you for afterslip east pattern,

00:26:05.410 --> 00:26:10.200
which is – which is basically
same as the coseismic pattern.

00:26:11.420 --> 00:26:17.120
Now, we talked about coseismic.
Can we actually say that these

00:26:17.120 --> 00:26:20.670
are coseismic deformation
that are also going into

00:26:20.670 --> 00:26:24.970
same direction of the
coseismic to the west?

00:26:24.970 --> 00:26:30.930
Actually, this requires a lot
of careful looking at the data.

00:26:30.930 --> 00:26:35.440
Because, for example, if you look at
these stations, or all the other station,

00:26:35.440 --> 00:26:41.500
there could be some seasonal
component – seasonal deformation

00:26:41.500 --> 00:26:47.040
that can be mixed up here.
So, in order to do that in a robust way,

00:26:47.040 --> 00:26:52.231
I wanted to do it more in a quantitative
analysis and take account all the

00:26:52.240 --> 00:26:56.400
amplitude information and
time evolution information.

00:26:56.400 --> 00:26:59.800
And tease out these
different signals.

00:27:00.890 --> 00:27:06.160
This is real challenge because
maybe the seasonal signal is

00:27:06.160 --> 00:27:11.750
as big as what we want to
look for in the postseismic signal.

00:27:11.750 --> 00:27:19.040
So, in order to tackle that problem, I’m
using independent component analysis.

00:27:19.040 --> 00:27:23.210
So it’s just – you can basically
think about it as a singular value

00:27:23.210 --> 00:27:28.640
decomposition, so it’s like PCA,
but it doesn’t – you don’t have to

00:27:28.640 --> 00:27:31.960
have each component to
be orthogonal to each other.

00:27:31.960 --> 00:27:39.100
So each component can have
independent physical meaning to it.

00:27:40.220 --> 00:27:45.680
So it’s a way of decomposing
the signal into its different sources.

00:27:45.680 --> 00:27:52.100
So, for example, if you have this
raw signal from the GPS station,

00:27:52.110 --> 00:27:55.799
you might be able to tease out
the seasonal signal versus

00:27:55.799 --> 00:27:59.880
the earthquake-related
signal that you’re interested in.

00:27:59.880 --> 00:28:02.700
So that’s what I’m
going to use here.

00:28:05.570 --> 00:28:10.340
So I’m using all three components –
so vertical component is included.

00:28:10.350 --> 00:28:17.820
And, as I said, I’m including time period
up to, more or less, present from 2017.

00:28:17.820 --> 00:28:23.740
And I’m using stations basically
within 1,500 kilometer, and I’ve also

00:28:23.740 --> 00:28:28.020
looked at some of the much
farther stations just in case.

00:28:31.220 --> 00:28:35.160
Now, since this deep earthquake,
we are looking at really small signals,

00:28:35.160 --> 00:28:40.500
so we are doing some pre-processing
before even getting into ICA.

00:28:40.500 --> 00:28:45.520
So, by using trajectory model,
which is basically, again, decomposing

00:28:45.520 --> 00:28:51.580
GPS signal time series into, for
example, linear features and a lot of

00:28:51.580 --> 00:28:58.740
step features and seasonal signals and
earthquake-related postseismic signal.

00:28:59.380 --> 00:29:04.580
So basically, I do a feeding
of data to this kind of model,

00:29:04.580 --> 00:29:11.620
and I get these estimates
for the steps and linear trend.

00:29:11.620 --> 00:29:17.740
And coseismic step also. And then
I removed them from the signal.

00:29:17.740 --> 00:29:22.660
So, in the pre-processing step,
I’m removing not only the

00:29:22.670 --> 00:29:28.140
coseismic step and linear – but also
linear step and other steps related to

00:29:28.140 --> 00:29:32.580
maybe other earthquakes or
instrument issues and also

00:29:32.580 --> 00:29:36.820
sometimes postseismic signals
of other local earthquakes.

00:29:39.400 --> 00:29:44.710
So this is one example of trajectory
model fitting from the same station

00:29:44.710 --> 00:29:49.840
I showed you earlier. So this is
raw data on the east component.

