WEBVTT

00:00:03.000 --> 00:00:12.000
Hello and welcome back everyone and welcome back to the "Too many earthquake anniversaries" edition because you know what happened this year?

00:00:12.000 --> 00:00:13.000
Lots of interesting new earthquakes, including a very important sequence in Turkey and a very important sequence in Afghanistan.

00:00:13.000 --> 00:00:23.000
Do both of them dissolve their own session to look at them in depth? Yes, yes they do.

00:00:23.000 --> 00:00:29.000
Do we have time for that? No, no we do not. So instead, we are going to get 

00:00:29.000 --> 00:00:39.000
an excellent introductory talk on each of those topics, a doublet of introductory talks, if you will, from Nadine Reitman and Richard Walker.

00:00:39.000 --> 00:00:42.000
And then we're going to talk about what do we know about earthquake sequences? We're going to talk about what it means from an earthquake side.

00:00:42.000 --> 00:00:51.000
From the ground motion side, from a hazards side, and from a risk side. And to take us on this.

00:00:51.000 --> 00:01:03.000
multiplicative journey of multiplets we have the unmatched moderators of Noha Farghal and Kevin Milner. [woohoo, applause]

00:01:03.000 --> 00:01:12.000
Thanks, Sarah. So yeah, I think this is gonna be a really exciting session. Our second to last one as we bring the

00:01:12.000 --> 00:01:19.000
workshop to a close. So I'm Kevin Milner with the USGS Geological Hazard Science Center.

00:01:19.000 --> 00:01:24.000
I'll be co-moderating with Noha Farghal at Moody's RMS.

00:01:24.000 --> 00:01:41.000
We'll just jump straight into it. The first talk is from Nadine Reitman on the 2023 Turkey earthquakes, response, surface rupture, and hazard implications. Take it away.

00:01:41.000 --> 00:01:49.000
Thank you for inviting me to give this talk. My name is Nadine Reitman and I'm a research geologist at the USGS Geologic Hazard Science Center in Golden, Colorado.

00:01:49.000 --> 00:01:57.000
I'm giving a talk today about the 2023 earthquakes in southeastern Turkey in terms of our response surface rupture and some hazard implications.

00:01:57.000 --> 00:02:05.000
And this work is truly a collaboration between many, many co-authors whose names you'll see later in this presentation.

00:02:05.000 --> 00:02:13.000
The outline follows the title of this talk with a quick background on the earthquakes and then moves into our response in terms of rapid satellite mapping.

00:02:13.000 --> 00:02:23.00n
Talks about the Narli fault story in terms of surface rupture and then moves into northern California earthquake hazard parallels.

00:02:23.000 --> 00:02:29.000
So you're probably familiar with the earthquakes that happened on February 6, 2023 in southeastern Turkey.

00:02:29.000 --> 00:02:37.000
The magnitude 7.8 Pazarcik earthquake was followed 9 hours later by the magnitude 7.5 Elbistan earthquake.

00:02:37.000 --> 00:02:48.000
The Pazarcik earthquake nucleated southeast of the East Anatolian fault ruptured northward on the Narli fault and then had bilateral rupture on the East Anatolian fault for about 345km 

00:02:48.000 --> 00:03:04.000
with a maximum displacement of about 7m. The 7.5 Elbistan earthquake nucleated in the center of the Çardak fault and ruptured bilaterally for about 175km with maximum displacement of about 8.5m.

00:03:04.000 --> 00:03:15.000
These earthquakes were devastating for the region with more than 50,000 deaths, over 200,000 buildings collapse or substantially damaged and more than $34 billion in damages.

00:03:15.000 --> 00:03:24.000
In terms of regional tectonics, the East Anatolian fault is a major plate boundary fault that accommodates westward extrusion of the Anatolian block from Arabian Eurasian collision.

00:03:24.000 --> 00:03:31.000
The East Anatolian fault has a slip rate that decreases from about 10 to 4mm per year from east to west.

00:03:31.000 --> 00:03:37.000
And it has a history of early historic earthquakes that range in magnitude from about 6.8 to 7.2.

00:03:37.000 --> 00:03:43.000
Prior to the 2023 earthquakes the largest earthquake considered for hazard was magnitude 7.5.

00:03:43.000 --> 00:03:55.000
Response used high-resolution worldview satellite images to rapidly map detailed surface rupture. This included a large team of people mapping in the office listed over here.

00:03:55.000 --> 00:04:03.000
Remote sensing help from USGS colleagues and field correspondents Sinan Akciz (CSU Fullerton) and Rich Koehler (UNR, Reno).

00:04:03.000 --> 00:04:11.000
Our motivation for rapid satellite mapping is that earthquake surface ruptures are perishable data that can quickly be erased from the landscape by both human and natural causes.

00:04:11.000 --> 00:04:14.000
For example, road repair precipitation and rebuilding can obscure the location, size, and character of surface ruptures and displacements.

00:04:14.000 --> 00:04:28.000
And in Turkey, road repair started within hours to days after the earthquake, and precipitation in the form of snow and clouds obscured the surface rupture in some of the satellite images.

00:04:28.000 --> 00:04:35.000
Furthermore, at long surface ruptures may be difficult to rapidly map on the ground due to damaged infrastructure and ongoing humanitarian crises.

00:04:35.000 --> 00:04:44.000
And if the earthquake occurred off of the map fault, such as the nucleation of the 7.8 on the Narli fault, the location of the surface rupture may be unknown initially.

00:04:44.000 --> 00:04:57.000
Thus a combined response including both remote and on the ground analyses is typically needed to quickly document surface rupture parameters such as fault length, displacement magnitude, complexity, and distribution of surface faulting.

00:04:57.000 --> 00:05:13.000
I'm now going to step through the first week of our rapid response. On the first day after the earthquake February 7th worldview and GOI images became available shown by these gray boxes and we were able to map about 25km of discontinuous surface structure on the eastern Anatolian fault.

00:05:13.000 --> 00:05:18.000
Though clouds obscured some of the rupture, seeing in red here.

00:05:18.000 --> 00:05:28.000
Also on this first day we were able to identify the first offset features, which is a road offset about 1.5m on the eastern Anatolian fault.

00:05:28.000 --> 00:05:34.000
On the second day after the earthquake on February 8th, the first ALOS interferometric pair was processed.

00:05:34.000 --> 00:05:37.000
And pixel correlation of range and azimuth offsets image portions of both the EAF and Çardak faults.

00:05:37.000 --> 00:05:57.000
Though both ruptures clearly continue beyond the scene extent of this ALOS swath. Also on the second day after the rupture, we tweeted the initial observations from the USGS Quake's account, including both the ALOS satellite data as well as the initial mapping and surface structure observations.

00:05:57.000 --> 00:06:03.000
Although these data got out to the public quickly, they weren't downloadable in their native format.

00:06:03.000 --> 00:06:09.000
On February 9th, we continued mapping, detailed surface structures seen in red on more worldview images that came available.

00:06:09.000 --> 00:06:21.000
And we started defining what we called a simple fault trace based on the A-list data. On February 10th, 4 days after the earthquake, Sentinel1 enabled complete coverage of both earthquakes from these swaths.

00:06:21.000 --> 00:06:32.000
The ruptured traces mapped from Sentinel1 pixel correlation, image about 340km of rupture on the EAF and about 145km of rupture on the Çardak fault.

00:06:32.000 --> 00:06:39.000
Although the two ruptures do not appear to connect. On February 10th this is what the extent of our mapping looked like; 

00:06:39.000 --> 00:06:47.000
we now have a simple fault trace outlined in black for both ruptures as well as a bit more detailed mapping from worldview images.

00:06:47.000 --> 00:06:55.000
Also on February 10th we released the mapping to the public via an interactive web map where users can download the data in their native formats.

00:06:55.000 --> 00:07:06.000
This is a first for USGS earthquake response and we were able to get these data out quickly under emergency provisional authority with a plan to release them as a peer reviewed USGS data release in the future, 

00:07:06.000 --> 00:07:15.000
which is now available at this QR code. These data were then updated daily or weekly as needed as new images and data came in.

00:07:15.000 --> 00:07:33.000
It was clear from the beginning that rapidly releasing these data to the public was very useful. These data were used not just internally by the USGS National Earthquake Information Center response products such as ShakeMap, Pager and the finite fault model, but they also guided teams on the ground such as GEER teams and others doing field work.

00:07:33.000 --> 00:07:34.000
They were used in early versions of dynamic rupture models, slip distributions, and stress drop models, 

00:07:34.000 --> 00:07:46.000
and they were picked up by news organizations. Additionally, they allow analysis of where lifelines may intersect with surface rupture 

00:07:46.000 --> 00:07:51.000
and down the line, they may be useful in probabilistic fault displacement hazard analysis.

00:07:51.000 --> 00:08:05.000
One week after the earthquake on February 13th, this is what coverage looked like. We had about two, over three of the EAF rupture mapped in detail shown in red and we clearly lacked a lot of coverage on the Çardak fault with only a bit of coverage on the western end, 

00:08:05.000 --> 00:08:16.000
and this situation didn't change really until the end of March. So from this summary you can see that the coverage within the first week covered about two-thirds of the EAF.

00:08:16.000 --> 00:08:24.000
And then it took a long time for the rest of those holes to be filled in and we capped our project at the end of March with no further data updates.

00:08:24.000 --> 00:08:41.000
Another piece that was critical to our response was being in contact with people on the ground in the field. Thanks to Sinan Akciz, we were able to define the eastern end of the Çardak rupture where there was no satellite imagery and the pixel correlation data were noisy and decorrelated.

00:08:41.000 --> 00:08:53.000
So ground reconnaissance was necessary. On the western end of the Çardak fault, there was satellite coverage which helped guide Sinan to this rupture in a snowy field which otherwise would be difficult to see.

00:08:53.000 --> 00:09:07.000
And similarly, this is in a place in the radar data where there is no signal. Putting these data sets together, the high-res optical, the field data and the radar pixel correlation were all required to define the rupture and points and lengths,

00:09:07.000 --> 00:09:16.000
which led to about 345km of primary rupture on the East Anatolian Fault and 175km of primary rupture on the Çardak fault.

00:09:16.000 --> 00:09:32.000
These earthquakes are two of the largest recorded continental strike-slip ruptures. Shown here in context of the other Wells and Coppersmith datasets in terms of surface structure length and maximum displacement relative to magnitude, they plot within the cloud of large strike slip earthquakes.

00:09:32.000 --> 00:09:42.000
Since the data were originally released in emergency provisional format, after the rapid mapping phase was complete, we released them as a peer-reviewed USGS science-based data release with versioning.

00:09:42.000 --> 00:09:50.000
We will soon be releasing an update of this data release, and we also published a TSR manuscript on this process.

00:09:50.000 --> 00:09:53.000
So now we're going to switch gears and talk about the Narli fault and a remote and on the ground quest to define the Narli-EAF intersection 

00:09:53.000 --> 00:10:08.000
and this part of the talk also represents the work of many, many people listed here. It was initially unclear what fault the magnitude 7.8 epicenter occurred on.

00:10:08.000 --> 00:10:19.000
It appeared to be on what was mapped as a discontinuous set of faults at the extreme northern end of the Dead Sea fault, known as the Narla Fault Zone, seen here in the red circle.

00:10:19.000 --> 00:10:27.000
The first Sentinel1 images that came in hinted at a north-south oriented Narli fault seen here in this pixel tracking result.

00:10:27.000 --> 00:10:36.000
Worldview images and a GEER field team documented about 15 kilometers of surface structure on the southern portion of the Narli fault seen here in this worldview image, 

00:10:36.000 --> 00:10:43.000
and a Canadian field team found possible surface structure near the town of Tetirlik, which we later confirmed on Worldview images, 

00:10:43.000 --> 00:10:48.000
and this is east of the proposed north-south orientation of the northern Narli.

00:10:48.000 --> 00:11:00.000
So there are two possibilities for the northern Narli fault orientation and its intersection with the EAF, and this question of which one of these the rupture took can't be resolved remotely and requires fieldwork.

00:11:00.000 --> 00:11:30.000
Our team did fieldwork in June 2023 and documenting this intersection became one of our primary targets. Fieldwork focused on reconnaissance, surface rupture mapping, displacement measurements along both the EAF and the northern Narli fault to help define this intersection.

00:11:37.000 --> 00:11:46.000
And here we are showing the remote surface rupture mapping in black. Field surface rupture mapping in blue and areas where we documented no rupture in red. We spent quite a bit of time walking around the northern extent of this north-south orientation to see if we could see rupture here, which we did not. Here is some of the surface rupture that we did find on the northern Narli fault.
We documented primarily narrow fault zones with one or two primary fault strands, while continuous in the aggregate they tended to be stepping on the ground.

00:11:46.000 --> 00:11:56.000
And we only found evidence of lateral rupture here, no vertical signal. This is an offset hillside which is why it looks like it has a vertical start.

00:11:56.000 --> 00:12:04.000
We also measured displacements along the northern Narli and EAF. Here we see a rock wall that's offset about half a meter measured in the field.

00:12:04.000 --> 00:12:09.000
When we look at this same rock wall in the worldview imagery, it's ambiguous whether or not there is offset there and if it's measurable.

00:12:09.000 --> 00:12:19.000
And similarly in optical pixel correlation of planet data. It's too noisy to define a displacement on the Narli fault.

00:12:19.000 --> 00:12:33.000
So these small offsets require fieldwork to determine their magnitude and location. Based on our  fieldwork with mapping surface rupture and displacement measurements, we indicate that the northeastern section of the Narli fault is the one that ruptured.

00:12:33.000 --> 00:12:45.000
This is based both on finding surface rupture along that northeastern portion of the Narli fault, as well as correlating the intersection of the Narli fault with changes in displacement along the East Anatolian fault.

00:12:45.000 --> 00:12:53.00
Pixel correlation of planet and Sentinel2 to optical images also support the Northeastern Narli fault orientation and its intersection with the EAF.

00:12:53.000 --> 00:13:05.000
The planet optical pixel correlation here shows a zone of distributed and diffuse faulting before the Narli fault intersects with the EAF and then becomes a zone of localized displacement after the intersection.

00:13:05.000 --> 00:13:18.000
Similarly in the Sentinel2 correlation, it's a bit noisier, but there's either a very well correlated zone of noise or a potential Narli fault projection here.

00:13:18.000 --> 00:13:30.000
The remote mapping field mapping, field displacement measurements and optical pixel correlation together constrain the northern Narli fault location to the northeastern orientation seen here, as well as its intersection with the EAF up here.

00:13:30.000 --> 00:13:44.000
We're updating the data release with this northeastern Narli fault location. However, there is a zone in here where we weren't able to document surface rupture on the northern Narli fault, and it may be very distributed there.

00:13:44.000 --> 00:13:51.000
It's also important to note that the Narli fault cuts across topography as well as preexisting fault mounts.

00:13:51.000 --> 00:13:59.000
So it's not a fault that would be expected to have a large hazard. Now let's change gears and look at some parallels to northern California.

