WEBVTT
Kind: captions
Language: en-US

00:00:01.360 --> 00:00:04.560
Hello. Thank you for the
opportunity to speak here today.

00:00:04.560 --> 00:00:07.840
Looking forward to telling you
about our project looking at

00:00:07.840 --> 00:00:11.120
aseismic deformation along the
central San Andreas and Calaveras

00:00:11.120 --> 00:00:15.520
Faults from repeat differencing
of airborne Lidar topography.

00:00:15.520 --> 00:00:18.880
This is a project I’ve been
working on with my collaborators,

00:00:18.880 --> 00:00:22.080
Steve DeLong and
Ramon Arrowsmith.

00:00:22.080 --> 00:00:25.440
By way of introduction, measuring
creep along the San Andreas Fault

00:00:25.440 --> 00:00:28.960
system indicates the moment
accumulation that is available

00:00:28.960 --> 00:00:33.576
for destructive earthquakes
versus as less-damaging creep.

00:00:33.600 --> 00:00:38.080
In this project, we’ve constrained
the aseismic fault creep rate along

00:00:38.080 --> 00:00:43.360
the central San Andreas and Calaveras
Faults by differencing repeat airborne

00:00:43.360 --> 00:00:47.360
Lidar topography data sets.
We constrain spatially dense

00:00:47.360 --> 00:00:52.080
creep rates every 400 meters and
examine approaches for decreasing

00:00:52.080 --> 00:00:56.960
noise in the resulting creep rates.
We show that portions of the central

00:00:56.960 --> 00:01:01.360
San Andreas Fault creep at over
rates of 30 millimeters per year.

00:01:01.360 --> 00:01:04.080
And this indicates limited
strain accumulation.

00:01:04.080 --> 00:01:06.000
Finally, we show,
at Mustang Ridge,

00:01:06.000 --> 00:01:11.600
the active fault trace is offset from
the geomorphically mapped fault.

00:01:11.600 --> 00:01:17.600
This shows a map of central California.
Here is the San Andreas Fault

00:01:17.600 --> 00:01:23.040
and the Calaveras Fault.
The color of both faults represent

00:01:23.040 --> 00:01:27.360
the creep rate determined from
Lidar topographic differencing.

00:01:27.360 --> 00:01:29.520
[sounds of birds in background]
Slip rate has been measured

00:01:29.520 --> 00:01:33.360
with a variety of instrumentation,
however this instrumentation

00:01:33.360 --> 00:01:40.240
typically lacks sensitivity to
the dense 3D deformation field.

00:01:40.240 --> 00:01:44.000
Lidar topographic differencing
can fill this gap, as it resolves

00:01:44.000 --> 00:01:48.376
deformation along and
adjacent to active faults.

00:01:48.400 --> 00:01:51.840
One application that we’re particularly
excited about is performing

00:01:51.840 --> 00:01:57.336
a slip inversion to solve for
creep along strike and with depth.

00:01:57.360 --> 00:01:59.360
To perform the 3D
topographic differencing,

00:01:59.360 --> 00:02:03.576
we use the iterative closest point,
or ICP, algorithm.

00:02:03.600 --> 00:02:08.320
This approach has been applied to
many earthquakes over the past decade.

00:02:08.320 --> 00:02:13.520
Originally, it gained a lot of attention
because it could resolve deformation

00:02:13.520 --> 00:02:17.200
along an adjacent active fault,
where other geodetic data sets

00:02:17.200 --> 00:02:19.816
typically lack
spatial resolution.

00:02:19.840 --> 00:02:25.336
The first step is to take a windowed
subset of the Lidar topography.

00:02:25.360 --> 00:02:27.840
This window – the length of
the window is the resolution,

00:02:27.840 --> 00:02:32.136
which is typically on the
order of 30 to 50 meters.

00:02:32.160 --> 00:02:37.120
The next step is to apply
a 3D rigid body deformation,

00:02:37.120 --> 00:02:43.120
which is used to align the point clouds.
And the rotation and displacement

00:02:43.120 --> 00:02:50.000
that’s required to align these data sets
is soon to then be the 3D deformation

00:02:50.000 --> 00:02:55.336
that occurred either during the
earthquake or from fault creep.

00:02:55.360 --> 00:03:00.480
And the final step is to calculate
uncertainty on the 3D displacement.

00:03:00.480 --> 00:03:03.040
For the differencing,
we used four data sets.

00:03:03.040 --> 00:03:10.320
This includes the 2005 B4,
2007 EarthScope, 2018 Salinas,

00:03:10.320 --> 00:03:15.280
and Parkfield. Here are abbreviated
acknowledgements for these data sets.

