WEBVTT
Kind: captions
Language: en-US

00:00:02.320 --> 00:00:05.760
Thank you very much to the organizers
for inviting me to give this talk.

00:00:05.760 --> 00:00:08.720
Today I’m going to be speaking about
ground shaking and hazards from

00:00:08.720 --> 00:00:12.640
Cascadia subduction zone earthquakes
with a focus on northern California.

00:00:12.640 --> 00:00:15.520
And I just want to acknowledge
my collaborators Alex Grant,

00:00:15.520 --> 00:00:18.447
Art Frankel,
and Nasser Marafi.

00:00:19.360 --> 00:00:22.560
Subduction zones are hosts to
the world’s largest earthquakes,

00:00:22.560 --> 00:00:25.440
and in addition to ground shaking,
they also produce devastating

00:00:25.440 --> 00:00:27.280
tsunamis and
ground failures.

00:00:27.280 --> 00:00:31.600
This is a map of recorded and
well-constrained historical earthquakes

00:00:31.600 --> 00:00:36.080
with magnitudes greater then 8.5,
and the colors indicate the tsunami

00:00:36.080 --> 00:00:41.096
maximum water height, which you can
see, in some cases, exceeds 30 meters.

00:00:41.120 --> 00:00:44.400
Now, in Cascadia, where the Juan de
Fuca Plate is subducting beneath North

00:00:44.400 --> 00:00:48.560
America, we haven’t experienced a large
megathrust earthquake in recent history.

00:00:48.560 --> 00:00:51.120
We know from a variety
of paleoseismic evidence

00:00:51.120 --> 00:00:53.440
that the subduction zone
is capable of generating

00:00:53.440 --> 00:00:56.936
magnitude 8 to 9
earthquakes.

00:00:56.960 --> 00:01:00.000
So this is a figure from Walton and
Staisch and others highlighting the

00:01:00.000 --> 00:01:04.720
multitude of paleoseismic evidence
at sites, both onshore and offshore,

00:01:04.720 --> 00:01:09.520
for Cascadia megathrust earthquakes.
It’s a bit hard to see, but these horizontal

00:01:09.520 --> 00:01:14.240
bars represent turbidite age ranges,
and the downward arrows and waves

00:01:14.240 --> 00:01:18.056
are dated estimates of coastal
subsidence and tsunami inundation

00:01:18.080 --> 00:01:21.760
at corresponding sites.
And thus, these vertical gray bars

00:01:21.760 --> 00:01:27.176
here are interpreted as ages of
previous Cascadia earthquakes.

00:01:27.200 --> 00:01:31.200
So, for margin-wide magnitude 9
ruptures, such as shown in this figure,

00:01:31.200 --> 00:01:34.560
the paleoseismic record suggests
a recurrence interval of roughly

00:01:34.560 --> 00:01:37.656
every 500 to 530 years.

00:01:37.680 --> 00:01:41.600
Now, for shorter partial-margin
ruptures on the order of magnitude 8,

00:01:41.600 --> 00:01:44.960
paleoseismic evidence suggests that
the ruptures may be more frequent

00:01:44.960 --> 00:01:50.001
and closer to about 300 years
or so in southern Cascadia.

00:01:50.800 --> 00:01:54.160
So I’m sure many of you have heard me
talk about the M9 Project, and the work

00:01:54.160 --> 00:01:57.760
I’m going to show you today is
an extension of those simulations.

00:01:57.760 --> 00:02:02.160
So, just as a reminder, as part of the
M9 Project, we ran 3D kinematic

00:02:02.160 --> 00:02:06.960
simulations of 50 M9 Cascadia
earthquake scenarios using a compound

00:02:06.960 --> 00:02:10.560
rupture model where low-frequency
energy comes from the steep

00:02:10.560 --> 00:02:14.560
background slip, and high-frequency
energy is radiated from distinct

00:02:14.560 --> 00:02:18.136
sub-events on the deeper
portions of the fault.

00:02:18.160 --> 00:02:20.960
So these 3D simulations
are run up to 1 hertz.

00:02:20.960 --> 00:02:24.640
And, for higher frequencies, we used a
stochastic approach and assumed that

00:02:24.640 --> 00:02:29.269
the high frequencies are coming from
these high-stress drop sub-events.