00:13:59.000 --> 00:14:06.000
The Turkey earthquake shown on the left are plotted at roughly the same scale as the northern California faults shown on the right.

00:14:06.000 --> 00:14:14.000
One earthquake hazard parallel to northern California is the nucleation of a large plate boundary earthquake from a low-slip-rate system of discontinuous faults.

00:14:14.000 --> 00:14:24.000
In northern California just disconnected faults in the Bennett Valley or Collayoni could lead to rupture of the Maacama or Hayward-Rodgers Creek fault systems.

00:14:24.000 --> 00:14:33.000
Both are possible in the current NSHM trim or the quick rupture forecast. Well, the Bennett Valley faults commonly participate in earthquake connections, the Collayomi faults do not.

00:14:33.000 --> 00:14:43.000
However, both of these faults are closer to their plate boundary counterparts than the distance between the magnitude 7.8 epicenter and the EAF in Turkey.

00:14:43.000 --> 00:14:53.000
Another potential parallel is a second very large strike-slip rupture happening hours after an initial very large strike-slip rupture on a different unconnected fault.

00:14:53.000 --> 00:15:02.000
In Turkey the 7.8 and 7.5 ruptures are about 17 km apart at the closest point though clearly linked in time.

00:15:02.000 --> 00:15:11.000
This would be somewhat similar to the 1857 San Andreas earthquake, followed by the 1952 Kern County earthquake on the White Wolf fault.

00:15:11.000 --> 00:15:18.000
Clearly there are opportunities for improved forecasting of independent but time correlated earthquakes.

00:15:18.000 --> 00:15:31.000
Finally, let's look at response parallels to a major California rupture. In both Turkey and California, immediate, remote and on the ground responses would be critical as rapid information dissemination both across teams and with the public.

00:15:31.000 --> 00:15:40.000
In Turkey, coordination between remote mapping, field mapping, remote sensing, the National Earthquake Information Center, and their pager team, etc.

00:15:40.000 --> 00:15:47.000
was critical. Sharing data between these teams and between the field and the office in real-time was necessary, but it was often difficult.

00:15:47.000 --> 00:16:00.000
And science took a back seat while we were in rapid response mode. Limitations of a remote response include weather such as clouds or snow, urbanization and forest obscuring rupture on satellite optical images.

00:16:00.000 --> 00:16:13.000
As well as the waiting time for a post-event satellite radar pass for interferometry. We need immediate data collection, multiple satellite passes, potentially aerial passes, and weather windows, and on-the-ground response.

00:16:13.000 --> 00:16:26.000
Limitations on the ground response may include humanitarian crises, lifeline, and infrastructure disruptions, and private property, which in California would be much more of an issue than in Turkey.

00:16:26.000 --> 00:16:34.000
I'll leave it here for now. Thank you for listening and I'm ready to take questions.

00:16:34.000 --> 00:16:45.000
Well, thank you very much, Nadine, for this excellent talk. Our next speaker is Richard Walker from the University of Oxford.

00:16:45.000 --> 00:16:54.000
He will discuss the seismotectonic and societal aspects of the 2023 October Herat Afghanistan earthquake sequence.

00:16:54.000 --> 00:17:05.000
Okay, so, I'm gonna be talking about the sequence of earthquakes that occurred in October of last year in Herat, Afghanistan, poor earthquakes, magnitude 6.3,

00:17:05.000 --> 00:17:16.000
occurring over a period of about a week or so. We'll talk about the tectonic setting, we'll talk about what we know about the sources of these earthquakes.

00:17:16.000 --> 00:17:19.000
This is work that's ongoing, but we'll we'll show you where we are at the moment.

00:17:19.000 --> 00:17:26.000
We'll also talk a little bit about the societal aspects of these earthquakes, what they did.

00:17:26.000 --> 00:17:37.000
Okay, just to start off a little bit of tectonic settings, so Herat is in northwest Afghanistan, so just over here.

00:17:37.000 --> 00:17:48.000
And the deformation that we see there is part of the Arabia, Eurasia collisions zone, so we have Arabia moving northwards relative to a fixed Eurasia.

00:17:48.000 --> 00:18:00.000
Rates of up to about 25mm per year as shown by these loss velocity relative to Eurasia.

00:18:00.000 --> 00:18:12.000
Those velocities decrease eastwards towards the border of Iran and Afghanistan. They also decreased northwards towards the border with Turkmenistan, Eurasia.

00:18:12.000 --> 00:18:25.000
Okay, so you end up with shear right-lateral north-south directed shear across the eastern parts of Iran and just into the edges of Afghanistan.

00:18:25.000 --> 00:18:35.000
You also end up with a lot of thrust faulting shortening on east-west Plains. There's a little bit of a complication once you get up into northeastern Iran.

00:18:35.000 --> 00:18:49.000
You get a lot of East-West faults that are also involved left-lateral faulting and so these are thought to accommodate this north-south shearing by radiation about vertical axes, right.

00:18:49.000 --> 00:18:56.000
So the left-lateral faulting plus a rotation allow you to accommodate that overall tectonic motion.

00:18:56.000 --> 00:19:06.000
Herat just here in northwest Afghanistan really sees a bit of all of these, right? So you get right- lateral faulting on more north-south 

00:19:06.000 --> 00:19:15.000
oriented planes, left-lateral faulting on Moyes West Plains and also thrust faulting as well.

00:19:15.000 --> 00:19:31.000
But there's very little within this genus velocities, that suggest that Herat is a zone of active faulting when a lot of the activity appears to end very close to the eastern border of the country.

00:19:31.000 --> 00:19:41.000
You get the same situation as well if you look at instrumental seismicity. So, pre-October of last year, instrumental seismicity.

00:19:41.000 --> 00:19:48.000
White dots here showing instrumental earthquakes, instrumentally recorded earthquakes, and there's very little 

00:19:48.000 --> 00:20:00.000
in the Herat region. There's a lot more, however, if you look at historical earthquakes. Historical earthquakes were represented by red dots in this compilation here.

00:20:00.000 --> 00:20:08.000
We know of significant earthquakes in Herat's in medieval times, also in the early 20th century, right.

00:20:08.000 --> 00:20:24.000
So these are important points. First of all, if you just rely on recent seismicity or GNSS velocities there's not that much that tells you that Herat's is even a zone of earthquake hazard.

00:20:24.000 --> 00:20:34.000
However, if you go into historical records or as we'll show as well in the next few slides, geological studies looking at the geomorphology mapping active faults.

00:20:34.000 --> 00:20:44.000
It's actually very clear that Herat is the region of significant earthquake hazard. This is what the Herat region looks like.

00:20:44.000 --> 00:20:51.000
This is Herat here. Okay. And you have a major three going strikes that's called the Herat fault.

00:20:51.000 --> 00:21:02.000
Which is this one, tacing through the image and actually carries on a lot further than I've mapped it here off to the top left-hand of this image.

00:21:02.000 --> 00:21:11.000
To the eastern side of the image it continues for a longer way. So we have this main strike-slip fault running along through the edges of these mountains.

00:21:11.000 --> 00:21:24.000
We also have east-west reverse faults branching away from it. In particular, just to the north of Herat we have this really rather clear, north-dipping thrust 

00:21:24.000 --> 00:21:36.000
that bounds the edges of these mountains just to the north of Herat itself, but also this is going to become significant when we look at what happened in the earthquakes 

00:21:36.000 --> 00:21:46.000
of October last year. We see some fault scarps running through the hanging wall of this main Harat thrust.

00:21:46.000 --> 00:21:59.000
And these actually look as though they may be significant faults in their own right and actually we think it may be these ones that were responsible for the slip that occurred in that earthquake.

00:21:59.000 --> 00:22:00.000
Looking at that very prominent thrust just north of Herat, it comes to the surface, right?

00:22:00.000 --> 00:22:12.000
So this is the fault scarp along the south. You can see a very prominent hanging wall anticline.

00:22:12.000 --> 00:22:25.000
To the north of that fault, you can see the truncation beds within the truncation of bedrock as well as these basin full sediments that have been uplifted.

00:22:25.000 --> 00:22:45.000
Going along to where, Herat City is itself this is, city and oblique view looking eastwards and there's a very sharp edge to the mountains at the northern edge of Herat and in fact you can see sharp scarps in older alluvial terrorist deposits here in the outskirts of Herat.

00:22:45.000 --> 00:22:54.000
So this is a fault that breaks the surface, and seems to be quite significant in terms of the the local geomorphology.

00:22:54.000 --> 00:23:11.000
The earthquakes had occurred in October. We're in the hanging wall of that thrust. Okay, so if this is Herat city, this is that thrust here are the earthquakes mainshocks, large amount of significant aftershocks as well.

00:23:11.000 --> 00:23:41.000
And if we look at those in a little bit more detail, all of these were magnitude 6.3 there were also some very large aftershocks as well. The first of those 6.3 events occurred local time, 11:11 in the morning on the 7th of October, the second one occurred half an hour later 11:42a.m.   Now they were, especially the first one was particularly devastating.

00:23:42.000 --> 00:23:50.000
It led to, I mean, I think the death toll is still debated, but several thousands of deaths.

00:23:50.000 --> 00:24:03.000
Because of the time of day most of those that were killed or injured were women and children. So most of the men are outside in the fields, the women and children were inside.

00:24:03.000 --> 00:24:20.000
And you had large amounts of building collapse and so unfortunately this has led to you know as I say a you know something about 90% of of fatalities within those parts of the population.

00:24:20.000 --> 00:24:28.000
The third event occurred on the eleventh of October, again in the early morning and 15th of October again in the morning.

00:24:28.000 --> 00:24:48.000
These were still damaging to structures, they were less damaging in terms of loss and life and injury because most people were living outside damage was heavy in rural communities close to the epicenters there was also damage in the city of Herat itself.

00:24:48.000 --> 00:24:54.000
It's in terms of what we know about these earthquakes, these are solutions from interferograms.

00:24:54.000 --> 00:25:01.000
Sadly, we don't have interferograms between these two events. So very closely spaced in time.

00:25:01.000 --> 00:25:04.000
We also don't have an interferograms between these two events on the 11th and the 15th of October.

00:25:04.000 --> 00:25:20.000
So we're only able to show models of best-fitting models for the cumulative between the two events on the on the 7th and the two later events as well.

00:25:20.000 --> 00:25:36.000
What we can say is that those events on the 7th seem to be pretty much close located on a single plane we can't resolve and any differences between them in terms of their positive structure.

00:25:36.000 --> 00:25:50.000
The later two events had a seem to have somehow overall seem to have had a different strike. They seem to have had somewhat a shallower depth extent.

00:25:50.000 --> 00:25:59.000
Both of these didn't reach the surface, both of them are blind earthquakes, they cause folding at the surface but no discrete surface rapture.

00:25:59.000 --> 00:26:09.000
But there seems to be a hint of more complexity in this region here of the two later earthquakes.

00:26:09.000 --> 00:26:14.000
But there's nothing in here. In the geormophology and in the source models that would give us a clue that this kind of behavior might be expected in some ways, right?

00:26:14.000 --> 00:26:28.000
So there's nothing that shows that this area is anomalous in some way.

00:26:28.000 --> 00:26:37.000
We said that all the earthquakes were blind. However, some service deformations were mapped.

00:26:37.000 --> 00:26:47.000
Lot of surface deformations were mapped along the edges of this significant river valley. And were parallel to that River Valley.

00:26:47.000 --> 00:26:59.000
So here are some of these, right? So examples of cracking, etc, which we would assume to be related to gravitational effects in the margins of that river valley.

00:26:59.000 --> 00:27:21.000
However, surface cracks were also identified running east-west so parallel to the structures. And in proximity to the epicentral region of those shallower events, those later ones in the sequences I'm going to show on the second slide here, so east-west running 

00:27:21.000 --> 00:27:31.000
fractures, fishes, which could be related to fold growth or, you know, movement along bedding planes.

00:27:31.000 --> 00:27:36.000
They may have a more tectonic or structural control.

00:27:36.000 --> 00:27:50.000
In terms of the damage from these earthquakes, we already mentioned that they were particularly damaging in the rural areas adjacent or within the epicentral zones.

00:27:50.000 --> 00:27:59.000
These show the typical dwellings so you have these little domed dwellings arranged around a courtyard.

00:27:59.000 --> 00:28:08.000
Okay. And in the post-earthquake image here you can see, well in this village seemingly almost total 

00:28:08.000 --> 00:28:19.000
collapse. These are efforts done by the United Nations Satellite Center you know said they did quite a lot of regional mapping of damage across the across the epicentral zones.

00:28:19.000 --> 00:28:31.000
Looking at that from a field photo point of view, you see the typical constructions, these four square walls, a heavy dome of earth, adobe bricks over the top of it.

00:28:31.000 --> 00:28:35.000
When the earthquakes hit, typically the walls are sheared, the dome collapses and so this is what we can often relate to the very high fatality rates.

00:28:35.000 --> 00:28:51.000
These are very heavy constructions and you know in some of the villages almost complete destruction leads to very high death tolls for anyone who is inside.

00:28:51.000 --> 00:28:55.000
Slightly different types of construction here again earth and construction and then some of these, yeah, almost complete collapse.

00:28:55.000 --> 00:29:07.000
The only thing is still remaining upright in this village here at the door frames everything else has gone, collapsed.

00:29:07.000 --> 00:29:14.000
Yeah, we're still working on the earthquakes that occurred in Herat, still a lot to find out from them.

00:29:14.000 --> 00:29:16.000
It's worthwhile however, spending a bit of time looking at some other earthquake sequences that have occurred within the region.

00:29:16.000 --> 00:29:46.000
In recent decades and also in history, this is one example from 2017, December, close to Kaman, eastern Iran not too far away from the Herat region where you had a very complex sequence of earthquakes occurring all, you know, roughly magnitude 6 earthquake one occurring on a north-dipping thrust followed by earthquake 2, almost parallel plane separated by just a few kilometers and with

00:29:52.000 --> 00:30:00.000
depth, limited slip on both of those. A third event occurred, above where the slip in those two events was and with the opposite sense of dip.

00:30:00.000 --> 00:30:11.000
South-dipping event which ruptured to the surface. There seems to be in some ways a structural control on the

00:30:11.000 --> 00:30:15.000
depths limit of folding within those arrests in the earlier two and then eventual rupture through that.

00:30:15.000 --> 00:30:34.000
At the third event, we have inferred that there might be some structural control, older structural control in the bedrock that might explain why you have this complex arrangement of the first two events.

00:30:34.000 --> 00:30:40.000
Also, Safida Bay in 1994 a sequence of several earthquakes occurring within a region again with a lot of bedrock structure within there, right?

00:30:40.000 --> 00:30:59.000
So older, mainly volcanic rocks, aligned with the active structure, and there's the interferogram, the best fitting solution for that is this sequence of on echelon offset,

00:30:59.000 --> 00:31:09.000
reverse faults. Okay, so again it seems that individual earthquakes have ruptured individual pots or individual short segments of fault offset from each other.

00:31:09.000 --> 00:31:21.000
And again, the inference there is that we're looking at bedrock control. Right, so we're looking at these older layers, which, you know, the active faults are following.