00:03:15.280 --> 00:03:21.336
I invite you to look at our
publication for the full citations.

00:03:21.360 --> 00:03:25.920
I would like to acknowledge
the foresight and coordination

00:03:25.920 --> 00:03:31.120
that it took to collect, particularly,
the B4 and EarthScope data sets

00:03:31.120 --> 00:03:34.320
over a decade ago.
And, critical to performing this

00:03:34.320 --> 00:03:40.160
differencing was collecting
both the 2018 Parkfield

00:03:40.160 --> 00:03:45.840
and the 2018 Salinas data set.
One of the challenges of performing

00:03:45.840 --> 00:03:52.000
differencing is that a decade or so
of creep produce significantly less slip

00:03:52.000 --> 00:03:56.720
than the magnitude 7 or so earthquakes
that this method has been applied to.

00:03:56.720 --> 00:04:00.560
So we looked at several approaches
for decreasing noise and ultimately

00:04:00.560 --> 00:04:03.816
being able to better resolve
slip along the fault.

00:04:03.840 --> 00:04:08.560
A primary source of noise was
changes in vegetation between

00:04:08.560 --> 00:04:11.840
the two acquisitions.
To mitigate the source of noise,

00:04:11.840 --> 00:04:16.080
we used ground points only.
There were a number of landslides

00:04:16.080 --> 00:04:20.240
along the central San Andreas Fault.
For example, here you can see,

00:04:20.240 --> 00:04:24.320
in the topographic hill shade,
landslide in this area.

00:04:24.320 --> 00:04:28.720
And here is the displacement field,
in this case, in the northeast,

00:04:28.720 --> 00:04:35.280
or fault-perpendicular, direction.
We see a very clear landslide signal

00:04:35.280 --> 00:04:42.160
here. Ultimately, masked areas that
had these very pronounced landslides.

00:04:42.160 --> 00:04:47.920
And finally, flight line errors caused
dominant source of noise in the

00:04:47.920 --> 00:04:53.200
northeast and vertical directions.
And so we focused our interpretation

00:04:53.200 --> 00:04:58.136
on the northwest, or fault-parallel,
displacement signals.

00:04:58.160 --> 00:05:00.880
So here is a resulting
displacement field.

00:05:00.880 --> 00:05:04.936
This is showing the northwest,
or fault-parallel, displacement.

00:05:04.960 --> 00:05:09.280
And we see here, along the southwest
side of the fault, that the area is

00:05:09.280 --> 00:05:14.240
moving positive, or to the northwest.
On the opposite side of the fault,

00:05:14.240 --> 00:05:17.840
you see the opposite sense of
motion with a clear,

00:05:17.840 --> 00:05:22.800
very sharp boundary along the fault.
The next thing we’re interesting in

00:05:22.800 --> 00:05:27.496
was calculating the fault creep
from the displacement field.

00:05:27.520 --> 00:05:32.640
And that is shown here in this cartoon,
where we essentially do a displacement

00:05:32.640 --> 00:05:38.696
discontinuity approach.
So we take the – calculate the average

00:05:38.720 --> 00:05:44.080
displacement on each side of the fault,
weighted by the error in the

00:05:44.080 --> 00:05:48.056
displacement in each of the
displacement measurements.

00:05:48.080 --> 00:05:53.496
And then, to calculate the fault creep,
we subtract, or find the difference,

00:05:53.520 --> 00:05:58.240
in creep across the fault.
The uncertainty in the creep rate

00:05:58.240 --> 00:06:04.000
is the second norm of the
weighted uncertainty across

00:06:04.000 --> 00:06:10.856
each side of the fault.
And then, to calculate the creep rate,

00:06:10.880 --> 00:06:14.800
we divide the – or, we take the
fault creep and we divide that

00:06:14.800 --> 00:06:20.080
by the time between
the Lidar acquisitions.

00:06:20.080 --> 00:06:25.200
So we apply this method to the
entire fault to calculate the creep rate,

00:06:25.200 --> 00:06:28.880
and the results are shown here.
This bottom plot shows distance

00:06:28.880 --> 00:06:33.200
northeast versus distance from
Parkfield. And the color that you’re

00:06:33.200 --> 00:06:39.176
seeing is the northwest, or on average,
right-lateral displacements.

00:06:39.200 --> 00:06:45.096
In the top plot, we are showing
creep rate in millimeters per year.

00:06:45.120 --> 00:06:49.440
And I’d like you to pay attention to
the blue, which is the right-lateral

00:06:49.440 --> 00:06:53.440
San Andreas Fault creep rate,
as well as the green, which is the

00:06:53.440 --> 00:06:58.320
right-lateral Calaveras Fault creep rate.
I will discuss some of the key features

00:06:58.320 --> 00:07:02.696
that we can observe from
this pattern of fault creep.

