WEBVTT 00:00:00.000 --> 00:00:25.000 Greetings! Are we ready for the final show, where we bring everything together in the grand unified theory of national seismic hazard model and the grand Noha Farghal and Grace Parker. 00:00:25.000 --> 00:00:31.000 Thanks, Sarah, and thanks everyone, and welcome to the last session of the workshop. 00:00:31.000 --> 00:00:38.000 The future is soon Northern California updates to the National Seismic Hzzard Model and beyond the UCERF. We're gonna have five talks in the session. 00:00:38.000 --> 00:00:56.000 The first 4 will be focused on the earthquake source models and the last for Mark Peterson will be focused on updates to ground motion modeling and changes in the hazard, all with a special focus on northern California. 00:00:56.000 --> 00:00:57.000 As usual feel free to ask your questions in the chat. 00:00:57.000 --> 00:01:03.000 But keep in mind that we will have 30 min for discussion. 00:01:03.000 --> 00:01:23.000 At the end of the talks. So if you have any particularly detailed or juicy questions, feel free to save them, and ask them in person after the talks and same for the speakers, if you see a question, you would prefer to answer after in-person feel free to hold off on in the 00:01:23.000 --> 00:01:25.000 chat. And so now I'd like to introduce my co- 00:01:25.000 --> 00:01:32.000 moderator, Noha Farghl, who's a seismic hazard modeler at RMS 00:01:32.000 --> 00:01:35.000 to introduce the first speaker. 00:01:35.000 --> 00:01:36.000 Thank you, Grace. Hello, everyone, and thank you for joining our session. 00:01:36.000 --> 00:01:44.000 I would like to welcome our first speaker, Alex Hatem from the USGS, who will speak on the northern California earthquake 00:01:44.000 --> 00:01:49.000 geology input data for use in the 2023 US 00:01:49.000 --> 00:01:50.000 National Seismic Hazard Model. Alex. 00:01:55.000 --> 00:01:56.000 Hi Everyone! My name is Alex Hatem. I'm a research geologist with the USGS 00:01:56.000 --> 00:02:01.000 sitting in golden Colorado. I'm gonna talk to you. 00:02:01.000 --> 00:02:09.000 about the earthquake geology, input data that we're going to use in the 2,023 update for the National Seismic Hazard Model. 00:02:09.000 --> 00:02:14.000 Before we begin, we need to acknowledge all of the collaborators and colleagues that have made this work possible. 00:02:14.000 --> 00:02:27.000 Starting with a Strong Group of us in the IMW Earthquake Geology Group back in Golden with Camille, Rich and Ryan, as well as input from Pacific Northwest and Steve Angster, and of course, guidance from Ned Field and the ERA 00:02:27.000 --> 00:02:29.000 Team. We collaborated extensively with State Geological Surveys and other earthquake geology groups within the USGS. 00:02:29.000 --> 00:02:38.000 So a big thank you to all the people that were a part of this project. 00:02:38.000 --> 00:02:44.000 Okay. So a little bit of background in prior iterations of the National Seismic Hazard Models, faults were only included 00:02:44.000 --> 00:02:51.000 if they had a site specific slip rate. And that, of course, would exclude many known Quaternary active faults. In the update from UCERF2 to UCERF3 that was released in 2014, 00:02:51.000 --> 00:03:07.000 there was a lot of effort to improve the fault database done by Tim Dawson and Ray Weldon that included faults for which they had no estimates of geologic slip rates, so shown in the map, on the right 00:03:07.000 --> 00:03:10.000 we have all the fault sections included in UCERF3 00:03:10.000 --> 00:03:24.000 for California and the green circles represent locations of slip rate cities. So you can see that not every fault has one of those cities we're not going to discuss the deformation models which Fred will discuss in a subsequent talk for how to derive slip rates on faults 00:03:24.000 --> 00:03:36.000 for which we don't have rates. The main point that I'm trying to get across here is that we're going to include faults that have variable degrees of rate study, but have good constraints on their geometry. 00:03:36.000 --> 00:03:46.000 The whole goal here is to bring the rest of the Western United States up to the standard that was set in California after UCERF3. 00:03:46.000 --> 00:03:49.000 Here's just a comparison on the left 00:03:49.000 --> 00:03:53.000 the known Quaternary faults for which we have about 2,000 across the Western US 00:03:53.000 --> 00:03:54.000 Vs. the faults that were considered in prior iterations of the NSHM, 00:03:54.000 --> 00:04:07.000 Which include about 640 faults across Western US. So spatially, we see that there is a disagreement between what we know to be true and what we include for hazard models. 