00:29:49.840 --> 00:29:53.290
And I’m removing the linear
trend and all these steps.

00:29:53.290 --> 00:29:58.640
So this is actually timing of the
earthquake that we are interested in.

00:29:58.640 --> 00:30:03.510
And this magenta line is a step
that’s related to instrument issue.

00:30:03.510 --> 00:30:08.330
So I’m removing all the signals,
and I’m left with this signal.

00:30:08.330 --> 00:30:12.590
So here, I have presumably
still seasonal signal embedded

00:30:12.590 --> 00:30:16.510
in here and also
postseismic signal.

00:30:16.510 --> 00:30:20.880
You don’t want to actually remove
seasonal signal from here because

00:30:20.880 --> 00:30:27.380
trajectory model assumes, again,
this sinusoidal pattern, and actual

00:30:27.380 --> 00:30:32.600
seasonal signal are not
actually looking like sinusoids.

00:30:33.680 --> 00:30:37.840
And I do the same for north and
vertical component and left with

00:30:37.840 --> 00:30:46.660
these kind of signals. And these are
the data that I use to put into ICA.

00:30:48.230 --> 00:30:53.300
So that’s what I just talked about, doing
a pre-processing of trajectory model.

00:30:53.310 --> 00:30:55.420
And do an ICA there.

00:30:55.420 --> 00:31:00.040
I actually had to do a lot of
iterations of this because

00:31:00.040 --> 00:31:05.710
the problem was quite tricky.
And so basically, what I did was repeat

00:31:05.710 --> 00:31:12.960
this until I actually get the robust
seasonal signals from the first ICA.

00:31:15.260 --> 00:31:20.060
This is necessary because, for example,
the trajectory model, as I mentioned,

00:31:20.060 --> 00:31:25.250
assumed sinusoidal seasonal.
And that can also affect other

00:31:25.250 --> 00:31:29.940
estimates like coseismic offset.
So we have to be careful

00:31:29.940 --> 00:31:33.160
in terms of getting good
estimate of coseismic

00:31:33.160 --> 00:31:37.700
and other ones here and
then do the first ICA.

00:31:39.440 --> 00:31:43.340
Another thing I did to stabilize
this problem is to impose maybe

00:31:43.340 --> 00:31:48.620
robust linear velocities,
for example, from [inaudible].

00:31:48.620 --> 00:31:54.390
That’s basically the velocity that’s
coming from tectonic processes.

00:31:54.390 --> 00:31:59.730
So using a velocity that’s estimated
using really long time series can be

00:31:59.730 --> 00:32:04.660
more robust than the just looking at
the time window that I’m looking at.

00:32:07.180 --> 00:32:15.390
So basically, I get – in the first ICA, I
get the robust seasonal signal from this.

00:32:15.390 --> 00:32:20.990
And then actually subtract the
seasonal component from the data.

00:32:20.990 --> 00:32:24.660
And then do an ICA
again on the residuals.

00:32:24.660 --> 00:32:29.820
So that’s where I actually
recovered a postseismic signal.

00:32:30.929 --> 00:32:38.300
So let me show you the preliminary
results I have from the first ICA – ICA I.

00:32:38.300 --> 00:32:43.390
So I do extract the seasonal signal,
which is one of the component.

00:32:43.390 --> 00:32:47.660
And, as you can see, this
doesn’t really look like sinusoids.

00:32:47.660 --> 00:32:50.710
But you can see
that it’s seasonal.

00:32:50.710 --> 00:32:57.490
So basically, this is the ICA
independent component I recovered.

00:32:57.490 --> 00:33:03.320
And, on the bottom, it’s showing you,
for each station time series, how much

00:33:03.320 --> 00:33:10.669
contribution of this is in each station.
So you can think about it as – this as

00:33:10.669 --> 00:33:18.010
eigenvalue and that as the eigenvector
to construct the time series here.

00:33:18.010 --> 00:33:23.850
So this is horizontal scale.
So most of the horizontal arrows are

00:33:23.850 --> 00:33:28.440
small here, but actually, if you look
at vertical, they are very significant.

00:33:28.440 --> 00:33:32.110
And the middle value is
about 2 centimeters or so.