00:31:21.000 --> 00:31:22.000
We can also make use of the very long and rich historical record in Iran to go further back in time.

00:31:22.000 --> 00:31:28.000
This is a particularly striking

00:31:28.000 --> 00:31:35.000
example from the city of Neyshabur. Again, not at all far away from Herat here.

00:31:35.000 --> 00:31:46.000
And we have a single large thrust fault running to the northern side of Neyshabur. Which ruptured in four large earthquakes, 1209,

00:31:46.000 --> 00:31:52.000
1270, 1389 and 1405, each of which with has a estimated magnitude of about magnitude 7.

00:31:52.000 --> 00:32:15.000
And again there's nothing here in the geomorphology, the structure that tells us that this is complex in some ways, okay, and yet you have a cluster of events all seemingly rupturing the same or closely spaced parts of this north Neyshabur fault system.

00:32:15.000 --> 00:32:25.000
However, this is not occurring over days, but it's occurring over decades or in this case even centuries.

00:32:25.000 --> 00:32:34.000
So if we summarize the main points, from the Herat earthquakes, there's no obvious feature in the geological structure that would suggest complex rupture.

00:32:34.000 --> 00:32:40.000
Right, there's no clue here that this might have occurred prior to the earthquakes occurring.

00:32:40.000 --> 00:32:47.000
It's not an isolated example, we have other recent historical examples. That show behaviors that occur over days to decades or even centuries.

00:32:47.000 --> 00:33:06.000
A final point, I think an important one here is that the main Herat thrust fault reaches to the surface and actually cuts through the northern side of Herat City itself.

00:33:06.000 --> 00:33:15.000
This did not rupture in the earthquake and I think is something that we should really try and understand exactly how 

00:33:15.000 --> 00:33:26.000
do the earthquakes of October last year relate to that fault, whether they connected in some ways structurally, were they a completely separate reverse fault cutting through its hanging wall as I've kind of suggested here that's our running hypothesis.

00:33:26.000 --> 00:33:38.000
How might they influence the probabilities of ruptures on them on the main Herat fault in future.

00:33:38.000 --> 00:33:47.000
I'll leave it there. Thank you very much for your attention. Bye, bye.

00:33:47.000 --> 00:33:54.000
Alright, thanks for a great talk, Richard. Our next speaker is Jeanne Hardbeck from USGS talking about stress shadows, insights into the physics of earthquake triggering.

00:33:54.000 --> 00:34:03.000
Take it away, Jean.

00:34:03.000 --> 00:34:17.000
Hi, today I'm going to talk about stress shadows. Some of the other talks in this session have been talking about recent earthquake sequences where we've had a number of large events in a row and we're asking ourselves the question, "why are we having so many big earthquakes?"

00:34:17.000 --> 00:34:22.000
Stress shadows is really about when we're asking ourselves the opposite question when we're asking ourselves, "why haven't we had so many big earthquakes?"

00:34:22.000 --> 00:34:34.000
And this is an example following the 1857 Fort Tejon earthquake in southern California.

00:34:34.000 --> 00:34:49.000
Where for about 50 years after that earthquake there weren't very many moderate-sized earthquakes in the southern California region and in particular there were no earthquakes in areas where there had been a decrease in stress due to that 1857

00:34:49.000 --> 00:35:00.000
earthquakel, so that was termed a stress shadow, an area where a stress decrease causes there to be no earthquakes happening for a while.

00:35:00.000 --> 00:35:06.000
So, when we talk about these stress changes, we're talking about static Coulombs stress changes.

00:35:06.000 --> 00:35:13.000
Static means that these are the stress changes that are permanent that stick around after the seismic waves have already passed.

00:35:13.000 --> 00:35:20.000
And the Coulomb stress change is the projection of the shear and normal stress change onto a receiver folk.

00:35:20.000 --> 00:35:25.000
And many of you have probably seen many figures like the map on the left. Which shows a map view of these calculated static flow stress changes 

00:35:25.000 --> 00:35:40.000
and the red areas being where there was an increase in static Columb stress and the purple areas being where there was a decrease and it's these purple areas that we're considering the stress shadows.

00:35:40.000 --> 00:35:50.000
And you would not just expect to see as many aftershocks occurring in a stress shadow or as many subsequent large events occurring in a stress shadow.

00:35:50.000 --> 00:35:59.000
And in fact, according to the rate and state friction model, there should actually be an earthquake rate decrease in the stress shadows.

00:35:59.000 --> 00:36:12.000
So, these stress shadows can help us differentiate between different models for what may be triggering aftershocks and subsequent large events after a large earthquake.

00:36:12.000 --> 00:36:30.000
And they can particular they can help us tell the difference between a static stress triggering model, as I said, the triggering from the stress that's permanent after the passage of seismic waves and the dynamic stress triggering which is triggering by those transient stresses caused by the passing seismic waves.

00:36:30.000 --> 00:36:39.000
This is because only the static stress changes produce shadows. That is, there can be either positive or negative permanent stress change.

00:36:39.000 --> 00:36:46.000
However, the dynamic stress change because of the oscillatory nature of the passing seismic waves, the largest stress changes always going to be positive.

00:36:46.000 --> 00:36:59.000
It's going to be positive everywhere. So there's going to be no stress shadows. So the fact that we see stress shadows indicates the importance of the static stress triggering model.

00:36:59.000 --> 00:37:15.000
And we do see stress shadows in a few different cases. One is the case I showed earlier of a very large magnitude mainshock like that 1857 event in southern California or the 1906 earthquake here in the Bay Area.

00:37:15.000 --> 00:37:17.000
Following that, there's a decrease in regional rate that can last for decades. And that is a function of these 

00:37:17.000 --> 00:37:27.000
big stress shadows from these large earthquakes.

00:37:27.000 --> 00:37:31.000
Stress shadows are a little bit harder to see following moderate magnitude main shocks, but there are a couple of cases where they are observed.

00:37:31.000 --> 00:37:42.000
This is a special case where a stress shadow overlaps an area that already has an ongoing aftershock sequence.

00:37:42.000 --> 00:37:50.000
And these are a couple of examples where an aftershock sequence starts and either days later or months later, there's another large nearby earthquake.

00:37:50.000 --> 00:38:01.000
And that earthquake projects a shadow into some of the area of the ongoing aftershock sequence. And if you look at that area where the stress shadow was, you can see a decrease in rate.

00:38:01.000 --> 00:38:12.000
And of course aftershock sequences always have an ongoing decrease in rate but the decrease in rate is beyond what we would expect just from the decay of aftershock sequences.

00:38:12.000 --> 00:38:22.000
So that looks like where these stress shadows overlap an ongoing aftershock sequence, they do in fact decrease the rate of aftershocks.

00:38:22.000 --> 00:38:34.000
It's a little harder to find evidence of stress shadows just looking in general at an aftershock sequence without this special case of overlapping an ongoing sequence.

00:38:34.000 --> 00:38:41.000
So these are a number of attempts to look for that. Look for the stress shadows in all of these plots on the x-axis.

00:38:41.000 --> 00:38:58.000
Is shear stress change with positive stress change to the right and negative to the left. And on the y-axis is a measure of the significance of the seismicity rate change with a positive rate change upwards and a negative rate change downwards.

00:38:58.000 --> 00:39:11.000
And you can see that there are a few stress shadows detected in some of these, some of these cases, particularly for the Lander's earthquake, that is, there are particular locations each dot indicates a location.

00:39:11.000 --> 00:39:20.000
There are some locations falling in that lower-left quadrant of seeing a stress decrease and a corresponding decrease in rate.

00:39:20.000 --> 00:39:23.000
However, you can see that most of the stress decreased regions actually correspond with either an increase of rate or no detectable decrease of rate.

00:39:23.000 --> 00:39:37.000
So for these moderate magnitude mainshocks, therefore these predicted rate decreases are rarely seen. So we looked at 

00:39:37.000 --> 00:39:45.000
this slightly differently. Now, on the x-axis again is stress change, but this is the absolute value of stress change 

00:39:45.000 --> 00:39:54.000
and we're plotting both the stress increase and stress decrease regions together. The stress increase regions in red and decrease in blue.

00:39:54.000 --> 00:40:02.000
And on the y-axis is just the change in rate before and after the mainshock. And, we can see.

00:40:02.000 --> 00:40:04.000
first of all, from the blue line, it barely ever goes below that one to one line.

00:40:04.000 --> 00:40:22.000
So we're barely ever seeing the predicted rate decrease in the stress shadows. However, if we look at the two lines, we can see that that blue line representing the stress shadows is almost always quite a bit below the red line like an order of magnitude.

00:40:22.000 --> 00:40:30.000
And that shows that there's an order of magnitude lower aftershock rate in the stress shadows compared to the stress increased regions.

00:40:30.000 --> 00:40:39.000
So that implies that there really is a stress shadow there that really is a difference in effect between having a positive and negative stress change.

00:40:39.000 --> 00:40:52.000
So, all of this shows that stress, the shadows are observed regionally with these big rate decreases suppression of ongoing aftershocks sequences if they're overlapped by a stress shadow.

00:40:52.000 --> 00:41:00.000
And lower aftershock rate in the shadows than in the stress increase areas. This implies an importance of that static stress triggering.

00:41:00.000 --> 00:41:05.000
However, aftershock still are occurring in the shadows and the question is why? There's a couple of possibilities.

00:41:05.000 --> 00:41:09.000
One, is simply some modeling uncertainty in calculating the static stress. And the other is that there's other triggering processes also in play.

00:41:09.000 --> 00:41:23.000
In particular, we're going to hypothesize that it's the dynamic stress changes, so the stress changes from the passing seismic waves.

00:41:23.000 --> 00:41:29.000
So first let's ask the question could the aftershocks just appear to be in the shadows because of any accuracy in the stress change calculations.

00:41:29.000 --> 00:41:55.000
And I don't think so. We did this exercise where we took a couple of different aftershock sequences and we not only calculated the stress change once for each aftershock, we calculated over multiple realizations where we're changing a lot of the inputs to the modeling, changing the mainshock model, the receiver fault orientation, coefficient of friction, and doing this calculation over and over

00:41:55.000 --> 00:42:02.000
again. And the color of the earthquake indicates basically what fraction of these realizations we got a positive Coulomb stress change.

00:42:02.000 --> 00:42:10.000
And the events in red are pretty clearly with a stress increase, but we also can see quite a few events, particularly even off fault events, in a dark blue color 

00:42:10.000 --> 00:42:18.000
indicating that sort of however you do the modeling, these events always end up in the stress shadows.

00:42:18.000 --> 00:42:24.000
So it does look like there are events that are pretty stably within the stress shadows and it's not just a modeling problem.

00:42:24.000 --> 00:42:32.000
We could also ask aftershocks in the shadows just be carrying on faults with different orientations though what we assumed for the model receiver vaults.

00:42:32.000 --> 00:42:41.000
And in particular, are they occurring on these unexpected fault orientations that received an increase in static Coulomb stress.

00:42:41.000 --> 00:42:59.000
This is possible. So looking at, some nice machine learning focal mechanism catalogs for both of those earthquake sequences, we can look at the PT and B access orientations for the background earthquakes and then aftershocks occurring in the stress increased regions and in the stress shadows.

00:42:59.000 --> 00:43:16.000
And what we see is that the mechanisms in the shadows appear to be more diverse than the events, the background events or other events in the stress increase regions and also less consistent with the background stress which is what's shown in the arrows there.

00:43:16.000 --> 00:43:22.000
So it does seem like these aftershocks in the shadows are occurring on sort of strange orientations.

00:43:22.000 --> 00:43:28.000
However, they don't really seem to be receiving a positive static, a static column stress change.

00:43:28.000 --> 00:43:39.000
And this is now modeling the Coulomb stress change on the actual planes of those vocal mechanisms and again doing multiple realizations with different modeling assumptions.

00:43:39.000 --> 00:43:45.000
And these histograms show the probability of a positive Coulomb stress change on those focal mechanisms.

00:43:45.000 --> 00:43:52.000
And you can see that quite a lot of these events, particularly for Ridgecrest, show a low probability of having it had a stress increase on the actual earthquake focal mechanism.

00:43:52.000 --> 00:44:02.000
So this doesn't seem to be the explanation for this events. So is dynamic stress triggering a good explanation?

00:44:02.000 --> 00:44:08.000
Well, here's a plot where we look again at that rate change of aftershocks compared to background events.

00:44:08.000 --> 00:44:15.000
Now is a function of distance. Again, with the red symbols being the stress increased regions and the ability of the shadows.

00:44:15.000 --> 00:44:24.000
And you can see that the aftershocks of the shadows are fit well by a one over distance squared spatial decay, which is consistent with dynamic triggering by near field body waves.

00:44:24.000 --> 00:44:32.000
How about the temporal decay? On the top in red is the temporal decay in the stress increase regions, which follows in a more decay like we would expect.

00:44:32.000 --> 00:44:42.000
And in the bottom and blue is in the stress decrease regions. And what we see there is an initial burst over a few days followed by return to a constant rate.

00:44:42.000 --> 00:44:52.000
And you can see that in the little inset there that burst over the first few days. And an initial burst like that is really consistent with a transient process like dynamics stress triggering.

00:44:52.000 --> 00:45:09.000
So we can estimate how many, how, what fraction of the aftershocks are dynamically triggered by just assuming that those all those earthquakes in the shadows are dynamically triggering and assuming that there's an equal number of dynamically triggered events also in the stress increase regions.

00:45:09.000 --> 00:45:17.000
And the histogram shows the results of the estimated fraction dynamically triggered again over a bunch of realizations with different modeling assumptions.

00:45:17.000 --> 00:45:26.000
And we come out with the results that suggest about a third of the earthquakes, third of the aftershocks are dynamically triggered.

00:45:26.000 --> 00:45:28.000
And if we make a hybrid model with two-thirds of aftershocks triggered by static stress and one third triggered by dynamic stress.

00:45:28.000 --> 00:45:51.000
This looks a lot more like the distribution of aftershocks than static metal alone. And then the maps here, the leftmost one is the static only and you can see in purple those deep stress shadows as well as in the observed in the center in the hybrid model on the right you really don't see those deep stress shadows.

00:45:51.000 --> 00:46:18.000
And this result is also consistent with other results by Nicholas van der Elst and Emily Brodsky, where they estimate this fraction by a totally different method by looking at farfield triggering which is all dynamic and projecting it into the nearfield and estimating what fraction events should should be dynamically triggered and they get 15% to 60% dynamically triggered which is you know almost perfect agreement with most of the power in our

00:46:18.000 --> 00:46:32.000
histogram so this is reassuring to get the same answer basically from two entirely different approaches. So, just to summarize then, we see that stress shadows are observed.

00:46:32.000 --> 00:46:44.000
We see a regional earthquake rate decrease that lasts for decades after a large May track. We see suppression of ongoing aftershock sequences where those sequences are overlapped by a stress shadow.

00:46:44.000 --> 00:46:57.000
And we also see lower aftershock rates in shadows than in stress increase areas. So these observation of stress shadows indicate that there that static stress triggering is really important in generating aftershocks.