00:04:07.000 --> 00:04:17.000 But again, there's a good reason why that was because a lot of the faults in the Q faults database don't have good constraints of rate, and here they are on the left piled on top of each other where Q 00:04:17.000 --> 00:04:21.000 faults are in the cream color, and the NSHM 00:04:21.000 --> 00:04:36.000 2018 faults are shown in blue. So overall there's relatively great agreement in California, the focus of this workshop, of course, northern California and that good agreement is thanks to the hard work that was done in UCERF3, but you look outside of California and you see that there is quite a bit 00:04:36.000 --> 00:04:40.000 of mismatch across the rest of the Western US. So again, this was our goal to to add detail and improve upon the already excellent fault 00:04:40.000 --> 00:04:50.000 sections database that was included in prior NSHMs. 00:04:50.000 --> 00:04:53.000 Here's a little bit about the database construction. 00:04:53.000 --> 00:04:56.000 So one improvement that we made over prior models is that we decoupled the fault 00:04:56.000 --> 00:05:06.000 section geometries from the parameters that govern their activity, so that again allowed us to keep the geometries and the rate separate. 00:05:06.000 --> 00:05:17.000 So there's a fault sections database here idealized in orange, where we have the fault trace as well as some basic 3D information like lower seismogenic depth, and the dip degree dip direction. 00:05:17.000 --> 00:05:26.000 We also include rate, and then for the rate information, we include a geologic slip rates in the database we call the Eq. 00:05:26.000 --> 00:05:40.000 Go dB earthquake challenge database. So currently it contains only geologic slip rates at points as well as categorical rates to help guide the deformation models where we don't have good estimates of the geologic slopes so for quick 00:05:40.000 --> 00:05:44.000 Consumption. It was used for geodetic deformation models in the long term. 00:05:44.000 --> 00:05:53.000 We hope it'll be used by the broader community to provide constraints on slippery rates in any region across Western us. 00:05:53.000 --> 00:05:56.000 Okay. So a little bit about the criteria of inclusion for prior models, versus what we're doing now. 00:05:56.000 --> 00:06:05.000 So in in the past, in 2,014, 2,018, and prior faults were only included with a known geologic slip rate. 00:06:05.000 --> 00:06:10.000 There really wasn't a hard restriction on quaternary activity, and there was no length requirement on faults in 2,023 we changed that a little bit. 00:06:10.000 --> 00:06:18.000 We mandated that there must be unequivocal. 00:06:18.000 --> 00:06:21.000 Or unequivocal evidence of quaternary deformation. 00:06:21.000 --> 00:06:37.000 That's tectonic. Faults were no longer excluded for lack of geologic, slippery information, and the fault length had to be greater than 7 kilometers in length, and it should also add that we grandfathered in all the data, that was included in prior models 00:06:37.000 --> 00:06:45.000 but from this point, going forward, we're only including data that's available and peer-reviewed, publicly available studies 00:06:45.000 --> 00:06:49.000 Okay, so how did we go from the queue faults? 00:06:49.000 --> 00:06:51.000 Database to a hazards fault database Q. 00:06:51.000 --> 00:07:05.000 Fs. Database has a lot more detail than what what is required or possible to use, in a hazard model. So here I'm just going to step through an example from the canyon ferry fault in Montana, where on this first column, we have the Qeue Fault 00:07:05.000 --> 00:07:11.000 Representation, where all the white dots were present, individual nodes or clicks. 00:07:11.000 --> 00:07:17.000 If you will, if you're digitizing the fault, and indeed those are individual points, but it actually looks like it's a proper line. 00:07:17.000 --> 00:07:19.000 There's so many nodes on these cue faults. 00:07:19.000 --> 00:07:24.000 They're highly detailed, and they come out of and and are used for a different purpose than for hazard modeling. 00:07:24.000 --> 00:07:26.000 So we need to adapt those faults to what we need for hazard models. 00:07:26.000 --> 00:07:47.000 So Camille Colette led this effort to simplify the Q faults, database and simplify them, and then provide sort of a baseline geometry that we, as geologists, could review by eye connect as needed and generalized to be a hazard appropriate fault. 00:07:47.000 --> 00:07:49.000 So here we're going from the Q fault geometries to the smooth geometries, and then to the eventual NSHM 00:07:49.000 --> 00:07:59.000 23 fault section geometry, where please zoom in 00:07:59.000 --> 00:08:03.000 we can see that even when we smooth it's still a little bit too complicated, 00:08:03.000 --> 00:08:08.000 so the final hazard version is extremely simplified, and we are trying to generalize what we think 00:08:08.000 --> 00:08:11.000 the seismogenic fault might look like at depth. 