00:33:32.110 --> 00:33:38.260
So this is very huge vertical seasonal
deformation that we are seeing here.

00:33:38.260 --> 00:33:42.000
And you can tell that this is clearly
not postseismic signal just by

00:33:42.000 --> 00:33:50.080
looking at this. And actually, these
seasonal signals can also be additional

00:33:50.080 --> 00:33:55.980
source of information on rheology,
and I’ll come back to this later.

00:33:59.560 --> 00:34:03.600
This is second component
of seasonal again.

00:34:03.600 --> 00:34:09.020
You see this pattern here.
That’s, again, not exactly sinusoidal.

00:34:09.020 --> 00:34:15.420
So, in this case, it’s basically the
one that’s going to the east motion.

00:34:17.360 --> 00:34:22.849
And then there’s another seasonal
component, that’s Seasonal #3,

00:34:22.849 --> 00:34:27.580
that’s going to southwest
direction, more or less.

00:34:29.820 --> 00:34:36.700
And, after removing these three
seasonal signals, I am then putting in

00:34:36.700 --> 00:34:41.340
the postseismic signal.
But them I’m getting the –

00:34:41.340 --> 00:34:47.660
then I’m doing the ICA round 2 and
actually get the postseismic signal.

00:34:48.360 --> 00:34:55.080
If I put some uncertainty on these
arrows, it looks something like this.

00:34:55.080 --> 00:35:02.059
And note that, if you look at this
postseismic signal, there’s no reversal

00:35:02.059 --> 00:35:08.430
in velocity yet. So maybe this tells
us something about asthenosphere.

00:35:08.430 --> 00:35:12.609
For example, it might be not be
extremely weak asthenosphere.

00:35:12.609 --> 00:35:16.339
Because we still haven’t
seen the reversal.

00:35:16.339 --> 00:35:19.839
Or we might not see the reversal,
which, in that case, might mean

00:35:19.840 --> 00:35:22.680
that we don’t
have asthenosphere.

00:35:23.700 --> 00:35:27.560
Now let me just focus
on the east-west component.

00:35:27.560 --> 00:35:34.260
And I’m going to color in each station
where east is red and west is blue.

00:35:34.269 --> 00:35:39.930
And if it’s something like this,
in the middle, not east or west,

00:35:39.930 --> 00:35:42.400
then I’ll put in gray color.

00:35:44.100 --> 00:35:48.460
So if I look at them in map view,
just the polarity of them,

00:35:48.460 --> 00:35:50.860
it looks something like this.

00:35:51.820 --> 00:35:58.720
And now I can compare this observation
with the forward calculations I’ve been

00:35:58.720 --> 00:36:05.640
showing you. For example, I can bring in
the afterslip pattern we looked at earlier.

00:36:06.480 --> 00:36:11.060
It’s actually quite a good match.
I was quite surprised by this.

00:36:11.900 --> 00:36:15.520
You see the red and blue
aligning with each other.

00:36:16.420 --> 00:36:21.500
And, as I said, this afterslip pattern
is same as coseismic pattern.

00:36:21.500 --> 00:36:27.140
And you always expect the
afterslip to be in the same direction

00:36:27.150 --> 00:36:32.530
with the coseismic.
So, if you see anything that we

00:36:32.530 --> 00:36:36.880
see in observation, these dots,
that are having opposite sign

00:36:36.880 --> 00:36:40.950
from the background here –
the forward calculation here,

00:36:40.950 --> 00:36:45.580
that means we have detected
the viscoelastic relaxation.

00:36:45.580 --> 00:36:51.120
So, for example, if you look at this
station, you are expecting blue,

00:36:51.130 --> 00:36:55.359
but you have red here,
which is east direction.

00:36:55.359 --> 00:37:01.299
So this might be a hint of viscoelastic
relaxation that we are looking at.

00:37:01.299 --> 00:37:04.750
But it’s maybe hard to tell
because they’re around

00:37:04.750 --> 00:37:07.460
this boundary region here.

