WEBVTT 00:00:05.000 --> 00:00:15.000 Good morning, everyone! Day 2 of 3 of the NorCal Workshops! Thank you all for coming. We are going to kick things off with a bang, literally, because it's Thunder Talk time. 00:00:15.000 --> 00:00:21.000 Once again, record number of Thunder Talks! Yay! But record number of Thundertalks. 00:00:21.000 --> 00:00:27.000 Do as many questions as you can in the chat, because we're not going to have a lot of time for Q&A at the end. Without further ado, 00:00:27.000 --> 00:00:31.000 Grace Parker, take us away! 00:00:31.000 --> 00:00:40.000 Thanks, Sarah! I think we can get right into it. As a reminder, each talk will be 3 minutes, and they'll go directly from speaker to speaker, so speakers, when you see your slides, 00:00:40.000 --> 00:00:42.000 Go for it. 00:00:42.000 --> 00:01:06.000 And take it away, Sue. 00:01:06.000 --> 00:01:09.000 You're muted, Sue. 00:01:09.000 --> 00:01:22.000 Actually, she's not muted, but we can't hear her. 00:01:22.000 --> 00:01:23.000 Yes, we can hear you. 00:01:23.000 --> 00:01:24.000 Okay, can you hear me? Sorry, that was an IT glitch on my end. So, good morning, honored to kick things off with a bang. 00:01:24.000 --> 00:01:29.000 Yes. 00:01:29.000 --> 00:01:39.000 Uh, if you heard me say yesterday that there was a sizable earthquake on the creeping section in the 19th century, I was wrong, I take it back. I accept Topazade's conclusion it was a big earthquake. 00:01:39.000 --> 00:01:48.000 I don't accept the creeping zone location. Um, but next slide, please. Last year, at the workshop, I talked about legacy macro-seismic data. 00:01:48.000 --> 00:01:55.000 Before there was Did You Feel It? We had postcards to collect earthquake reports. They were… summarized in great literature reports. 00:01:55.000 --> 00:01:59.000 with pages that look like the page on the right. Next slide. 00:01:59.000 --> 00:02:07.000 Um, so you can take those pages, you can mine them by hand, and last year I talked about the 1954 Fickle Hill. 00:02:07.000 --> 00:02:12.000 Earthquake. Um, Bill Ellsworth suggested looking at the 1957 Daly City earthquake. 00:02:12.000 --> 00:02:17.000 Uh, which I did. Turns out there are over 2,000 reports for that earthquake. 00:02:17.000 --> 00:02:25.000 mining and geolocating the reports by hand would be excruciating, so… AI to the rescue large language models. 00:02:25.000 --> 00:02:38.000 Uh, get it to use their powers for good and not evil, and we're able to bring these accounts to life, or this earthquake with the help of a couple of interns, including a very bright high school student, uh. 00:02:38.000 --> 00:02:45.000 Arnaf Agrawal. And next slide. So, quickly, 1957, here's my typo, it's not 1857. 00:02:45.000 --> 00:02:55.000 Uh, we can produce a quite modern-looking shake map for this earthquake, which was widely felt in the Bay Area. It's one of the more interesting earthquakes through the 20th century. 00:02:55.000 --> 00:03:01.000 Um, so intensity map on the left, constrained by 2,000-odd reports. 00:03:01.000 --> 00:03:09.000 Intensity residuals on the right. Uh, the intensities overall are quite well fit by a modern intensity prediction equation. 00:03:09.000 --> 00:03:15.000 You see that red halo? That can be explained by a combination of two things. One, directivity towards the north-northwest. 00:03:15.000 --> 00:03:20.000 The other is post-critical Moho reflections, which were first identified. 00:03:20.000 --> 00:03:26.000 in the community, only in the 80s, but the signature is totally in the… in these data. Next slide. 00:03:26.000 --> 00:03:38.000 So just some, uh, summary, uh, that I already mostly talked about. And lastly, next slide, I wanted to start and with a teaser, speaking of macro-seismic data, I've been looking at a different type. 00:03:38.000 --> 00:03:47.000 Recently, um… broken trees, or trees that are broken by earthquakes, and using them to figure out. 00:03:47.000 --> 00:03:58.000 how we can constrain near-field shaking from. a large, early instrumental or pre-instrumental events. So, uh, if you're going to go to SSA, I'll be talking about this. 00:03:58.000 --> 00:04:14.000 There. Thank you. Perfect. 00:04:14.000 --> 00:04:20.000 Okay, um, I will just dive right in then, I guess. Thanks very much, Sue, for nicely teeing 00:04:20.000 --> 00:04:23.000 up, um, the… 00:04:23.000 --> 00:04:28.000 1957 earthquake, I want to talk a little bit about some work that I did with 00:04:28.000 --> 00:04:35.000 Peggy Helwig and Doug Drager, and a whole bunch of other great, um, collaborators on the 1954 earthquake a little bit further up the coast. 00:04:35.000 --> 00:04:40.000 And how we'd like to extend that, um, to the 1957 earthquake. 00:04:40.000 --> 00:04:44.000 So, just to… 00:04:44.000 --> 00:04:48.000 cover some of what Sue already said. Um, the… oh, we went… 00:04:48.000 --> 00:05:00.000 forward. The 54 Cotton Northern California earthquake was estimated at, um, a 6.5, and this was in the Eureka Arcata-Corbel area in December of 1954. 00:05:00.000 --> 00:05:03.000 Uh, those are these pictures on the left here. 00:05:03.000 --> 00:05:13.000 Uh, there is a decent amount of some structural damage done, and it was hundreds of thousands of dollars damage, estimated as millions today. On the other hand, 00:05:13.000 --> 00:05:19.000 Three years later, there was an earthquake that was estimated at 5.7 in San Francisco, or the Daily City earthquake. 00:05:19.000 --> 00:05:23.000 this one, and Sue, I'm sure, could clarify even more. 00:05:23.000 --> 00:05:33.000 is estimated, or was estimated, at 5.7. Um, did non-structural damage and structural damage, but again, around about a million dollars of damage. And it's of interest to quantify these. 00:05:33.000 --> 00:05:42.000 So, one thing that we did for 1954 is we looked at strong motion data, um, that was triggered and recorded at Eureka. 00:05:42.000 --> 00:05:45.000 Um, and we were able to… 00:05:45.000 --> 00:05:50.000 not only scan those images, but use software to digitize them, 00:05:50.000 --> 00:05:53.000 into high-resolution SAC time series. 00:05:53.000 --> 00:06:00.000 and be able, from there, to combine that information, that three-component strong-motion digitized data, 00:06:00.000 --> 00:06:06.000 with First Motions picked from old scanned seismograms, and altogether put that 00:06:06.000 --> 00:06:13.000 into a focal mechanism, um, fitting exercise, such that we can at least constrain 00:06:13.000 --> 00:06:16.000 Um, the solution 00:06:16.000 --> 00:06:22.000 to being a preferred, uh, mechanism along the Cascadia interface, rather than 00:06:22.000 --> 00:06:30.000 A gorter plate or aino type, which, of course, are also potentials given the complex tectonics in this area. 00:06:30.000 --> 00:06:36.000 Um, this does require a good amount of digging through old archives, and 00:06:36.000 --> 00:06:42.000 rather than looking at the old words, like, um, Sue was talking about, in those wonderful old seismic reports, we are trying to 00:06:42.000 --> 00:06:46.000 trace over those old wiggles, um, after we find the data. And so this is 00:06:46.000 --> 00:06:56.000 What some of the… a taste of what we can do, uh, given this sort of analysis with 1954, and something we would like to very much go through with 1957. 00:06:56.000 --> 00:07:02.000 We started kind of looking at some of these things, and already we do know that there is data, uh, it's a matter of starting to track it down and… 00:07:02.000 --> 00:07:07.000 go through the same digitization process. 00:07:07.000 --> 00:07:13.000 Do I still have time left? 00:07:13.000 --> 00:07:21.000 That's 30 seconds. 00:07:21.000 --> 00:07:31.000 Can everybody hear me? So, I'm Fabio Terra, and I did this work with Lisa Schleicher at the USGS while I was at Berkeley. 00:07:31.000 --> 00:07:40.000 So, um, everybody uses shake maps. Um, they're an important part of our science process. Um, in Northern California. 00:07:40.000 --> 00:07:49.000 We have 3 different agencies that issue shape maps. Um, California Geological Survey, the Global ShakeMap folks at the NEIC, and, um. 00:07:49.000 --> 00:07:55.000 the Northern California Seismic System, which is composed of UC Berkeley and the USGS offices at Moffitt. 00:07:55.000 --> 00:08:01.000 Um, and this is an example from Kate Mendeshino offshore earthquake in 2024. 