00:02:30.080 --> 00:02:34.400
Now, 30 of those earthquake scenarios
were derived from a logic tree that

00:02:34.400 --> 00:02:38.240
varied multiple parameters.
So we varied the down-dip rupture

00:02:38.240 --> 00:02:42.560
extent, so whether it was completely
offshore, kind of brushing the coastline,

00:02:42.560 --> 00:02:45.440
or if the down-dip rupture limit
extended eastward beneath the

00:02:45.440 --> 00:02:49.760
coastline. We varied the hypocenter
location, the slip distribution was

00:02:49.760 --> 00:02:52.880
randomly generated,
and we also varied the locations

00:02:52.880 --> 00:02:56.010
of those high-stress drop
sub-events.

00:02:56.400 --> 00:02:59.680
So last year, we expanded upon
the M9 Project simulations and

00:02:59.680 --> 00:03:04.376
used them to develop what we
refer to as ensemble ShakeMaps.

00:03:04.400 --> 00:03:08.560
So, for each of the 30 M9 Cascadia
earthquake scenarios that were derived

00:03:08.560 --> 00:03:12.160
from that logic tree, we developed
a single-scenario ShakeMap,

00:03:12.160 --> 00:03:15.360
and then we combined those to make
a median ensemble ShakeMap,

00:03:15.360 --> 00:03:20.216
as well as plus or minus 1 or 2 sigma
ensemble ShakeMaps.

00:03:20.240 --> 00:03:23.200
So these ensemble ShakeMaps have
a few important differences from

00:03:23.200 --> 00:03:26.800
traditional scenario ShakeMaps.
They include additional epistemic

00:03:26.800 --> 00:03:30.160
uncertainty, such as considering
different down-dip rupture limits,

00:03:30.160 --> 00:03:33.896
while in a traditional single ShakeMap,
you’d assume a single fault plane.

00:03:33.920 --> 00:03:37.120
In addition, because our ensemble
ShakeMaps are based on a 3D –

00:03:37.120 --> 00:03:40.960
on 3D numerical simulations, that
means that, for periods greater than

00:03:40.960 --> 00:03:44.720
1 second, we should be able to
better capture regionally specific

00:03:44.720 --> 00:03:48.080
3D wave propagation effects that
might not be well-represented

00:03:48.080 --> 00:03:52.080
in ground motion models.
So a good example of this is in the

00:03:52.080 --> 00:03:56.160
Seattle Basin where we know that
simulated ground motions do a better

00:03:56.160 --> 00:04:01.496
job of matching observational data
compared to ground motion models.

00:04:01.520 --> 00:04:04.720
Since I know everyone here is interested
in California, I do want to point out

00:04:04.720 --> 00:04:07.920
something important about
the extent of these ShakeMaps.

00:04:07.920 --> 00:04:12.080
So the M9 Project 3D earthquake
simulations only covered the extent

00:04:12.080 --> 00:04:16.400
that is shown in this green box here.
Our original intent was not to expand

00:04:16.400 --> 00:04:19.040
the ShakeMaps to
a larger geographical extent.

00:04:19.040 --> 00:04:22.720
But, towards the end of our project,
the state agencies and FEMA partners

00:04:22.720 --> 00:04:27.200
that we were working with requested
this larger extent shown in the blue box

00:04:27.200 --> 00:04:31.520
here in order to facilitate state-
and margin-wide risk analyses.

00:04:31.520 --> 00:04:34.448
[dog barking briefly]

00:04:36.080 --> 00:04:39.680
So, to do this, we extended the
simulated ground motions at the

00:04:39.680 --> 00:04:44.000
edges of our model box – again,
shown in green, using a fall-off rate

00:04:44.000 --> 00:04:48.720
that was – a fall-off rate with distance
that was derived from the subduction

00:04:48.720 --> 00:04:53.200
zone ground motion models used in the
2018 National Seismic Hazard Maps.

00:04:53.200 --> 00:04:56.800
So we know that areas outside of
this green box don’t have any

00:04:56.800 --> 00:05:00.080
local 3D wave propagation
effects included, and that is

00:05:00.080 --> 00:05:03.656
an obvious place for
future improvement.

00:05:03.680 --> 00:05:07.440
So, to make these ensemble ShakeMaps,
we had to apply some site corrections.