00:37:07.460 --> 00:37:13.339
Well, let me compare this to other
viscoelastic relaxation models

00:37:13.339 --> 00:37:16.589
right after the earthquake.
So this is the pattern that’s expected

00:37:16.589 --> 00:37:20.280
from this two-layer –
simple two-layer model.

00:37:20.280 --> 00:37:22.860
And you see that it really
completely doesn’t fit.

00:37:22.860 --> 00:37:27.780
It’s like opposite sign
from the observation.

00:37:27.780 --> 00:37:31.680
however, if you put in
asthenosphere and just think about

00:37:31.690 --> 00:37:35.390
this four-layer model,
you start to fit this data

00:37:35.390 --> 00:37:39.840
in the middle except for
these two stations here.

00:37:41.109 --> 00:37:44.900
Now let’s look at the
north component as well.

00:37:44.900 --> 00:37:49.800
So that was the quadrant feature
that we expect from afterslip.

00:37:49.800 --> 00:37:55.540
And actually, you see these
two stations that are definitely

00:37:55.540 --> 00:37:58.160
opposite sign from
the background.

00:37:58.160 --> 00:38:02.630
So, actually, these two stations are
really strong evidence that we are

00:38:02.630 --> 00:38:08.040
seeing viscoelastic relaxation
here, at least some of it.

00:38:11.240 --> 00:38:18.500
Then, if I compare this to, again,
the simple model, it doesn’t explain

00:38:18.500 --> 00:38:24.400
all parts of the data, but it
does match this part, actually.

00:38:26.180 --> 00:38:28.800
This is, again,
the four-layer model.

00:38:28.800 --> 00:38:33.680
And, yeah, you don’t
explain this part, for example.

00:38:34.880 --> 00:38:38.740
Let me compile
all this in one slide.

00:38:38.749 --> 00:38:43.859
So we’ve just looked at east
and north direction for afterslip

00:38:43.860 --> 00:38:47.860
and two viscoelastic
relaxation models.

00:38:47.860 --> 00:38:56.099
And you see that, if you look at
this station – actually, I have compared

00:38:56.100 --> 00:39:02.380
this to other forward modeling from
other complex structures like this,

00:39:02.380 --> 00:39:06.460
and I haven’t actually found
one yet that can explain these

00:39:06.460 --> 00:39:10.859
streaks of red
in the middle.

00:39:10.859 --> 00:39:14.500
And especially this station.

00:39:14.500 --> 00:39:21.239
So that can mean that, if some pattern
cannot be explained by any viscoelastic

00:39:21.239 --> 00:39:27.529
prediction, then that can be an evidence
for the existence of afterslip after

00:39:27.529 --> 00:39:33.760
this big, large, deep earthquake.
So that’s quite exciting.

00:39:36.280 --> 00:39:43.060
Now I have been, again, comparing this
to other synthetics, as I just mentioned,

00:39:43.060 --> 00:39:46.230
for the complex –
more complex structures.

00:39:46.230 --> 00:39:52.239
And it seems like we have
better fit as we include asthenosphere

00:39:52.240 --> 00:39:54.860
and also slab as well.

00:39:55.680 --> 00:40:01.779
So basically, I’m finding that
the weak viscosity structures

00:40:01.779 --> 00:40:06.410
like asthenosphere and
3D structures like slab

00:40:06.410 --> 00:40:12.040
are quite important in
explaining the data we have.

00:40:12.740 --> 00:40:16.780
I’m working on this still
and trying to look at more detailed

00:40:16.789 --> 00:40:20.710
structures like what
I just shown you earlier.

00:40:20.710 --> 00:40:27.210
And also thinking about slab
viscosity and slab dip and thickness

00:40:27.210 --> 00:40:30.710
and asthenosphere thickness
and viscosity, obviously.

00:40:30.710 --> 00:40:36.150
And also half slab, which is
basically coming from this map.

00:40:36.150 --> 00:40:40.599
This is quite complex
subduction zone structure.

00:40:40.599 --> 00:40:43.910
You don’t have, actually, slab
that’s going all the way through.

00:40:43.910 --> 00:40:46.700
You have slab that’s
ending around here.

00:40:46.700 --> 00:40:49.430
And there’s actually even
another slab coming in

00:40:49.430 --> 00:40:52.940
from the opposite
directions like that.