00:08:01.000 --> 00:08:11.000 And often, there are differences in the shake maps, and um… the… this is a rapidly evolving software, and when we go from version to version, we don't. 00:08:11.000 --> 00:08:20.000 Not necessarily configure or have all the same. data inputs when we issue shake maps from agency to agency. 00:08:20.000 --> 00:08:25.000 Carol? Next slide, please. 00:08:25.000 --> 00:08:35.000 See those little… that right there. Got it. Okay. Now, no. 00:08:35.000 --> 00:08:43.000 Sorry, observational ditches. Okay. 00:08:43.000 --> 00:08:51.000 Okay, so the first thing we did with the shake maps is we just listed what the parameters are for one shape map. This is an example from. 00:08:51.000 --> 00:08:56.000 Berkeley earthquake in 2025, and we have four major observations. So. 00:08:56.000 --> 00:09:05.000 Sometimes different stations were missing between versions. Um, there's variations in the ground motion outputs, they're minor, but there's variations. 00:09:05.000 --> 00:09:13.000 Um, there's also variations in what version of the shape map throws out stations from being included in the shape map, and then. 00:09:13.000 --> 00:09:19.000 Um, different versions have different numbers of did you feel it data points. 00:09:19.000 --> 00:09:32.000 So, to go from place to place… So, um, some observations. We… we saw differences in the number of, um, observations, especially at state borders. 00:09:32.000 --> 00:09:35.000 State boarders don't stop the shake map ground motions. 00:09:35.000 --> 00:09:38.000 Second, we see a little bit of variation in the ground motions. 00:09:38.000 --> 00:09:47.000 Third, um, we looked at heat maps between, um, the different versions of shape maps and the very strong heat maps show that there's, um. 00:09:47.000 --> 00:09:53.000 Variation in the configurations, um, throw out certain stations from, um. 00:09:53.000 --> 00:09:58.000 being included in the versions of the shake map, so we see those variations, and then we see. 00:09:58.000 --> 00:10:07.000 Minor variations, and the pink circles. In the next slide, and those are because of did you feel at duty of points that may be included or not included in the shape maps. 00:10:07.000 --> 00:10:12.000 So, I'll, um, get into some detail about one of these elements. 00:10:12.000 --> 00:10:19.000 So this is an example of the heat map between the NCSS version of the shake map and the NCIC. 00:10:19.000 --> 00:10:28.000 And we will see, once the screen comes back, that, um… We met with the NEIC, and the reason the. 00:10:28.000 --> 00:10:37.000 shake maps were different was because the, um… differences in cloud computing and the transition from. 00:10:37.000 --> 00:10:43.000 Um, moving to Moffitt. And I'll stop there, because I have no time left. 00:10:43.000 --> 00:10:47.000 But we… we can answer your questions better with a tool like this. 00:10:47.000 --> 00:11:02.000 Thanks. 00:11:02.000 --> 00:11:14.000 Good morning, everyone! My name is Anoushka. I'm an undergrad at UC Berkeley, um, and I'm going to give a brief overview of our work for ground mushroom analysis in the Mendocino fracture zone. 00:11:14.000 --> 00:11:22.000 Okay, um, oh. 00:11:22.000 --> 00:11:33.000 The medicinal fracture zone, as many of you know, is California's most seismically active region, and earthquakes along this region, um, produce really strong ground motions, um, like seen in the magnitude 7 Cape Mendocino earthquake in 2024. 00:11:33.000 --> 00:11:44.000 The region is really unique in that in the south, you have the, um, northwest moving Pacific plate, um, you have the eastward moving Gorda plate in the north, and then the North American plate to the east. 00:11:44.000 --> 00:11:55.000 The boundary is really challenging to model with ground motion modeling, um, but that's part of the reason why it's very unique, and it is difficult because they, uh, have both crestal and subduction-like qualities. 00:11:55.000 --> 00:12:01.000 Um, so that's ultimately our motivation for the project, along with the fact that the Mendocino, um, Magnitude 7 Mendocino. 00:12:01.000 --> 00:12:08.000 Mendocino earthquake in 2024, um, provides us, like, steps towards going, um. 00:12:08.000 --> 00:12:15.000 to characterize the region for hazard and safety purposes. 00:12:15.000 --> 00:12:26.000 So our data self itself is adapted from the 2025 Lomax High Precision Earthquake Relocation Catalog. Um, we reconsolidated the, um, dataset by placing a 5-record minimum. 00:12:26.000 --> 00:12:31.000 Um, and this ultimately resulted in around 25,000 records and 486 events. 00:12:31.000 --> 00:12:39.000 Our catalog is mostly, um, magnitude events below magnitude 4, but we have 23 events magnitude 5 and above. 00:12:39.000 --> 00:12:47.000 Um, with the most notable being the, um, Cape Mendocino 2024 earthquake, the 2022 Ferndale earthquake, and the 2021 Petrolia earthquake. 00:12:47.000 --> 00:12:53.000 So our plan is to use, um, this data to look at the NGA West 2 and NGA subduction models. 00:12:53.000 --> 00:12:59.000 Um, and look at potential trends with, um. Spatial trends and, um, explanatory variables. 00:12:59.000 --> 00:13:04.000 And we're using linear mixed effects models to investigate the event, site, and path effects more, um. 00:13:04.000 --> 00:13:10.000 like, specifically. 00:13:10.000 --> 00:13:18.000 So, from our preliminary results, uh, we noticed that our data shows indicators, um, that there is stronger distance attenuation, um, than the models can predict. 00:13:18.000 --> 00:13:25.000 Um, and we try to rectify this, uh, with a distance magnitude scaling, um, to counter-effect, uh, potential trade-offs. 00:13:25.000 --> 00:13:35.000 Um, and as you can see, with the distance magnitude scaling, the spatial coherence decreases, but it's still possible that we're experiencing certain trade-offs, so we're still going to investigate that. 00:13:35.000 --> 00:13:40.000 Um, we also, uh, are noticing interesting trends with the event terms. 00:13:40.000 --> 00:13:52.000 But, yeah. Oh, okay, great. Um, so we're noticing that to the east of the transform, uh, you have more negative event terms, and to the west of the transform, um, there's more positive event terms. 00:13:52.000 --> 00:14:02.000 Um, and this might indicate something about the crustal and subduction properties. Um, so our plan is to look at the event site and path effects, um, in which plate they originate in and travel through. 00:14:02.000 --> 00:14:21.000 Um, to see what they might highlight about potential physical takeaways for the region. 00:14:21.000 --> 00:14:30.000 Okay, uh, good morning. My name is Tara Nye. I'm with USGS, and I'm going to briefly discuss some implications of attenuation observed in the San Francisco Bay Area. 00:14:30.000 --> 00:14:38.000 Many earthquake studies rely on ground motion models. Empirically derived models, which are often developed using multi-regional datasets. 00:14:38.000 --> 00:14:46.000 If we compare these average model predictions with observed ground motions, we can see that the Bay Area tends to be more attenuating than these average models. 00:14:46.000 --> 00:14:53.000 So this is shown for three example earthquakes in the Bay Area. The blue line is the original average ground motion model prediction. 00:14:53.000 --> 00:14:59.000 Using a large data set of Bay Area ground motions, we calculated attenuation adjustments for two ground motion models for the Bay Area. 00:14:59.000 --> 00:15:06.000 And so, and shown in red is the adjusted ground motion model, which provides a better fit to what we actually observe. 00:15:06.000 --> 00:15:10.000 If you want to learn more about the actual details of how we calculated these attenuation adjustments. 00:15:10.000 --> 00:15:15.000 You can read our publication, which as of yesterday was accepted by SRL. 00:15:15.000 --> 00:15:21.000 But here, I just want to highlight some interesting implications of using these Bay Area regional models. 00:15:21.000 --> 00:15:27.000 So, um, probabilistic seismic hazard assessments rely heavily on ground motion models. 00:15:27.000 --> 00:15:37.000 As a simple illustration of the impact of using different ground motion models in seismic hazard, Alan Poulos helped me code up a toy hazard model, which is just a simple model here with a linear fault. 00:15:37.000 --> 00:15:41.000 An array of seismic stations running perpendicular to the fault. 