WEBVTT

00:00:00.000 --> 00:00:00.000
[Okay, I'm ready when you're ready.

00:00:00.000 --> 00:00:03.000
Here we go. Recording now.]

00:00:03.000 --> 00:00:07.000
I hope everybody had a good lunch or whatever meal is most appropriate for your time zone 

00:00:07.000 --> 00:00:13.000
and welcome back. We will continue right now with the "fault not in our stars, but over here, 

00:00:13.000 --> 00:00:23.000
maybe there." And to get us started, we have two excellent moderators in the form of Evan Hirokawa and Artie Rogers. [Woo-hoo!]

00:00:23.000 --> 00:00:43.000
Okay, so yeah, this session is going to be kind of about fault geometries and potential hazards associated with faults. The first talk is going to be by Scott Marshall, the SCEC Statewide Community Fault Model, CFM 7.0.

00:00:43.000 --> 00:00:48.000
Hello, everyone. My name is Scott Marshall and I'm excited today to talk to you about the SCEC Community Fault Model, 

00:00:48.000 --> 00:00:56.000
which is now statewide. It covers Northern California and beyond. I'm also going to give a plug for some of the other SCEC community models in this talk,

00:00:56.000 --> 00:01:11.000
but first, I got to give a shout out to my collaborators, Andreas Plesch and John Shaw, who've been building this community fault model at SCEC for well over 20 years, and this project obviously would be nothing without their expertise and countless years of time they put into this.

00:01:11.000 --> 00:01:18.000
So SCEC is now the statewide California Earthquake Center, and we've been the Statewide Center for a little over a year at this point 

00:01:18.000 --> 00:01:24.000
and our community models we refer to as Community Earth Models. And so we call those CEMs.

00:01:24.000 --> 00:01:37.000
The idea behind CEMs is we're trying to leverage the expertise of the SCEC Community and beyond to build these models that quantify key physical parameters related to earthquake science.

00:01:37.000 --> 00:01:51.000
So under our community earth models umbrella or CEM umbrella, we have a fault model, a geodetic model, a rheology model, a stress model, a thermal model, a velocity model, and then a software package that delivers those velocity models in a uniform format.

00:01:51.000 --> 00:02:04.000
So we have quite a bit of stuff available. If you go and look for these models during my talk or after, just know that we've only been a Statewide Center for about a year. So we have some statewide models and Northern California models available,

00:02:04.000 --> 00:02:11.000
but we're currently in the process of updating the other ones to expand their geographic scope.

00:02:11.000 --> 00:02:19.000
Community models cannot be successful, in my opinion, without a healthy and collaborative relationship with IT professionals.

00:02:19.000 --> 00:02:27.000
And the SCEC IT and web team has been fantastic to work with. Tran Huynh, Phil Maechling, Edric Pauk, and Mei-Hui Su.

00:02:27.000 --> 00:02:31.000
And a special shout out here to Mei Wei Su, who is our community model software engineer.

00:02:31.000 --> 00:02:40.000
She's built these CFM Explorer that you're going to see a lot and all the model explorers in collaboration with the scientists that develop these models.

00:02:40.000 --> 00:02:44.000
So I hope that you like these explorers that we've built.

00:02:44.000 --> 00:02:50.000
What we've been doing recently is we've been focusing a lot of effort, obviously, on expanding the geographic scope of our community earth models.

00:02:50.000 --> 00:03:01.000
But also making sure our models meet FAIR software standards. And if you're not in the software world, you might not know what FAIR stands for. So I'll give you the very short version of what it means. We don't want to get too much in the weeds here.

00:03:01.000 --> 00:03:09.000
But "F" is findable. Can I find these community models? Are they posted at an address that you can find? Are they Google searchable?

00:03:09.000 --> 00:03:16.000
Accessible means once I find them, can I actually download and interact these models or do I need some special code to do it or something like that?

00:03:16.000 --> 00:03:23.000
Interoperable means, is the data in a format that I can use and work in my workflow?

00:03:23.000 --> 00:03:34.000
And reusable is similar, but it basically means, are these models sufficiently well documented so that we can use these and reuse them in future work and maybe even build upon them.

00:03:34.000 --> 00:03:43.000
So we hope that you'll find our models FAIR compliance. And if not, please let us know so we can improve in those respects. But let me tell you how we've tried to meet these challenges.

00:03:43.000 --> 00:03:51.000
So the findable and accessible thing we've posted the community models at the SCEC homepage. If you go to the SCEC homepage, you'll see something very similar to this.

00:03:51.000 --> 00:04:00.000
If you click on research or computing. A pop-up will come and you'll see community earth models are listed right there,

00:04:00.000 --> 00:04:07.000
under research or computing. So if you click on the community earth models link, what you'll get taken to is the community earth models homepage.

00:04:07.000 --> 00:04:13.000
Which looks like this. You'll see that the main feature of the homepage is this little gray widget here.

00:04:13.000 --> 00:04:19.000
That allows you to scroll through all the community models that we have available really quickly with the left and right arrows.

00:04:19.000 --> 00:04:25.000
And it gives you quick links to their documentation homepage, their online web-based explorer,

00:04:25.000 --> 00:04:29.000
and there's Zenodo archive if they have one. So we hope that this makes it really, really simple.

00:04:29.000 --> 00:04:38.000
Overall, we have eight items in our widget. So these are our six community models and we also have a geologic slip rate database that's available.

00:04:38.000 --> 00:04:47.000
And a precariously balanced rocks database that's available. So again. Please use these models. They're really great. We put a lot of effort into them. And if you don't find it in Northern California yet 

00:04:47.000 --> 00:04:53.000
stay tuned. We're trying to expand all these models to cover the entire San Andreas plate boundary.

00:04:53.000 --> 00:04:58.000
So let's talk about the CFM while you're here. The CFM widget looks like this.

00:04:58.000 --> 00:05:04.000
Here's our link to our model homepage. This is a link to our explorer and this is linked to our Zenodo archive, just like the other pages.

00:05:04.000 --> 00:05:16.000
Our homepage looks like this. You'll notice that it's at a southern.scec.org address at the end of the Southern California Earthquake Center, we sort of archived and froze all the web pages,

00:05:16.000 --> 00:05:25.000
but just due to limited personnel time, we haven't updated all these pages for the Statewide Center, but all this documentation for the community models is still correct and valuable.

00:05:25.000 --> 00:05:32.000
Just know that we're now in a statewide fault model, but all the rest of the information is correct.

00:05:32.000 --> 00:05:41.000
Hopefully we'll update those in the upcoming months. If you click on the Explorer, you're going to be taken to a web-based tool that looks like sort of like a Google map. This is a leaflet based interface 

00:05:41.000 --> 00:05:47.000
that we've built that allows you to quickly browse this model with no specialized software required at all.

00:05:47.000 --> 00:05:53.000
You can click on the faults to select them and download them. You can also change the base maps to a whole bunch of different options.

00:05:53.000 --> 00:05:59.000
You can load in relocated seismicity from Waldhauser and Hauksson
that we've combined into a single catalog.

00:05:59.000 --> 00:06:11.000
And look at significant events since 1900; that are magnitude six and above. It also hosts version 7, version 6.1, and version 5.3 of the CFM, all in this interface.

00:06:11.000 --> 00:06:20.000
The cool thing too is you can even select faults and then click plot 3D down here and actually view them in 3D on your computer with no specialized software at all.

00:06:20.000 --> 00:06:28.000
So this is a great thing for people that want to quickly look at the model and see if it's something they want to use in their work.

00:06:28.000 --> 00:06:32.000
Okay, so the CFM has these cool tools, but what is the CFM?

00:06:32.000 --> 00:06:40.000
Well, the CFM is this object-oriented, fully three-dimensional geometric representation of faults in California covering the San Andreas System.

00:06:40.000 --> 00:06:50.000
The developers are there. I've already mentioned them. It's myself, Andreas Plesch, and John Shaw. And what you see here is an oblique view of the model rendered using GOCAD.

00:06:50.000 --> 00:06:58.000
Now, there are different fault models available for users to download out there, and these are distinct products. And so I want to make sure we're all on the same page here.

00:06:58.000 --> 00:07:08.000
On the left, I've plotted the USGSQ faults. This is no longer actively maintained. The state surveys are going to take this over, but I don't think there's any active development yet.

00:07:08.000 --> 00:07:14.000
This is essentially like an encyclopedia of everything that's been published and mapped on faults ever.

00:07:14.000 --> 00:07:26.000
Up to maybe about 10 years ago. It's amazingly comprehensive, but it's solely two-dimensional. It's just fault traces. Sometimes you get information about DIP, but basically it's just fault traces. So this is sort of like an encyclopedia.

00:07:26.000 --> 00:07:32.000
The National Seismic Hazard Model was based on the UCERF model, which was based on the community fault model,

00:07:32.000 --> 00:07:47.000
but it covers a much larger footprint now. And it's geometrically smooth compared to the CFM. And also it has no variations in down dip angle. So Alex Hatem, the primary author on this, would like to say it's 2.5D.

00:07:47.000 --> 00:07:52.000
It's a three-dimensional model, but there's no variations down dip, only a long strike variations.

00:07:52.000 --> 00:07:57.000
The CFM, however, is fully three-dimensional and we have variations in strike, dip,

00:07:57.000 --> 00:08:05.000
irregular tip lines, faults cutting other faults, very complicated model. It's as complicated as the data needs.

00:08:05.000 --> 00:08:13.000
So to get an idea of what these complicated fault surfaces look like, here I've rendered one using a MATLAB script that's freely available on our homepage.

00:08:13.000 --> 00:08:21.000
What you see down here is a histogram of all the dip angles of all these triangles and a rose diagram of all the strikes. So you can see it's non-planar.