00:40:52.940 --> 00:40:58.599
So I’ve also trying to test the slab
that’s just going through the half

00:40:58.599 --> 00:41:02.920
of the domain here.
And maybe that can explain

00:41:02.920 --> 00:41:07.599
some of the observation
on the one side and the other side

00:41:07.600 --> 00:41:11.400
of the trench and also
above the slab trench.

00:41:13.360 --> 00:41:17.680
And then, after exploring all that,
I can maybe say something

00:41:17.680 --> 00:41:24.020
more strong about potential
afterslip following this earthquake.

00:41:27.100 --> 00:41:32.099
Now there’s some additional
constraints to rheology structure here,

00:41:32.100 --> 00:41:35.820
which I have mentioned earlier,
using seasonal terms.

00:41:37.420 --> 00:41:43.520
So this is a study – recent study
that actually have looked at Earth

00:41:43.529 --> 00:41:49.809
response to the surface loading.
So basically, these seasonal signals

00:41:49.809 --> 00:41:54.640
are there because the Earth is
responding to the surface load.

00:41:54.640 --> 00:41:57.460
And, depending on the
rheological structure,

00:41:57.460 --> 00:42:01.549
as you expect, the response
becomes different.

00:42:01.549 --> 00:42:08.240
So the gray bars here with the error bars
are basically data you have from GPSs.

00:42:08.240 --> 00:42:15.100
And the red lines are what’s expected
from completely elastic Earth.

00:42:15.109 --> 00:42:18.880
And all the other colored
lines are basically showing you

00:42:18.880 --> 00:42:22.339
different rheological structure.

00:42:22.339 --> 00:42:26.739
So you can – by trying to match
the data, you can actually get

00:42:26.740 --> 00:42:31.740
some information about the
Earth rheological structure here.

00:42:31.740 --> 00:42:36.999
Although the caveat here is that, again,
this wouldn’t be really sensitive to

00:42:37.000 --> 00:42:41.859
deeper viscosity structure like we
were looking at for deep earthquakes.

00:42:42.420 --> 00:42:47.940
So, for example, I can bring back this
seasonal term I showed you earlier.

00:42:47.940 --> 00:42:50.180
So that’s my data.

00:42:50.180 --> 00:42:54.960
And I try to
compare that to this,

00:42:54.960 --> 00:43:01.460
which is expected pattern from
completely elastic Earth.

00:43:01.460 --> 00:43:10.579
And first thing to notice here is that,
if I compare these two time series,

00:43:10.579 --> 00:43:14.690
they’re actually offset in time,
but if I try to overlap them,

00:43:14.690 --> 00:43:18.960
there’s some phase lag
in between the two.

00:43:20.130 --> 00:43:24.140
If you look at the amplitude,
we are looking at the order of

00:43:24.150 --> 00:43:29.640
2 centimeter here, while the
prediction is only 1-centimeter level.

00:43:29.640 --> 00:43:33.229
So there’s mismatch of
2 times of amplitude, maybe,

00:43:33.229 --> 00:43:35.640
and two months
of phase shift.

00:43:35.640 --> 00:43:38.950
So we might be able to
find a rheological model

00:43:38.950 --> 00:43:42.160
that can explain
this misfit here.

00:43:44.420 --> 00:43:51.440
So let me just conclude and say that,
most importantly, we actually have

00:43:51.440 --> 00:43:56.280
detected the postseismic deformation
from these deep earthquakes.

00:43:56.280 --> 00:44:00.680
And our predictions show
that afterslip pattern would be

00:44:00.680 --> 00:44:04.800
expected to be exactly
the same as coseismic.

00:44:04.800 --> 00:44:09.700
And the viscoelastic relaxation
modeling has shown us that we are

00:44:09.700 --> 00:44:17.160
sensitive to deep mantle rheology
and also slab rheological structures.

00:44:18.700 --> 00:44:22.860
Our observation and using ICA,
we could actually recover the

00:44:22.869 --> 00:44:28.489
postseismic and also the seasonal terms,
where the seasonal terms can be

00:44:28.489 --> 00:44:32.839
additional constraints
on the rheological structure.