00:15:41.000 --> 00:15:49.000 At three of these stations, I've shown what the hazard curves would look like, so in each panel, the blue line is the hazard curve using one of our original ground motion models. 00:15:49.000 --> 00:16:01.000 The red line is the hazard curve using our Bay Area adjusted ground motion model. And so what we can see is, as we increase in distance from the fault, we see a greater deviation between those hazard curves. 00:16:01.000 --> 00:16:10.000 To quantify this deviation at each of these stations, I calculated the percent decrease in the 50-year, 2% PGA exceedance levels. 00:16:10.000 --> 00:16:14.000 And there's an increase in this… or there's a greater decrease. 00:16:14.000 --> 00:16:21.000 at the farthest stations, ranging up to about 20-30% for this specific scenario. 00:16:21.000 --> 00:16:26.000 Earthquake early warning also, um, is impacted by the use of ground motion models. 00:16:26.000 --> 00:16:36.000 So during the event of an earthquake, ground motions are estimated using real-time source characterization combined with ground motion models, and these ground motions are converted to MMI. 00:16:36.000 --> 00:16:42.000 And then, uh, way alerts are issued to people located within the MMI 4 and above contours. 00:16:42.000 --> 00:16:46.000 So here I've shown the shake map for the 2014 Napa earthquake. 00:16:46.000 --> 00:16:49.000 The shaking intensity shown is for MMI 4 and above. 00:16:49.000 --> 00:16:55.000 The solid circle shows what the original alert area would be using one of the original ground motion models. 00:16:55.000 --> 00:16:59.000 The dashed line shows the new alert area using our Bay Area model. 00:16:59.000 --> 00:17:07.000 And we can see that it better fits the observed intensities, and it helps to reduce the alert area and also helps to reduce over-alerting. 00:17:07.000 --> 00:17:31.000 These are just some simple examples, but they do help to highlight that the high attenuation that we observe here in the Bay Area should be considered not only for long-term hazard forecast, but also for these real-time earthquake alerts. 00:17:31.000 --> 00:17:38.000 Alright. Um… Hi, everybody. It's great to see you all, and I'm going to be talking very similar. 00:17:38.000 --> 00:17:51.000 topic to the last couple of presentations. Um… We're looking at more complex regions in California that have ground motions that might be a little different than we would expect. 00:17:51.000 --> 00:17:59.000 From our global ergotic ground motion models, so… Um, this has all been a study that was really kicked off after the. 00:17:59.000 --> 00:18:10.000 Magnitude 4.8 Isla Vista earthquake. in 2013, and so you see there, it was the 29th of May, 2013. There was this. 00:18:10.000 --> 00:18:18.000 You know, moderate earthquake off the coast of Santa Barbara, about 7 kilometers west of UC Santa Barbara's campus. 00:18:18.000 --> 00:18:26.000 Uh, which has 3 seismometers, um… There are strong motion stations right there on the campus. 00:18:26.000 --> 00:18:38.000 Uh, two CGS ones, you can see one up towards the northwest in that map, Station 25091. There's Station 25392 in the middle, and then. 00:18:38.000 --> 00:18:48.000 Down towards the right, Station USB. Now, in this… Same earthquake event that happened, uh, 2013. 00:18:48.000 --> 00:19:07.000 Obviously, all these stations shared the same source. They shared practically the same path, uh, the… especially 25392 and USB are only separated by about 350 meters, which is about a 2 degree difference with respect to the epicenter. 00:19:07.000 --> 00:19:12.000 Um, of the earthquake, and both of these sites have. 00:19:12.000 --> 00:19:16.000 What we would characterize as the same site conditions. 00:19:16.000 --> 00:19:21.000 At least, from an engineering seismology standpoint, they have, uh. 00:19:21.000 --> 00:19:33.000 about the same thickness of marine terrace deposits overlying Sisquac Formation Silkstone, with almost exactly the same VS30 between these two stations. 00:19:33.000 --> 00:19:43.000 Uh, but what we saw was that in the same earthquake, and there were no errors in the recording instruments, this was double-checked by Jamie Steidele. 00:19:43.000 --> 00:19:52.000 Uh, the PGA at 25392 was .08G, and the PGA at USB was .26G, about 3 times as large. 00:19:52.000 --> 00:20:01.000 So, there's got to be something going on with some kind of site condition, right? That's messing with it, so I gathered a bigger data set. 00:20:01.000 --> 00:20:08.000 of earthquakes in the region and around both Recorded and Santa Barbara and at stations all over California. 00:20:08.000 --> 00:20:13.000 To be able to try and get a robust mixed effects regression going. 00:20:13.000 --> 00:20:25.000 And there's some interesting… overall biases in this data set that I've been finding, and there's still a lot of work that has to be done in this, but. 00:20:25.000 --> 00:20:33.000 My thinking is that there's actually, because of the tortured geology and complex faulting and folding in the Santa Barbara region, there seems to be. 00:20:33.000 --> 00:20:40.000 some kind of different path attenuation as waves travel out of the Santa Barbara region from earthquakes source there out to the rest of California. 00:20:40.000 --> 00:20:47.000 And then within the Santa Barbara region itself, there's also a lot of highly variable. 00:20:47.000 --> 00:20:53.000 Um, site responses you get. So, more work to be done on this, but I think there's going to be some interesting. 00:20:53.000 --> 00:21:07.000 changes that we want to see in the prediction of ground motions here. 00:21:07.000 --> 00:21:11.000 Good morning. Thank you. 00:21:11.000 --> 00:21:12.000 people hear me? 00:21:12.000 --> 00:21:14.000 Yep. 00:21:14.000 --> 00:21:17.000 I'm going to proceed. Yes, thanks. 00:21:17.000 --> 00:21:24.000 So, my name's Mitch Craig, uh, I'll be talking about near-surface velocity in the East Bay and Delta. 00:21:24.000 --> 00:21:32.000 San Francisco East Bay and Sacramento Delta. My co-authors are Mike Romero, also Cal Cities Bay, and Koichi Hayashi. 00:21:32.000 --> 00:21:35.000 Um, Kyoto University. 00:21:35.000 --> 00:21:39.000 So, uh, we'll be, uh, I'll be showing you results from a… 00:21:39.000 --> 00:21:46.000 about 50 sites, soft sediment sites, about half of them are along the East Bay Plain or shoreline. 00:21:46.000 --> 00:21:49.000 And the other half are in the Sacramento Delta. 00:21:49.000 --> 00:21:56.000 And thanks, uh, good fortune to have, uh, let me see if I can advance these. Yes, it's advancing… 00:21:56.000 --> 00:21:58.000 Um, 00:21:58.000 --> 00:22:03.000 Can… giving… explaining what VS30 is. 00:22:03.000 --> 00:22:08.000 Anyway, um, the goal of this study… well, what we did is we used 00:22:08.000 --> 00:22:11.000 Both combined active and passive seismic surface wave methods, 00:22:11.000 --> 00:22:16.000 Uh, to measure upper near-surface S wave velocity, 00:22:16.000 --> 00:22:23.000 And we… our products include… our primary product is velocity profiles, so that's… 00:22:23.000 --> 00:22:25.000 Uh, S wave velocity versus depth. 00:22:25.000 --> 00:22:34.000 Um, from that, we compute VS30, which is of interest to engineering geologists. It's used in calculating ground motion. 00:22:34.000 --> 00:22:40.000 Estimates, um, and then also there's another product, the site class, which is using building codes. 00:22:40.000 --> 00:22:46.000 Um, and the latest versions are NEHERP 2020 and ASCII. 00:22:46.000 --> 00:22:51.000 We did a few surveys, uh, using two-dimensional arrays. 00:22:51.000 --> 00:23:02.000 So, either triangular or L-shaped arrays to complement the bulk of our work was done using linear arrays. 00:23:02.000 --> 00:23:05.000 saying, going forward, 00:23:05.000 --> 00:23:09.000 Trying to advance the slide. 00:23:09.000 --> 00:23:12.000 What do I do to advance the slide? 00:23:12.000 --> 00:23:14.000 There, there it goes. Okay. 00:23:14.000 --> 00:23:18.000 Oops, I'm backing up on… 00:23:18.000 --> 00:23:20.000 Okay, here's the summary, uh, showing 00:23:20.000 --> 00:23:23.000 Um, the distribution of sites, 00:23:23.000 --> 00:23:27.000 Um, so, uh, these are the new… 00:23:27.000 --> 00:23:29.000 me, her scheme, 00:23:29.000 --> 00:23:34.000 So, um, the center average was in the Class DE, 00:23:34.000 --> 00:23:37.000 Which is centered in about 175 00:23:37.000 --> 00:23:40.000 meters per second, uh, 30 seconds left. 00:23:40.000 --> 00:23:45.000 Um, some of our slides were the low velocity ones, those tend to be more in the delta. 