00:07:02.720 --> 00:07:06.000
First order, the creep
rate is asymmetric.

00:07:06.000 --> 00:07:12.000
We see higher creep rates here along
the southeastern side of the creeping

00:07:12.000 --> 00:07:16.000
portion of the San Andreas Fault.
The highest creep rate of about

00:07:16.000 --> 00:07:22.616
40 millimeters per year occurs near
Slack Canyon in this area of the fault.

00:07:22.640 --> 00:07:28.720
As we go towards Parkfield, there’s a
very sharp decrease in creep rate –

00:07:28.720 --> 00:07:35.040
a rate of about 8 millimeters – so, a rate
of about, yeah, 8 millimeters per year.

00:07:35.040 --> 00:07:40.640
At Dry Lake Valley, which is here,
we see a rate of about 20 millimeters

00:07:40.640 --> 00:07:45.816
per year. And this is a local
minimum in the creep rate.

00:07:45.840 --> 00:07:49.680
And here, where the San Andreas
and Calaveras Faults overlap,

00:07:49.680 --> 00:07:56.296
we see that there is a sharp decrease in
creep rate along the San Andreas Fault,

00:07:56.320 --> 00:08:00.800
likely relating to the partitioning
of the plate boundary deformation

00:08:00.800 --> 00:08:03.656
along those
two active faults.

00:08:03.680 --> 00:08:07.040
We then asked the question of whether
the topographic differencing rates

00:08:07.040 --> 00:08:12.320
are similar to the geodetic rates.
The conclusion is yes, when the

00:08:12.320 --> 00:08:16.216
difference in aperture of the
instrumentation is accounted for.

00:08:16.240 --> 00:08:20.320
This table shows the other geodetic
rates that we compared to the

00:08:20.320 --> 00:08:24.773
topographic differencing from a
creepmeter with a 30-meter aperture

00:08:24.805 --> 00:08:30.456
to Geodolites with apertures
typically greater than 15 kilometers.

00:08:30.480 --> 00:08:34.400
This plot shows creep rate
versus distance from Parkfield.

00:08:34.400 --> 00:08:38.400
And, for the most part, we see
that the topographic differencing,

00:08:38.400 --> 00:08:42.536
shown in black, are consistent
with the other measurements.

00:08:42.560 --> 00:08:47.600
Largest difference is the Geodolite
instrumentation, but likely those

00:08:47.600 --> 00:08:52.376
data sets are larger because of their
larger aperture, which means

00:08:52.400 --> 00:08:58.376
these instruments are measuring
elastic strain accumulation.

00:08:58.400 --> 00:09:01.680
This plot shows topographic
differencing in creep rate versus

00:09:01.680 --> 00:09:07.416
creep rate from other data sets.
This is the 1-to-1 line.

00:09:07.440 --> 00:09:10.400
There is a aperture dependence
on the agreement between the

00:09:10.400 --> 00:09:14.800
creep rate from the different data sets.
For example, the topographic

00:09:14.800 --> 00:09:19.496
differencing rates are higher
than all of the creepmeter rates.

00:09:19.520 --> 00:09:23.920
And, in contrast, the topographic
differencing rates are lower than

00:09:23.920 --> 00:09:28.640
many of the Geodolite
[audio cuts out].

00:09:28.640 --> 00:09:32.320
We then looked at Mustang Ridge
to learn what topographic differencing

00:09:32.320 --> 00:09:35.680
indicates in regions of
complex deformation.

00:09:35.680 --> 00:09:39.120
This region was first mapped by
Rymer et al., who interpreted

00:09:39.120 --> 00:09:43.680
distributed faulting as being responsible
for creep rates that vary with aperture.

00:09:43.680 --> 00:09:48.160
It was then mapped by DeLong et al.,
who interpreted that this distributed

00:09:48.160 --> 00:09:52.960
faulting here produces an
en échelon pattern of faults.

00:09:52.960 --> 00:09:57.360
Topographic differencing is consistent
and adds to these interpretations.

00:09:57.360 --> 00:10:01.840
First, we see that, although there is
about 10 kilometers of along-strike

00:10:01.840 --> 00:10:07.896
en échelon faults in the
geomorphology, the active deformation

00:10:07.920 --> 00:10:13.440
spans about 4 kilometers along strike.
These black and red lines indicate

00:10:13.440 --> 00:10:18.240
the active portion of the fault
mapped by DeLong et al. as well as

00:10:18.240 --> 00:10:23.280
the USGS and CGS.
However, topographic differencing

00:10:23.280 --> 00:10:28.640
shows that the active fault trace
is not here, but instead along this

00:10:28.640 --> 00:10:34.056
portion of the fault, is located about
half a kilometer to the northeast.