00:44:34.180 --> 00:44:39.660
Now, we have some preliminary
implications, which are that we

00:44:39.660 --> 00:44:46.989
are looking at viscoelastic relaxation
just because the afterslip pattern

00:44:46.989 --> 00:44:53.800
cannot explain the reverse –
the different polarity.

00:44:53.800 --> 00:44:58.540
And actually the weak rheological
structure and 3D structures,

00:44:58.540 --> 00:45:01.440
like slab, are quite important.

00:45:02.100 --> 00:45:08.900
And we might actually be detecting
the post potential afterslip there too.

00:45:09.840 --> 00:45:14.540
Lastly, I would like to
just quickly introduce something

00:45:14.540 --> 00:45:18.180
that’s completely off-topic.

00:45:18.180 --> 00:45:23.960
Because I thought that would
be interest of this audience.

00:45:23.960 --> 00:45:29.619
And that’s something that
I haven’t even properly named yet,

00:45:29.619 --> 00:45:34.289
but I’m just calling it
3D printing seismology.

00:45:34.289 --> 00:45:38.999
So the idea here is that we have
the seismic velocity model.

00:45:38.999 --> 00:45:44.849
And, using the 3D printing technology,
we can print a physical model in

00:45:44.849 --> 00:45:50.720
a such way that it exactly represents
the realistic seismic velocity.

00:45:50.720 --> 00:45:55.719
Then we can use that to do an
experiment on it by putting

00:45:55.719 --> 00:45:59.569
source and receiver –
in this case transducers –

00:45:59.569 --> 00:46:03.690
and actually generate
the wavefield and record it.

00:46:03.690 --> 00:46:08.560
This becomes very important
for seismic hazard assessment,

00:46:08.560 --> 00:46:11.440
as you can imagine.

00:46:12.140 --> 00:46:16.559
For example, we can print
mini southern California,

00:46:16.560 --> 00:46:20.400
with the size of,
say 50 by 50 centimeters.

00:46:20.400 --> 00:46:25.359
We’ll be able to explore the effect
of small-scale topography and

00:46:25.359 --> 00:46:31.920
basin interfaces on ground shaking,
which is a really unfeasible problem

00:46:31.920 --> 00:46:36.329
to tackle using conventional
numerical simulation due to

00:46:36.329 --> 00:46:39.040
the high computational cost.

00:46:40.420 --> 00:46:46.460
So I think the subject has a lot of
potential future research directions,

00:46:46.469 --> 00:46:51.829
and I’ll actually be talking about
this at a upcoming SCEC meeting,

00:46:51.829 --> 00:46:55.680
if any of you are attending it.
And that’s it.

00:46:55.680 --> 00:46:59.960
I’ll leave the conclusion slide here.
Thank you very much.

00:47:04.140 --> 00:47:05.840
- Great.

00:47:05.840 --> 00:47:08.480
Thank you, Sunny,
for a great talk.

00:47:10.120 --> 00:47:13.749
If anybody has questions,
you can either type them into the chat

00:47:13.749 --> 00:47:18.120
or unmute yourself to
ask Sunny your questions.

00:47:20.840 --> 00:47:24.980
[Silence]

00:47:24.980 --> 00:47:28.769
- Hello. This is Fred.
That was a great talk.

00:47:28.769 --> 00:47:33.619
And you’re taking on a really
challenging problem, trying to separate

00:47:33.619 --> 00:47:39.630
viscous relaxation from afterslip
and extract a signal from the data.

00:47:39.630 --> 00:47:42.780
So I had a couple of questions.
The first was …

00:47:45.460 --> 00:47:50.280
… in the Tonga region, have you
looked at the effects of the 2009

00:47:50.280 --> 00:47:53.680
magnitude 8 earthquakes that
happened there to see if they

00:47:53.680 --> 00:48:00.720
perturbed the response
that you’re measuring?

00:48:00.720 --> 00:48:05.520
- Are you talking about the
intermediate-depth earthquake?

00:48:05.520 --> 00:48:11.340
- Oh, those were fairly shallow.
- Like, 50-kilometer depth, maybe?