00:05:07.440 --> 00:05:11.360
So the original M9 Cascadia
simulations did not include site effects.

00:05:11.360 --> 00:05:14.880
So Vs30 in the Puget Lowland
was about 600 meters per second.

00:05:14.880 --> 00:05:18.616
Vs30 elsewhere was
about 800 meters per second.

00:05:18.640 --> 00:05:22.000
So, to model site response,
my collaborator, Alex Grant,

00:05:22.000 --> 00:05:26.960
took 580 Pacific Northwest shear wave
velocity profiles that were collected

00:05:26.960 --> 00:05:30.240
by Sean Ahdi and others for
the NGA subduction project.

00:05:30.240 --> 00:05:33.760
We did remove the profiles from the
Fraser Delta, which were very deep

00:05:33.760 --> 00:05:37.040
and didn’t capture the shallow velocity
gradients that we were interested in

00:05:37.040 --> 00:05:41.680
for the broader Pacific Northwest.
So, using the Ahdi et al. data set,

00:05:41.680 --> 00:05:44.800
Alex then solved for shear wave
velocity as a function of depth

00:05:44.800 --> 00:05:49.496
and Vs30 for all of the shallow
profiles across the Pacific Northwest.

00:05:49.520 --> 00:05:51.760
Some examples of this
are shown in this figure.

00:05:51.760 --> 00:05:57.120
So, for example, for Vs30s between
120 and 160 meters per second,

00:05:57.120 --> 00:06:00.856
the Ahdi et al. profiles
are shown in gray.

00:06:00.880 --> 00:06:03.440
The median of those profiles
are shown in black.

00:06:03.440 --> 00:06:08.320
And our Pacific Northwest model,
based on this equation, is shown in blue.

00:06:08.320 --> 00:06:11.840
So you can see that our model,
the blue line, is doing a good job

00:06:11.840 --> 00:06:15.976
of tracking the median
across a range of Vs30s.

00:06:16.000 --> 00:06:18.800
An equivalent model that was
developed by Shi and Asimaki

00:06:18.800 --> 00:06:22.776
using data from central and southern
California is shown in red.

00:06:22.800 --> 00:06:25.040
And you can see that there’s
a clear regional difference.

00:06:25.040 --> 00:06:28.320
And this is what motivated us to do this
analysis ourselves through the Pacific

00:06:28.320 --> 00:06:34.000
Northwest instead of just borrowing
from the existing work in California.

00:06:34.000 --> 00:06:37.440
So, to estimate site response,
we computed amplification factors

00:06:37.440 --> 00:06:41.920
using equivalent linear analysis and
PI SRA and running 30 examples

00:06:41.920 --> 00:06:45.816
for every input PGA-Vs30
pair in our sample space.

00:06:45.840 --> 00:06:49.920
So the resulting amplification or
de-amplification for all combinations

00:06:49.920 --> 00:06:54.280
is shown here. So, for example,
you can see that, for a site with Vs30

00:06:54.280 --> 00:06:58.160
around 400 meters per second,
the ground motions are typically

00:06:58.160 --> 00:07:03.336
amplified over the original
M9 Project ground motions.

00:07:03.360 --> 00:07:07.920
And sites with very low Vs30,
so less than 200 meters per second,

00:07:07.920 --> 00:07:12.820
often de-amplify due
to nonlinear site behavior.

00:07:13.840 --> 00:07:17.920
So, once we’ve extended the model
region and applied the site corrections,

00:07:17.920 --> 00:07:22.320
this is the resulting median ensemble
ShakeMap based on our 30 M9

00:07:22.320 --> 00:07:26.400
Cascadia earthquake scenarios.
So, in general, we note shaking

00:07:26.400 --> 00:07:31.120
intensity levels around MMI 8,
so severe shaking for coastal areas,

00:07:31.120 --> 00:07:36.400
including in northern coastal California.
This decreases to MMI 6 or 7 –

00:07:36.400 --> 00:07:41.840
so strong or very strong shaking around
the inland I-5 corridor, and to MMI 4

00:07:41.840 --> 00:07:48.752
or 5 in eastern California, and
around MMI 4 in the Bay Area.