00:08:21.000 --> 00:08:28.000
This is the Mount Diablo thrust. And if I rotate it around in 3D, by the power of movies 

00:08:28.000 --> 00:08:36.000
you can see that this is a complicated surface that would not be representable with a simple 2.5D representation of here's a trace and here's a dip angle.

00:08:36.000 --> 00:08:41.000
You can see that near the surface of the Earth, the dips are steep, over 80 degrees,

00:08:41.000 --> 00:08:49.000
and as you get deeper and deeper, the fault shallows out. It's like a listric fault geometry, and the dips at the bottom get to almost 10 degrees.

00:08:49.000 --> 00:08:56.000
So the CFM can represent this kind of geometry. The challenge of the CFM is uncertainty 

00:08:56.000 --> 00:09:02.000
because if you say, well, what's the dip uncertainty? Well, every single triangle would have to have an uncertainty. And if you change one triangle,

00:09:02.000 --> 00:09:09.000
it's dependent upon the dip of the next triangle because if you just change one triangle and not the one next to it, you're going to have a hole in your fault.

00:09:09.000 --> 00:09:14.000
So for that reason, we don't provide a quantitative estimate of error with the CFM,

00:09:14.000 --> 00:09:24.000
but what we do provide is where there are published data to suggest different structural interpretations that are significantly different we provide those in our what's called alternatives model,

00:09:24.000 --> 00:09:32.000
and so you'll see that some faults have drastically different representations depending on whose publication you use.

00:09:32.000 --> 00:09:40.000
CFM is now out. The statewide model, it's version 7.0. The Northern California portion should be considered preliminary. It's brand new.

00:09:40.000 --> 00:09:51.000
This model has no changes in Southern California from the last version. Our entire effort over the last two years or so has been to make these 113 new representations in the model.

00:09:51.000 --> 00:09:58.000
This model and the past models of CFM have been peer reviewed. We're currently starting up the peer review process right now.

00:09:58.000 --> 00:10:01.000
And I'll talk about that at the very end of this talk. We need your help.

00:10:01.000 --> 00:10:15.000
We have a two-dimensional representation for people that just want traces, and one of the new changes this year is our KML version for Google Earth or QGIS now has all the attributes loaded in.

00:10:15.000 --> 00:10:21.000
As official attributes, so QGIS can see them. Whereas before they were just sort of just fault traces.

00:10:21.000 --> 00:10:28.000
So this is the East Bay area. And if you look at the Hayward Fault, if you were to load this into Google Earth and click on the Hayward Fault.

00:10:28.000 --> 00:10:33.000
You get a nice little webpage that tells you all the metadata available for the community fault model and that particular object.

00:10:33.000 --> 00:10:39.000
So it's pretty cool. We hope you try it out. Let us know if you like it or have any issues.

00:10:39.000 --> 00:10:45.000
The complete archive is available for download at Zenodo and you can download all the past model versions as well.

00:10:45.000 --> 00:10:56.000
But you're probably saying, well, what's in the archive? Well, some stuff's in the archive that's not available through the web-based tools because the web-based tools are meant to be just simple tools for allowing you to browse the model.

00:10:56.000 --> 00:11:02.000
So what is the directory structure of the CFM archive actually look like? Well, it looks like this.

00:11:02.000 --> 00:11:14.000
In the root directory, you're going to have two directories called doc and obj. Well, the documentation directory contains a reference list in Word format and documentation in Excel format. It's a spreadsheet.

00:11:14.000 --> 00:11:20.000
There's 19 columns here of metadata that include things like references, average strike and dip, area.

00:11:20.000 --> 00:11:31.000
The object directory contains all the fault objects. We have three separate models, the alternatives, the preferred model, and the ruptures model, which is just faults that have had historic ruptures.

00:11:31.000 --> 00:11:37.000
Each of these has subdirectories, but what I've done is I've just blown up one of them here to avoid cluttering the screen.

00:11:37.000 --> 00:11:54.000
Each model directory is going to have the three-dimensional objects in the native mesh 2000 meter, 1,000 meter, and 500 meter meshes. I'll show you those in a minute. And all the traces in different format, everything from plain text to QGIS KML, to ArcGIS shape files.

00:11:54.000 --> 00:11:59.000
Again, if there's some format you need that we don't provide, let us know. We might be able to convert it for you quickly.

00:11:59.000 --> 00:12:05.000
So what are these 3D native 2,000 meter meshes. What's that all about?

00:12:05.000 --> 00:12:10.000
Well, what I'm showing here is, again, I'm using that MATLAB script that's publicly available.

00:12:10.000 --> 00:12:15.000
This is the Northridge Thrust in Southern California that was host of the 1994 6.7 event.

00:12:15.000 --> 00:12:18.000
What you can see is there's very dense triangles in here.

00:12:18.000 --> 00:12:23.000
And this is the region where there's a lot of data. This is the relocated seismicity from Carena & Suppe.

00:12:23.000 --> 00:12:31.000
So the native mesh shows you that this is well constrained because there's a lot of elements. Out here where the triangles are very sparse 

00:12:31.000 --> 00:12:43.000
that's based on a dislocation model from Hudnut and others where there's less constraint. So the native mesh allows users to quickly look at a surface and know where we have data and where we don't have data, at least in a qualitative sense.

00:12:43.000 --> 00:12:49.000
Now, if you're a computational modeler, you're looking at this going, I don't want to model that.

00:12:49.000 --> 00:13:01.000
We know. So that's why we provide these regularized meshes. They're not exactly regularized, but they're pretty close to having equilateral triangles and same surface area per element.

00:13:01.000 --> 00:13:07.000
So this is the 500 meter mesh and you'll see that the triangle size is very consistent. We hope these are modeler ready.

00:13:07.000 --> 00:13:14.000
So if not, try them out, let us know. You might be thinking now, well, but what format is this data in?

00:13:14.000 --> 00:13:24.000
Well, it's in Gocad format. And the number one misconception we hear with the community fault model is I can't use the community fault model because I don't have Gocad. Gocad's proprietary.

00:13:24.000 --> 00:13:32.000
That's true. GOCAD files are not proprietary. Gocadzz files are open source. They're plain text. Anything can read them.

00:13:32.000 --> 00:13:37.000
A lot of CAD software packages read them natively. So that's why we distribute in this format.

00:13:37.000 --> 00:13:47.000
To create a three-dimensional fault object, basically what you need are what are the XYZ locations of the vertices. Here's just three triangles from that Northridge thrust.

00:13:47.000 --> 00:13:52.000
And you'll see that this is a snapshot of what the Gocad file looks like in a text editor.

00:13:52.000 --> 00:13:57.000
And you'll see it's just a bunch of vertices. They're X, Y, Z coordinates, and then how they're connected in 3D space.

00:13:57.000 --> 00:14:14.000
That's it. It's really basic. So you can write scripts to parse this stuff. CAD software can read it. You don't have to have Gocad. I don't have Gocad And I'm a developer of the community fault model. So please don't think that that's a hindrance. This is open source stuff that we're providing here.

00:14:14.000 --> 00:14:20.000
The CFM is only a geometric model. We do not provide estimates of slip rates.

00:14:20.000 --> 00:14:32.000
But SCEC does have a slip rate database that is publicly available. It's derived from the NSHM model, and we're hoping to update that this year, and it has a really cool web tool that makes it really easy with clickable references.

00:14:32.000 --> 00:14:38.000
We have used the community fault model in lots of different crustal deformation models, my group especially 

00:14:38.000 --> 00:14:53.000
and Michelle Cook's group has done a lot of this. But our most recent paper comes from my students, Savannah and Hugh, where they actually simulated 83 faults full three-dimensional slip distributions on these faults, and these are some really complicated non-planar faults in Southern California.

00:14:53.000 --> 00:15:01.000
And what we found was we were able to predict slip rates that fit the existing geologic data really well with a low RMS error.

00:15:01.000 --> 00:15:09.000
So check out that paper. It's an SRL. If you're interested about slip rates in a full three-dimensional sense.

00:15:09.000 --> 00:15:19.000
The next question becomes, well, how can we evaluate if the CFM is actually complete or not? And that's almost a philosophical question, but I guess one question we could ask that's quantitative is.

00:15:19.000 --> 00:15:28.000
given the current relocated seismicity catalog, how many of those events can be statistically associated with a fault object in the community fault model?

00:15:28.000 --> 00:15:36.000
And so this paper was published by Walker Evans and others in 2020 that laid out a statistical method that used an independent data set to test and verify this.

00:15:36.000 --> 00:15:46.000
And this has been running on Caltech servers for about five years and anybody can go and sign up for this. We call it the CFM Association Service.

00:15:46.000 --> 00:15:53.000
Whereas any magnitude three and above event in Southern California, you get an email within about 10 or 15 minutes of that coming out.

00:15:53.000 --> 00:16:00.000
That tells you which fault it's most likely associated with or if it's not associated with any faults that we know about.

00:16:00.000 --> 00:16:06.000
And this has been great for people that have to talk to the media in rapid response situations.

00:16:06.000 --> 00:16:12.000
It's proven to be quite useful and we're currently in the stages of implementing this statewide.

00:16:12.000 --> 00:16:25.000
So stay tuned for that. But this is running currently on an old version of the CFM, so we're working to update it to the new one. And we asked the question, is the new CFM actually better than the old CFM? And what about NSHM?

00:16:25.000 --> 00:16:35.000
So what we did is we ran these through the algorithm and we updated the seismicity catalog to have the Hauksson and Yang focal mechanisms and Hauksson locations 

00:16:35.000 --> 00:16:39.000
all the way up to 2023, these papers have updates every year.

00:16:39.000 --> 00:16:45.000
And cropped it to just the area of Southern California where we think the community fault model was complete.