00:46:57.000 --> 00:47:06.000
However, we also do see that some aftershocks continue to occur in the shadows, and it doesn't appear that this is just modeling uncertainty.

00:47:06.000 --> 00:47:15.000
And these events, both their spatial and temporal decay away from the mainshock are really consistent with what we would expect from a dynamic stress triggering.

00:47:15.000 --> 00:47:25.000
So, all of that points to the importance of dynamic stress changes in the nearfield as a part of the triggering of aftershocks.

00:47:25.000 --> 00:47:32.000
And we find that if we take a hybrid model, static and dynamic triggering, that's about two-thirds static and one-third dynamic 

00:47:32.000 --> 00:47:49.000
this really best matches the aftershocks. So we can see then that stress shadows can be helpful in understanding the physics of earthquake triggering and I'm very happy to take any questions you might have.

00:47:49.000 --> 00:47:53.000
Thank you.

00:47:53.000 --> 00:47:59.000
Oh, thank you very much, Jeanne, for a fantastic talk. Our next speaker is Annemarie Baltay from the USGS.

00:47:59.000 --> 00:48:05.000
Will be sharing insights on comparing doublets and aftershocks.

00:48:05.000 --> 00:48:11.000
Great. Thanks so much to the organizers for having me and to all of you for being here this afternoon.

00:48:11.000 --> 00:48:17.000
So I'll talk about mistaken ground motion identity, comparing ground motion between doublets and aftershocks.

00:48:17.000 --> 00:48:23.000
So Jeanne probably just told us about the interactions or lack thereof during doublet and aftershock sequences.

00:48:23.000 --> 00:48:28.000
And now I'll try to break it down as to how much ground motion they each produce.

00:48:28.000 --> 00:48:40.000
Let's start out with defining aftershock and doublets. So a doublet doesn't really have a strict definition, but let's consider it to be two or more earthquakes occurring close in space and time with similar magnitudes.

00:48:40.000 --> 00:48:45.000
Such that their centroids are closer than their rupture length. And the timeframe is shorter than a typical recurrence time.

00:48:45.000 --> 00:48:57.000
So today I'll talk about two different sequences with doublets. The first is the 2023 Turkey sequence with a M7.8 on the East Anatolian fault followed by a magnitude M7.5 several hours later to the north.

00:48:57.000 --> 00:49:08.000
The second sequence is the 2019 Ridgecrest sequence with the magnitude 7.1 which was preceded by the M6.4 about a day and a half on the perpendicular fault.

00:49:08.000 --> 00:49:13.000
For aftershocks, however, we'll use a strict definition employed by the NGA West 2 project.

00:49:13.000 --> 00:49:26.000
All events in the NGAs to database are classified as either Class 1 mainshocks. Or they could be off-fault aftershocks or Class 2 on fault aftershocks which occur after and are associated with a larger Class 1 event.

00:49:26.000 --> 00:49:37.000
And the centroid of that on-fault aftershock has to be within some distance, some cutoff distance to the surface projection of the Class 1 the mainshock fault plane.

00:49:37.000 --> 00:49:43.000
So the organizers asked me to answer this question. Are ground motions from doublets and aftershocks different?

00:49:43.000 --> 00:49:49.000
So, just in case you don't wanna pay attention or you get something else to do, I'll cut right to the chase.

00:49:49.000 --> 00:50:00.000
For doublets, we see that no ground motion seems to be independent and well modeled by current ground motion models and is not different from one another.

00:50:00.000 --> 00:50:12.000
Aftershocks on the other hand do seem to have different ground motions as compared to their mainshock. So we see that high frequency ground motion from smaller on fault aftershocks is less than the main shock events so far.

00:50:12.000 --> 00:50:20.000
And especially as we move as. Seismic hazard analysis moves away from the cross on assumption. We should consider this difference in ground motion models,

00:50:20.000 --> 00:50:30.000
for hazard and for forecasting. So this afternoon, I'll walk us through a few examples of these two cases using both ground motion observations and of course stress drop.

00:50:30.000 --> 00:50:36.000
So I know you didn't ask me to talk about stress drop, but that's what you get.

00:50:36.000 --> 00:50:42.000
Alright, so let's define what we're gonna talk about when we say ground motion. So on the left here are ground motion observations shown in the blue dots that's peak ground velocity PGV.

00:50:42.000 --> 00:50:56.000
For the Chino Hills magnitude 5.4 earthquake as a function of distance. So the first thing we typically do in a ground motion analysis is consider a ground motion model.

00:50:56.000 --> 00:51:06.000
So here's a typical ground motion model. In this case it's the board all, 2014 ground motion model plotted for a magnitude 5.4 event PGV as a function of distance.

00:51:06.000 --> 00:51:15.000
So we can see that while the general attenuation or distance decay features of the ground motion data are well matched, the overall amplitudes are under predicted.

00:51:15.000 --> 00:51:22.000
So in this case we might consider that simply shifting the curve upward would match the data on a whole better.

00:51:22.000 --> 00:51:28.000
So, we would say that this earthquake has a positive event term residual. So the ground motion event term is the excess ground motion on average for each earthquake.

00:51:28.000 --> 00:51:37.000
So this event is energetic and has a high stress drop.

00:51:37.000 --> 00:51:50.000
Here's an example from another event M5.1 Alum Rock earthquake. Again, the ground motion data shown in the blue and green dots is less than the typical ground motion curve; the green curve for magnitude 5.1.

00:51:50.000 --> 00:51:58.000
So it's perhaps better fit by this red dashed line. I mean, we would say that this event has a negative event term.

00:51:58.000 --> 00:52:06.000
I'll touch briefly on stress drop and I just want to make the point here that the stress drop I'm talking about is a high frequency spectral stress drop.

00:52:06.000 --> 00:52:11.000
That's something like moment times the corner frequency cubed. So it's dependent on that corner frequency.

00:52:11.000 --> 00:52:19.000
And it's a model parameter that relates the low frequency or the moment part of the spectrum to the high frequency ground motion through that corner frequency.

00:52:19.000 --> 00:52:26.000
On the bottom, we have two example broom source models for displacement on the left and acceleration on the right for two different magnitudes 

00:52:26.000 --> 00:52:32.000
and we can see in both cases as we increase that corner frequency, if we step the corner frequency out to the right 

00:52:32.000 --> 00:52:44.000
how that generates a higher stress drop and particularly on the acceleration spectrum as we increase the corner frequency by moving it to the right, we see now that that higher corner frequency and hence the higher stress drop 

00:52:44.000 --> 00:52:55.000
generates a higher high frequency ground motion, so the difference between the acceleration level for a 1 MPA stress drop event as compared to a 25 MPA stress drop event is significant.

00:52:55.000 --> 00:53:02.000
So that gives us an idea of how theoretically the stress drop is related to the high frequency ground motion through the corner frequency.

00:53:02.000 --> 00:53:19.000
And we can also see empirically that they're related. On the right is a correlation plot between on the y-axis the event terms for PGA from the Ridgecrest sequence and a measure of the stress drop the areas intensity stress drop on the x-axis and we can see that they're correlated.

00:53:19.000 --> 00:53:31.000
Alright, so let's talk about ground motion from aftershocks. So the first piece of evidence we have here is from the NGA West to ground motion model by Abrahamson, et al., 2014.

00:53:31.000 --> 00:53:40.000
And through careful modeling and robust modeling, Abrahamson et al., observe a decrease in the peak ground acceleration for these Class 2 PGA aftershocks when you look within 15km.

00:53:40.000 --> 00:53:47.000
So it is aftershocks that are occurring within 15km of the main fault plane. The figure on the bottom shows the coefficient difference between the Class 2 aftershocks and the Class 1 mainshocks.
 
00:53:47.000 --> 00:53:58.000
And for the short period, the high frequency, we can see that that coefficient is negative about -0.3.

00:53:58.000 --> 00:54:07.000
So for PGA, -0.3 implies about a 35% decrease in the ground motions at these frequencies for the Class 2 aftershocks.

00:54:07.000 --> 00:54:14.000
Now this is a fairly robust result. The NGA-West 2 database covers about 500 earthquakes globally in active tectonic environments.

00:54:14.000 --> 00:54:26.000
The Class 1 vs. Class 2 events are classified throughout, so it contains many sequences, and so this is seen as an effect on average over the whole data base.

00:54:26.000 --> 00:54:33.000
We can take the same database and actually look at stress drop, so the Arias intensity is a measure of stress drop that's related to the brood stress drop 

00:54:33.000 --> 00:54:39.000
and again, we can take all of the Class 1 mainshocks compared to the fault aftershocks.

00:54:39.000 --> 00:54:50.000
On the left we see the box plots showing the difference in the average stress drop. Where the aftershocks have about 30% lower stress drop as related to the Class 1 mainshocks.

00:54:50.000 --> 00:54:55.000
The right plot is just showing as we move away from the fault, so if it has a distance of 0, it means the aftershock is exactly on the fault 

00:54:55.000 --> 00:55:08.000
and as we move out, it means they're moving perpendicularly off the fault and so for all of the aftershocks that are within 15km, we see on average the stress drop is low, 

00:55:08.000 --> 00:55:17.000
and as we start to move to 15 and 20kms away from the fault plane, we see an increase of the stress drop, which maybe implies some spatial healing 

00:55:17.000 --> 00:55:21.000
as we move farther away from the rupture plane.

00:55:21.000 --> 00:55:30.000
And we can look within that NGAs two database at individual sequences where we've plotted the aftershocks for any one sequence, the stress drop of the aftershocks in blue,

00:55:30.000 --> 00:55:38.000
and the corresponding mainshock in the black triangle. So for the two sequences on the left, we can see that all of the aftershocks have stress drop 

00:55:38.000 --> 00:55:46.000
that is less than or equal to the mainshock stress drop. While for the two sequences on the right, we can see that the stress drops 

00:55:46.000 --> 00:55:53.000
are actually a little bit more all over the map. Some stress drops are higher than the stress drop for the mainshock, whereas some aftershock stress drops are lower.

00:55:53.000 --> 00:56:00.000
So we know that both mainshocks and here we're showing that classic on-fault aftershock stress drop is very variable,

00:56:00.000 --> 00:56:08.000
but again, on average when we look at many sequences together, the aftershocks are less energetic and have lower stress drop and hence less ground motion.

00:56:08.000 --> 00:56:18.000
So let's talk about some physical reasons for this. Imagine that a mainshock occurs and causes as a stress drop on the rupture plane, while the area outside of the rupture plane may experience a shear stress increase.

00:56:18.000 --> 00:56:26.000
In the middle is a schematic of the Ridgecrest showing fault slip contours overlain by all the aftershocks.

00:56:26.000 --> 00:56:38.000
So once the main shock has occurred, then these classic on fault aftershocks primarily ruptured areas of the fault that have already ruptured in the mainshock and because they experience a stress, that main track experience is stress drop in those areas.

00:56:38.000 --> 00:56:54.000
The aftershocks will also have reduced stress drop or energy, but obviously aftershocks occur all over the fault plane and some of these aftershocks can rupture either in places that haven't already ruptured outside of the slip areas or can re-rupture high-stress asperities resulting in high stress drops.

00:56:54.000 --> 00:57:02.000
So on fault aftershocks show a wide range of stress drop and hence generate a wide range of high frequency ground motion,

00:57:02.000 --> 00:57:14.000
but on average, the stress drop and the ground motion is lower. So we'd like to understand also is there a connection between the slip and energy and the mainshock and where or when or how energetic the aftershocks are.

00:57:14.000 --> 00:57:18.000
So let's look at a little bit more detail at the Ridgecrest sequence.

00:57:18.000 --> 00:57:33.000
So the left here shows a spatial map of Ridgecrest aftershocks colored by their time relative to the magnitude 7.1 so the blue events occur before the 7.1 that's 6.4 and it's aftershocks whereas the red ones are associated with the 7.1.

00:57:33.000 --> 00:57:42.000
And on the right here I'm showing two different ways of estimating stress drop. The first is from the Ridgecrest Community Stress route validation project with which I'm involved.

00:57:42.000 --> 00:57:51.000
In which we invited the community to estimate stress drop for these events and we have many submissions in some cases, 30 or 40 different submissions of stress drop for each earthquake.

00:57:51.000 --> 00:58:02.000
Here we're just looking at a corner frequencies for magnitude over three. And for each case, a D-trended  the corner frequencies for the magnitude, the depth, and based on the author.

00:58:02.000 --> 00:58:09.000
And then we average those corner frequencies into a stress drop in the Madariaga stress drop over all the submissions.

00:58:09.000 --> 00:58:16.000
So I'm showing two cases. The first is when we have 4 or more submissions for a single earthquake and the second is when we have 10 or more submissions.

00:58:16.000 --> 00:58:23.000
And we can initially also just see that when we have 10 or more submissions, the overall variability in stress drop is lower.

00:58:23.000 --> 00:58:32.000
But comparing the Class 1 and the Class 2, we can see that the Class 2 aftershocks actually have slightly lower stress drop in both cases than the associated mainshock.

00:58:32.000 --> 00:58:37.000
Now this result might not be statistically significant, but it's still indicating that this behavior is present.

00:58:37.000 --> 00:58:42.000
The second stress drops we can look at are the the Arias stress drops are related to a brune stress drop.

00:58:42.000 --> 00:58:49.000
From the Parker et al., 2020 Ridgecrest study, and again, we can see the same thing here and now looking at two different cutoff distances.

00:58:49.000 --> 00:58:54.000
So both 5km cutoff for aftershocks to be included versus 10km and we can see the same thing.

00:58:54.000 --> 00:59:00.000
And here we see that the difference is on the same order as what we saw for the NGA-West 2, about 30%.

00:59:00.000 --> 00:59:06.000
So again here we're seeing a smaller stress drop for the on-fault aftershocks as compared to the mainshocks.

00:59:06.000 --> 00:59:27.000
Let's take a look at some of the same data, but in-ground motion space. So now we're looking at the ground motion event residuals for aftershocks within 5km of the event compared to mainshocks, the main events or the off-fault aftershocks. So now we're not just looking at one ground motion model to create these event turns, but we're looking at three ground motion models.

00:59:27.000 --> 00:59:37.000
So the top two panels are showing the YA15 ground motion model for PGA and PGV and then the bottom three panels are showing the BSSA ground motion model PGA and PGV and the ASK14 

00:59:37.000 --> 00:59:43.000
ground motion model. So there's subtle differences between the ground motion model and one to see how this would affect these results.

00:59:43.000 --> 00:59:56.000
So for the Jenny and Atkinson model on the top we do indeed see that the aftershocks are showing slightly reduced ground motion event residuals as compared to the other events the off-fault aftershocks or the main events,

00:59:56.000 --> 01:00:07.000
but not significantly. For the BSSA14 and the ASK14 models, we actually see that the on-fault aftershocks are showing higher ground motion than the other events.

01:00:07.000 --> 01:00:16.000
So we see there's a lot of variability within this Ridgecrest sequence and some ground motion models have aftershocks with less ground motion than mainshocks and others are opposite.