00:10:34.080 --> 00:10:37.760
So this shows that topographic
differencing is very important

00:10:37.760 --> 00:10:43.256
at indicating which faults in
the geomorphology are active.

00:10:43.280 --> 00:10:44.960
Using the topographic
differencing results,

00:10:44.960 --> 00:10:47.976
we can look at
off-fault deformation.

00:10:48.000 --> 00:10:52.320
At Dry Lake Valley, the deformation
is highly localized to the fault.

00:10:52.320 --> 00:10:57.360
Consistent with this current work,
Scott et al., in a JGR paper

00:10:57.360 --> 00:11:02.240
from last year, also showed
that deformation is highly localized

00:11:02.240 --> 00:11:04.696
to the fault at
Dry Lake Valley.

00:11:04.720 --> 00:11:08.720
And this was shown by comparing
surface fractures to topographic

00:11:08.720 --> 00:11:14.480
differencing rates. And the localization
of deformation was attributed to

00:11:14.480 --> 00:11:19.656
the fault geometry, which is very
well-aligned for right-lateral slip.

00:11:19.680 --> 00:11:22.800
At Mustang Ridge, as we
discussed on the last slide,

00:11:22.800 --> 00:11:28.800
deformation is highly distributed.
Along the central creeping San Andreas

00:11:28.800 --> 00:11:33.760
Fault, we conclude that 20 to 50%
of the deformation is accommodated

00:11:33.760 --> 00:11:37.920
beyond a 30-meter aperture.
This amount of high off-fault

00:11:37.920 --> 00:11:44.296
deformation is highly unexpected
for a mature and non-creeping fault.

00:11:44.320 --> 00:11:49.016
We do see this clear
dependence on fault geometry.

00:11:49.040 --> 00:11:53.416
In addition, it’s possible that
the high off-fault deformation

00:11:53.440 --> 00:11:56.960
reflects a difference in
mechanics dynamics

00:11:56.960 --> 00:12:01.176
between a creeping
and seismogenic fault.

00:12:01.200 --> 00:12:04.320
I will now discuss applications
and future work of these

00:12:04.320 --> 00:12:08.391
topographic differencing results.
We are particularly interested in

00:12:08.391 --> 00:12:12.560
looking at strain accumulation.
For example, by conducting a joint

00:12:12.560 --> 00:12:17.360
slip inversion to quantify fault slip
and depth-dependent changes in slip,

00:12:17.360 --> 00:12:20.320
and topographic differencing
would add a very important

00:12:20.320 --> 00:12:23.416
intermediate-scale aperture.

00:12:23.440 --> 00:12:27.600
We’re also interested in mapping
active fault traces and comparing

00:12:27.600 --> 00:12:33.096
the location of the actively creeping
fault to Quaternary fault traces.

00:12:33.120 --> 00:12:37.280
These displacements could also
be used to better understand

00:12:37.280 --> 00:12:42.160
fault displacement hazard.
And finally, the fourth is application is

00:12:42.160 --> 00:12:47.256
looking at the mechanical properties
of the crust, sediment cover, and soil.

00:12:47.280 --> 00:12:52.945
For example, does shallow rheology
control the distribution of deformation?

00:12:54.480 --> 00:12:58.880
So, in conclusion today,
shown that we have constrained

00:12:58.880 --> 00:13:02.880
creep rates at a dense spatial resolution
along the central San Andreas

00:13:02.880 --> 00:13:07.576
and Calaveras Faults from
Lidar topographic differencing.

00:13:07.600 --> 00:13:12.776
The topographic differencing rates are
consistent with other geodetic data sets,

00:13:12.800 --> 00:13:18.616
particularly given the different
apertures of the instruments.

00:13:18.640 --> 00:13:23.976
Discontinuities in the displacement
fields indicate the active fault trace.

00:13:24.000 --> 00:13:27.440
We see that, at Mustang Ridge,
the active fault trace is not always

00:13:27.440 --> 00:13:34.936
consistent with the location of the fault
mapped by the USGS and the CGS.

00:13:34.960 --> 00:13:39.200
And finally, 20 to 50% of
the deformation occurs

00:13:39.200 --> 00:13:44.560
off the principal fault. So thank you
for hearing this presentation today.

00:13:44.560 --> 00:13:51.656
I’d like to acknowledge funding from
the NSF, ASU, NEHRP, and NASA.

00:13:51.680 --> 00:13:57.016
And we are very excited
to continue this work into 2021

00:13:57.040 --> 00:14:00.880
with funding from NEHRP.
Thank you.