00:16:45.000 --> 00:16:59.000
And this was the hardest thing to script I've ever done. It looks like fruity pebbles, but The dots are color coded to the fault that those earthquakes are associated with so all these little pink events up here are associated with that pink fault and blue faults there. If you see black

00:16:59.000 --> 00:17:10.000
dots, that means that earthquake was not associated with the CFM fault. So what percentage of the magnitude three and larger events were associated with something in the CFM 6

00:17:10.000 --> 00:17:17.000
89.7%. I think that's pretty good. Well, how does that compare to the past models? Well, if you look at CFM5 in gray,

00:17:17.000 --> 00:17:27.000
the NSHM model, which is provided as just traces in a dip angle, so I wrote a script that just meshes that all in 3D and then ran it through the algorithm, 

00:17:27.000 --> 00:17:33.000
and then the current version of the CFM, CFM 6, because remember 7 is in Northern California and that's not complete.

00:17:33.000 --> 00:17:40.000
So what you can see here is for all magnitude ranges, magnitude three to five, five to six, or six plus or all.

00:17:40.000 --> 00:17:48.000
The CFM-6, or the current version of the CFM in Southern California, seems to have the most events associated with the fault.

00:17:48.000 --> 00:17:56.000
And what we take this to mean is that we've done a reasonably decent job of responding to seismic events as they happen.

00:17:56.000 --> 00:18:09.000
Like Ridgecrest, for example, was a surprise to many people. So after that event, we created a detailed representation of Ridgecrest. So we think that the model is reasonably complete, about 90% or so of the seismicity is explainable.

00:18:09.000 --> 00:18:13.000
Again, read the paper for the details. It's a fascinating paper on this algorithm.

00:18:13.000 --> 00:18:20.000
But now is where we need your help. So we're currently evaluating the Northern California portion of the CFM. It's brand new.

00:18:20.000 --> 00:18:34.000
It's probably missing some stuff, so we need your expertise. So please contact me if you're willing to help us out evaluate this model. All past versions of the CFM have been evaluated and it's critical that we evaluate this new portion of the model.

00:18:34.000 --> 00:18:42.000
This kind of thing can't be done without contributors. So here's a list in tiny font of everybody that's contributed over the years. Thank you so much.

00:18:42.000 --> 00:18:47.000
And past CFM evaluators that evaluated CFM5. Thank you so much.

00:18:47.000 --> 00:18:52.000
And if you're willing to evaluate CFM7, please let us know. It'd be super-duper helpful.

00:18:52.000 --> 00:18:57.000
And with that, I leave you with this slide. Thank you so much for having me.

00:18:57.000 --> 00:19:05.000
I hope you found this interesting. Bye-bye.

00:19:05.000 --> 00:19:13.000
Okay, thanks, Scott. That was awesome talk. I'm writing some notes down, so I think a lot of people might have some questions.

00:19:13.000 --> 00:19:19.000
Now the next talk is going to be by Vera Schulte-Pelkum and Debi Kilb.

00:19:19.000 --> 00:19:26.000
"Imaging Northern California faults" and I'll just let them read the rest. Go ahead.

00:19:26.000 --> 00:19:33.000
Hi everybody, we're going to do some deep fault imaging in Northern California.

00:19:33.000 --> 00:19:40.000
Debbie and I contributed equally to this talk. And we'd like to acknowledge our collaborators.

00:19:40.000 --> 00:19:51.000
And also the funding agencies. Alright, so the goal here is to image fault-related features in situ at depth.

00:19:51.000 --> 00:19:59.000
Those might include material contrasts across faults.

00:19:59.000 --> 00:20:05.000
We might see a record of off-fault deformation or fault-related fabric.

00:20:05.000 --> 00:20:20.000
And also possibly an interaction with pre-existing fabric And we'd also maybe like to see faults in the brittle depth range root into shear zones when you cross into the ductile 

00:20:20.000 --> 00:20:30.000
zone. So the study areas we're going to look at are we're going to focus on central and northern California, San Andreas 

00:20:30.000 --> 00:20:36.000
Zones here as well as the Hayward-Calaveras junction in the Bay Area.

00:20:36.000 --> 00:20:45.000
We have some ongoing work in Mendocino that I'm not going to talk about today and at the end I'm going to compare to some locations in Southern California.

00:20:45.000 --> 00:20:56.000
So we have a transform plate boundary going on here. One point that will become more important later is that this was formerly a subduction zone.

00:20:56.000 --> 00:21:02.000
Okay, so we're going to accomplish this by using a receiver function method.

00:21:02.000 --> 00:21:13.000
So we have seismic stations here and we have teleseismic P-waves incident at a fairly steep angle so they eliminate a narrow cone under the station.

00:21:13.000 --> 00:21:26.000
If these P waves encounter a contrast in rock fabric they might convert to S waves that will vary in appearance with back azimuth.

00:21:26.000 --> 00:21:32.000
So such contrast could occur across faults or across the boundaries of shear zones.

00:21:32.000 --> 00:21:45.000
We've published other applications using this method for regional and fault fabrics in Southern California and Colorado, weave image detachments faults like the main Himalayan thrust in the Appalachian decollement,

00:21:45.000 --> 00:21:53.000
and it turns out that volcanic areas have very strong fabric that gets picked up by this method as well.

00:21:53.000 --> 00:22:03.000
Okay, so let's look at waveforms from one station. This one is very close to San Andreas surface trace near Hollister.

00:22:03.000 --> 00:22:16.000
Here, this panel shows radial component, standard receiver functions sorted by and bin by Bakazima. So these guys come in from the north, east, south, and so forth.

00:22:16.000 --> 00:22:21.000
And these little numbers represent the number of events that went into that bin 

00:22:21.000 --> 00:22:38.000
waveform and you can see so this top trace is the average overall bins, right. You might look at some of these and think, okay, maybe Hirosh is the Moho and you see the waveform appearance changes quite dramatically as you go around in back azimuth.

00:22:38.000 --> 00:22:48.000
And there's also a very strong arrival early on around one second that even changes polarity. If you look at the transverse component receiver function in the second panel,

00:22:48.000 --> 00:23:03.000
you see a very strong polarity changing arrival also at near one second So that means there's a lot of conversion happening that goes out of plate. If this was isotropic and flat layered, we wouldn't see anything on this panel.

00:23:03.000 --> 00:23:24.000
So I'll point you to this paper and other papers cited in there as far as the theory behind this but we can basically extract the out of plane signal conversions on the radio component by subtracting the azimuthal average, and on the transverse component to match the radial, we shift this in back azimuth.

00:23:24.000 --> 00:23:30.000
Now we have these arrivals impinging from the same back azimuth we can stack everything,

00:23:30.000 --> 00:23:46.000
which conveniently gets rid of these holes in back azimuthal coverage. Okay, so now we have isolated the out of plane conversions and we take a moving time window and do a four-year harmonic decomposition 

00:23:46.000 --> 00:23:58.000
at each time step. This is the first Azimuthal harmonic amplitude and we see there's a very large arrival close to one second.

00:23:58.000 --> 00:24:05.000
The time of this tells us how deep the converter was. In this case, it's about 6 kilometers depth.

00:24:05.000 --> 00:24:17.000
The amplitude of this peak tells us how strong the fabric contrast was at that depth where it changes polarity that outlines defoliation strike 

00:24:17.000 --> 00:24:26.000
and the maximum amplitude gives us an indication what the dip sense is, which way it's dipping.

00:24:26.000 --> 00:24:35.000
Alright, this is what it might look like for stations sitting over some fabric contrasts.

00:24:35.000 --> 00:24:42.000
In the rest of this talk, we'll focus on the biggest arrival, the biggest A1 amplitude at each station.

00:24:42.000 --> 00:24:58.000
So the timing gives us the depth and we're going to show this in Map view. So in Map view you'll see in the next few plots a bubble that's plotted at the station location, so right under this guy, the color will be the depth.

00:24:58.000 --> 00:25:08.000
The size of this circle will scale with the amplitude, which scales with the fabric contrast strength The red bar is the strike 

00:25:08.000 --> 00:25:14.000
of the fabric and the tick mark will usually point down dip.

00:25:14.000 --> 00:25:20.000
Okay, so let's look at what that looks like for the station that I showed you the waveforms for earlier.

00:25:20.000 --> 00:25:40.000
That stage is located right here. So now we're looking at the San Andreas up here near the Calaveras junction and the strike that we image here at about 6 kilometers depth is parallel roughly to the surface trace of the San Andreas.

00:25:40.000 --> 00:25:51.000
You can see some similar kind of largish arrivals from various depths, this color is the depth scale near the fault in other locations.

00:25:51.000 --> 00:26:13.000
This is double difference relocated seismicity and it's showing up at slightly shallower depths than what these image contrasts are. The fact that this is offset from the surface trace by a little bit means that the San Andreas here might dip to the southwest, provided this is on the same fault plane.

00:26:13.000 --> 00:26:31.000
When we go farther away from the fault we see much smaller arrivals from receiver functions. These A1 arrivals are not nearly as big and the strikes are more misoriented relative to the fault.

00:26:31.000 --> 00:26:35.000
Okay, so let's look at this in 3D view now.

00:26:35.000 --> 00:26:46.000
So this is that same station we looked at. The green surfaces are from the SCEC community fault model.

00:26:46.000 --> 00:27:02.000
That was recently expanded to all of California. So you can see the local southwest tip of the San Andreas these little gray dots are seismicity so you can maybe see a little bit better now that these imaged contrast faults slightly below

00:27:02.000 --> 00:27:08.000
the seismicity volume.