01:00:16.000 --> 01:00:22.000
We also see previously looking at the overall NGA-West2 database that different sequences may behave differently.

01:00:22.000 --> 01:00:26.000
So the jury is still out here.

01:00:26.000 --> 01:00:33.000
So in the future overall, understanding the relationship between the mainshock structure pattern aftershock stress drop, ground motion occurrence and rates is really interesting.

01:00:33.000 --> 01:00:38.000
So here's a spatial map of all the stress drops from the community stress drop experiment.

01:00:38.000 --> 01:00:46.000
And so trying to understand how the details of these stress drops vary depending on what happened in the mainshock is really interesting and something we're going to be looking at 

01:00:46.000 --> 01:00:53.000
a little bit more. Let's touch quickly on the doublets. So this is an example from the 2019 Ridgecrest sequence.

01:00:53.000 --> 01:01:02.000
We're looking at the magnitude 7.1 on the top and the 6.4 on the bottom with the peak ground acceleration in the left panels and the peak ground velocity on the right panels.

01:01:02.000 --> 01:01:16.000
And all the ground motion observations are in black triangles and the ground motion model in red. And now unlike those two examples I showed you earlier, the Chino Hills in the Alum Rock, here the ground motion is very well modeled by the ground motion model.

01:01:16.000 --> 01:01:21.000
So the red curve fits the black data really well. We don't have to adjust it up and down.

01:01:21.000 --> 01:01:29.000
So we would say that both these events show average source behavior about near 0 event terms for the high frequency PGA and PGV, 

01:01:29.000 --> 01:01:40.000
and if we look actually at all periods, we see that while the event terms do change, the behavior between the magnitude 7.1 and the 6.4 the two doublets is very similar so there's no bias between them.

01:01:40.000 --> 01:01:50.000
So we would say that these doublets are acting differently than the aftershocks, whereas the on-fault, smaller aftershocks had reduced ground motion here, the doublets have very similar ground motion to each other,

01:01:50.000 --> 01:01:56.000
so they're acting as independent events, and we can see the same thing from the 2023 Turkey sequence.

01:01:56.000 --> 01:02:08.000
Again, the larger event, the M7.8 on the top row and the smaller event the 7.4 on the bottom row where all the ground motion observations are in green and the average ground motion model is the red dashed line.

01:02:08.000 --> 01:02:15.000
And again, we can see that overall these events are behaving very similarly to what we would expect with the ground motion models 

01:02:15.000 --> 01:02:21.000
and there isn't a big difference between how the aftershock the 7.5 is behaving compared to the mainshock.

01:02:21.000 --> 01:02:29.000
So in this case for the double events we again think that these are independent fault ruptures from the point view of the ground motion genesis.

01:02:29.000 --> 01:02:36.000
Now obviously for risk or impact, these were very coupled events and I think we'll hear more about that later this afternoon.

01:02:36.000 --> 01:02:43.000
But as from the point of view of how we should model these events and they both seem to be behaving similarly.

01:02:43.000 --> 01:02:47.000
Alright, so some conclusions and implications. Whereas on-fault, Class 2 type aftershocks seem to show that high frequency is about 30% lower on average 

01:02:47.000 --> 01:03:05.000
and for some sequences, the doublets seem to show independent ground motions and are dissimilar from aftershocks and therefore we should continue to be modeling the doublets as independent events for ground motion, 

01:03:05.000 --> 01:03:14.000
but we might want to consider that these on-fault aftershocks have reduced ground motion. So open questions about longer period motion here 

01:03:14.000 --> 01:03:23.000
we only talked about high frequency ground motion and we also have the NGA-West 3 database coming up and hopefully we'll see those classified into these different 

01:03:23.000 --> 01:03:35.000
mainshock aftershock sequences and see what's happening there. So for implications in the National Seismic Hazard Model in 2023 for the first time included full, not declustered seismicity rates in the grid seismicity models.

01:03:35.000 --> 01:03:38.000
So therefore, aftershocks are included in the rate model. But the NSA time in 2023 does not treat aftershock ground motion any differently.

01:03:38.000 --> 01:03:57.000
So we might want to consider if this should be included. Furthermore, for operational aftershock forecasting, ShakeAlert, and ShakeMap, we could consider the difference to account for different ground motions in some aftershock sequence and thus reduce the ground motion there.

01:03:57.000 --> 01:04:01.000
Thanks so much.

01:04:01.000 --> 01:04:12.000
Alright. Thank you, Annemarie. That was really interesting and it's generated a lot of interesting discussion in the chat that we'll also come back to in the discussion section at the end.

01:04:12.000 --> 01:04:22.000
Next is going to be Rob Graves from the USGS talking about some ground motion simulations for one of these potential doublet scenarios 

01:04:22.000 --> 01:04:26.000
on the Hayward and Calaveras. Take it away, Rob.

01:04:26.000 --> 01:04:30.000
Hi, I'm Robert Graves. I'm going to be talking about broadband ground motion simulations for the Hayward and Calaveras faults.

01:04:30.000 --> 01:04:43.000
This is work done in collaboration with Jia Wang-Connelly of Cal OES, some folks from the Golden, Colorado USGS office and U.C. Berkeley.

01:04:43.000 --> 01:04:54.000
The motivation for this work the occurrence and impact of the February 2023 earthquake doublet in Turkey begs the question, what if such a sequence occurred in California?

01:04:54.000 --> 01:05:03.000
Here we consider a doublet scenario on the Hayward and Calaveras faults. The first phase involves characterizing the fault ruptures and simulating the ground motions.

01:05:03.000 --> 01:05:16.000
This presentation, and then the next phase involves analyzing the potential impacts on the built environment using the simulated ground motions and that's ongoing and will be presented at a later date.

01:05:16.000 --> 01:05:26.000
The scenario characterization is based on fault locations and orientations taken from the 2023 National Seismic Hazard Model.

01:05:26.000 --> 01:05:34.000
The map on the right shows the locations of the two faults. The black traces are for Hayward and the blue are for the Calaveras.

01:05:34.000 --> 01:05:41.000
We consider bilateral magnitude about 7 rupture in both cases. The epicenters are indicated by the stars.

01:05:41.000 --> 01:05:51.000
In the scenario, the Hayward ruptures first and then the Calaveras several hours later.  However, explicit triggering mechanisms are not considered.

01:05:51.000 --> 01:06:04.000
That is, these are dissimulated as two separate events, but in doing so, we will analyze the accumulated impacts in the region experiencing strong shaking in both events.

01:06:04.000 --> 01:06:14.000
The simulations are done using a hybrid technique where low frequency and high frequency are simulated  separately and then some together into the broadband motions.

01:06:14.000 --> 01:06:23.000
At low frequencies a complete finite fault retro description, full theoretical wave propagation for 3D velocity structures are used.

01:06:23.000 --> 01:06:30.000
I use a finite difference modeling approach with a grid space of 80m, minimum velocity of 400m per second, 

01:06:30.000 --> 01:06:41.000
and this has been referred to as a deterministic calculation. At high frequencies it's a stochastic approach with a simplified finite fault rupture description.

01:06:41.000 --> 01:06:52.000
Source, radiation has stochastic phase and average radiation pattern and simplified greens functions that include travel times, geometric spreading, attenuation, and impedance effects.

01:06:52.000 --> 01:06:59.000
In addition, the high frequencies have sites specific Vs30 adjustment factors.

01:06:59.000 --> 01:07:14.000
For the deterministic calculation, the 3D velocity structures taken from the San Francisco Bay region community you've lost in model version 21.1 In this slide, I'm showing cross sections of shear-wave velocity through that model.

01:07:14.000 --> 01:07:24.000
The cross sections are indicated in map view on the right; and then on left are the panels of the shear- wave velocity as a function of depth.

01:07:24.000 --> 01:07:28.000
You can see in the near surface there's quite a bit of lateral variation in the model.

01:07:28.000 --> 01:07:42.000
This includes sedimentary basins but also boundaries due to changes in geology. Additionally, I've indicated the major fault, San Andreas, Hayward, and Calaveras, and in many cases these faults 

01:07:42.000 --> 01:07:47.000
form media boundaries within the model.

01:07:47.000 --> 01:07:56.000
Some more details about the rupture characterization for the Hayward. It includes the Hayward South and Hayward North sections.

01:07:56.000 --> 01:08:06.000
In the model, I use five segments near vertical orientation, but with slight changes in stripe to represent this.

01:08:06.000 --> 01:08:11.000
There's a slight dip to the east on the fault, it's 95km long,

01:08:11.000 --> 01:08:22.000
and 14km wide. For the Calaveras, I use essential northern sections this is modeled with 6 segments again with some slight deviations in the strike.

01:08:22.000 --> 01:08:28.000
But again, your vertical total length of 100km and 14km down dip.

01:08:28.000 --> 01:08:39.000
The scenario ruptures are created using the Graves-Pitarka rupture generator. For these particular faults, we know that they exhibit creep.

01:08:39.000 --> 01:08:53.000
So how do we handle this? The figure at right is from Tsusard and it shows their model estimates of the for each of the faults.

01:08:53.000 --> 01:09:09.000
The red colors are getting up in to about 10mm per year. The top view is looking at the Hayward fault because so we're looking towards the northeast and then in the bottom view that slipped 180 degrees so the Calaveras is in the front.

01:09:09.000 --> 01:09:19.000
So the strongest creep occurs in this southern Hayward and down into the central Calaveras.

01:09:19.000 --> 01:09:22.000
I'd say it's an open question on how to handle this for a kinematic or even dynamic ruptures.

01:09:22.000 --> 01:09:35.000
What I've done here is a pretty simple approach. I just taper the coseismic slip in these creeping zones using the ratio of the long-term creep rate to the long-term fault slip-rate.

01:09:35.000 --> 01:09:43.000
And that taper is shown by the blue curve that's meant to average this aseismic factor in the upper 10km.

01:09:43.000 --> 01:09:49.000
So how does this look, when applied to the faults? So I first generate a rupture, assuming no creep, that is the 

01:09:49.000 --> 01:09:58.000
aseismicity factor is 0. That's the panel showing at top. And then I apply the slip taper.

01:09:58.000 --> 01:10:05.000
Everything else in the rupture is the same. For the Hayward case, the aseismic factor varies from about point 2 to 3.

01:10:05.000 --> 01:10:10.000
You can see that there's a significant decrease in the slip in the upper part of the fault.

01:10:10.000 --> 01:10:20.000
Below 10km the slip is unchanged the impact in terms of the total magnitude reduces by about point 0 5 units.

01:10:20.000 --> 01:10:26.000
The magnitude used for the simulation is 7.14. Here's a plot showing the 

01:10:26.000 --> 01:10:30.000
approach has applied to the Calaveras. Again, significant reduction of slip in the upper part of the fault.

01:10:30.000 --> 01:10:36.000
For the Calaveras, the aseismicity factor varies from 0 in the far north, which is on the right of this plot.

01:10:36.000 --> 01:10:50.000
Up to about 0.4 in the central region which is to left. Again, the reduction in total moment or total magnitude is about 0.5 units.

01:10:50.000 --> 01:10:55.000
So the simulated magnitude is 7.17.

01:10:55.000 --> 01:11:03.000
So here I'm gonna show some animations. First, the Hayward fault on the left-side panel.

01:11:03.000 --> 01:11:10.000
The rupture begins in the middle and propagates towards both ends. You can see very strong directivity effects.

01:11:10.000 --> 01:11:20.000
Propagating off the ends of the faults. In addition, they're strong coupling into the sedimentary basins down near San Jose, Santa Clara region.

01:11:20.000 --> 01:11:27.000
And then also quite significantly just east of Hayward into the Livermore Basin. We move to the right panel.

01:11:27.000 --> 01:11:35.000
This is the Calaveras fault. Again, there's very strong directivity. It's in your vertical strike-slip fault.

01:11:35.000 --> 01:11:50.000
The coupling into the basins is also apparent, not quite as strong in the Santa Clara region, but much stronger in the Livermore region because the fault is right along the edge of the base.

01:11:50.000 --> 01:11:59.000
So for the output, I say three component broadband waveforms for each rupture scenario on a 1.2km

01:11:59.000 --> 01:12:06.000
z1.2km grid that covers the model the main. That's 240x340km.

01:12:06.000 --> 01:12:13.000
It's indicated in the red box on the right. There's over 50,000 waveforms that are saved.

01:12:13.000 --> 01:12:20.000
Then I extract PGA, PGV and pseudospectral acceleration and these will be used to generate ShakeMaps.

01:12:20.000 --> 01:12:30.000
Additionally, three component broadband waveforms are also saved at a selected number of building locations that will be used later in dynamic structural analyses.

01:12:30.000 --> 01:12:37.000
This plot just shows an example from one location of the output.

01:12:37.000 --> 01:12:42.000
The black traces are for the Hayward scenario and the red are for the Calaveras.

01:12:42.000 --> 01:12:48.000
Acceleration at the top, ground velocity in the middle and ground displacement at the bottom.

01:12:48.000 --> 01:12:58.000
And then the three components of motion and the three columns. This site is very close to both faults, so their ground motions are pretty healthy.

01:12:58.000 --> 01:13:05.000
For both scenarios getting up 0.5g acceleration. Strong pulse-like motions and velocity.

01:13:05.000 --> 01:13:21.000
Peak values above .50cm per second and then pretty strong ground displacement showing also the static deformation due to the fault dislocation.

01:13:21.000 --> 01:13:32.000
As a point of comparison, this plot shows the simulated motions with respect for NGA west to ground motion models.

01:13:32.000 --> 01:13:43.000
This is just a subset of points. The yellow dots come from the simulation. I'm only looking at locations here where the Vs30 between 2 entered and 450m per second.

01:13:43.000 --> 01:13:49.000
For the ground motion models, I'm just using a reference condition of 315m per second.

01:13:49.000 --> 01:13:59.000
And this comparison is just really to make sure that there aren't any obvious pledges or artifacts and just make sure we're in the right ballpark.

01:13:59.000 --> 01:14:10.000
There are some features in the simulations that deviate from ground motion models, for example, in peak ground velocity, we see that there's a cluster of points here that are

01:14:10.000 --> 01:14:19.000
noticeably higher, this is basin response. There's also a cluster point out points out here further distance and this is a rupture directivity.

01:14:19.000 --> 01:14:34.000
You'll also notice, particularly at the shorter periods, these two panels here that in close the simulations are well below the ground motion models and that's due to the slip tapering that we apply.

01:14:34.000 --> 01:14:41.000
So that's for the Hayward, on the right now, similar display for the Calaveras PGV,

01:14:41.000 --> 01:14:45.000
PGA and then the three periods of pseudospectral acceleration. Similar types of features.

01:14:45.000 --> 01:15:02.000
There's, again, a generally favorable comparison. However, in the case of the Caleveras, particularly these longer periods, notice that in close we have some pretty high simulated motions and these are coming from the northern portion of the fault where we don't apply that 

01:15:02.000 --> 01:15:06.000
slip tapering for the creeping zones.