00:07:50.080 --> 00:07:53.040
As I mentioned earlier, we also
developed ensemble ShakeMaps

00:07:53.040 --> 00:07:56.960
for the plus or minus 1 or 2 sigma
shaking intensity levels.

00:07:56.960 --> 00:08:01.280
So there’s quite a bit of variability,
but just to focus on California,

00:08:01.280 --> 00:08:06.560
at minus 1 sigma, we see MMI 6 or 7
in coastal northern California,

00:08:06.560 --> 00:08:12.560
MMI 3 in the Bay Area. For the
plus 1 sigma, we see MMI 8 or 9

00:08:12.560 --> 00:08:18.586
around coastal northern California
and MMI 4 or 5 in the Bay Area.

00:08:19.520 --> 00:08:23.200
So how does our median ensemble
ShakeMap compare to the previous

00:08:23.200 --> 00:08:26.960
Cascadia ShakeMap that was based
exclusively on empirical

00:08:26.960 --> 00:08:30.160
ground motion models?
So these maps show the difference

00:08:30.160 --> 00:08:34.400
in MMI, PGA, and PGV.
So red means our ensemble

00:08:34.400 --> 00:08:37.920
ShakeMaps are higher than the
previous GMM-based ShakeMap.

00:08:37.920 --> 00:08:41.096
And blue means our ensemble
ShakeMaps are lower.

00:08:41.120 --> 00:08:46.000
And these plots show the GMMs that
are used to develop the legacy Cascadia

00:08:46.000 --> 00:08:51.280
ShakeMap in orange, the GMMs used
in the 2018 National Seismic Hazard Maps

00:08:51.280 --> 00:08:56.216
in blue, and the gray dots are data
from our ensemble ShakeMap.

00:08:56.240 --> 00:09:00.080
So, for the purpose of this talk,
I’ll just speak about California.

00:09:00.080 --> 00:09:05.760
So, in general, we see higher MMI,
PGA, and PGV in northern coastal

00:09:05.760 --> 00:09:09.840
California with our new ensemble
ShakeMaps and slightly lower shaking

00:09:09.840 --> 00:09:15.016
intensity as we move closer to
the Bay Area, so further south.

00:09:15.040 --> 00:09:18.560
These higher ground motions along
the coast, which occur throughout

00:09:18.560 --> 00:09:22.640
the margin, are due to two things.
So one is the high-stress drop

00:09:22.640 --> 00:09:26.320
sub-events in our rupture model
that place the high-frequency ground

00:09:26.320 --> 00:09:28.720
motions in distinct patches
near the coastline.

00:09:28.720 --> 00:09:32.136
So that contributes to this
increase around the coast.

00:09:32.160 --> 00:09:36.480
But also the coastal ground motions
are also being influenced by our

00:09:36.480 --> 00:09:41.680
consideration of multiple M9 Cascadia
earthquake scenarios, 30% of which had

00:09:41.680 --> 00:09:46.320
a down-dip rupture limit that is deeper
than that used for the legacy ShakeMap.

00:09:46.320 --> 00:09:49.840
And this was following the
weighting of the national maps.

00:09:49.840 --> 00:09:54.376
The lower shaking intensities as you
move further south in California,

00:09:54.400 --> 00:09:57.840
we think that’s because the GMMs
are based on the closest distance

00:09:57.840 --> 00:10:01.200
to the fault. So they might be
underestimating ground motions

00:10:01.200 --> 00:10:04.640
towards the middle of the fault where
sites are likely to see incoming waves

00:10:04.640 --> 00:10:09.360
from a larger portion of the rupture.
Conversely, the GMMs might be

00:10:09.360 --> 00:10:12.960
overestimating ground motions towards
the end of the fault where sites are

00:10:12.960 --> 00:10:15.520
impacted by a smaller
portion of the rupture.

00:10:15.520 --> 00:10:19.099
And our 3D simulations
would have captured that.

00:10:19.680 --> 00:10:23.360
So I think there’s still a lot of open
questions related to this work and to

00:10:23.360 --> 00:10:27.896
ground shaking in California from
Cascadia subduction zone earthquakes.

00:10:27.920 --> 00:10:31.600
So one open question is, how does 3D
wave propagation in the Central Valley

00:10:31.600 --> 00:10:35.120
impact ground shaking from Cascadia
subduction zone earthquakes?