00:27:08.000 --> 00:27:18.000
Okay, so let's move slightly farther south to the Parkfield area now there's a very nice temporary array here with dense station coverage 

00:27:18.000 --> 00:27:26.000
and we see quite large A1 arrivals, similar or slightly larger than what we saw near Hollister.

00:27:26.000 --> 00:27:35.000
Much more dense station coverage here now and we can see the strikes again are largely parallel to the surface trace of the San Andreas here.

00:27:35.000 --> 00:27:48.000
Interestingly, the dips are all to the northeast. Just like in our previous example, so here is a cross section of this box.

00:27:48.000 --> 00:28:01.000
So we've projected just everything within this box, onto this profile here. So here's the surface the fault model 3D surface of the San Andreas. Here is the seismicity. These are receiver function A1 arrivals,

00:28:01.000 --> 00:28:07.000
and the dip sense points to the northeast on both sides of the fault 

00:28:07.000 --> 00:28:20.000
and these contrasts kind of outline a volume that dips away from the fault. Larger amplitudes on the northeast side and kind of an intermediate dip.

00:28:20.000 --> 00:28:27.000
Structure outlined there. So what could be going on here and why is this not symmetric about the fault?

00:28:27.000 --> 00:28:34.000
One interpretation could be that the whole region has recorded northeast 

00:28:34.000 --> 00:28:51.000
dipping fabric back from when the subduction was happening. Then the incipient San Andreas, when the whole thing turned into transform exploited some of that fabric, it amplified the rock fabric contrasts here, and then over time might have 

00:28:51.000 --> 00:28:58.000
gradually change to a vertical geometry because that's energetically more favorable.

00:28:58.000 --> 00:29:09.000
So here's a deep review, so the same cross-section button projecting the entire array. Here is the San Andreas in green, this is the seismicity.

00:29:09.000 --> 00:29:18.000
These are all our receiver function A1 arrivals, and here we've also plotted the LFEs from the Shelley catalog 

00:29:18.000 --> 00:29:27.000
and Hardebeck and these guys plot to the northeast side of the surface trace as well.

00:29:27.000 --> 00:29:35.000
Okay, last location we're going to look at in detail. Let's move to the Hayward-Calaveras

00:29:35.000 --> 00:29:41.000
fault junction in the Bay Area. So this is the Hayward fault. This is the Calaveris fault.

00:29:41.000 --> 00:30:04.000
Here are our receiver function bubbles and strikes, largely parallel to the fault trend in the area. There are some that are almost sort of orthogonal to the fault trends and here we've plotted The strikes off the nodal planes of the focal mechanisms from Rob's catalog.

00:30:04.000 --> 00:30:09.000
If we zoom in on this then station 9.

00:30:09.000 --> 00:30:14.000
Here is a depth cross section. So here's the Hayward fault 

00:30:14.000 --> 00:30:28.000
from the fault model, merging with the Calaveras. This is seismicity and the fault model here continues to quite large depths where it's probably not a single surface 

00:30:28.000 --> 00:30:45.000
down to 30 kilometers down but what's interesting is that the receiver function conversions kind of outline the downward continuation of the brittle fault as shown by the seismicity. So these image contrasts extend below the brittle ductile transition.

00:30:45.000 --> 00:30:58.000
Okay, so let's zoom out a little bit and have a regional view. So the size of these bubbles are now the receiver function A1 conversion amplitudes across the whole region.

00:30:58.000 --> 00:31:09.000
You have the San Andreas here, Garlock here, the Sierra Nevada, so these very large amplitudes these big bubbles occur along the San Andreas.

00:31:09.000 --> 00:31:15.000
A little bit smaller up here in the Bay Area on other faults of the San Andreas System.

00:31:15.000 --> 00:31:35.000
There are other faults like the White Wolf fault here that have large amplitudes this might be because there's a decent amount of vertical offset on this vault and it juxtaposes basement over sediment overthrust which might contribute to this large amplitude here. Ridgecrest

00:31:35.000 --> 00:31:39.000
which kind of looks interesting we want to look at that in more detail in the future.

00:31:39.000 --> 00:31:49.000
And then the third area that has very large amplitudes is Mammoth Lakes here which is volcanic.

00:31:49.000 --> 00:31:57.000
Everywhere else, we see some increased amplitudes on top of faults, but they're not as large as along the San Andreas System.

00:31:57.000 --> 00:32:21.000
Okay. We see some smaller amplitudes near the San Andreas too, but when you only have one or two stations that are kind of close to the fault it's hard to draw any systematic conclusions so here we've compared The amplitudes measured with dense fault crossing arrays at four different locations. Parkfield, which we looked at.

00:32:21.000 --> 00:32:33.000
Hayward-Calaveras, San Jacinto Fault Zone has some dense fault causing arrays, and then we also had a station line in the Mojave in the eastern California shear zone that crosses several minor faults 

00:32:33.000 --> 00:32:38.000
in the area. So what do we see here in terms of systematics?

00:32:38.000 --> 00:32:46.000
San Andreas has the largest cumulative offset. It's the most mature fault, we see large amplitudes there,

00:32:46.000 --> 00:33:06.000
and the A1 receiver function conversion. Hayward-Calaveras and San Jacinto have still substantial but still less large offset than the San Andreas and we see about half the amplitude there in the receiver functions, and these little faults in the eastern California shear zone

00:33:06.000 --> 00:33:13.000
no significant offset that's accumulated show maybe a quarter of the amplitude that we see 

00:33:13.000 --> 00:33:28.000
at the very mature fault trend. Okay, so the conclusions we are drawing from this is that when you look at fault fabric the strength seems to correlate to the maturity of the fault.

00:33:28.000 --> 00:33:35.000
The fabric tends to parallel faults within an area of several kilometers away from the fault.

00:33:35.000 --> 00:33:40.000
Regionally, we see consistent dips to the northeast 

00:33:40.000 --> 00:33:49.000
In most places in California. So this might be because of structural inheritance stemming back to subduction zone times 

00:33:49.000 --> 00:33:59.000
after a long subduction. And we can trace this fault fabric to below sometimes well below the brittle ductile transition.

00:33:59.000 --> 00:34:14.000
Thanks for listening and I'd be happy to answer any questions live.

00:34:14.000 --> 00:34:20.000
Okay, thanks a lot, Vera. So let's move to the last talk 

00:34:20.000 --> 00:34:29.000
of this session. So it's going to be by Scott Callaghan talking about "CyberShake results for Northern California."

00:34:29.000 --> 00:34:37.000
Hi everyone, I'm Scott Callahan and I'm going to be talking to you today about our most recent CyberShake results for Northern California.

00:34:37.000 --> 00:34:44.000
First, a brief overview of CyperShake. I don't know if many of you are familiar with it, but it's SCEC's 3D physics-based PSHA platform.

00:34:44.000 --> 00:34:48.000
CyberShake works by starting with an earthquake rupture forecast or ERF 

00:34:48.000 --> 00:35:00.000
that has a list of specific events and the associated probabilities. It then uses a reciprocity-based approach. So we run wave propagation simulations to calculate string green tensors.

00:35:00.000 --> 00:35:06.000
We then convolve the string green tensors with individual slip time histories for each event that we want to simulate.

00:35:06.000 --> 00:35:11.000
And we end up with hundreds of thousands of seismograms for every site that we're interested in.

00:35:11.000 --> 00:35:27.000
From those seismograms, we can then derive a variety of intensity measures. These can be combined with the probabilities from the ERF to produce site-specific hazard curves. And then we can interpolate those to produce a hazard map like the one you see over here on the right for Southern California.

00:35:27.000 --> 00:35:44.000
We also have the option of performing stochastic high frequency simulations, so then we can produce an overall CyberShake broadband hazard model. What I want to talk about today is our most recent CyberShake results, which are from study 24.8.

00:35:44.000 --> 00:35:51.000
To this was done during the fall, and we focused here on Northern California. We selected 315 sites.

00:35:51.000 --> 00:36:03.000
Kind of in the greater San Francisco Bay Area and we use a set of parameters that were largely similar to our most recent Southern California study, study 2212, Because we wanted to support easy comparisons between the two studies.

00:36:03.000 --> 00:36:11.000
So we use the use of two derived ERF. The Graves and Pitarka rupture generator. It produces about 200,000 events per site of interest in Northern California.

00:36:11.000 --> 00:36:23.000
And we did one hertz deterministic simulations supplemented with 50 hertz stochastic simulations, which were generated using the Graves and Pitarka module from the SCEC broadband platform.

00:36:23.000 --> 00:36:32.000
A couple of things that were different with this study. One is that we did not include any events in the southern San Andreas. So these are events that have a hypocenter south of Parkfield.

00:36:32.000 --> 00:36:47.000
In our previous CyberShake study in Northern California, which was study 18.8, we did include these events, but they come with a cost, which is that some of these ruptures end up rupturing all the way down to Bombay Beach in the ERF 

00:36:47.000 --> 00:36:52.000
and so then you need a simulation volume, which essentially encompasses the entire state.

00:36:52.000 --> 00:36:56.000
So we took a look to see if those events matter for hazard.

00:36:56.000 --> 00:37:13.000
It turns out both looking at the Cybershake results and looking at ASK, that these events do not have a major contribution to hazard. Most of the time it's well less than tenths or hundreds of a percent. The largest that we saw was maybe a percent or so.

00:37:13.000 --> 00:37:19.000
And so we felt comfortable excluding these events from our hazard calculations.

00:37:19.000 --> 00:37:28.000
We therefore saved about 40% in terms of the volume size when we run our simulations, and that helps reduce the computational cost.

00:37:28.000 --> 00:37:36.000
Our minimum VS reduced it from 500 to 400 meters per second to try to better capture some of the soft sediments around the San Francisco Bay.

00:37:36.000 --> 00:37:40.000
We made some changes to the velocity model. I'll get into details on that in a minute.