00:08:11.000 --> 00:08:12.000 We are not interested in curvy geometries, or really high resolution 00:08:12.000 --> 00:08:17.000 details. 00:08:17.000 --> 00:08:18.000 Okay, so to let the cat out of the bag here we have on the left in blue the NSHM 00:08:18.000 --> 00:08:37.000 1418 fault sections again, I said, they're about 640 faults in that database, and on the right we have the updated fault sections database and orange for 2023. So on the right, we have about a 1,000 faults. 00:08:37.000 --> 00:08:49.000 Here they are, overlaying with each other. So the blue again is the old and the orange is the new, so overall we went from again about 640 falls to over a 1,000, 00:08:49.000 --> 00:08:58.000 58% increase in California. We didn't see that many faults added, and that is due to the great prior work that was done in UCERF3. 00:08:58.000 --> 00:09:13.000 So again, you can visualize this update process that happened from UCERF2 to UCERF3 in the IMW, the Intermountain in the West and the Pacific Northwest, where we saw a large amounts of increase in the fault network. 00:09:13.000 --> 00:09:18.000 Just a note about the faults that we added. Again previous NSHM 00:09:18.000 --> 00:09:35.000 did a great job of including the main players, so to speak, and including the faults that are responsible for most of the hazard across the western US however, there were many faults that were there locally originally important, but maybe aren't very fast slipping. 00:09:35.000 --> 00:09:42.000 So the majority of the faults that we added are short slowly slipping normal faults across basically the Intermountain West 00:09:42.000 --> 00:09:51.000 and the Eastern Pacific Northwest, and here on the bottom, I'm just showing fault rate, and highlighting the fact that most of the faults in the NSHM 00:09:51.000 --> 00:10:08.000 in the fault sections data base are indeed normal faults written in this light blue color in comparison to our study of the this workshop mostly strike-slip environment with some oblique faults. 00:10:08.000 --> 00:10:12.000 Okay, so really quick about the EQGeoDB, the Earthquake Geology Database, 00:10:12.000 --> 00:10:15.000 again, we compile the slip rates at points for both the geologic point-based studies, as well as these categorical slip rates. 00:10:15.000 --> 00:10:37.000 Both of those were used varying degrees in the deformation models, and we hope to augment this database soon with some paleo seismic information, recurrence information as well as slip per event, and provide a fuller definition of what is earthquake geology right 00:10:37.000 --> 00:10:42.000 now it's just the slip rates, but this will be augmented through time. 00:10:42.000 --> 00:10:43.000 Okay. And so now I just want to zoom in for a little bit on n orthern California. 00:10:43.000 --> 00:10:56.000 This is the focus here. So on the left, we have the 2014, 2018 fault sections database in blue, and on the right we have the 2023 in orange so off the bat 00:10:56.000 --> 00:11:05.000 they actually look pretty similar, and they should because again, the fault database was in good shape 00:11:05.000 --> 00:11:11.000 from UCERF3. Okay, we'll just zoom in on a few examples here, and highlight 00:11:11.000 --> 00:11:17.000 some changes that did occur I'm in northern California from 2014 to 2023. 00:11:17.000 --> 00:11:18.000 So first of all, we're no longer including alternative fault models. 00:11:18.000 --> 00:11:27.000 Which was a logic tree branch in UCERF3 where there's fault. 00:11:27.000 --> 00:11:36.000 Model 1 and fault model 2. We are not implementing alternative fault models at this time, but it's certainly something we're considering for the next time around. 00:11:36.000 --> 00:11:44.000 And once such place, where there was alternative fault models, was right in this region, with the Contra Costa Shear zone as well as Mount Diablo. 00:11:44.000 --> 00:11:50.000 So you can see that there are many blue faults for which now there's only a handful of orange faults. 00:11:50.000 --> 00:11:57.000 And so we've simplified this area for Contra Costa, for instance, with just the Franklin and Southampton faults. 00:11:57.000 --> 00:12:05.000 Now there are two Mount Diablo faults that are in the model at the same time as opposed to all the alternative geometries from 2014. 00:12:05.000 --> 00:12:18.000 We also added the West Chase and Midway Faults, and implemented a connection of the Calaveras-Hayward system down here, based on published literature. 00:12:18.000 --> 00:12:24.000 Now moving further north in California to highlight some more examples. 