00:23:45.000 --> 00:23:50.000 Um, so, these, uh, classification schemes are based… 00:23:50.000 --> 00:23:54.000 Um, solely on velocity, on the VS30, 00:23:54.000 --> 00:23:58.000 Um, now there's another category that I'm not going to go into right now, but it's Class F. 00:23:58.000 --> 00:24:03.000 So if there's something that's potentially liquefiable or peat, it gets thrown into F. 00:24:03.000 --> 00:24:10.000 So, a lot of our sites in the Delta should actually be Class F, because they do have more than a 2 or 3 meter thick peat layer. 00:24:10.000 --> 00:24:15.000 So, I'm gonna finish with just a comparison 00:24:15.000 --> 00:24:19.000 Um, of… if I can get my last slide to come up. 00:24:19.000 --> 00:24:24.000 It's taking a minute. 00:24:24.000 --> 00:24:27.000 trying to jump to the last slide. 00:24:27.000 --> 00:24:32.000 Um, nope, you need to come back one, sorry about that. 00:24:32.000 --> 00:24:38.000 There we go. Um, so we compared results from 2D arrays and 1D arrays. 00:24:38.000 --> 00:24:41.000 Um, at, um… 00:24:41.000 --> 00:24:45.000 two sites in the East Bay, and the results were almost 00:24:45.000 --> 00:24:50.000 identical. Um, comparison with the VS30 map of Wills et al., 00:24:50.000 --> 00:24:53.000 are very good at those East Bay sites as well. 00:24:53.000 --> 00:25:04.000 Um, comparisons, um, with the Wills map are not quite so good in the San Ramon Valley, um, or the Delta, so I'll end with that. 00:25:04.000 --> 00:25:11.000 Good morning, everyone. My colleagues and I examined variations in ground shaking near the Hayward Fault using data we recorded in 2013. 00:25:11.000 --> 00:25:19.000 In August 2013, CSU East Bay imploded a 13-story building they deemed unsafe because it was only 600 meters from the Hayward Fault. 00:25:19.000 --> 00:25:23.000 The building weighed approximately 11 million kilograms. 00:25:23.000 --> 00:25:26.000 It presented an excellent known source. 00:25:26.000 --> 00:25:30.000 to evaluate ground shaking as it radiated from the source and across faults. 00:25:30.000 --> 00:25:35.000 In the East Bay, we used multiple arrays of 600 seismometers. 00:25:35.000 --> 00:25:38.000 Uh, to evaluate the ground shaking. 00:25:38.000 --> 00:25:40.000 The main array is shown here. 00:25:40.000 --> 00:25:46.000 And it consisted of 2 kilometer-long radio arrays emanating from the source. 00:25:46.000 --> 00:25:52.000 outward, with 200 meter spacing between seismometers centered on the explosion. 00:25:52.000 --> 00:25:57.000 There were 30 of these, uh, radio arrays going out at 12 degrees apart. 00:25:57.000 --> 00:25:59.000 And this formed a 360… 00:25:59.000 --> 00:26:04.000 degree coverage, so that we could look at propagation. 00:26:04.000 --> 00:26:08.000 This radio array configuration also forms 10… 00:26:08.000 --> 00:26:12.000 constituent circles around the implosion. 00:26:12.000 --> 00:26:15.000 Looking at one of those circular arrays, 00:26:15.000 --> 00:26:21.000 Shown here, which is 1,000 meters from the source. 00:26:21.000 --> 00:26:25.000 Crosses over into the East Bay, playing across the Hayward Fault. 00:26:25.000 --> 00:26:27.000 If we look at the plot of… 00:26:27.000 --> 00:26:29.000 PGV versus… 00:26:29.000 --> 00:26:36.000 the angle of this plot. Essentially, we unwrap it from 0 to 360 degrees. 00:26:36.000 --> 00:26:39.000 What you see is that the East Bay Plain, we get… 00:26:39.000 --> 00:26:41.000 Uh, sediment amplification. 00:26:41.000 --> 00:26:45.000 Relative to the hills, except for a couple places where we see 00:26:45.000 --> 00:26:48.000 Some topographic amplification, we know those are on hills. 00:26:48.000 --> 00:26:55.000 Another major propagation effect that we observe is what I call fault shielding. Perhaps it's more important than the… 00:26:55.000 --> 00:26:57.000 actual amplification in the basin. 00:26:57.000 --> 00:27:02.000 The inset shows, uh, we emanate, uh, energy from… 00:27:02.000 --> 00:27:05.000 the source out across default. 00:27:05.000 --> 00:27:07.000 into the East Bay Plain. 00:27:07.000 --> 00:27:10.000 And if we look at the plot of PGG versus… 00:27:10.000 --> 00:27:19.000 distance, what we see is near the source, the energy is very high. Each time it comes across a fault, the energy steps down dramatically. 00:27:19.000 --> 00:27:26.000 And we get minor amplification out in the basin, but it's minor relative to the steps across the fault. 00:27:26.000 --> 00:27:29.000 Looking to the northeast across the Chabot Fault. 00:27:29.000 --> 00:27:34.000 We see that same effect. Essentially, you get this fault shielding, where… 00:27:34.000 --> 00:27:39.000 energy drops dramatically across the Chabot Fault. 00:27:39.000 --> 00:27:44.000 There was also an earthquake on the San Andreas Fault while the array was out, and we recorded that data. 00:27:44.000 --> 00:27:47.000 And we can also see these effects. 00:27:47.000 --> 00:27:51.000 Uh, from the earthquake data as they cross the fault. 00:27:51.000 --> 00:27:54.000 These are only two of many effects that we observed. 00:27:54.000 --> 00:27:55.000 Thank you for your time. 00:27:55.000 --> 00:28:11.000 Yep. 00:28:11.000 --> 00:28:12.000 Yes. 00:28:12.000 --> 00:28:22.000 Hello, can everybody hear me? Awesome, awesome. Good morning, everybody. Um… Today, I want to talk to you about some of the work that we're doing down here in Southern California, um, related to… and sometimes we'll be doing another food ready to… 00:28:22.000 --> 00:28:30.000 Um, physics-based earthquake simulations. We, um… let me just jump into it so we can go ahead and start, make sure the slides are going. 00:28:30.000 --> 00:28:34.000 So, uh, so the first thing I want to talk about is what's the problem and why? 00:28:34.000 --> 00:28:42.000 why we're trying to validate physics-based earthquake simulations using observant, um… response metrics. So we know that we've had over. 00:28:42.000 --> 00:28:46.000 We've had over, um… the years, we've had multiple events. 00:28:46.000 --> 00:28:51.000 that have occurred, um, and each time a large event occurs, we learn something new. 00:28:51.000 --> 00:28:57.000 Um, because prior to that, we did not have the data in that fashion to be able to understand what's going on. 00:28:57.000 --> 00:29:04.000 And as the time proceeds, we expect this to continue to happen. 00:29:04.000 --> 00:29:10.000 Um, as we know, geology doesn't stop for anybody, so we're going to keep having these major earthquakes over time, maybe not. 00:29:10.000 --> 00:29:13.000 As frequently, but you can see the rate at which it's happening. 00:29:13.000 --> 00:29:20.000 So this is the kind of problem, is that we have these earthquakes happen, and often we don't have the data, essentially, to assess or analyze. 00:29:20.000 --> 00:29:27.000 the consequences of these events. As a result. 00:29:27.000 --> 00:29:32.000 If I can get my slide to move… Oh, boy. 00:29:32.000 --> 00:29:41.000 Okay. So, the question is, why use simulations, and why validation is critical? Well, the main thing is simulations give us an opportunity to be able to. 00:29:41.000 --> 00:29:46.000 to essentially pro… prepare or generate data that has not happened yet. 00:29:46.000 --> 00:29:49.000 For some reason, the GIF is not playing, but that's okay. 00:29:49.000 --> 00:29:54.000 Um, but again, these simulations give us the opportunity to essentially. 00:29:54.000 --> 00:30:00.000 generate earthquakes that have not happened, that are of critical nature, such as, you know, a large earthquake on the Raymond Fault. 00:30:00.000 --> 00:30:03.000 Large earthquake on the San Andreas large on the Haywood Fall. 00:30:03.000 --> 00:30:07.000 Um, and the reason why we're doing this, essentially, is to. 00:30:07.000 --> 00:30:14.000 be able to have the… characterized demand. Of what we're going to be experiencing so that we can prepare and build our building safely. 00:30:14.000 --> 00:30:17.000 Um, so that we can avoid a situation where. 00:30:17.000 --> 00:30:26.000 Um, this happens, like what happened in Turkey. But before this is done, we need to validate these things against observed motions. 