00:37:40.000 --> 00:37:48.000
And we also generated some new data products. So the big one is that we now have three component seismograms. So we included the vertical component with this study.

00:37:48.000 --> 00:37:55.000
So we also then give vertical response spectra. We calculated period dependent durations as well.

00:37:55.000 --> 00:38:05.000
So talk about the velocity model a little bit. The model that we needed to construct for our meshes is larger than the region that's covered by any one individual model.

00:38:05.000 --> 00:38:10.000
So as is often the case for these large CyberShake studies, we tiled multiple models together.

00:38:10.000 --> 00:38:17.000
If you look here, this is a sample simulation volume, and so we started with the green region, which is the USGS SFCVM.

00:38:17.000 --> 00:38:24.000
Second was the CCA 06 model, which is a tomography model for Central California.

00:38:24.000 --> 00:38:36.000
And then lastly, the gray areas here are 1D background model that we generated by extending a version of the Sierra Geologic region in the SFCVM. So essentially, it's similar to kind of extending this edge out.

00:38:36.000 --> 00:38:39.000
There are very few sources in the gray area. I think just Walker Lane.

00:38:39.000 --> 00:38:44.000
So basically the gray area is there to fill out the rectangle.

00:38:44.000 --> 00:38:48.000
We did some other things as well to modify the velocity model.

00:38:48.000 --> 00:38:55.000
We modified the San Leandro Gabbro geologic region because it has very high near surface velocities.

00:38:55.000 --> 00:39:02.000
We applied smoothing 20 kilometers from all of these model interfaces to minimize reflection and refraction artifacts.

00:39:02.000 --> 00:39:10.000
We populated the surface point of the model not by querying each of these individual models at the surface, but at a depth of 20 meters.

00:39:10.000 --> 00:39:18.000
And we also found that around the San Francisco Bay, there are some locations where the VPVS ratio is very high, up to about 15.

00:39:18.000 --> 00:39:23.000
And so that led to a problem when applying a VS4 of 400 meters per second.

00:39:23.000 --> 00:39:34.000
So we ended up capping those ratios at four. The other thing we did to model is we applied the merge taper approach. This is the same thing that we did in study 2212 in Southern California.

00:39:34.000 --> 00:39:39.000
The idea is that we apply a taper, which is based on the Thompson VS30 values.

00:39:39.000 --> 00:39:47.000
To grid points in our mesh in the top 700 meters, but we only apply the taper if it results in a VS value which would be lower.

00:39:47.000 --> 00:39:52.000
And so you can see the impact of this depends on the location in the model.

00:39:52.000 --> 00:39:59.000
There are some sites like the one down here. So this is, I think in the San Leandro Gabbro region.

00:39:59.000 --> 00:40:06.000
Where after the taper, you see a sharp reduction in VS. I should mention that VS, the orange traces here.

00:40:06.000 --> 00:40:14.000
There are other sites like Livermore and downtown San Francisco. We're applying the taper leads to a modest change in VS 

00:40:14.000 --> 00:40:18.000
and there are other sites like Moffett Field where there's basically no impact at all.

00:40:18.000 --> 00:40:24.000
Overall, the plot here on the left is a surface VS plot before we apply the taper.

00:40:24.000 --> 00:40:27.000
And then the plot on the right is after we apply the taper.

00:40:27.000 --> 00:40:34.000
And so you can see that it reduces a lot of these velocities that are 1,200 meters per second or higher.

00:40:34.000 --> 00:40:44.000
We also get a lot of heterogeneity out here even in the 1D model region, and that reflects the  heterogeneity and the VS30,

00:40:44.000 --> 00:40:55.000
the Thompson VS30 model that we're using here. Getting the velocity model ready was a lot of the run-up to the study, once we had a model that we were comfortable with.

00:40:55.000 --> 00:40:57.000
We went ahead and ran the study. It took us 45 days.

00:40:57.000 --> 00:41:06.000
We used a combination of Frontier at Oak Ridge to do the AWP wave propagation simulations. These were GPU based.

00:41:06.000 --> 00:41:14.000
And then our CPU codes, which did the low frequency size program synthesis, as well as the stochastic calculations were run on Frontera at the Texas Advanced Computing Center.

00:41:14.000 --> 00:41:18.000
It was a modest study by kind of in the middle by CyberShake study sizes,

00:41:18.000 --> 00:41:27.000
180,000 node hours. We did get up to using almost half of Frontier at one time. Frontier is the biggest open science system in the world at the time we did this study 

00:41:27.000 --> 00:41:36.000
so that was a neat accomplishment. We used our standard set of workflow tools to manage all of the jobs. Also kept track of about a petabyte of data.

00:41:36.000 --> 00:41:47.000
In the end, we produced 36 terabytes of data mostly three component seismograms, but also a large number of intensity measures as well.

00:41:47.000 --> 00:41:52.000
And so this is the kind of final output data products that we have from this study.

00:41:52.000 --> 00:41:56.000
What I want to do now is show you some results.

00:41:56.000 --> 00:42:03.000
So first, let's take a look at some low frequency hazard maps. We have results here at 2, 3, 5, and 10 seconds.

00:42:03.000 --> 00:42:09.000
For each one, we have three maps. The map on the left are the new CyberShake Study 24.8 results.

00:42:09.000 --> 00:42:14.000
The one in the center is a ratio between the CyberShake results and the GMMs,

00:42:14.000 --> 00:42:22.000
4NGA West 2s. Yellow is a ratio of one. So the greens are where CyberShake is lower, the red is where CyberShake is higher.

00:42:22.000 --> 00:42:25.000
And then the one on the right is ratio of the new to old CyberShake.

00:42:25.000 --> 00:42:30.000
I should mention these are all Rothie 50; 2% in 50 years.

00:42:30.000 --> 00:42:45.000
Just a couple of things to draw your attention to. You notice, for example, that the new CyberShake results compared to the old, you have kind of this warmer spot near the San Pablo Bay. That's because of some changes made to the underlying velocity model there.

00:42:45.000 --> 00:42:53.000
You'll notice that here at the shorter periods, we tend to see CyberShake as maybe slightly lower hazard than the GMMs, but less true at five seconds.

00:42:53.000 --> 00:42:58.000
Also, you probably noticed this big hotspot here out near Livermore at 10 seconds.

00:42:58.000 --> 00:43:06.000
That's also due to changes in the velocity model. The previous USGS velocity model here, the CENCAL model.

00:43:06.000 --> 00:43:15.000
Has looked like this. And then the current model, you'll see that we have a low velocity region that extends deeper.

00:43:15.000 --> 00:43:25.000
And it seems like the resonance there is around 10 seconds, and that's why we see higher ground motions now than we did previously.

00:43:25.000 --> 00:43:35.000
We did a couple other analyses as well that I'd like to share. One, Kevin Miller helped us out with a Z-score distribution plots. So we compare the CyberShake results to the GMMs from the J West 2.

00:43:35.000 --> 00:43:41.000
We calculate a Z-score and then we bin them and plot them up and compare them to a normal distribution.

00:43:41.000 --> 00:43:49.000
And so what you see here is that our means and variability at two and three seconds are very similar between CyberShake and the GMMs.

00:43:49.000 --> 00:43:52.000
At 5 and 10 seconds, our means shift a little higher,

00:43:52.000 --> 00:43:57.000
but very similar width. So we're capturing the variability here well.

00:43:57.000 --> 00:44:01.000
Another analysis that we like to do, Xiaofeng helped us out with this one.

00:44:01.000 --> 00:44:08.000
Is that we use the CyberShake results to produce essentially a CyberShake derived ground motion model.

00:44:08.000 --> 00:44:19.000
And then that allows us to do things like look at distance and magnitude scaling or breakdowns of variance. And so we can do comparisons between CyberShake and the GMMs in other ways.

00:44:19.000 --> 00:44:35.000
So some interesting things come out of this. So for example, if you look at the magnitude scaling, you'll notice that the new CyberShake results in the black line are pretty similar to the GMMs. The colored lines here at low magnitudes, but as you get to higher magnitudes, they tend to get higher.

00:44:35.000 --> 00:44:46.000
We're still investigating why that is, but this is why this kind of analysis is useful, right. So we can see some of these trends and start to ask some interesting questions about the CyberShake results.

00:44:46.000 --> 00:44:50.000
We can also use the CyberShake derived GMM to do things like pull site terms.

00:44:50.000 --> 00:44:54.000
This is just a comparison in the left column is the old CyberShake results.

00:44:54.000 --> 00:44:57.000
Center column are the new and then the right column is the difference.

00:44:57.000 --> 00:45:06.000
And so you see some of the same trends, right. You see up by San Pablo Bay that the new site terms are a little higher, again, reflecting changes in the velocity model.

00:45:06.000 --> 00:45:14.000
You notice there's kind of some some both higher and lower, some strange stuff going on here south of San Francisco Bay.

00:45:14.000 --> 00:45:21.000
We believe that's because in the old CyberShake simulations, we had a velocity model interface there, which might have been causing some artifacts.

00:45:21.000 --> 00:45:28.000
But with our most recent study, we pushed the velocity model interface farther to the south so we wouldn't have that impact.

00:45:28.000 --> 00:45:33.000
You also can see the Livermore Basin light up again here with the site terms from the most recent study 

00:45:33.000 --> 00:45:42.000
just like we saw in the hazard maps. On the broadband side, here on the left is a one-second hazard map with the CyberShake results.

00:45:42.000 --> 00:45:46.000
And then in the center, a ratio between CyberShake and the GMMs.

00:45:46.000 --> 00:45:51.000
You notice that overall the CyberShake results, the hazard here is lower.