00:12:24.000 --> 00:12:32.000 We made some revisions in the in the fold, and thrust belt sort of revising default geometries, as well as adding a few faults in here. 00:12:32.000 --> 00:12:49.000 We added some new faults in the Redding area, thanks to some work from publish work from Steve Angster, and we also added in the zones of distributed deformation shown here in the pink areas. We implemented these really idealized proxy faults shown here in 00:12:49.000 --> 00:12:56.000 orange that represent sort of an average of where we think this train might be localized, 00:12:56.000 --> 00:13:02.000 but we fully understand that the strain up in this part of the world is not localized, and it's very distributed. 00:13:02.000 --> 00:13:06.000 It's extremely challenging to put a single fault geometry on the map. 00:13:06.000 --> 00:13:10.000 So that's why there represented as proxy faults with these aerial sources, these are yet to be implemented, and we'll see how it goes. 00:13:10.000 --> 00:13:17.000 See how this goes. Currently, they're used as fault sections. 00:13:17.000 --> 00:13:18.000 And finally, just to touch on the rate information which you'll hear about next from Fred. 00:13:18.000 --> 00:13:26.000 The green circles are the the geologic slip rates. 00:13:26.000 --> 00:13:40.000 There are many, many observations of geologic slip rates, but of course there are many more faults, and there are observations, both the geologic and the geodetic deformation models are required to interpolate between these observations of geologic slip rates as well 00:13:40.000 --> 00:13:53.000 as GPS measurements for the geodetic models to determine the rates on faults without field studies, and then finally, in the blue stars, something I haven't really spoken too much about the paleo-seismic recurrent sites these are used as a constraint on rupture 00:13:53.000 --> 00:14:11.000 rates in the inversion approach. These were compiled and revised by Devon McPhillips during the compilation period, and these are not an ingredient in the deformation models, but are an ingredient in the inversion process itself. So that summarizes 00:14:11.000 --> 00:14:25.000 all of the earthquake geology update process as well as some examples from northern California. So I'm happy to take any questions. Thanks very much. 00:14:25.000 --> 00:14:29.000 Thanks Alex for that great talk. Our next speaker will be Fred. 00:14:29.000 --> 00:14:37.000 Pollitz from the USGS. 00:14:37.000 --> 00:14:38.000 I'm Fred Pollitz. I'm going to discuss the contribution of deformation models to the 2023 update to the US 00:14:38.000 --> 00:14:51.000 National Seismic Hazard Model. This has been a large collaborative effort with members of the Tectonic Geodesy Disciplinary group. 00:14:51.000 --> 00:14:55.000 The deformation models were first used in 2013 for the uniform California earthquake rupture forecast. 00:14:55.000 --> 00:15:15.000 And the 2014 National Seismic Hazard Model, and they play a central role in the National Seismic Hazard Model, contributing directly to the earthquake rate model as well as potentially physics-based simulator models. The goals of the 00:15:15.000 --> 00:15:30.000 Deformation Models are to determine on fault, slip rates, as well as off faults, moment rates as well as discriminate between seismic and aseismic slip on creeping faults as well as the Cascade Subduction Zone. 00:15:30.000 --> 00:15:34.000 All of the deformation models are using a GPS data set which has been updated this time around has almost 5,000 horizontal velocity 00:15:34.000 --> 00:15:45.000 vectors, and that's 1,800 more than were used in the 2014 NSHM. 00:15:45.000 --> 00:15:49.000 And these record diverse tectonic signals, including right lateral shear across the San Andrea system, eastern California. 00:15:49.000 --> 00:15:58.000 Sheer zone, Walker Lane, normal faulting in the Basin Range, 00:15:58.000 --> 00:16:17.000 Intermountain seismic bell for real Grand valley in the Front range as well as east-west contraction of the Pacific northwest coastal region in rotation of the Oregon coastal block, and a couple of these things are shown in the strain rate maps so this illustrates the 00:16:17.000 --> 00:16:23.000 East, West, contraction in the Pacific Northwest, near the subduction zone fronts, and on the right 00:16:23.000 --> 00:16:34.000 these are clear indicators of the strong right lateral shear across the San Andrea system. 00:16:34.000 --> 00:16:36.000 as well as Walker Lane, in the eastern California 00:16:36.000 --> 00:16:39.000 Shear Zone. The geologic deformation model 00:16:39.000 --> 00:16:48.000 it used published rate on about 600 faults of the Quaternary fault in fold database. 00:16:48.000 --> 00:16:51.000 It's re-evaluated QFFD 00:16:51.000 --> 00:16:54.000 slip rate bins for about 400 faults. 