01:15:06.000 --> 01:15:20.000
Okay, so scenario ShakeMaps. These have been produced for both scenarios. Using the same methodology that's used in the generation of shake naps as a USGS product.

01:15:20.000 --> 01:15:30.000
So Hayward on the left, we can see, and what I'm showing here is just for PGV, but we have also other, and metrics as well.

01:15:30.000 --> 01:15:36.000
You can see, starting with a Hayward, there's a very noticeable elongation to the north and the south.

01:15:36.000 --> 01:15:43.000
This is due obviously to the fault geometry, but also due to the rupture directivity effects.

01:15:43.000 --> 01:15:52.000
In close we're getting peak motions that exceed over 50cm per second 

01:15:52.000 --> 01:16:07.000
right along the fault. On the right side, the Calaveras. Similar types of features this elongation is also present but you'll notice in this region here just east of Hayward, the Livermore Basin gets very strong amplifications.

01:16:07.000 --> 01:16:19.000
In this case, exceeding a 100cm per second. So, summary, broadband, ground motion simulations have been run for a Turkey-like doublet on the Hayward and Calaveras faults.

01:16:19.000 --> 01:16:25.000
Ruptures were generated using the Graves-Pitarka method with near-surface slip tapered within creeping zones.

01:16:25.000 --> 01:16:35.000
Broadband emotions, so generally favorable agreement, ground motion models, there are some deviations, so the simulations show strong rupture directivity and basin response effects.

01:16:35.000 --> 01:16:45.000
And then your fault simulated motions along the creeping sections generally are smaller than the ground motion models. Waveforms have been saved on a 1.2km square grid 

01:16:45.000 --> 01:16:55.000
to generate ShakeMaps, scenario ShakeMaps. These ShakeMaps are to be used for has its calculations to estimate impacts on the built environment,

01:16:55.000 --> 01:17:02.000
and the fragilities will be adjusted to account for accumulated impacts in the region, a strong shaking in both events.

01:17:02.000 --> 01:17:11.000
Finally, waveforms, I've been saying that additional locations that will be used later in dynamic structural analyses of selected buildings.

01:17:11.000 --> 01:17:15.000
That's all I have. Thanks. Much for your attention.

01:17:15.000 --> 01:17:38.000
Thank you. Thank you very much, Robert, for your insightful presentation and really cool animations. Our last speaker for this session is Jessica Velasquez from Moody's RMS who will look at a cluster of events from a slightly different angle by navigating the complexities of damage attribution, hours clauses, and progressive damage.

01:17:38.000 --> 01:17:43.000
I am Jessica Velasquez representing Moody's RMS and the team here that you see listed below.

01:17:43.000 --> 01:17:59.000
And I'm going to bring up some discussion points on some of the ways that we talk about and think about modeling and communicating clustered events within the risk modeling space that are a little different 

01:17:59.000 --> 01:18:06.000
from the way that we think about and talk about them in the hazard space.

01:18:06.000 --> 01:18:14.000
First, I want to say that as earth scientists, clustered events are not a surprise to us. We know that earthquakes can trigger other earthquakes.

01:18:14.000 --> 01:18:25.000
We know that aftershocks happen. We know that aftershocks can be large. We know that the first event in a sequence is not always the largest one that it could be followed by an even larger earthquake.

01:18:25.000 --> 01:18:35.000
It's not a matter of knowing that they happen, it's how or whether we should model them for various use cases and how we communicate about it.

01:18:35.000 --> 01:18:42.000
For example, we have a lot of statistical tools in our toolbox in the form of ETOS type modeling and other statistical modeling models.

01:18:42.000 --> 01:18:53.000
They can tell us about how a sequence can unfold, but there's a lot of uncertainties in terms of how the magnitudes are assigned 

01:18:53.000 --> 01:19:02.000
and pretty much as many ETOS models as there are out there, there's that many different spatial kernels to plug into them.

01:19:02.000 --> 01:19:16.000
Another tool that we have in our toolbox is Coulomb stress transfer modeling. We see this after many of the large magnitude events, but how do we do it prospectively and do it in a reliable way?

01:19:16.000 --> 01:19:23.000
Just because they're not a surprise doesn't mean that they're easy to model. It seems like every earthquake and every sequence is unique.

01:19:23.000 --> 01:19:41.000
There have been attempts to characterize global earthquake sequences, but even when you zoom in to the same geographic regions, same tectonic environment, you still see differences in the earthquake sequences one from another.

01:19:41.000 --> 01:19:49.000
Okay, so let's go into the risk challenges. So the first challenge that I want to talk about is damage attribution.

01:19:49.000 --> 01:19:57.000
Before a qualified individual like an engineer or claims adjuster can go out into the field and visit a site and assess the damage 

01:19:57.000 --> 01:20:05.000
they have to pull information about that building. They have to do their research before they even step foot on that site.

01:20:05.000 --> 01:20:15.000
So there's a flowchart here of the process that engineers had to go through to assess bridge damage after the Croatia earthquake, 

01:20:15.000 --> 01:20:21.000
and you can see how long and involved this is, now imagine they have to do that for every single bridge.

01:20:21.000 --> 01:20:42.000
It can take months to years to understand the final damage and loss information. So I have a plot here that was put together by one of my colleagues, Manabo Matsuda, for the recent Japan earthquake and it shows the number of days versus the number of reported damaged buildings.

01:20:42.000 --> 01:20:45.000
And what we see, which is fairly common after large earthquakes is that you have kind of a flat line where you have very little damage 

01:20:45.000 --> 01:21:00.000
reports coming in the initial days after the event because the municipality is everyone is just really focused on life safety, 

01:21:00.000 --> 01:21:11.000
they're trying to make sure that the population has what they need to continue to survive this disaster. So often the hardest hit regions are the last ones to report their damage.

01:21:11.000 --> 01:21:18.000
Claims data, therefore, includes both progressive damage, which I'll talk about in a couple slides 

01:21:18.000 --> 01:21:25.000
and aftershocks because of this time delay.

01:21:25.000 --> 01:21:36.000
Another complication in these risk modeling is the hours clause. If you're a primary and you insure a house that's rattled by one earthquake or 100 earthquakes within the hours clause,

01:21:36.000 --> 01:21:47.000
then it's all considered the same event. Same with a reinsure if you have a portfolio that's damaged by one earthquake or 100 as long as it's within the hours cause it's considered the same event 

01:21:47.000 --> 01:21:56.000
there's no differentiation between events. So if you have your typical aftershock decay, you have your hours clause.

01:21:56.000 --> 01:22:05.000
What's within this hour's clause is covered and damage and loss sustained outside of the hours clause may be covered differently or not covered at all.

01:22:05.000 --> 01:22:15.000
In this example it's not covered at all. So for earthquake sequences, how do we categorize events for reinsurance and insurance purposes?

01:22:15.000 --> 01:22:24.000
And the last challenge that I'll bring up is progressive damage. Once damage a building or its contents can lose resistance to subsequent damage in subsequent events.

01:22:24.000 --> 01:22:36.000
This shows an example here of a building that went through the New Zealand, 2010, 2011 sequence of events and how it sustained damage as the sequence went along.

01:22:36.000 --> 01:22:49.000
To model this requires conditional vulnerability functions and or a shift to fragility functions, which would completely rewrite the way that we currently model risk in the framework that we have.

01:22:49.000 --> 01:22:56.000
And what about things like business interruption or builders risk? For example, if you have a hotel, it sustains damage in the main event.

01:22:56.000 --> 01:23:02.000
Okay, even if it doesn't sustain damage in any of the aftershock events who's going to want to stay there in the midst of an earthquake sequence?

01:23:02.000 --> 01:23:15.000
So you still have business interruption despite their not being physical building, damage to the building after the mainshock.

01:23:15.000 --> 01:23:25.000
This is a Northern California Workshop, so I've thrown in some northern California case studies. These are meant to kind of highlight some of the discussion points that we've had,

01:23:25.000 --> 01:23:31.000
these are not in any way an answer to any of the questions I've brought up so far.

01:23:31.000 --> 01:23:40.000
For the first case study I used OpenSHA to simulate 10,000 five-year periods. Of events on known mapped faults only.

01:23:40.000 --> 01:23:46.000
So this does not include background seismicity. This would preclude any occurrence of a Northridge type event.

01:23:46.000 --> 01:23:54.000
This is in other words a rather conservative simulation. Assume the BPT model for the time dependence within OpenSHA.

01:23:54.000 --> 01:24:06.000
Within these simulations when I talk about a Bay Area rupture, I mean any rupture that ruptures any of these segments in bold on this map.

01:24:06.000 --> 01:24:14.000
Okay, so over two-thirds of the 10,000 periods have more than one fault-based event in them.

01:24:14.000 --> 01:24:24.000
This is important for things like California wide portfolios. So you have sustained more than one California fault-based events,

01:24:24.000 --> 01:24:30.000
within these 5 years.

01:24:30.000 --> 01:24:38.000
Thirty of the 10,000 5-year periods have more than one event on faults in the Bay Area.

01:24:38.000 --> 01:24:52.000
Most of these fault-based events rupture either Hayward plus San Andreas or Hayward plus Calaveras and that's probably not a surprise to anyone familiar with the Bay Area.

01:24:52.000 --> 01:24:58.000
Of the periods with multiple fault-based events in the Bay Area. So, of those 30 in the previous bullet point.

01:24:58.000 --> 01:25:05.000
I'm going to focus on just these red diamonds. So I guess I should explain this figure.

01:25:05.000 --> 01:25:15.000
So these are the Bay Area ruptures within one year of one another. So these within one year I figured was the most relevant for progressive damage.

01:25:15.000 --> 01:25:24.000
If you go beyond a year you're likely to have repairs already done and so your progressive damage is no longer as influential.

01:25:24.000 --> 01:25:34.000
So these are color coded by what period they happen and so some of them happen very very close to each other and some of them happen in more spaced out from one another.

01:25:34.000 --> 01:25:54.000
I chose to focus in on just one example. And so in this example you have a magnitude 8 San Andreas to Calaveras rupture and you have a magnitude 7 Calaveras and Contra Costa rupture and these happen about 3.5 months apart.

01:25:54.000 --> 01:25:59.000
So my assumption is that with 3.5 months difference with all of the other damage that would have occurred within magnitude 8 

01:25:59.000 --> 01:26:10.000
San Andreas rupture, it's unlikely that there would have been repairs already started on on these damage structures.

01:26:10.000 --> 01:26:24.000
I use a simple progressive damage factors from Goda, 2012 it's a very, very simplified approach and just applying factors to the to the damage.

01:26:24.000 --> 01:26:31.000
His world conference on engineering on earthquake engineering paper if anyone's interested in looking.

01:26:31.000 --> 01:26:38.000
So there is modeling framework, we have a stochastic event set of earthquakes and their rates.

01:26:38.000 --> 01:26:48.000
Applied ground motions on top of that. We have industry exposure database. In these examples I'm using economic exposure so not just insured building stock.

01:26:48.000 --> 01:26:54.000
And then calculate the the damage associated with it. And then we go all the way to, you know, applying contracts and treaties and all the financial layers on top of that.

01:26:54.000 --> 01:27:05.000
I'm not going to talk about that in this talk. I'm going to stop at the mean damage ratio.

01:27:05.000 --> 01:27:11.000
The mean damage ratio is exactly what it sounds like. It's the average floss over the replacement value.

01:27:11.000 --> 01:27:18.000
And in this study it includes ground shaking, landslide, and liquefaction.

01:27:18.000 --> 01:27:29.000
Okay. So we have our magnitude 8 large San Andreas structure coming up into the Bay Area and jumping from the San Andreas up onto the Calaveras.

01:27:29.000 --> 01:27:38.000
We have our magnitude 7 Calaveras structure. So we have the same segment of the Calaveras being ruptured in both events.

01:27:38.000 --> 01:27:51.000
So this is what the mean damage ratios look like for the mainshock this is without progressive damage this is just if this event had happened in isolation and there are some very large caveats here.

01:27:51.000 --> 01:28:02.000
For example, if you look at these regions right here, this is sustained high damage ratios. These may be considered complete losses.

01:28:02.000 --> 01:28:10.000
And so having any subsequent damage to them in an aftershock doesn't really change anything about their loss.

01:28:10.000 --> 01:28:16.000
So if they have already been deemed a complete loss and tear down, it needs to be rebuilt.

01:28:16.000 --> 01:28:25.000
If it sustains you know, even more damage. It doesn't change the fact that it needs to be turned down and rebuilt.

01:28:25.000 --> 01:28:38.000
And also keep in mind that if it had been any longer than 3.5 months, you might have started some repair work here but We chose this because it was so closely spaced in time.

01:28:38.000 --> 01:28:46.000
Okay, so this is that same figure from the previous slide without progressive damage, and this is with applying that progressive damage factor.

01:28:46.000 --> 01:28:59.000
And you can see how areas particularly along the fault really light up. And you have MDR changes of up to 31% increases when you apply a progressive damage.

01:28:59.000 --> 01:29:11.000
So this area here I already said, you know, it may not play much of a role in increasing the losses, but if you're to look down here in the South Bay, for example, You go from a mean damage ratio of maybe 10 to 15%.

01:29:11.000 --> 01:29:25.000
Up to a mean damage ratio of say, 25 to 30%. That's a large damage increase, especially considering the very high exposure values down there.

01:29:25.000 --> 01:29:32.000
Okay, and so the the other kind of case study that I wanted to point out is from the HayWired scenario.

01:29:32.000 --> 01:29:40.000
So the USGS within the HayWired Scenario identified 16 aftershocks between magnitude 5 and 6.4.

01:29:40.000 --> 01:29:49.000
RMS Moody's took those aftershocks or events similar to those aftershocks and ran them through our risk calculating software 

01:29:49.000 --> 01:30:14.000
to come up with how much loss would those aftershocks cause relative to the mainshock. In the study do not consider progressive damage and did not consider ours clauses and temporal aspects of claims like I said these are not meant to be answers to anything just discussion points. The total losses from the 16 aftershocks were almost two-thirds of the mainshock losses.

01:30:14.000 --> 01:30:21.000
Okay, you know, keeping in mind this is just one scenario of potential aftershocks from a Hayward event.

01:30:21.000 --> 01:30:28.000
There's underestimations and overestimations. So under estimate we're not accounting for progressive damage.

01:30:28.000 --> 01:30:39.000
Overestimates were not accounting for double counting of losses from the mainshock. So this is an example on the extreme end of you can't collapse a building twice.

01:30:39.000 --> 01:30:48.000
And then how this overestimate and underestimate balance out in the end depends on the overlap of the mainshock with subsequent events.

01:30:48.000 --> 01:30:59.000
Okay, so like I said, I have no conclusions, only discussions. So the occurrence of cluster events is no surprise and we partially know how to model them.

01:30:59.000 --> 01:31:09.000
It will take a cooperative effort to figure out how to fully model them from all parties involved and that is why I want to bring these kind of talking points to your attention.

01:31:09.000 --> 01:31:19.000
Modeling them will look different in the risk modeling space because of the challenges presented here. Retrospective and deterministic studies are worthwhile, interesting, and fun.