00:30:26.000 --> 00:30:28.000 to make sure that they essentially show the same. 00:30:28.000 --> 00:30:36.000 are so reliable, um, behaviors. That, uh, that we can use to ensure that they apply to the second hazard applications. 00:30:36.000 --> 00:30:40.000 So what do we do? We essentially decide to. 00:30:40.000 --> 00:30:45.000 take the observations that we have, the modern magnetic observations where we have a lot of observations. 00:30:45.000 --> 00:30:49.000 and simulated some observations, uh, simulated these, um, earthquakes. 00:30:49.000 --> 00:30:54.000 As soon as earthquakes over. Uh, over this, over the Southern California area. 00:30:54.000 --> 00:31:01.000 Then what we do is we do a simulation for 3D, for the full 3D physics-based simulation, and then we benchmark to 1D simulations. 00:31:01.000 --> 00:31:09.000 And then we conduct residual analysis. to essentially see what's going on with the bias. And what we found is that the three simulations do provide. 00:31:09.000 --> 00:31:11.000 A minimum buy… the lowest amount of buys, but there's. 00:31:11.000 --> 00:31:14.000 There are biases that are associated with the source. 00:31:14.000 --> 00:31:19.000 and the site that are not well accounted for due to limitations to velocity model, uh, due to limitations of how we essentially. 00:31:19.000 --> 00:31:25.000 identify paths or queue, and due to, like, resolutions of, um. 00:31:25.000 --> 00:31:37.000 Due to, um… simulation protocols, um, such as going to higher frequencies, and distribution of higher frequency content, so… Uh, in the next steps, what we're doing is that we're going to essentially apply this analysis that we're doing. We're currently doing that. 00:31:37.000 --> 00:31:41.000 I'm going to apply it to, uh, we've done it in Southern California, we just submitted our paper. 00:31:41.000 --> 00:31:46.000 We're going to apply this, um, to Northern California, and we're working on. 00:31:46.000 --> 00:31:49.000 From an engineering perspective. Um, validation in the Pacific Northwest. 00:31:49.000 --> 00:31:56.000 Um, to capture different things, and also trying to capture duration typography, because we… what I just talked about is focused on capturing. 00:31:56.000 --> 00:32:10.000 Um, uh, amplitude. Of Earthquake. That's pretty much it. Thank you so much. 00:32:10.000 --> 00:32:17.000 Okay, I want to introduce Seismicat, which is an R package for Seismicity Catalog analysis that I recently released. 00:32:17.000 --> 00:32:21.000 Um, Seismic has written in R, because it's free and open source, so anyone can use it. 00:32:21.000 --> 00:32:28.000 Um, RMAX documentation easy, so all the functions in sample datasets are fully documented and have examples that run in your browser. 00:32:28.000 --> 00:32:36.000 There's also an article as a vignette, it's more like a cookbook that shows you how to apply a seismic hat to a data set from start to end. 00:32:36.000 --> 00:32:40.000 Importantly, it includes the methods used to compute the earthquake rates in the NSHM. 00:32:40.000 --> 00:32:45.000 And so, by releasing this package and also a paper that Andrea Lenos and I recently released, 00:32:45.000 --> 00:32:48.000 Um, these methods are fully documented. 00:32:48.000 --> 00:32:51.000 And I will put, um, links to 00:32:51.000 --> 00:32:57.000 Um… those… the paper and the package in the chat after the talk. 00:32:57.000 --> 00:33:03.000 First thing you can do with Seismic Cat is ingest and manipulate catalogs. You can synthesize data and make exploratory plots. 00:33:03.000 --> 00:33:10.000 such as this, uh, map and magnitude time plot of earthquakes in Northern California last year. 00:33:10.000 --> 00:33:18.000 Once you've ingested data, you might want to compute Gutenberg-Richter B values, and we can do that in Seismicat with both traditional and B-positive methods. 00:33:18.000 --> 00:33:26.000 On the left, you saw magnitude frequency diagram where magnitudes are shown as the triangles, and magnitude differences are shown as the 00:33:26.000 --> 00:33:30.000 black rind, and we can do fits to those, and also do a stability. 00:33:30.000 --> 00:33:33.000 Versus magnitude of completeness on the right. 00:33:33.000 --> 00:33:43.000 Um, for the National Seismic Hazard Model, we compute B-value and rate joint distributions, so B-value on the x-axis, rate on the y-axis. 00:33:43.000 --> 00:33:52.000 We can do that for different epochs of a synthetic dataset in this case, and then do weighted averages or weighted mixture models to get a single distribution. 00:33:52.000 --> 00:33:54.000 See if that fits the data, we… 00:33:54.000 --> 00:34:01.000 um, project that distribution into the space of number of earthquakes on the y-axis versus magnitude. 00:34:01.000 --> 00:34:08.000 With Poisson variability, and that's shown as the red and black dotted lines are the quantiles. You can see the data 00:34:08.000 --> 00:34:11.000 Uh, fits pretty well here. 00:34:11.000 --> 00:34:18.000 And finally, for the National Seismic Hazard Model, we also want to compute logic tree branches that actually get used 00:34:18.000 --> 00:34:23.000 to compute the hazard, and we can do that with Seismicat in a few different ways. 00:34:23.000 --> 00:34:29.000 Um, those are shown as different color lines here of the mean, and then the quantiles of 00:34:29.000 --> 00:34:37.000 riding through branches. For this dataset, they all do about the same thing, but other datasets, there are choices between them. 00:34:37.000 --> 00:34:43.000 So, you know, a lot of people might be thinking, well, I don't use R, so I want to emphasize that it does include a step-by-step 00:34:43.000 --> 00:34:50.000 guide that includes some amount of advice on how to use R, some of the motivation, a lot of links to references. 00:34:50.000 --> 00:34:54.000 And everything from how to install it, to then step-by-step going through the process. 00:34:54.000 --> 00:35:00.000 of analyzing a dataset with examples, and then you can just take those example lines of code, 00:35:00.000 --> 00:35:12.000 And put them in your own code to modify. Thanks. 00:35:12.000 --> 00:35:13.000 Mark, go ahead. 00:35:13.000 --> 00:35:20.000 Morning, everyone. Uh, this is extremely recent work. Feel free to take it with a big grain of salt. 00:35:20.000 --> 00:35:26.000 Thanks to the organizers for allowing me to update my slides on Monday to include the most recent 00:35:26.000 --> 00:35:32.000 earthquakes. So I think many of you are probably aware with the history of earthquakes swarms in the San Ramon area. The plot on the 00:35:32.000 --> 00:35:36.000 Right? It's showing that, including the 2015 00:35:36.000 --> 00:35:41.000 um, sequence and the 2025, um, sequence now into 2026. 00:35:41.000 --> 00:35:46.000 toward the south. 00:35:46.000 --> 00:36:00.000 So, uh, I won't talk very much about the technique, because we don't have much time, but basically, I'm leveraging the routinely recorded earthquakes as waveform templates, detecting smaller events that are missed by the catalog, and then relocating 00:36:00.000 --> 00:36:03.000 everything with HypuDD. 00:36:03.000 --> 00:36:09.000 So here are the very preliminary results. We use almost 700 00:36:09.000 --> 00:36:13.000 uh… routinely cataloged earthquakes. Uh, we relocate 00:36:13.000 --> 00:36:15.000 a little bit more than 2200. 00:36:15.000 --> 00:36:19.000 earthquakes, uh, they should say November 9th through February 7th. 00:36:19.000 --> 00:36:25.000 On the right is the map view and the east-west and the north-south cross-sections. 00:36:25.000 --> 00:36:32.000 Uh, like we see in many earthquakes swarms, it's a very complicated fault structure that we see in this swarm. 00:36:32.000 --> 00:36:37.000 The deeper seismicity is… seems to be dominated by these more or less east-west striking structures. 00:36:37.000 --> 00:36:40.000 The shower seismicity seems to… 00:36:40.000 --> 00:36:45.000 be dominated more by, uh, north-south striking structures. 00:36:45.000 --> 00:36:51.000 We see migration, uh, generally from, uh, north to south, so the black activity 00:36:51.000 --> 00:36:56.000 initiates the swarm, then we see this shell or purple seismicity, 00:36:56.000 --> 00:37:00.000 deeper red seismicity, and finally in this most recent stage, we see 00:37:00.000 --> 00:37:03.000 um, shower seismicity that then extends 00:37:03.000 --> 00:37:06.000 North and South again. So we see these complex migration patterns. 