00:16:54.000 --> 00:17:02.000 The main thing is about 350 new faults have been added and more than 600 of those are more than half are in the lowest 00:17:02.000 --> 00:17:04.000 slip rate bin. 00:17:04.000 --> 00:17:10.000 There are 4 Western US Geodetic deformation models and I will go through these one at a time. 00:17:10.000 --> 00:17:24.000 The first is a block model, contributed by Eileen Evans, and in this model the continental crust is divided into 853 blocks, and these are bounded by more than 1,000 NSHM 00:17:24.000 --> 00:17:25.000 fault sections and a sparsity regularization is used to group blocks together. 00:17:25.000 --> 00:17:43.000 So the final model has fewer than 853 blocks. The on-fault slip rates are solved for, simultaneous with Cascadia backslip rates, and off-fault moment rates are contributed by slip rates on block boundaries that are not part of an 00:17:43.000 --> 00:17:46.000 NSHM fault section. 05:00:05.000 --> 05:00:19.00- The next model is a deep dislocation fault-based model, contributed by Yuehua Zeng and here the deformation models are the deeper extension of the NSHM 00:18:00.000 --> 00:18:09.000 faults that are defined in the upper crust, and a few additional blocks in California are added. 00:18:09.000 --> 00:18:20.000 They on-fault slip rates are solved for kinda simultaneously with a Cascadia backslip and the off-fault strain rates are determined from a couple of different sources. 00:18:20.000 --> 00:18:41.000 One, since the NSHM fault sections don't form a perfect block network when you slip all of the faults, including their deeper extensions at their full slip rates you're going to wind up with some additional strain at earth's surface and similarly there is going 00:18:41.000 --> 00:18:48.000 to be residuals to fitting the GPS velocity vectors with the initial model and a smooth velocity 00:18:48.000 --> 00:18:58.000 fields which is associated with another strain fields is fit to those residuals. So there are additional sources of strength. 00:18:58.000 --> 00:19:12.000 The next model is a Neokinema fault-based model, contributed by Zheng-Kang Shen and Peter Berg, and this is a finite element model of deformation of the Continental lithosphere, and it's informed by both the GPS vectors, and the 00:19:12.000 --> 00:19:20.000 horizontal principal stress orientations locking effects through the Cascadius slab seduction. 00:19:20.000 --> 00:19:30.000 Our accommodated using a correction with an Opera model, and that would be a model published by Rob McCaffrey and others in 2013. 00:19:30.000 --> 00:19:50.000 The fourth model is contributed by myself, its Viscoelastic model, and here we suppose that the Continental lithosphere can be subdivided into an elastic upper crust in a ductile lower crust and mantle and we need only assume that 00:19:50.000 --> 00:19:53.000 slip, whether it's seismic or aseismic 00:19:53.000 --> 00:20:04.000 occurs episodically, and when we can do that, then we can quantify the the contribution of an individual fault to the average interseismic 00:20:04.000 --> 00:20:10.000 deformation that occurs during a cycle. 00:20:10.000 --> 00:20:14.000 So we have 4 Western US geodetic deformation models. 00:20:14.000 --> 00:20:21.000 They share the same fault geometry and geologic slip rate constraints, and the same GPS velocity. 00:20:21.000 --> 00:20:25.000 They share the same corrections which I'll go through in a moment. 00:20:25.000 --> 00:20:35.000 The corrections are for an a-priori model of creep as well as earthquake cycle effects, which we also call ghost transients. 00:20:35.000 --> 00:20:40.000 The differences should then arise from the modeling techniques and assumptions. 00:20:40.000 --> 00:20:58.000 This slide provided by Jessica Murray shows details of the revised creep dataset that was assembled, and essentially the a wealth of new information provided by Sentinel 1 data. 00:20:58.000 --> 00:21:07.000 One InSAR data was appended to the earlier 2014 database. 00:21:07.000 --> 00:21:12.000 And so most of the data that contributes to the creep 00:21:12.000 --> 00:21:35.000 rate model is supplied by a InSAR, but there is also contributions from alignment arrays and creep meters, and so forth, and the forward model of creep that results from this results in the velocity vectors shown on the left or northern California and this velocity field 00:21:35.000 --> 00:21:51.000 is used as a correction to the observed GPS velocity fields, so that the corrected velocity field can be interpreted by the modelers in terms of long-term slip rates on the model faults without having to worry about the issue of how those slip 00:21:51.000 --> 00:21:57.000 rates are partitioned into seismic versus aseismic slip, and this shows the similar distribution of the velocity vectors in southern California. 