00:48:11.340 --> 00:48:13.099
- Yeah.

00:48:13.099 --> 00:48:18.190
So they affected a local area.
But some of your outliers around

00:48:18.190 --> 00:48:23.860
the latitude of 10 or 15 south
might be near that area.

00:48:25.100 --> 00:48:32.440
- So is this what you’re
talking about, maybe? Or no?

00:48:32.440 --> 00:48:35.020
- It happened in 2009.
I’m not sure.

00:48:35.020 --> 00:48:36.940
- Oh, okay.

00:48:37.760 --> 00:48:40.360
Yeah. So that
wouldn’t be this one.

00:48:40.360 --> 00:48:45.140
But, yeah, actually, this is very,
yeah, seismically active region.

00:48:45.140 --> 00:48:50.100
And there are not only that event –
I’m not specifically aware of the

00:48:50.109 --> 00:48:54.729
event that you’re mentioning,
but there are other major events

00:48:54.729 --> 00:49:00.289
that are happening throughout this –
the time of this analysis.

00:49:00.289 --> 00:49:05.329
So I have looked at them carefully, and,
for example, what’s happening for those

00:49:05.329 --> 00:49:12.859
cases is that, if that’s shallow –
let me see if I can bring up …

00:49:14.080 --> 00:49:19.200
So, for example, these are some of the
earthquakes that doesn’t include the

00:49:19.209 --> 00:49:24.059
event what you’re mentioning,
but that’s the one I was looking at.

00:49:24.059 --> 00:49:25.980
And there are other
significant earthquake

00:49:25.980 --> 00:49:29.650
that are happening
around the time here.

00:49:29.650 --> 00:49:33.349
For example, these are
all magnitude 7-pluses.

00:49:33.349 --> 00:49:37.519
So I have looked at that quite carefully.
And what happens for these shallow

00:49:37.519 --> 00:49:42.420
earthquakes, for example, is that
they don’t – they affect the stations

00:49:42.420 --> 00:49:47.119
around there in the vicinity,
but for example, not the ones here.

00:49:47.119 --> 00:49:54.930
So, by using ICA, I can actually tease
out those different signals and just focus

00:49:54.930 --> 00:49:58.969
on the signal that’s starting right
at the time of my earthquake.

00:49:58.969 --> 00:50:03.560
So that’s kind of the way that
I can get around that problem.

00:50:04.960 --> 00:50:09.140
- Okay. Yeah. I agree. Each one
would have just a local effect.

00:50:09.149 --> 00:50:12.530
- Mm-hmm.
- Second question was, can you look at

00:50:12.530 --> 00:50:17.400
GRACE gravity to extract a signal?
Is there any possibility?

00:50:17.400 --> 00:50:25.540
- Yes. So actually, the seasonal terms –
so this thing I’ve been showing you here,

00:50:25.549 --> 00:50:32.119
this expected elastic response of the
Earth – basically, when we do this,

00:50:32.119 --> 00:50:39.190
the surface loading is basically –
is estimated from GRACE signal.

00:50:39.190 --> 00:50:44.190
So that’s the way here
I’ve been using GRACE.

00:50:44.190 --> 00:50:49.569
But just, if you want to look at
anything like a – directly infer

00:50:49.569 --> 00:50:55.800
the postseismic deformation from
GRACE, it’s actually quite challenging.

00:50:55.800 --> 00:51:02.200
Well, it’s challenging for GPS, too,
but the reason why – it’s also really

00:51:02.200 --> 00:51:07.910
oceanic area, so we have to also
think about, say, like, oceanic

00:51:07.910 --> 00:51:14.450
water movement all the time.
And so, yeah, answer is, I haven’t really

00:51:14.450 --> 00:51:20.840
looked at GRACE just for postseismic –
probing postseismic directly.

00:51:20.840 --> 00:51:25.460
Sort of indirect way to
look at the seasonal, yes.

00:51:25.480 --> 00:51:28.460
- Okay. Thanks.
- Thank you.

00:51:31.400 --> 00:51:35.580
[Silence]

00:51:35.580 --> 00:51:40.200
- All right. are there
any other questions?