00:45:51.000 --> 00:45:56.000
We can see why that is if we take a look at the Z-score distribution here at the short periods.

00:45:56.000 --> 00:46:03.000
You'll notice that it means a little low at 0.1 and 0.2 seconds, but at 0.5 and 1 seconds, the means are pretty good.

00:46:03.000 --> 00:46:15.000
But the width of the distribution is quite a bit less with CyberShake than with the GMM and that's not surprising. You might remember that we saw something similar with study 2212 in Southern California 

00:46:15.000 --> 00:46:31.000
and that's because the graves in Pitarka broadband module that we used was developed to try to match high frequency means, but not necessarily match the whole range of variability that we see. And so we see that reflected here, right, where the variability isn't quite as much.

00:46:31.000 --> 00:46:36.000
And when you're looking at hazard, hazard is particularly sensitive to the long tails of the distributions.

00:46:36.000 --> 00:46:43.000
So the fact that the CyberShake tails are a little shorter means that we see significantly reduced hazard with CyberShake than with the GMMs.

00:46:43.000 --> 00:46:49.000
And so it's a good reminder that the hazard is going to be a little low if you look at overall hazard at the high frequencies with CyberShake.

00:46:49.000 --> 00:46:54.000
But if you want to go in and look at individual scenarios, magnitude 7 on the Hayward or something.

00:46:54.000 --> 00:47:02.000
You're going to end up with good and useful results. So where do we go from here?

00:47:02.000 --> 00:47:07.000
So in the short term, we'd like to continue our analysis of study 24.8.

00:47:07.000 --> 00:47:16.000
Answer some of these questions that we've raised here. We'd like to continue to work to improve community access to this rich set of data that we've just produced.

00:47:16.000 --> 00:47:24.000
We're planning to migrate this study to DesignSafe. To help support kind of easy querying and access to this data.

00:47:24.000 --> 00:47:32.000
We also are planning to calculate Fourier spectra, for all of these events because we know that's a useful data product for the engineering community.

00:47:32.000 --> 00:47:45.000
In the medium term, hopefully in 2025, we'd like to perform some two hertz tests. We'd like to push that minimum frequency higher. So pick a small region of interest and do some calculations. It will require some modifications to the code base, but

00:47:45.000 --> 00:48:02.000
we're looking forward to increasing the deterministic frequency. We're also interested in maybe doing some tests with reducing the minimum VS, especially in the San Francisco area, there are some very low VS values around the bay and it would be good to be able to capture that.

00:48:02.000 --> 00:48:09.000
In the long term, we're looking at how do we integrate nonlinear effects with the reciprocity-based approach that CyberShake uses.

00:48:09.000 --> 00:48:18.000
We know from simulations that, you know, including non-linearity can have a big impact, especially as you go to higher frequencies and so we know we need to account for that.

00:48:18.000 --> 00:48:27.000
And we'd also like to include topography. We know from simulations in Southern California especially that including topography can make quite a difference in the ground motions 

00:48:27.000 --> 00:48:31.000
and so that's another thing we'd like to come up with ways to integrate into CyberShake.

00:48:31.000 --> 00:48:35.000
Well, please let me know if you're interested in accessing any of this data.

00:48:35.000 --> 00:48:41.000
If you're interested in the velocity model, I can tell you that Mei has worked really hard to integrate it into UCVM.

00:48:41.000 --> 00:48:51.000
So we can help point you to that if you'd like to extract cross-sections or use it for another application.

00:48:51.000 --> 00:48:58.000
And also, I'm happy to help connect you with intensity measures or seismograms or anything else that would be useful.

00:48:58.000 --> 00:49:02.000
I'd like to thank the funding agencies and the compute centers that made this work possible.

00:49:02.000 --> 00:49:15.000
And please let me know if you have any questions. Thank you very much for the opportunity to speak to you today.

00:49:15.000 --> 00:49:26.000
Alright. Thank you, Scott, Vera and Scott for some great presentations. Thanks, everyone, for the thoughtful questions in the chat.

00:49:26.000 --> 00:49:37.000
Maybe we should start with Scott. And Scott, you answered a number of questions in the chat about Scott Marshall, that is.

00:49:37.000 --> 00:49:50.000
And the chat about use cases and evaluations and things is there anything in those questions that you want to bring to our attention or any themes there that you want to hit?

00:49:50.000 --> 00:50:01.000
I'm not sure. I guess a lot of the questions were about related to the geometry because, you know, the Skype community fault model is a geometric model.

00:50:01.000 --> 00:50:12.000
And I just want to emphasize that the community fault model is not built based on a data set. Like I think a lot of people think we take the seismicity and make faults out of it.

00:50:12.000 --> 00:50:23.000
We don't do that. We basically take every single kind of data that we can get our hands on and use that to build the faults in the best way possible. And this is what Andreas spends hundreds of hours doing.

00:50:23.000 --> 00:50:34.000
So some faults may be based on seismicity alone. Some may be based on focal mechanisms, some might not. Some might be based on 3D seismic profiles. Others might be based on gravity data.

00:50:34.000 --> 00:50:43.000
And that's sort of the goal of the SCEC model is to say, given all the data that we have as a scientific community, what's the best representation of three-dimensional geometry that we can come up with?

00:50:43.000 --> 00:50:49.000
The downside of that is the answer to that question ends up that faults are really complicated.

00:50:49.000 --> 00:51:11.000
Yeah. All right. We do have the opportunity for people to ask questions if they want to raise their hand or raise their hand or turn on their video With the absence of questions, maybe then Vera and or Debbie, is there anything you answered a number of questions in the chat or any themes or any points that you want to bring up?

00:51:11.000 --> 00:51:17.000
From from those.

00:51:17.000 --> 00:51:38.000
Yeah, so we had a number of questions comparing to detailed results in any particular location And to answer those, if you remember my cartoon about the station location and that narrow cone of you know sampling under the station we need to hit the structure you're interested in from different angles.

00:51:38.000 --> 00:51:48.000
And tell seism's coming pretty steeply, which is nice because you kind of have a spotlight on what's under you, but you need to really sit on top of what you're trying to image.

00:51:48.000 --> 00:52:02.000
So any particular parts you're interested in, feel free to email us and we are happy to go dig in our you know the awesome three models that Debbie constructed and see what's there. But we need to be sitting right on top of something with a

00:52:02.000 --> 00:52:14.000
Station that has three components and has run for ideate longer than a few months a year plus is great. And then we can try and tell you what we see in that particular spot.

00:52:14.000 --> 00:52:32.000
Yeah, so then I did look briefly at the, you know, the EOS article that came out about your GRL paper recently and that That includes like a systematic analysis of permanent stations, or at least you showed something towards the end

00:52:32.000 --> 00:52:38.000
Is that your analysis of the number of permanent stations in California?

00:52:38.000 --> 00:52:45.000
No, that was just what we looked at to look right under the Sierra Nevada.

00:52:45.000 --> 00:53:04.000
So we have, there is we have that 2020 paper that was in the slideshow that introduced the method and does sort of broad regional results for California. And then we've padded that in recent work with local arrays.

00:53:04.000 --> 00:53:12.000
All right, great. With that, I see Christy has raised her hand. So Christy, you want to ask your question?

00:53:12.000 --> 00:53:30.000
Yeah, hi. I've been thinking a lot about what is a fault because during an earthquake, a fault might be a single surface. And then geologically a fault is many intersecting surfaces in a volume with all kinds of other stuff going on. And so like all these different representations are really capturing

00:53:30.000 --> 00:53:46.000
Different sort of spatial definitions of what the fault zone is, but also temporal definitions or rate related distinctions and like which parts of the fault are active. So I don't think that there's anything that's like incompatible between all the talks that we saw today.

00:53:46.000 --> 00:53:55.000
This is more of a comment than a question. But I was thinking in particular for the CFM, like it's going to be used a lot for rupture.

00:53:55.000 --> 00:54:06.000
Modeling, but it also has slip rate tied to it, which is kind of convolves many different time scales.

00:54:06.000 --> 00:54:19.000
I don't know how that would be addressed in terms of a community product, like people are going to take this and run with it in lots of different directions, but it's something to think about in terms of slip rate versus strain rate versus volumes over which deformation takes place in the plate boundary.

00:54:19.000 --> 00:54:27.000
And then in the Bay Area, that is It's the same style of problem as in SoCal, but the geometry of it is different and the materials are different.

00:54:27.000 --> 00:54:32.000
So like the fabrics are much stronger in the Bay Area and a lot of those lithologies, for example.

00:54:32.000 --> 00:54:33.000
Yeah, so that might help distribute strain.

00:54:33.000 --> 00:54:39.000
Yeah, I can. I can comment on that. This is a really good point, Christy.

00:54:39.000 --> 00:54:57.000
That what is a fault? And for example, our association code doesn't answer that question. It just simply says, what's the statistical probability that an earthquake is associated with a fault zone. Is it on a splay? Is it in the damage zone? Is it on the main plane? The code doesn't assess that.

00:54:57.000 --> 00:55:07.000
But also, I wanted to make a couple points. One, you said that the faults are tied to a slip rate. The community fault model is not tied to a slip rate. The community fault model is just a geometry.

00:55:07.000 --> 00:55:08.000
And the reason…

00:55:08.000 --> 00:55:14.000
No, sorry, but you said that you go to the viewer and where there are slip rates and then like link it back to what fault the They're associated too.

00:55:14.000 --> 00:55:22.000
You can, but a lot of the work that I've been doing in the last couple decades has basically shown that slippery is just massively spatially variable.

00:55:22.000 --> 00:55:32.000
So if you use those slip rates, you have to take care to know that many of those slip rates might be totally correct but not spatially representative for that fault on a whole.