01:31:19.000 --> 01:31:26.000
We see them after every earthquake. However, we have a long way to make these into perspective forecasts.

01:31:26.000 --> 01:31:34.000
So what do we need to be able to do this? Besides a million years of observations.

01:31:34.000 --> 01:31:45.000
Do we need more analyses of sequences and what controls their behaviors in different places? Do we need improved fault databases so that we can better assess the Coulomb stress transfers, for example?

01:31:45.000 --> 01:31:53.000
Do we need more non-point source aftershock simulations? And I hope this list continues on and on as you guys add to it. [laughs]

01:31:53.000 --> 01:32:06.000
A side discussion that I wanted to point out really quickly and don't have time to get into. Catastrophe bonds are another insurance kind of product and they have different triggers.

01:32:06.000 --> 01:32:12.000
So there's law space triggers and then there's kind of event based parametric triggers.

01:32:12.000 --> 01:32:20.000
So loss base triggers would suffer from the same sort of damage distribution challenges that we mentioned before, and parametric triggers would consider events individually.

01:32:20.000 --> 01:32:31.000
Depending on the trigger could depend on how we choose to model them when designing them. So maybe we need different ways of modeling them.

01:32:31.000 --> 01:32:40.000
And I thank you very much and I look forward to further discussions and any questions that you may have.

01:32:40.000 --> 01:32:43.000
Thank you.




















01:32:43.000 --> 01:32:57.000
Okay, thank you Jessica and big thank you to all of our speakers. This has been really interesting session for speakers in mind turning your videos on at this time for the discussion.

01:32:57.000 --> 01:32:58.000
Will get going.

01:32:58.000 --> 01:33:09.000
There is a lot of really interesting chatter in the chat.

01:33:09.000 --> 01:33:20.000
Maybe a good place to start the discussion is during Annemarie's session. Annemarie, if you're still here and not tending to the swarm.

01:33:20.000 --> 01:33:30.000
I know, I thought the stats I view of doublets versus aftershocks, for shocks.

01:33:30.000 --> 01:33:44.000
Is one law to rule them all, do doublets have a separate. Class, could you just consider on fault on versus off aftershocks?

01:33:44.000 --> 01:33:45.000
Leave.

01:33:45.000 --> 01:33:51.000
I thought that kind of stimulated a bit of discussion. Do you want to summarize kind of where your response to that and where we ended up.

01:33:51.000 --> 01:34:00.000
Could I jump in just first? Because there's a semantics issue. It seems to me that It's not, whether it's a doublet.

01:34:00.000 --> 01:34:08.000
May not be as important as whether it's just an off fault. But then the issue is, are the, how much models used for anything?

01:34:08.000 --> 01:34:13.000
If the point is just to identify an aftershock and not bias a ground motion model, that's fine.

01:34:13.000 --> 01:34:18.000
But if you start talking about aftershock ground motion models. They're going to be used in any way.

01:34:18.000 --> 01:34:22.000
That's what I'm starting to think about potential issues.

01:34:22.000 --> 01:34:28.000
Yeah, so I guess I would be interested in hearing some clarification for you because I don't understand the.

01:34:28.000 --> 01:34:39.000
The misunderstanding, I guess, is that. We can use these if it aftershock is located to be on fault, then we use the on fault ground motion model.

01:34:39.000 --> 01:34:42.000
And if it's not, then we don't.

01:34:42.000 --> 01:34:46.000
Right. So yeah, the semantics don't matter, but it is an aftershock specific ground motion model going to be used in any way.

01:34:46.000 --> 01:34:56.000
That's what I don't know.

01:34:56.000 --> 01:34:57.000
Okay.

01:34:57.000 --> 01:35:02.000
Right, so right now it's not, but you could for early warning. The event or for a shake map you then use a different ground motion model or for aftershock forecasting you use a different ground motion model.

01:35:02.000 --> 01:35:11.000
And likewise in the hazard map, we Every event is assigned a different ground motion model that's then probabilistically combined together.

01:35:11.000 --> 01:35:16.000
So. I don't think it's a problem if we decide that this is a robust.

01:35:16.000 --> 01:35:22.000
Thing to do to assign. Those particular events to have. That particular ground motion model.

01:35:22.000 --> 01:35:24.000
Okay.

01:35:24.000 --> 01:35:29.000
It's not, in practice it's not being. Used right now, but that's sort of.

01:35:29.000 --> 01:35:31.000
An interesting, you know. Placed it.

01:35:31.000 --> 01:35:37.000
Yeah, if you if you use it, you're going to have to consider on versus off aftershocks.

01:35:37.000 --> 01:35:41.000
And maybe, yeah, did that under a

01:35:41.000 --> 01:35:42.000
Sure, yeah.

01:35:42.000 --> 01:35:44.000
Yeah.

01:35:44.000 --> 01:35:46.000
Andy, did you have a comment?

01:35:46.000 --> 01:35:54.000
Yeah, I was just gonna say that. You know, we do produce, well, Nicholas produces for the international earthquakes and we can do it as follow-on work for domestic ones.

01:35:54.000 --> 01:36:04.000
Aftershock hazard maps and there's absolutely no reason why we can't you know, draw an area on the maps and have the sources inside those.

01:36:04.000 --> 01:36:09.000
Those polygons use one ground motion model at high frequency and outside use another. I mean, we.

01:36:09.000 --> 01:36:14.000
We can do this and we can also use the logic tree. You, you could, you know, partially included, but.

01:36:14.000 --> 01:36:23.000
Yeah, it's really fascinating work and we absolutely can do it. You know, of course it is, and I think it's, really pointed out well.

01:36:23.000 --> 01:36:29.000
We've been discussing this for few weeks actually. Separately. Yeah, there's just a lot of variability.

01:36:29.000 --> 01:36:39.000
And so you might, for instance, in some sequences that she showed. You know, decided to turn off the aftershock model when you realize that now you're in a sequence where the aftershocks are in smaller, you know, lower stress route.

01:36:39.000 --> 01:36:46.000
But, you know, we do the same thing with other after sharp parameters. We update them as we observe the sequence.

01:36:46.000 --> 01:36:56.000
So yeah, I think there's a lot we can use there whether or not we Should do it and clearly we have to do it carefully so that we you know respect the limits of those.

01:36:56.000 --> 01:37:04.000
I could even see that being the case down the line for, you know, for the regular long term hazard models if we were.

01:37:04.000 --> 01:37:10.000
For example, to use kind of an earthquake simulator to generate a really long catalog and use that to generate.

01:37:10.000 --> 01:37:18.000
Hazard map, then you could assign different ground motion models to parameterize them differently depending on their.

01:37:18.000 --> 01:37:19.000
And aftershock.

01:37:19.000 --> 01:37:27.000
Well, actually just in the more standard NSA, if we're going to assign, we're including the full rate.

01:37:27.000 --> 01:37:33.000
And so we know that some percentage of that rate is after shocks. We Most of the aftershocks are going to be close to the main chock.

01:37:33.000 --> 01:37:41.000
So most of them are going to be what is termed on fault in the ground motion models. We can, we could apply a correction to the high frequency ground motions.

01:37:41.000 --> 01:37:52.000
For that excess rate that's above the depluster rate and we could do that right now. Just in the standard logic tree setting.

01:37:52.000 --> 01:37:53.000
Yeah.

01:37:53.000 --> 01:37:54.000
Robert.

01:37:54.000 --> 01:38:05.000
Yeah, just to throw out another. I'll take a pragmatic reason for, you know, flagging these is that in the development of the ground motion models.

01:38:05.000 --> 01:38:18.000
If there are some systematics for particular class of events that you can identify. It's actually worthwhile to separate those out because then that will reduce your variability in the ground motion models, right?

01:38:18.000 --> 01:38:32.000
So you don't want to if there's a, you know, certain class of aftershocks that have lower ground motions, you don't want to include those in the regular ground motion model.

01:38:32.000 --> 01:38:38.000
Thank you. Thank you very much. Any other comments about this topic?

01:38:38.000 --> 01:38:46.000
Alright, there is another conversation in the chat that was the question started with a question from Annemarie to Jessica.

01:38:46.000 --> 01:38:51.000
About hours clauses and how they're defined and you know, Terran made a couple of valuable comments.

01:38:51.000 --> 01:38:58.000
Jessica, would you like to start the conversation about that?

01:38:58.000 --> 01:39:02.000
Sure, so the question was, what's What's the what does an hour's calls look like?

01:39:02.000 --> 01:39:12.000
How is that days? Is it? Is it actually ours? So it's not standardized.

01:39:12.000 --> 01:39:23.000
It's It's variable and it depends on whether you're looking, I believe, and Taren, please feel free to unmute and jump in if I say anything wrong.

01:39:23.000 --> 01:39:29.000
But it's different for primary versus reinsurance contracts too. And there's kind of this average.

01:39:29.000 --> 01:39:40.000
I've seen it's about 7 days. Hours clause, but I, googled before my presentation trying to find out if there was like kind of a standard with some standard deviation around it.

01:39:40.000 --> 01:39:51.000
And I couldn't really come up with an agreement on what that hours clause generally looks like. If we had a standard hours clause, if we knew it was always 7 days.

01:39:51.000 --> 01:40:00.000
Oh, and make modeling these things so much easier. You knew that you could take all of the aftershocks within 7 days and love them together and come up with something, but.

01:40:00.000 --> 01:40:04.000
It's not standard and I don't know if you want to jump in add anything.

01:40:04.000 --> 01:40:09.000
Yeah, I totally agree. I would just say or add to it that. It becomes kind of a game because you can have different types of insurance coverages.

01:40:09.000 --> 01:40:21.000
So when we talk about Primary insurance, right, and things about in the Ca world, it's very different than we're talking about reinsurance.

01:40:21.000 --> 01:40:31.000
So their clients can benefit from multiple reinsurance recoveries. So you would want a very short hours class, right, if we're thinking about earthquake sequences that last.

01:40:31.000 --> 01:40:38.000
Maybe a year or so. But If not, you might want a longer hours clause, right?

01:40:38.000 --> 01:40:49.000
So you might want one big recovery and you can. Accumulate a lot of losses at once. So for us, it comes down within the legal contract and so they don't have a scientific.

01:40:49.000 --> 01:40:58.000
Specific definition to point to. So they stick with the 168 h. And Nobody's challenged it since then.

01:40:58.000 --> 01:41:13.000
So, you know, even just looking at things that have happened in New Zealand or Japan. We just haven't made any movement in the insurance world, but I'm trying very slowly with a good team here at Gay Carpenter.

01:41:13.000 --> 01:41:16.000
Thank you for the time.

01:41:16.000 --> 01:41:24.000
Thank you very much for your comments, Kevin.

01:41:24.000 --> 01:41:25.000
Yup, I'm here.

01:41:25.000 --> 01:41:29.000
Yeah. See, is Nadine still here? I had a question. I, I really enjoyed your talking. I had, I had a question.

01:41:29.000 --> 01:41:37.000
I was wondering how the how the speed at which you were able to. I mean, map all the surface rpture sense.

01:41:37.000 --> 01:41:51.000
Related to its location in Turkey and how that might be different for California. For a California quake, what information might we be getting at what speeds, for example, for a San Andreas event instead.

01:41:51.000 --> 01:41:52.000
Have you thought about that?

01:41:52.000 --> 01:42:10.000
Yeah, that's a great question. It also, so we were primarily mapping from worldview, high resolution optical satellite data in the office, but of course for California then we would have probably NASA collecting data, we would have worldview data coming in, we'd have lots of folks on the ground.

01:42:10.000 --> 01:42:24.000
And so the speed depends on how many people are involved and what kind of data are available. I also imagine the Turkish groups on the ground were flying drones as well and they've mapped the entirety of both ruptures with drones.

01:42:24.000 --> 01:42:42.000
Given how big those ruptures were, how long, that effort took months. And so a similar sized earthquake in California might also take a long time to get drone imagery for the whole.

01:42:42.000 --> 01:42:52.000
Another issue that the satellite data had in Turkey was that there was snow and precipitation immediately following the earthquakes.

01:42:52.000 --> 01:42:54.000
And so in parts of California that might be just as much an issue and in parts of California that might be less of an issue.

01:42:54.000 --> 01:43:08.000
But we would also potentially have the ability to get planes out when we knew that there were good weather windows in California.

01:43:08.000 --> 01:43:09.000
There was a question for you. Sorry, go ahead.

01:43:09.000 --> 01:43:14.000
Okay. Yeah, no, I was just gonna say we also had a huge team of people mapping in the office and there were lots of people in the field as well.

01:43:14.000 --> 01:43:27.000
And so the more people that are working on it, the faster to some extent we could be. Mapping and processing data.

01:43:27.000 --> 01:43:32.000
I don't know if Kristoff is still here. I apologize if I mispronounced your name.

01:43:32.000 --> 01:43:41.000
So you, there was a question for you, about maybe possible utilization of drones in the imagery.

01:43:41.000 --> 01:44:06.000
And I was exactly thinking that when you were talking about the snowy field that CNN had to walk through and I wonder if if you know Kristoff if you wouldn't like to join us and you know if this could be as kind of a supplementation or maybe a terrestrial counterpart let's say I wonder if you if you have comments about that

01:44:06.000 --> 01:44:07.000
Hey Dean.

01:44:07.000 --> 01:44:11.000
Yeah, I think you know in California we would absolutely be out there on the ground trying to fly drones.

01:44:11.000 --> 01:44:22.000
And I'm sure that there would be many people trying to do that. USGS, CGS, academic partners.

01:44:22.000 --> 01:44:47.000
And I think coordinating. Who is, flying drones where and having to put all of that data together is really going to be one of the biggest efforts with a major California rupture, both in terms of coordinating data sets but also in terms of coordinating field teams and coordinating surface structure and displacement measurements and mapping and having systems in place to be able to share those data.

01:44:47.000 --> 01:44:58.000
And put them together is going to make collecting them much, much. Easier.

01:44:58.000 --> 01:45:02.000
Thank you. That sounds great.

01:45:02.000 --> 01:45:10.000
Feel free to raise your hands if you have any. Any questions? We were lively in the chat but required.

01:45:10.000 --> 01:45:11.000
In the discussion.

01:45:11.000 --> 01:45:27.000
I actually have a question for Richard if that's okay. So touching on the societal aspects here, you know, people's intuition probably doesn't tell them that there will be aftershocks and maybe tell them that aftershocks will let's say have a smaller magnitude.

01:45:27.000 --> 01:45:48.000
So what do you think are the societal implications of having a sequence like that? You know, challenging, prevailing public perceptions and what might be the potential cycle implications of such a seismic pattern that let's say differs from those single large earthquake patterns that were used.

01:45:48.000 --> 01:46:00.000
Yeah, okay, so, I mean, as I mentioned, that It was the first earthquakes, the first one especially, that really caused the majority of deaths.

01:46:00.000 --> 01:46:06.000
You have the, you know. I guess in these kind of, you know, magnitude 6.3 type earthquakes in a place with, dominant Adobe construction such as those.