00:37:06.000 --> 00:37:12.000 Uh, from deeper to shallower to deeper, and back to shower again. 00:37:12.000 --> 00:37:16.000 Okay, so we could take a little bit, um, a look at 00:37:16.000 --> 00:37:26.000 a direction that aligns a little bit better with the seismicity. Here's looking 13 degrees. You can see this purple, shallower seismicity, um, aligns pretty well with that orientation. 00:37:26.000 --> 00:37:29.000 We can look, um, from the west, and then… 00:37:29.000 --> 00:37:35.000 73 degree orientation lines up for these, you know, close to east-west striking faults, the black. 00:37:35.000 --> 00:37:39.000 And the red faults on the deeper end there. 00:37:39.000 --> 00:37:47.000 Um, I pulled just a couple of, uh, focal mechanisms. 00:37:47.000 --> 00:37:53.000 So here's, uh, the start of the swarm, we have east-west faulting, 00:37:53.000 --> 00:37:55.000 Um, that early December part we have 00:37:55.000 --> 00:37:58.000 Close to north-south Falding, and then again… 00:37:58.000 --> 00:38:04.000 Uh, toward the end, we have close to east-west, or toward the… in late December, we have the east-west falling. 00:38:04.000 --> 00:38:07.000 I haven't looked more carefully at the most recent stage. 00:38:07.000 --> 00:38:15.000 Um, so we likely see this kind of conjugate fault structure, um, at the very end here, I threw in, optimistically threw in an animation that 00:38:15.000 --> 00:38:18.000 Probably won't play very well here on Zoom, but… 00:38:18.000 --> 00:38:21.000 I'll let this play until my time is up. 00:38:21.000 --> 00:38:32.000 Um, you'll see on February 2nd, in just a few moments here, there's a relatively large of magnitude 4.2 earthquake, which is the largest of the sequence, and you'll see this 00:38:32.000 --> 00:38:35.000 light up here in just a second. 00:38:35.000 --> 00:38:41.000 It's very complicated, lots of little faults, um, so there's just a heck of a lot going on there. 00:38:41.000 --> 00:38:58.000 Thanks very much. 00:38:58.000 --> 00:39:08.000 Alright, well, thanks today for the great introduction to the 2025 SWARM. Um, I'm going to be talking about a different aspect of the San Ramon swarms, and that's how we can use them to improve our earthquake forecast models. 00:39:08.000 --> 00:39:29.000 Um, so here, I'm just showing a map, uh, magnitude time clock, cumulative number of events over time, showing sort of the eight main San Ramon swarms, including the most recent one, um, since 1970, color-coded, and then on the right is a zoom-in of the most recent swarm. Unlike Dave, I did not, uh, bother updating my slides, so this only shows I've missed you through, uh, January 20th. 00:39:29.000 --> 00:39:34.000 So, unlike aftershock sequences, uh, swarms are driven by some external process, like. 00:39:34.000 --> 00:39:39.000 fluid flow, for example, and it can vary over time, which makes forecasting them much more challenging. 00:39:39.000 --> 00:39:57.000 A commonly used aftershock forecast models the epidemic type aftershock sequence, or ETOS model. Um, I don't know what happened to my slides, I didn't touch anything, somebody please help me. 00:39:57.000 --> 00:39:58.000 Hold on, I'll fix it, hold on. 00:39:58.000 --> 00:40:11.000 Um… Um… Um, I mean, I guess I could… The guys' rooms? 00:40:11.000 --> 00:40:18.000 Okay. Um… 00:40:18.000 --> 00:40:23.000 I don't seem to have control anymore. 00:40:23.000 --> 00:40:29.000 Um, okay, next slide, please. 00:40:29.000 --> 00:40:35.000 Um, okay, so, uh, unlike aftershocks, yeah, so where was I? 00:40:35.000 --> 00:40:40.000 So, yeah, so the ETOS models is a commonly used aftershock forecast model. 00:40:40.000 --> 00:40:50.000 Um, and in this framework, swarms can be represented by a time-varying background made new. So here on the left, I'm showing a cumulative number of events over time during the swarm, observed is in black. 00:40:50.000 --> 00:41:03.000 And if we do the standard practice of fitting an ETOS model to all the prior seismicity in the region that occurred, and then using that model to forecast through the swarm, that's the green line shown here. And you can see that it really underestimates the number of events. 00:41:03.000 --> 00:41:07.000 Um, and this comes up at the background rate of .02. 00:41:07.000 --> 00:41:13.000 I don't know if you try to just fit the… fit just the swarm with ETAS, um. 00:41:13.000 --> 00:41:16.000 the background rate jumps up by a vector 4 to something like .09. 00:41:16.000 --> 00:41:23.000 But the fact that we've had previous swarms in San Ramon, um, allows us to try and create a swarm model by fitting ETOS to, um. 00:41:23.000 --> 00:41:32.000 just the prior swarms, and you can see that the brown line here, uh, does a pretty decent job at fitting the swarm as well, with a background rate increase of a factor of 10. 00:41:32.000 --> 00:41:39.000 But in all of these models, you can generally see that the second half of the swarm isn't really fit as well as the first half. Next slide, please. 00:41:39.000 --> 00:41:47.000 Um, so this suggests that something in the model changes in the middle of the swarm, and we can see this if we try on the left fitting ETOX to just the first month of the swarm. 00:41:47.000 --> 00:41:53.000 Um, and then extrapolating that model through the second half, and you can see that it very much underestimates the rate. 00:41:53.000 --> 00:42:02.000 And we can also try to update the background rate periodically to the swerve, and so on the figure on the right, I'm showing the background rate estimates that are updated every 3 days using, um, different. 00:42:02.000 --> 00:42:08.000 lengths of look-back windows to estimate the rate, ranging from 2 to 10 days. And so you can see that. 00:42:08.000 --> 00:42:15.000 the second half of the swarm really seems to require a much larger background rate. Next slide, please. 00:42:15.000 --> 00:42:21.000 So incorporating these higher background rates can really improve our earthquake forecasts during the storm, which we've also seen, um. 00:42:21.000 --> 00:42:27.000 Four swarms in the Salton Trough in Southern California. So on the left, I'm showing 3-day forecasts of the number of earthquakes throughout the swarm. 00:42:27.000 --> 00:42:32.000 But he observed, um, in the white squares, and each of the different models that I've described are shown with different. 00:42:32.000 --> 00:42:37.000 marker. So green is sort of the low mu model to fit all prior seismicity. 00:42:37.000 --> 00:42:44.000 Greatest High View model, fit only to prior swarms, and then we have these three updating, um, new models with the different look-back windows. 00:42:44.000 --> 00:42:50.000 And so the likelihood scores are shown here in the legend, and also in the plot over time on the right, and. 00:42:50.000 --> 00:43:00.000 They, um, so then they show that the best forecasts, the best performing forecast really require these higher backward rates. And interestingly, and contrary to what Andy Michael and I found in our paper. 00:43:00.000 --> 00:43:05.000 on the 2015 swarm, the fixed swarm model seems to outperform all of the, um. 00:43:05.000 --> 00:43:20.000 updating, um, background rate models for reasons that we're still trying to understand, but… The main takeaway message here is that the more that these swarms in San Ramon, these regional swarms happen, the better we can model and understand their behavior, and the better informed their forecasts can be in the future. 00:43:20.000 --> 00:43:28.000 So, thank you. 00:43:28.000 --> 00:43:36.000 Um, good morning, everyone. It's my pleasure to be here. My name is Jay Tong, I'm a computational geophysicist from Texas Tech, so, um… 00:43:36.000 --> 00:43:42.000 Today, I would like to share a brief talk about how, if we can relate, um, the nearby sustainability 00:43:42.000 --> 00:43:47.000 And the essential for Crete to the nearby, um, geothermal energy production at Geyser. 00:43:47.000 --> 00:43:52.000 So, in the top left, um, subfigure shows the seismicity surrounding Kaiser. 00:43:52.000 --> 00:44:02.000 And they are either over the red dots, like the guys are near field, versus over the farther away for the geyser, such as Rogers Creek Fort. 00:44:02.000 --> 00:44:07.000 My common thoughts on boilers, springfort, and Green Valley Faults. 