00:51:43.300 --> 00:51:46.520
[Silence]

00:51:46.529 --> 00:51:53.200
I have one. This is a question from
a geodesist, so I don’t really know,

00:51:53.200 --> 00:52:00.520
but can you kind of talk about the
potential uncertainty in the – in the

00:52:00.520 --> 00:52:05.589
focal mechanisms for deep earthquakes?
And, like, if there is any sort of

00:52:05.589 --> 00:52:12.579
small change in angle, could it change
where the lobes of deformation appear?

00:52:12.579 --> 00:52:14.119
- Yes.

00:52:14.119 --> 00:52:19.359
So the orientation of focal
mechanism does affect the pattern.

00:52:19.359 --> 00:52:25.089
For example – yeah, both afterslip
and viscoelastic patterns.

00:52:25.089 --> 00:52:32.410
However, actually, one of the interesting
questions for deep earthquakes are that,

00:52:32.410 --> 00:52:38.500
do they really happen in, say,
horizontal or vertical plane –

00:52:38.500 --> 00:52:44.410
in between the two nodal planes?
So, for example, here, they often

00:52:44.410 --> 00:52:48.549
have this kind of mechanism
where you basically have,

00:52:48.549 --> 00:52:53.309
more or less, horizontal plane and
one, more or less, vertical plane.

00:52:53.309 --> 00:52:58.109
And even just using seismological data,
people have argued if it’s more vertical –

00:52:58.109 --> 00:53:02.069
actually happening in
vertical plane or horizontal plane.

00:53:02.069 --> 00:53:09.020
But, so I’ve tested that, and that
doesn’t change the prediction much.

00:53:09.020 --> 00:53:16.340
But changing the orientation – yeah,
it does change the pattern more, yeah.

00:53:19.500 --> 00:53:36.800
[Silence]

00:53:36.800 --> 00:53:39.080
- Any other questions?

00:53:41.820 --> 00:53:56.720
[Silence]

00:53:56.720 --> 00:54:02.160
Well – oh, there’s one on the –
on the chat. Yeah.

00:54:02.160 --> 00:54:08.560
So another question from Jeff McGuire.
If there are multiple sub-events with

00:54:08.560 --> 00:54:13.860
different strikes, is that enough to affect
the polarity of the postseismic response?

00:54:16.400 --> 00:54:18.400
[Silence]

00:54:18.540 --> 00:54:24.000
- I would say yes, if there’s
sub-event with different mechanisms.

00:54:24.000 --> 00:54:26.819
Yeah, that is possible.

00:54:26.820 --> 00:54:31.960
Although – so this event definitely 
attracted a lot of attention because

00:54:31.960 --> 00:54:34.180
it’s one of the
three large earthquakes.

00:54:34.180 --> 00:54:39.040
So this has been
studied using seismic data.

00:54:39.040 --> 00:54:47.599
And I don’t think I have seen, say,
like, two sub-events, for example,

00:54:47.599 --> 00:54:51.150
that’s almost similar energy
having different mechanism.

00:54:51.150 --> 00:54:56.979
It’s, like, one major event that has a
mechanism close to this, and I wouldn’t

00:54:56.979 --> 00:55:01.329
be worried too much about the
different sub-events in this case.

00:55:01.329 --> 00:55:06.160
But, yes, in general, it would be
important to think about that, yeah.

00:55:06.160 --> 00:55:09.040
And that would
change the pattern.

00:55:12.480 --> 00:55:19.960
- Okay. Well, if there are any
last questions, go ahead and ask.

00:55:19.960 --> 00:55:25.600
We have – Alan says, great talk.
[laughs]

00:55:25.600 --> 00:55:29.240
Best wishes on Chicago.
- [laughs]

00:55:29.240 --> 00:55:31.360
Thank you.

00:55:33.480 --> 00:55:37.180
- And if there aren’t any
other questions, we’ll –

00:55:37.180 --> 00:55:42.060
let’s thank our speaker again,
and see you guys next week.

00:55:44.800 --> 00:55:46.360
- Thank you.

00:55:48.180 --> 00:55:51.740
- Thank you.
- Thank you.

00:55:54.720 --> 00:56:00.020
[Silence]