01:46:06.000 --> 01:46:36.000
There's real, dwellings. You get very high levels of destruction in a in a small area that rapidly diminishes right so people didn't have a chance in the first one right in the later ones people were living outside they were living in tents so the destruction continued in those later events, they were sufficiently far away that they were hitting places, villages that weren't so badly damaged in the first

01:46:41.000 --> 01:47:08.000
ones. But the the death tolls did not rise significantly within their in terms of long term effects I don't know so much right so this is more more speculations but we've got 2 issues one is that in Afghanistan right now there's There's a lot of, you know, there's issues in terms of long-term famine and just you know within those areas that people were already in a very non

01:47:08.000 --> 01:47:32.000
resilient situation. In terms of the rebuilding, the local, UN offices have been very proactive in, in, in trying to rebuild and also trying to promote rebuilding in the vernacular style in the in the local styles of building but introducing some basic earthquake resilience measures.

01:47:32.000 --> 01:47:45.000
So, you putting in horizontal banding for instance a roof band to try and keep those walls together trying to make sure that the roof is actually remaining stable and not collapsing.

01:47:45.000 --> 01:47:57.000
So in those respects. You know, working in a very difficult situation, things have managed to work out.

01:47:57.000 --> 01:48:12.000
Br well in terms of the relief that's been put in but the people there are basically already in a non resilient situation and I can imagine we are going to have repercussions in the way that the people are living there.

01:48:12.000 --> 01:48:14.000
Well, yes, decades, right?

01:48:14.000 --> 01:48:23.000
Yeah, there's probably, you know, a heightened fear now that they know that one earthquake could be followed by one, another 1, 2, 3.

01:48:23.000 --> 01:48:33.000
Completely and one thing that I mean is always struck me I mean I showed examples from east of Iran earthquakes that have good in recent decades.

01:48:33.000 --> 01:48:40.000
And one thing is that there's always this process of rebuilding in that area, right?

01:48:40.000 --> 01:48:48.000
But somehow the lessons are in terms of, you know, the preparation of building. I know that these are very resource poor areas, right?

01:48:48.000 --> 01:48:59.000
But There's no building outwards from this. I mentioned that, you know, the UN offices are trying to rebuild in a, in a more resilient way in terms of the buildings.

01:48:59.000 --> 01:49:07.000
But Still, just done there, right? Where the villages were destroyed. But actually there's still a large part of that thrust system that has not ruptured.

01:49:07.000 --> 01:49:10.000
Likewise there are many other places around Afghanistan and Iran. In Iran that are, you know, still at risk from earthquakes.

01:49:10.000 --> 01:49:23.000
And, and I think one of the, one of the real difficulties, one thing that doesn't get done is to actually try and take those lessons and apply them elsewhere.

01:49:23.000 --> 01:49:31.000
You get the rebuilding in that place. But it doesn't, the lessons are not spread more widely.

01:49:31.000 --> 01:49:33.000
Thank you. Very much.

01:49:33.000 --> 01:49:36.000
Sue, did you have a question?

01:49:36.000 --> 01:49:41.000
Yeah, this is something different and I really just wanted to bump up a question that Brian Olsen asked to Rob.

01:49:41.000 --> 01:49:48.000
I didn't see an answer and it's something I've wondered myself having worked with Rob on these animations a little bit.

01:49:48.000 --> 01:49:50.000
Why do they always rotate?

01:49:50.000 --> 01:49:53.000
Yeah.

01:49:53.000 --> 01:49:56.000
Yeah, well, why not?

01:49:56.000 --> 01:49:58.000
To prove that you can rotate them that they're.

01:49:58.000 --> 01:50:07.000
Yeah, yeah, maybe so. Yeah, so, so it's,

01:50:07.000 --> 01:50:19.000
Really just visual stimulation to be honest. The the animation obviously has information behind it. And conveying that information is important.

01:50:19.000 --> 01:50:26.000
You know, that's that's what we do. That's our jobs. You know, it doesn't have to rotate.

01:50:26.000 --> 01:50:36.000
It could rotate the other way. You know, there's various ways to do it. So there's not a It's not something that, had focus groups to study, you know, what should be done.

01:50:36.000 --> 01:50:43.000
Just came out of my head and I will go with the consensus if folks don't like them rotating.

01:50:43.000 --> 01:50:44.000
I'll turn it off.

01:50:44.000 --> 01:50:50.000
It worked for me, but I'm gonna show one of your animations in my talk on Northridge.

01:50:50.000 --> 01:50:55.000
So before Brian asked me the same question I figured I would ask you. Thank you.

01:50:55.000 --> 01:51:05.000
So I have a question for Anne Marie.

01:51:05.000 --> 01:51:06.000
I'm here.

01:51:06.000 --> 01:51:07.000
Hey, Anne Marie is still around. I don't know. Maybe she. Okay, so and I and I did kind of touch on this in the in the chat.

01:51:07.000 --> 01:51:19.000
And it's, I'm just, I'm intrigued when you showed the ground motions for Turkey, PGA and PGV.

01:51:19.000 --> 01:51:26.000
And for both of the large events, the PGV. Just it looks larger than the ground motion model.

01:51:26.000 --> 01:51:38.000
PGAs are kind of fit, right? And that touches on the nature that. An event term isn't.

01:51:38.000 --> 01:51:39.000
Yeah.

01:51:39.000 --> 01:51:50.000
Necessarily just static across all frequencies, right? It can have a frequency dependence. Yeah, but, so I'm just Curious if you or someone else may be grace has looked into that anymore for those turkey events.

01:51:50.000 --> 01:51:57.000
You know, as to what might be going on is it directivity, this, you know, big large long period of ground motion, something like that.

01:51:57.000 --> 01:52:03.000
Yeah, so I'll say that, Grace has done all this work. So, she gets all the credit for this stuff.

01:52:03.000 --> 01:52:13.000
But I think, and when we look across all the periods for Turkey, and we looked at a whole suite of events that Brad Aguard processed, something like 65 events including the 2,010.

01:52:13.000 --> 01:52:25.000
6 point something, LASIK. And Yeah, the event terms are not constant across period.

01:52:25.000 --> 01:52:26.000
Okay.

01:52:26.000 --> 01:52:37.000
And I actually had a plot that I took out because I didn't wanna belabor the point but right the event terms can vary with with period right and so that's reflecting some you know frequency dependent difference in the spectrum right so we expect that right.

01:52:37.000 --> 01:52:55.000
But the, I guess let me just make the point that I was making here and Grace made in the chat is that there's no difference between the 7.5 and the 7.8 sort of in the in those biases right so I'll send it to you Rob but the You know, the PGVs are high for both of them and then the 1 s is low for both of them. Right?

01:52:55.000 --> 01:53:03.000
So there's not anything special about the 7.5, for example, being an aftershock, right?

01:53:03.000 --> 01:53:05.000
That was kind of the point I was trying to make there.

01:53:05.000 --> 01:53:11.000
Right, yeah, and that point is clear. I just, I was kinda going, beyond that.

01:53:11.000 --> 01:53:30.000
Sure, yeah, so, right, we have looked in more detail at, so now we have some like, 65 events, so decomposing all of the event terms and the site terms and the directivity and we'll send you the poster that we put together for AGU, but it looks like there is some directivity but that Seems like primarily the large ground motions are coming from the site

01:53:30.000 --> 01:53:34.000
amplification. And I don't know if there she wants to chime in on this, but.

01:53:34.000 --> 01:53:44.000
And especially along the, the southwestern portion of the rupture. There seemed to be a lot of large site amplification.

01:53:44.000 --> 01:53:45.000
Yeah, okay, okay, interesting.

01:53:45.000 --> 01:53:52.000
And it's particularly prevalent, in the PGV that PGB, I think, had the largest event term outlier.

01:53:52.000 --> 01:54:06.000
So for that for those events for that period that's where we see the largest amplification so yeah it's interesting and that's I think that's kind of the extent that we've looked at, but there certainly could be more.

01:54:06.000 --> 01:54:07.000
Work. Yeah, Grace, do you wanna comment?

01:54:07.000 --> 01:54:14.000
I, yeah, I agree with everything you said. Pgb and also the long response spectral period.

01:54:14.000 --> 01:54:28.000
Positive event turns so larger than average ground motions. Like 2 s and greater. And it's, I think what we're seeing is that it's really a combination of the directivity and side effects.

01:54:28.000 --> 01:54:34.000
In the areas off the ends of bolts, right? It's like, that classic.

01:54:34.000 --> 01:54:39.000
Double whammy case, right, where you have. Directivity into an area of soft sediments.

01:54:39.000 --> 01:54:50.000
Okay. Yeah, yeah, interesting. Okay, thank you both of you for the additional explanation.

01:54:50.000 --> 01:54:51.000
Are there any questions for any of our speakers?

01:54:51.000 --> 01:54:53.000
Good. Andy has his.

01:54:53.000 --> 01:55:01.000
Yeah, actually, this will go back to Richard. Perfect. Yeah, so.

01:55:01.000 --> 01:55:06.000
Okay, so here. The question I actually have is how people decided when to stop living intense.

01:55:06.000 --> 01:55:17.000
I mean, that's a very sensible, low impact thing to do and It was actually done in Italy 1 point where I think a local government just said, Hey, there's a pilot in front of city hall.

01:55:17.000 --> 01:55:25.000
Feel free to go, you know, take one home. And, one thing, the reason I'm asking is one thing we're wrestling with is whether or not to release or how to release and communicate forecasts.

01:55:25.000 --> 01:55:42.000
Outside of the US. Because we actually do provide forecast for some sequences. Inside the federal government for guide humanitarian response.

01:55:42.000 --> 01:55:46.000
But those forecasts currently are not available. Publicly. So, yeah, I'm sort of curious how people made that decision.

01:55:46.000 --> 01:56:00.000
They can certainly do it in lots of ways.

01:56:00.000 --> 01:56:01.000
Yeah, yeah.

01:56:01.000 --> 01:56:05.000
So this is the question was how long it takes until people actually start moving back into. Yeah. I doubt they have actually.

01:56:05.000 --> 01:56:16.000
I mean, you've got, remember there's a couple of competing things here though because they're in the middle of winter in quite a mountain as part of the middle of winter in quite a mountain as part of Afghanistan so it's very cold.

01:56:16.000 --> 01:56:25.000
Probably I don't know for sure, but it's probably, you know, it's not, exactly pleasant to be out in the tents as well in terms of the climatic conditions.

01:56:25.000 --> 01:56:31.000
My expectation, I don't know, my expectation is that people are still living in tents.

01:56:31.000 --> 01:56:47.000
And my expectation is that that will continue. For quite a long time using examples from Iran of the past few decades that actually the process of moving back into permanent dwellings has been a long one.

01:56:47.000 --> 01:57:13.000
The other thing as well is the people will. Yeah, so examples from Iran, again, where earthquakes have occurred and where the rebuilding has been done by central government, sometimes in a non in a way that is not so suitable for the local conditions you know small concrete dwellings or whatever in in desert environments the people had moved into those and then very rapidly they've moved out of them again.

01:57:13.000 --> 01:57:26.000
Right, so you go through this period that homes have been provided but they don't like them. Everyone moves out again and then they go through a process of rebuilding themselves.

01:57:26.000 --> 01:57:34.000
Typically in the villages that we destroyed. So they will go back to the areas and they will rebuild from the ruins to recreate their homes.

01:57:34.000 --> 01:57:43.000
So there's a big challenge not only in the process of persuading, you know, getting people to move back, but also doing it in a way that's appropriate for their way of life.

01:57:43.000 --> 01:57:49.000
And that's one of the big issues in these areas.

01:57:49.000 --> 01:57:52.000
Yeah, thank you so much. I appreciate the answer.

01:57:52.000 --> 01:57:53.000
Thank you.

01:57:53.000 --> 01:58:01.000
If there are no questions from, are there any questions from our audience?

01:58:01.000 --> 01:58:09.000
Okay, I actually have a question for Geen. Because I'm always thinking about for elastic effects.

01:58:09.000 --> 01:58:23.000
Would you please comment on what you think may be a result of the interplay between poor elastic effects and stress shadows and how that might influence the patterns and intensity of seismic events.

01:58:23.000 --> 01:58:28.000
Yeah, that's certainly something I kind of didn't have time to talk about in my talk, but.

01:58:28.000 --> 01:58:38.000
One of the earlier talks had this nice chart of all of the, you know, 8 different things that can, that can trigger aftershocks.

01:58:38.000 --> 01:58:47.000
And so I would assume that that things like poor elasticity. This go elasticity some of these would be kind of longer term.

01:58:47.000 --> 01:58:58.000
Processes that we'd see more over kind of months to years to maybe decades. So we're kind of outside of the immediate aftershock sequence.

01:58:58.000 --> 01:59:21.000
So we're kind of outside of the the immediate aftershock sequence that I was showing you but yeah there's certainly there's certainly a lot of literature on how, how these kind of longer term the you know as the crest readjusts after these stress changes either viscoelasticity viscoelastically or pore elastically can certainly continue continue the triggering process.

01:59:21.000 --> 01:59:32.000
Yeah, so that's that's that's definitely in there, but I think it's working on different timescales than the dynamic triggering which looks like from from what I was looking at the dynamic triggering maybe would last up to a few days or or a couple of weeks.

01:59:32.000 --> 01:59:43.000
And that's that's actually a little bit of a physical mystery to kind of what's the the process that allows that allows that delayed dynamic triggering.

01:59:43.000 --> 01:59:57.000
But that again could also be, you know, maybe poor fluid changes if the passing seismic waves like, you know, unclog some channels and then.

01:59:57.000 --> 02:00:03.000
You know, there's poor fluid diffusion that that kind of changes the stresses and changes the the poor pressure.

02:00:03.000 --> 02:00:08.000
So those those those processes could kind of immediately interact like that. Yeah, that's a good question.

02:00:08.000 --> 02:00:10.000
Thank you.

02:00:10.000 --> 02:00:12.000
Thank you so much. Great answer.

02:00:12.000 --> 02:00:23.000
Hmm. Okay, well we are out of time here. I'd like to just once again thank all of our all of our speakers for really excellent presentations.

02:00:23.000 --> 02:00:27.000
Simulated a lot of great discussion.

02:00:27.000 --> 02:00:35.000
Thank you all very much. Great cox and great audience and great discussion. And thank you to the organizers.

02:00:35.000 --> 02:00:41.000
Well, and thank you to all moderator within our speakers. That was wonderful. Thank you so much.

02:00:41.000 --> 02:00:49.000
And clearly we have launched so much and also unloaded so much because we apparently no longer know what earthquakes or doublets mean.

02:00:49.000 --> 02:00:50.000
So, this has been delightful. We have one more thing to delight you with after a brief break.

02:00:50.000 --> 02:01:03.000
So everybody take 15 min, come back at 2 30 PM. Except if you are in our next session, please hang around so we can help you get set up.

02:01:03.000 --> 02:01:06.000
But that would be Ward, McDawson, Hath Schwartz, Wine, Mcbride, Shelley, Stewart, Johnson, Askin.

02:01:06.000 --> 02:01:14.000
Everyone else? Take 15. Those are