00:44:07.000 --> 00:44:11.000 And so, the top central figure here, and showed that, um, 00:44:11.000 --> 00:44:16.000 The magenta line showing the increase in production volume since 1960s, 00:44:16.000 --> 00:44:24.000 And billions of barrels of water, um, steam have been removed from the underground. And I also… 00:44:24.000 --> 00:44:31.000 calculates the size of MISD rates, um, since the 1967, based on the Northern California Ag Lots. 00:44:31.000 --> 00:44:39.000 And, um, particularly, we see the red lines, uh, the red line showing the geyser-seism rate since 1967. 00:44:39.000 --> 00:44:48.000 And you start to take off after, um, the… the launch of the production activity. However, for those four field cycles, which shows in black, 00:44:48.000 --> 00:44:54.000 lines, um, they kind of tapered since the 1969th of Santa Rosa earthquakes. 00:44:54.000 --> 00:45:02.000 So, we just try to understand if, or figure it out, or explore if that's a possibility that the stress change caused by the geothermal energy productions 00:45:02.000 --> 00:45:05.000 my interference source activity. So, in the bottom… 00:45:05.000 --> 00:45:10.000 That figure shows, um, the final element model we built, try to, um, similarly… 00:45:10.000 --> 00:45:19.000 The decay those stress change in port pressure change, um, that, uh, that initiate by the guys that full exchange, 00:45:19.000 --> 00:45:25.000 Um, with the undergrounds, okay? So, we would try to incorporate other, um, 00:45:25.000 --> 00:45:27.000 potential, uh… 00:45:27.000 --> 00:45:30.000 factors, for example, the 3D printing structures, 00:45:30.000 --> 00:45:37.000 the cap roll of the reserve permability, and however the well depth is proprietary, so it's really hard to figure it out, but we will try to 00:45:37.000 --> 00:45:41.000 find data. Um, but the right-hand side shows the preliminary data, 00:45:41.000 --> 00:45:48.000 that, um, the Coulomb stress, um, projectors on a northwest… northwest trending fort. 00:45:48.000 --> 00:45:55.000 Um, in the regions. And you can tell that more than 80 kilo per scale of corn stress being accumulated every year. 00:45:55.000 --> 00:46:02.000 And so, um, but they are in different, um, dipoles of the current stress map. 00:46:02.000 --> 00:46:15.000 Right now, so it means that segment of nearby fort will either stabilize or destabilize by the stretchings, by the geyser. But one point I really want to highlight is, like, given that we accumulate 80 kilo per scale of 00:46:15.000 --> 00:46:22.000 stress every year, either negative or positive. Over 60 years, it's gonna be up to a mega per scale. 00:46:22.000 --> 00:46:26.000 of stress, which is 100 times after the triggering threshold, anyway. 00:46:26.000 --> 00:46:35.000 So, we will try to see if there is a possibility that those Falk Creek and Sessimistic could be related to the stress change. 00:46:35.000 --> 00:46:50.000 And we will try to improve the model as well. So, that's what I have today. Thanks. 00:46:50.000 --> 00:47:01.000 Good morning. Thanks for letting me talk here about some interesting observations that me and my colleagues have made at the Geyser's Geothermal Field following the 2014. 00:47:01.000 --> 00:47:12.000 magnitude 6 out of Napa earthquake. Um, the audience knows the location of both the geysers and South Napa, so we'll dive right in. In the first 12 hours. 00:47:12.000 --> 00:47:21.000 Following the passage of surface waves. Uh, from the 2014 South Napa earthquake seismicity lit up throughout the reservoir and beyond, shown here by the red stars. 00:47:21.000 --> 00:47:26.000 The blue line is generally, uh, the outline of the reservoir itself. 00:47:26.000 --> 00:47:33.000 The grey underlying dots are showing 6 years of seismicity from 2010 to 2016. There was a triggered. 00:47:33.000 --> 00:47:41.000 magnitudes M38 in the southern part of the reservoir did not seem to have an aftershock sequence. 00:47:41.000 --> 00:47:45.000 Uh, the analysts and operators noticed this sparkling of seismicity. 00:47:45.000 --> 00:47:52.000 or effervescent seismicity, as we call it. throughout the reservoir, which prompted our study to understand whether, one. 00:47:52.000 --> 00:47:59.000 This is, in fact, abnormal behavior for the Geyser's geothermal field, and two, what the likely causes are. 00:47:59.000 --> 00:48:09.000 To determine whether the observations of effervescent, um. seismicity is, in fact, abnormal behavior for the guys as we board spatial statistics from. 00:48:09.000 --> 00:48:17.000 Astrophysics, namely the Ripley's K and L functions. The Ripley's functions test for spatial homogeneity of each event relative to all others. 00:48:17.000 --> 00:48:22.000 And in the top plot, we show the distribution of the Ripley function for each. 00:48:22.000 --> 00:48:31.000 12-hour segment in that 7 years of data in the catalog. A flat line near the surface would indicate complete spatial randomness. 00:48:31.000 --> 00:48:38.000 And, um… the curves that dip lower from horizontal indicate increasing. 00:48:38.000 --> 00:48:45.000 Quite heterogeneous patterns of seismicity. The dark blue line at the top indicates the 12 hours following South Napa. 00:48:45.000 --> 00:48:52.000 clearly an outlier, and so the first question is answered that this is, in fact, abnormal behavior. 00:48:52.000 --> 00:49:01.000 For, um, uh, the geysers. I won't talk much about the plots in the second and third row, but there are enough people that are currently in review. 00:49:01.000 --> 00:49:09.000 So then, finally, we're trying to understand what the possible mechanisms for the spatial homogeneity or effervescent seismicity. 00:49:09.000 --> 00:49:23.000 might be, and so we, uh, on the top plot, we show the actual metric, uh, singular metric of, um, spatial homogeneity. It's largest for the 12 hours following the South Napa event. All the other dots are. 00:49:23.000 --> 00:49:30.000 moderate and large regional events. The second part is be positive value, it doesn't seem to have any correlation with. 00:49:30.000 --> 00:49:39.000 possibly elevated pull pressure. effects in the reservoir, tidal strains in the third, um. 00:49:39.000 --> 00:49:44.000 Ro also don't seem to be elevated, neither volumetric nor deviatoric. 00:49:44.000 --> 00:49:49.000 However, peak strain measured during the passage of the surface wave is significantly elevated. 00:49:49.000 --> 00:49:56.000 And so we suggest that peak strain is the likely cause for this diffuse seismicity in the geothermal field. 00:49:56.000 --> 00:50:06.000 And the implications of this. Well, the take-home message is that significant strain was exerted on it, the strain was large enough to trigger seismicity. 00:50:06.000 --> 00:50:14.000 in areas that are not hugely affected by injection and production, and therefore might indicate a strain threshold needed to trigger a large portion of faults and fractures in the area. 00:50:14.000 --> 00:50:19.000 Thank you. 00:50:19.000 --> 00:50:26.000 Right, good morning, everyone. Um, so my name's Rob Scoble. I'd like to talk about how we can improve focal mechanisms with DAS. 00:50:26.000 --> 00:50:32.000 Specifically, that we can include SP amplitude ratios. into our solutions. 00:50:32.000 --> 00:50:37.000 Um, so the data that I'll be talking about today come from the GORDAS experiment, which is located in Northern California. 00:50:37.000 --> 00:50:42.000 Um, so we're considering a couple hundred earthquakes within 100 kilometers of the STES. 00:50:42.000 --> 00:50:49.000 Array. On the right here is a zoom-in of that cable. Runs from Arcata down to the town of Eureka. 00:50:49.000 --> 00:50:56.000 about 14 kilometers long. And, uh, co-located with that DAS array are nodal instruments, represented by those. 00:50:56.000 --> 00:51:01.000 white triangles. So the idea behind this experiment is that we know how to compute SPRAMP. 00:51:01.000 --> 00:51:06.000 SP, amplitude ratios on the nodal instruments. So let's make those measurements, and then compare those. 00:51:06.000 --> 00:51:14.000 with the amplitude measurements we get from the adjacent DAS channel. 00:51:14.000 --> 00:51:24.000 So these are the raw results from these instruments. On the x-axis are the SP ratios measured on the nodal instruments. 00:51:24.000 --> 00:51:28.000 along the y-axis are the ratios measured from that adjacent DAS. 00:51:28.000 --> 00:51:33.000 Uh, channel. Um, so if you're not familiar with ISP ratios, this is actually a really good correlation. 