00:55:32.000 --> 00:55:43.000
And so you have to use those things with care. And that's one of the reasons why the SCAT community fault model doesn't provide a slip rate on every little triangle in the model because it's unconstrainable with the current amount of data that we have.

00:55:43.000 --> 00:55:56.000
And second, about the fault zone stuff. Skek has a community rheology model And I think that group is starting to think about making a fault zone rheology model, but we're in the early stages of putting something like that together.

00:55:56.000 --> 00:56:08.000
When we discussed that at the meeting, I just like saw problems materializing out of the atmosphere in the room. That's such a like an interesting philosophical problem that needs to be grappled with.

00:56:08.000 --> 00:56:09.000
But, yeah. Thank you.

00:56:09.000 --> 00:56:11.000
Yeah, it's going to be challenging.

00:56:11.000 --> 00:56:29.000
Well, maybe that gets into the question or the issue of evaluators, right? Scott, there were some questions about that and, you know, the use cases for the community of these products. So maybe can you… Talk a little bit about that process. Some people were interested in participating, it sounded like.

00:56:29.000 --> 00:56:32.000
The use cases or the evaluation that we're going through.

00:56:32.000 --> 00:56:40.000
Well, I think the use, well, the evaluation specifically but uh because of the diversity of use cases, you can see lots of different perspective.

00:56:40.000 --> 00:56:49.000
Yeah, so past What is meant by community and community models means different things to different groups.

00:56:49.000 --> 00:56:57.000
But for the community fault model, what it's meant to us is that every few years when we have a major revision to the fault.

00:56:57.000 --> 00:57:11.000
We put out an open call to the earthquake community to evaluate our model. And that evaluation has looked different in different years. But typically what we do is we have people look at faults that are in question and we have them answer some questions. We provide them what the fault is based on.

00:57:11.000 --> 00:57:26.000
And maybe if there's alternatives, what that's based on. And then people rate them. We usually have a quantitative like a Likert scale like one to five And then whichever rating ends up highest is what ends up as the preferred representation.

00:57:26.000 --> 00:57:38.000
And so since the Northern California portion of the community fault model is brand new this year, we recognize that it's likely that we're missing something or have something wrong. So it's really critical that people that are here now that have expertise

00:57:38.000 --> 00:57:50.000
To volunteer for this. And what we're going to do is we're going to separate it into regional groups. And each regional group is going to sort of look at the faults in that region and evaluate whether they're complete or whether we have things wrong.

00:57:50.000 --> 00:57:56.000
But the use cases of the CFM are broad. I mean, you heard about the NSHM, the National Seismic Hazard Model.

00:57:56.000 --> 00:58:03.000
That model is, at least in Southern California, is heavily based on the community fault model. So they're quite similar there.

00:58:03.000 --> 00:58:08.000
But people use it in crustal deformation models. People use them in seismic hazard models.

00:58:08.000 --> 00:58:14.000
Yeah, Scott Kelly, Callahan's talk was, I'm sure, the fault sources they used partially were derived from the CFM.

00:58:14.000 --> 00:58:18.000
So there's a broad range of uses for these models.

00:58:18.000 --> 00:58:31.000
Yeah. And on that, as, you know, SCC moves further north into Northern California now overlaps with the region that Crescent is evaluating.

00:58:31.000 --> 00:58:45.000
What are your thoughts on the overlap region and how you know these two efforts should uh should possibly collaborate or at least treat the overlapping region.

00:58:45.000 --> 00:59:02.000
So we are, Andreas, myself, and John Shaw were funded to collaborate with the Crescent Group right now to ensure compatibility between our two models in that region but Since we have to remember that Skek's been around for a long time. So we have this long legacy of products, whereas Crescent is pretty new.

00:59:02.000 --> 00:59:09.000
Their first pass community fault model is going to be the National Seismic Hazard Model, and that's already released in beta form.

00:59:09.000 --> 00:59:14.000
So as they start to build on that model and we start to build on our model in Northern California.

00:59:14.000 --> 00:59:20.000
Constantly in talks about making sure the models are compatible and in the same region.

00:59:20.000 --> 00:59:26.000
Again, we have a funded proposal to look at that area right now.

00:59:26.000 --> 00:59:35.000
All right. I'll just remind everyone that the speakers are available for your questions. So if you have questions, feel free Vera has a question.

00:59:35.000 --> 00:59:59.000
Yeah, a quick question to Scott Marshall. I think Skick has a geological framework as well right So can those So with these big strike slip falls, obviously they're going to juxtapose different blocks from you know different you know the same chunk in the geological model is going to appear in different places along a strike slip wall say so

00:59:59.000 --> 01:00:12.000
Are you going to, or maybe you already have a tie-in from the 3D fault model to the geological framework model like let's say i want to see what kind of blocks are juxtaposed along a specific part of a vault.

01:00:12.000 --> 01:00:23.000
Yeah, so Andreas plush is here right now. I see him on the video screen and he was one of the people that helped with the computer aspects of that framework model. So Andres, could you comment on that?

01:00:23.000 --> 01:00:39.000
Yeah, so yeah, that's exactly right. The geological framework model is actually based on the main faults in the CFM. So there should be a direct compatibility between these two models.

01:00:39.000 --> 01:01:00.000
But the actual content of the blocks and the framework model is, you know, sort of independently constrained and so it doesn't take into account you know strike slip or any other displacements or history a thing at this point.

01:01:00.000 --> 01:01:14.000
It's just lithology in a column. In these blocks in the framework model and the boundaries themselves of the boundaries themselves follow the main CFM faults.

01:01:14.000 --> 01:01:21.000
And their geometry. So there should be a good consistency.

01:01:21.000 --> 01:01:25.000
We have a question from Mark.

01:01:25.000 --> 01:01:39.000
Yeah, and I'm curious if the Northern California fault model will include the subducting border plate and the faults that are involved in that with the produces most of the seismicity.

01:01:39.000 --> 01:01:42.000
North of Mendocino triple junction.

01:01:42.000 --> 01:01:48.000
The subducting like the mega thrust or the

01:01:48.000 --> 01:01:54.000
No, in the Gorda plate, the oceanic plate, the subducting that has the faults that are producing much of our seismicity that area.

01:01:54.000 --> 01:01:57.000
Oh, um… That's actually been an active conversation with the Crescent Group as well.

01:01:57.000 --> 01:02:03.000
Yeah.

01:02:03.000 --> 01:02:04.000
Andres, maybe you want to comment?

01:02:04.000 --> 01:02:14.000
Yeah, no. So the megaverse is certainly part of that, but in the upper plate the faults which are involved in these large earthquakes.

01:02:14.000 --> 01:02:34.000
Are harder to constrain and then i guess at this point but consensus model, which can be easily included What we certainly have are the surface faults, which in the upper plate have traces and tracers also offshore traces.

01:02:34.000 --> 01:02:56.000
Sort of in between. Yeah, there's… Not sure if there's a good way to represent those faults, then probably needs to be more work on those in the community So at this point they are in not part but they obviously big sources of earthquakes so they're very important faults.

01:02:56.000 --> 01:03:02.000
Yeah. Thank you.

01:03:02.000 --> 01:03:10.000
We haven't had a question for Scott Callahan on the CyberShake work. Does anybody have a question for Scott?

01:03:10.000 --> 01:03:12.000
Scott Callahan again. Or, yeah.

01:03:12.000 --> 01:03:15.000
Poor Debbie. Right.

01:03:15.000 --> 01:03:23.000
Well, I have a question for Scott Callahan. It looks like, and maybe this loops Evan in, I'm not sure.

01:03:23.000 --> 01:03:37.000
The changes in the livermore Basin seemed pretty significant and it seems like that area is deserving of some further attention to try to resolve the geometry there. What are your thoughts?

01:03:37.000 --> 01:03:38.000
Both of the effect.

01:03:38.000 --> 01:03:52.000
Yeah, as I tried to show in the slides, there's you know being a moderate change there going from the previous iteration of the USGS Bay Area model to the current version.

01:03:52.000 --> 01:04:08.000
And so then, you know, running the grave propagation calculations through there, we see, you know, we see it light up, we see significantly larger ground motions But as I showed kind of right around that, that's in 10 second interval. I know Evan and I actually, we talked about this thing at the SCAC meeting because I

01:04:08.000 --> 01:04:14.000
We'd seen this and someone went and they said, hey, Evan, do you know anything about this? And Evan said, oh yeah, yeah, we tweaked the basin out there.

01:04:14.000 --> 01:04:23.000
So it's not too surprising. So it certainly seems to be very reflective of what is in the model. Now, of course, then there are questions about how do we feel about the model.

01:04:23.000 --> 01:04:34.000
But yeah, I think is maybe the largest or at least the most eye-catching change between the previous cyber state calculation six years ago in the Bay Area and the results now.

01:04:34.000 --> 01:04:36.000
I'll let you talk about it a little too, Evan.

01:04:36.000 --> 01:04:55.000
Yeah, so did you, you, you didn't really see much change around like three seconds, two seconds. It's more Yeah, so that update from the last that from the last update of the velocity model, we lowered the velocities In the Livermore Basin.

01:04:55.000 --> 01:05:02.000
Mainly to better fit ground most simulated ground motions are and like Two seconds.

01:05:02.000 --> 01:05:17.000
Half a hertz so it's yeah i mean it is possible that we lowered it too much at depth. But yeah, we need to kind of go back and validate maybe um lower frequencies but

01:05:17.000 --> 01:05:23.000
Yeah.

01:05:23.000 --> 01:05:35.000
I was going to say, although you had that question earlier about potential kind of collaborations with Crescent and pushed in the chat One thing with Cybershake that would be interesting to look at is subduction zone events. So far, the ERF that we've been using have been

01:05:35.000 --> 01:05:50.000
User VRFs which don't include the subduction zone but obviously those have potential impact hazard up in Northern California. And so that's, I think, another potential area for collaboration is to work with them on including simulations of some of the events up there and see what impact they would have.