00:10:35.120 --> 00:10:38.640
So recall that the 3D earthquake
simulations didn’t extend south

00:10:38.640 --> 00:10:40.400
of the Mendocino
Triple Junction.

00:10:40.400 --> 00:10:45.576
So any local 3D wave propagation
effects in California were not included.

00:10:45.600 --> 00:10:48.560
Another open question is,
how do differences in near-surface

00:10:48.560 --> 00:10:51.920
geology along the Cascadia
margin impact site effects?

00:10:51.920 --> 00:10:55.360
So here we used all the Pacific
Northwest velocity profiles

00:10:55.360 --> 00:10:58.400
in aggregate. But in some work
we’re doing now, we’re instead

00:10:58.400 --> 00:11:02.960
breaking those profiles up into geologic
regions such as those profiles in the

00:11:02.960 --> 00:11:07.040
Seattle Basin, in the Puget Lowland,
those on hard rock, and you can

00:11:07.040 --> 00:11:12.296
imagine that there might be along-strike
differences in Vs profiles as well.

00:11:12.320 --> 00:11:15.040
Another important question is,
what is the range of ground shaking

00:11:15.040 --> 00:11:18.400
from partial margin ruptures?
So here I only talked about M9

00:11:18.400 --> 00:11:22.320
Cascadia subduction zone earthquakes,
but this is a question that we’re about

00:11:22.320 --> 00:11:25.440
to start working on as part of
a new NSF-funded project,

00:11:25.440 --> 00:11:29.600
which I’ll talk about in a moment.
And then also, as part of that project,

00:11:29.600 --> 00:11:33.520
we’re going to be looking at, how does
ground shaking, tsunami inundation,

00:11:33.520 --> 00:11:36.880
and ground failure from Cascadia
subduction zone earthquakes

00:11:36.880 --> 00:11:41.600
interact in space and time?
So, for my last slide, I just want to

00:11:41.600 --> 00:11:45.040
briefly mention this newly funded
project by the National Science

00:11:45.040 --> 00:11:48.856
Foundation called the Cascadia
Coastlines and Peoples Hub.

00:11:48.880 --> 00:11:53.600
So the CoPe Hub is a five-year,
$19 million project to improve coastal

00:11:53.600 --> 00:11:57.920
resilience to hazards in the Pacific
Northwest through targeted scientific

00:11:57.920 --> 00:12:01.440
advances and modeling, which are
co-produced in collaboration with

00:12:01.440 --> 00:12:04.560
impacted coastal communities.
And you can see all the range

00:12:04.560 --> 00:12:08.536
of institutions that are involved
in this project down here.

00:12:08.560 --> 00:12:13.280
So a big part of this project includes
modeling of tectonic geohazards,

00:12:13.280 --> 00:12:17.840
both regionally and with
a specific focus on a few specific

00:12:17.840 --> 00:12:19.920
coastal communities that
are shown in this map,

00:12:19.920 --> 00:12:24.376
so including Humboldt County
in northern California.

00:12:24.400 --> 00:12:27.360
So some of the modeling that’s
going to be coming out of this project

00:12:27.360 --> 00:12:31.600
includes coupled Cascadia subduction
zone earthquake-tsunami simulations.

00:12:31.600 --> 00:12:33.840
These will be kinematic,
but they will be using the same

00:12:33.840 --> 00:12:36.560
earthquake sources for
both the earthquake and

00:12:36.560 --> 00:12:39.896
tsunami simulations, which
we haven’t done previously.

00:12:39.920 --> 00:12:43.200
We’re also going to be looking at
a wider range of magnitudes based on

00:12:43.200 --> 00:12:47.016
the logic trees that are in the
National Seismic Hazard Maps.

00:12:47.040 --> 00:12:49.840
And other researchers will be
looking at multi-hazard impact,

00:12:49.840 --> 00:12:53.280
so shaking and inundation
on coastal infrastructure, as well as

00:12:53.280 --> 00:12:56.936
tsunami debris forecasting and
vulnerability assessments.

00:12:56.960 --> 00:12:59.520
And so we look forward to
sharing some of these results

00:12:59.520 --> 00:13:03.176
with you as they come out
over the next few years.

00:13:03.200 --> 00:13:05.680
So that’s all I have.
Thanks for your time.