00:51:33.000 --> 00:51:37.000 you kind of had this inherent scatter of a factor of 2 to 3. 00:51:37.000 --> 00:51:44.000 Um, so this is very promising results there. I should also mention that these are the raw results, so it's customary to do some kind of station correction. 00:51:44.000 --> 00:51:48.000 But we think there's value, even if you don't do the station correction. 00:51:48.000 --> 00:51:57.000 Um, for these measurements. Another test you can do is compare your ratios against what you would theoretically expect to get if your focal sphere is evenly sampled. 00:51:57.000 --> 00:52:01.000 So this kind of orange line is that theoretical distribution. 00:52:01.000 --> 00:52:06.000 And on top of that, I'll plot our ratios from the nodal and the DAS instruments. 00:52:06.000 --> 00:52:11.000 Um, so once again, we see a pretty good correlation between the nodal and DAS instruments. 00:52:11.000 --> 00:52:15.000 But we also see that it matches this theoretical distribution pretty well. 00:52:15.000 --> 00:52:18.000 The one exception is that we have this kind of deficiency. 00:52:18.000 --> 00:52:25.000 And these highest SP amplitude ratios. Um, if you want to know why that is, you have to attend the workshop next year. 00:52:25.000 --> 00:52:30.000 For the purposes of this talk, we know it's not a limitation of DAS, because both the inertial seismometers. 00:52:30.000 --> 00:52:42.000 And the DAS, um, show the same patterns. Finally, in the time I have left, I'd like to talk about one example that we have. 00:52:42.000 --> 00:52:47.000 If it ever changes… 00:52:47.000 --> 00:52:53.000 There we go. Okay, um, so this study is also published in GRL, so if you'd like to take a closer look, um, you're more than welcome to do that. 00:52:53.000 --> 00:52:59.000 But this earthquake that I'll be showing was about 20 kilometers northwest of the Dass Aray. 00:52:59.000 --> 00:53:05.000 I'm making the measurement is pretty simple. So these are the waveforms. You just take the peak amplitude and the S. 00:53:05.000 --> 00:53:11.000 little window, and the P window, divide those to get the SP ratio. 00:53:11.000 --> 00:53:16.000 these mechanisms in the top right, the blue ones are the best matching solutions. 00:53:16.000 --> 00:53:19.000 computed just using the SP ratios from the DAS cable. 00:53:19.000 --> 00:53:22.000 And we're comparing those solutions against the moment tensor in green. 00:53:22.000 --> 00:53:27.000 Um, so, you know, there's… it's kind of demonstrating that there's promise with these ratios. 00:53:27.000 --> 00:53:31.000 You should not just use the desk cable alone, right? This is a very poorly constrained solution. 00:53:31.000 --> 00:53:35.000 What you're supposed to do is consider, you know, your entire network together. 00:53:35.000 --> 00:53:37.000 Which is what these kind of purple solutions down there are. 00:53:37.000 --> 00:53:42.000 And when you consider these NAS cables, it does help improve the mechanism quite a bit. 00:53:42.000 --> 00:53:52.000 Right? So, uh, thanks a lot. 00:53:52.000 --> 00:54:02.000 Hello, my name is Teresa Sawy, I'm a Mendenhall postdoc at the USGS. Today, I'm excited to talk about our research, Locating Earthquakes on short DAS cable subsegments, 00:54:02.000 --> 00:54:12.000 for earthquake early warning, and I'd like to thank my collaborators. 00:54:12.000 --> 00:54:15.000 Um… 00:54:15.000 --> 00:54:17.000 Next slide, please. 00:54:17.000 --> 00:54:25.000 Perfect. Um, one of the outstanding challenges in Earthquake early warning systems is locating earthquakes that are outside of the network footprint. 00:54:25.000 --> 00:54:30.000 Beamforming with seismic arrays is a well-established method for approaching this problem. 00:54:30.000 --> 00:54:35.000 And work has been done on beamforming on DAS for earthquake early warning. 00:54:35.000 --> 00:54:43.000 In this work, we're going to be beamforming on multiple earthquakes on multiple DAS cables in both rural and urban settings, 00:54:43.000 --> 00:54:46.000 Using just 1.5 seconds of P wave data. 00:54:46.000 --> 00:54:53.000 And our research question is, is beamforming on small DAS arrays a viable method for constraining back azimuth 00:54:53.000 --> 00:55:00.000 And apparent slowness. Next slide, please. 00:55:00.000 --> 00:55:07.000 So our method is this. We take a, um, DAS cable, shown in this black zigzag line here, 00:55:07.000 --> 00:55:19.000 And we're going to take short sub-segments of this cable, they're about 1-5 kilometers long. Each one of these segments is an array, so when an earthquake occurs, each one of these arrays is going to record a different wave field. 00:55:19.000 --> 00:55:23.000 We're gonna perform beamforming on these, uh, rays. 00:55:23.000 --> 00:55:26.000 And each one is going to result in these polar plots here. 00:55:26.000 --> 00:55:33.000 Where the hot colors show where the highest normalized beam power is, the radial tick marks are showing apparent velocity. 00:55:33.000 --> 00:55:37.000 And the back azimuth is going around the circle, with North at the top. 00:55:37.000 --> 00:55:46.000 The white line shows the catalog back azimuth, and the green dashed line is what we calculate. The star is the catalog apparent velocity. 00:55:46.000 --> 00:55:51.000 Let me see if I can do the next slide. 00:55:51.000 --> 00:56:07.000 Cool. Um, so our data come from 3 DAS cables. One is the Mendocino Triple Junction, that's about 15 kilometers long. One is from San Francisco, and one is in Berkeley, spanning into Oakland. 00:56:07.000 --> 00:56:16.000 First, we'll start with the San Francisco cable. This was a magnitude 3.5. We're using 4 subarrays. They're each about 1-2 kilometers long. 00:56:16.000 --> 00:56:21.000 Um, shown by these four waveform arrays here. These are just selections of waveforms. 00:56:21.000 --> 00:56:26.000 We have Form B forming results, which we add together to get our final beamforming result. 00:56:26.000 --> 00:56:35.000 Which is about 6 degrees from the catalog back azimuth, and we see it has a reasonable apparent velocity. 00:56:35.000 --> 00:56:42.000 In Oakland, we're looking at the magnitude 4.3 back in September. We used four, uh, sub-segments of the cable. 00:56:42.000 --> 00:56:50.000 And again, they're calculated back azimuth is about 6 degrees from the back catalog back azimuth. 00:56:50.000 --> 00:57:01.000 Finally, we look at the magnitude 7 from December 2024. We use 4 dash subarrays. We do beamforming on all of them, combine them, and get one final. 00:57:01.000 --> 00:57:08.000 beam-forming result, where the calculated back azimuth is 2 degrees from the catalog back azimuth, and we get a reasonable apparent velocity. 00:57:08.000 --> 00:57:16.000 So that brings us back to our research question, and we find that, indeed, beamforming on small DAS arrays is a viable method for constraining back azimuth. 00:57:16.000 --> 00:57:21.000 and apparent slowness using just 1.5 seconds of paint wave data. 00:57:21.000 --> 00:57:31.000 Uh, thank you so much. 00:57:31.000 --> 00:57:38.000 Okay, that brings us to the end of our Thundertalk session. Thanks again to all of the great speakers. 00:57:38.000 --> 00:57:45.000 Um, we have time for one or two burning questions, if anyone has them. 00:57:45.000 --> 00:57:52.000 Uh, and if you do, please raise your hand. You can ask a question to any of the 16 speakers. 00:57:52.000 --> 00:58:00.000 From that session. 00:58:00.000 --> 00:58:02.000 And if no one has… 00:58:02.000 --> 00:58:06.000 a burning question, then we can… 00:58:06.000 --> 00:58:10.000 go into break, but I'm gonna give folks a few more… 00:58:10.000 --> 00:58:14.000 Seconds to raise their hand. 00:58:14.000 --> 00:58:24.000 I'm sure any of our speakers would love to answer a question? 00:58:24.000 --> 00:58:28.000 Okay, then I think I'll turn it back to Sarah and John. 00:58:28.000 --> 00:58:34.000 Oakley-dokely, then we are on break until 1045 Pacific Time, unless… 00:58:34.000 --> 00:58:40.000 You are a speaker in the next exciting news and events update session, so that means Sarah McBride, 00:58:40.000 --> 00:58:48.000 Um, I sent Rob and Harold and Ahmed, and Albert, and Chris, and Jill, stick around and check your, uh… 00:58:48.000 --> 00:58:51.000 Um, microphone and camera. Otherwise, everyone else? 00:58:51.000 --> 00:59:05.000 escape and be back at 1045, please.