01:05:50.000 --> 01:05:59.000
What are the northernmost events that Cybershake considers? It doesn't go up towards it.

01:05:59.000 --> 01:06:00.000
We have San Andreas events that go all the way offshore up there.

01:06:00.000 --> 01:06:02.000
The triple junction or

01:06:02.000 --> 01:06:06.000
But nothing beyond that. Yeah, taking on that.

01:06:06.000 --> 01:06:23.000
Okay. So does that consider like historic scenarios like the Ferndale, like those kind of geometries or are those more obscure.

01:06:23.000 --> 01:06:39.000
So it's the current it's the current Results I presented today are the use of two geometries, which I think are pretty cleaner. So I don't think we have anything particularly odd or complex up there in cybershake. I mean, I know people have done simulations of that but

01:06:39.000 --> 01:06:41.000
Yeah.

01:06:41.000 --> 01:06:44.000
All right. Scott Marshall. Yeah.

01:06:44.000 --> 01:06:46.000
Marshall has a question.

01:06:46.000 --> 01:06:56.000
God's going to ask Scott. Hey, Scott, so I saw that you'd tapered different velocity models. There's been several people that have been working on methods to combine different velocity models.

01:06:56.000 --> 01:07:03.000
I'm curious if you tested how different tapering methods between the velocity models affected your results.

01:07:03.000 --> 01:07:09.000
Like Patricia Prasad's worked on this and Yehuda's had some students working on this. Did it matter?

01:07:09.000 --> 01:07:15.000
So the tapering that we apply isn't necessarily across the smoothing boundaries. It's just the top 700 meters.

01:07:15.000 --> 01:07:16.000
Oh.

01:07:16.000 --> 01:07:26.000
And then it is applied uniformly. The merging of the models that we use is just an averaging within That's moving within 20 kilometers of the model interface.

01:07:26.000 --> 01:07:45.000
However, I know that there's a lot of interesting work being done looking at, as you mentioned, you had a British persona and I think a couple of others are looking at better ways to integrate velocity models I think it's especially interesting for regions where we have maybe higher resolution basin models that we kind of like to drop into a lower resolution but regional model.

01:07:45.000 --> 01:07:57.000
And there's been some good progress there in how do you kind of preserve the characteristics and you know, maybe smoothing is not necessarily the way to go because you might really have sharp boundaries that you want to preserve, right?

01:07:57.000 --> 01:08:08.000
So, you know, certainly kind of in a future cybersec study, I think it would be good to use one of these more advanced methods for merging these multi-scale velocity models.

01:08:08.000 --> 01:08:31.000
And seeing what impact that has on hazard. But for this calculation, what we did is we just did the averaging across the boundaries and then the taper that we applied is to the top 700 meters. And then that was based on work from Kim Olson's group who took a look at applying near surface VS30 based tapers to different depths and found that 700 to 1,000 meters was optimal for best reproducing historic events.

01:08:31.000 --> 01:08:41.000
So that's why for both the last Southern California Cyberishing Study and then for this one I presented today, we used that taper in the top 700 meters.

01:08:41.000 --> 01:08:42.000
Great. Yeah, thanks.

01:08:42.000 --> 01:09:00.000
Yeah, and just two points. Oh, sorry. But two points on community velocity models. One is that um This report that Brad Agard is a champion following up a SCC community seismic velocity or seismic wave speed model workshop in March of 2024.

01:09:00.000 --> 01:09:16.000
It's getting close to being released and Sarah mentioned that. And then Greg mentioned in his SCEC overview, there is an effort There will be a workshop For the April timeframe.

01:09:16.000 --> 01:09:23.000
To start to look at a larger effort for seismic wave speed models.

01:09:23.000 --> 01:09:35.000
Including what Scott talked about merging models and combining models. Before we get too far on, Debbie, is there anything that you want to add and reflect? And then we'll go to Zia's question.

01:09:35.000 --> 01:09:42.000
No, I was more the dots on plots earthquake girl so I think that's pretty straightforward. We'll move on to Zia.

01:09:42.000 --> 01:09:45.000
All right.

01:09:45.000 --> 01:10:00.000
Hey, thanks. Some of the plots you were showing for the cybershake results You mentioned that you use VS 30 of 400 meters per second, but then you were showing plots of 500 meters per second.

01:10:00.000 --> 01:10:01.000
Oh, okay.

01:10:01.000 --> 01:10:11.000
So our minimum VS was 400 meters per second. Yeah, the VS30 that we use for the taper comes from the Thompson model. And so I'd have to take a look. I don't remember exactly how low those values get.

01:10:11.000 --> 01:10:24.000
Because of the way we apply the taper though, because we only use the taper if it gets us a lower VS, there are definitely going to be places in the model where the Surface VS is down to 400 and the VS, if you were to calculate

01:10:24.000 --> 01:10:29.000
Kind of simulation VS30 in those regions, you'd probably see something close to 400 as well.

01:10:29.000 --> 01:10:32.000
Okay, thanks.

01:10:32.000 --> 01:10:33.000
All right, we're drawing.

01:10:33.000 --> 01:10:44.000
Scott, how much do you think you're… Underestimating ground motions by stopping at 400 because you know around the bay there's some places that are like 100 meters per second

01:10:44.000 --> 01:10:58.000
Yeah, Kim Olson's group ransom Simulations to take a look at that. So I picked a bunch of sites and ran at 400 and then I think down to 100 or something meters per second. And we saw that we're underestimating

01:10:58.000 --> 01:11:13.000
Probably in some places 30 to 40%, but only around two seconds at longer periods, we're getting most of we're we're getting basically all of it. So it's one of the reasons why one of the follow-ups we'd like to do is to lower that minimum VS.

01:11:13.000 --> 01:11:23.000
You know run a few, probably not the whole study again, but run a few targeted sites and take a look in cybershake, what impact do we see? Is it on par with kind of what they saw as well?

01:11:23.000 --> 01:11:34.000
Because, yeah, we know we're missing some of it, you know, given how low I think it goes down to 80 meters per second in some regions in the SGS model.

01:11:34.000 --> 01:11:53.000
Yeah, but of course there's two effects of low wave speeds at the surface on ground motion. There's the amplification due to elastic propagation but If you're close to the fault, like the Hayward Fault is closed to these very low wave speeds on the East Bay.

01:11:53.000 --> 01:12:09.000
The nonlinear site response can drive down the high frequencies so you know we we looked at this with our simulations of Hayward Fault earthquakes and looked at the effect of the minimum shear wave speed and tried to quantify that and

01:12:09.000 --> 01:12:20.000
That's available i mean we're available where hopefully as time progresses, honoring these models and the physics more closely.

01:12:20.000 --> 01:12:37.000
We're just about to wrap up. But there's probably time for one more question if anybody has a question in the audience or wants to emphasize the point.

01:12:37.000 --> 01:12:38.000
Here I am.

01:12:38.000 --> 01:12:45.000
If nobody else has a question, a quick one to Scott Marshall. What does the fault evaluation entail?

01:12:45.000 --> 01:12:54.000
What kind of people are you looking for and what kind of, you know, processes?

01:12:54.000 --> 01:12:59.000
Sorry, I was pasting a link. And then it said the link's too long. So I just got denied.

01:12:59.000 --> 01:13:11.000
We're looking for a wide range of expertise. Anybody that is knowledgeable or has direct experience working on these fault systems, especially in regards to their geometry.

01:13:11.000 --> 01:13:21.000
In Northern California. And again, what it's going to entail is we're going to separate these into each group will be given a specific region to look at that they have interest or experience in.

01:13:21.000 --> 01:13:26.000
And we'll be giving them a guided series of sort of questions or tasks to answer.

01:13:26.000 --> 01:13:32.000
And if you're worried, if you're thinking, oh, I want to do that, but I'm not sure, I don't have any software, dah, dah, dah.

01:13:32.000 --> 01:13:51.000
We're going to do that all through our CFM Explorer software. I was trying to post a link, but there's a There's a character limit in Zoom and it won't let me post a link Because it's too long. Otherwise, you can actually save direct links and be able to look at stuff directly in 3D with no specialized software. So we're trying to make it easy on people.

01:13:51.000 --> 01:14:02.000
And we're trying to keep the time commitment reasonable, but it will be an effort because it's an important model and we want to make sure people put time into making sure that it's correct.

01:14:02.000 --> 01:14:18.000
Does that kind of answer the question? I mean, it's a little bit vague because it depends on the region.

01:14:18.000 --> 01:14:37.000
All right. Rachel, just to answer your question, I think nonlinear soil response And the algorithm I used, it's based on the five hertz for a spectral amplitude. So it generally is impacting a few hertz and higher.

01:14:37.000 --> 01:14:50.000
And more pronounced at higher frequencies. All right. Well, I think that brings this session to a close. Thanks to the speakers again for great talks and answering questions. And thanks all for the discussion.

01:14:50.000 --> 01:14:55.000
I guess we need Sarah right now.

01:14:55.000 --> 01:15:01.000
Thank you everyone for a fantastic session. That was really exciting.

01:15:01.000 --> 01:15:15.000
Okay, so we have to squeeze in a late breaking session on a Cape Mendocino earthquake so Unless you are giving a talk in part one you get to get a 15 minute break but If you're giving a talk in port one session.

01:15:15.000 --> 01:15:21.000
Please stick around. Laurie, Vulcan, Bo, Clara, David. How and Bob. Everybody else?

01:15:21.000 --> 01:15:23.000
Staying on the ground. Not a lot of people.

01:15:23.000 --> 01:15:28.000
Take 15