00:21:57.000 --> 00:22:07.000 Primarily around the Imperial fault. 00:22:07.000 --> 00:22:14.000 We're also making a correction for ghost transients in that basically means correcting for time dependent 00:22:14.000 --> 00:22:28.000 displacements during an earthquake cycle. So the average interseismic velocity around a vertical strike slip faults would follow the pattern shown here. 00:22:28.000 --> 00:22:47.000 However, after a large event, early in the cycle, due to viscous relaxation of the lower crust and mantle, the deformation rates are elevated and so we would have a positive perturbation, and we could call that a right-lateral ghost transient However, 00:22:47.000 --> 00:23:01.000 late in the cycle the displacements are more subdued, and we would have a negative perturbation, and we would call that a left-lateral ghost transient, and we want to correct for those effects 00:23:01.000 --> 00:23:24.000 so the the GPS velocity fields, after the correction, is what we would expect for the cycle average contributed by all of the models faults because all of the interpretational methods of the models assume that we're observing the cycle average. So this is what 00:23:24.000 --> 00:23:32.000 the correction looks like for various regions, and I would just point out in northern and southern California the ghost transient correction is left-lateral 00:23:32.000 --> 00:23:44.000 and so when you make the correction, and then in interpret the corrected GPS velocity fields in terms of slip rates, you are going to raise the slip rates along the San 00:23:44.000 --> 00:23:50.000 Andrea System, and elsewhere, potentially by several millimeters per year. 00:23:50.000 --> 00:24:01.000 So now I'll go through some deformation model results, including on-fault, deformation, off-fault, deformation, and the budget of on-fault, and off-fault moment. 00:24:01.000 --> 00:24:10.000 Two of the models which are the Pollitz and Evans models that have more extreme variations with respect to the geologic deformation model. 00:24:10.000 --> 00:24:38.000 So here model-slip rates that agree perfectly with the geologic deformation model plot with the value of one and are in green, whereas slip rates that fall very far below or above that are in dark blue or purple so these differences are by design because the Pollitz and 00:24:38.000 --> 00:24:45.000 Evan's models do allow for more extreme variations with respect to the geologic rate 00:24:45.000 --> 00:24:50.000 then to the Zeng and Shenbird models. 00:24:50.000 --> 00:25:09.000 This shows that the geodetic deformation models tend to agree well with the geologic deformation model on the higher slip-rate faults, however, they tend to be higher than the geologic deformation model rates on the very low slip rate . 00:25:09.000 --> 00:25:31.000 faults. So a tenth of a millimeter per year or less, and that could simply be due to the errors in the geodetic models being at the point 1 per year level, or higher, they're simply greater than the slip rates themselves, or could reflect the fact 00:25:31.000 --> 00:25:42.000 that geodetic strain in some sense it's higher than the geologic strain, and that has to lead to geodesic rates 00:25:42.000 --> 00:25:48.000 being higher than geologic rates. Somewhere in the spectrum of slip rates. 00:25:48.000 --> 00:26:06.000 As I mentioned the the ghost transient across correction has a potentially large effect, and this shows the difference between slip rate inversions where we include this correction and inversions where we don't. 00:26:06.000 --> 00:26:13.000 So without the correction, we get this pattern of slippery when we include the correction, we get a revised pattern. 00:26:13.000 --> 00:26:24.000 If we toggle back and forth, we do see that there are differences in the plate interior as well as in California, and so in general, the geodesic slip rates across the Western US 00:26:24.000 --> 00:26:35.000 are impacted by the correction for earthquake cycle effects 00:26:35.000 --> 00:26:40.000 And this is further and simplified in the figure from SRL 00:26:40.000 --> 00:27:02.000 Paper by Eileen Evans, where the effects on our absolute scale is most pronounced in California, where slip rates after the correction, go up by several millimeters per year, and this can be looked at for the individual model so for the Pollitz and Evan's models. For the high slippery 00:27:02.000 --> 00:27:20.000 faults that are identified here the effect of the ghost trains, the incorrection is to raise those slip rates by several millimeters per year, and we see the same tendency for the Zeng and Shenbird models, and it also brings those geodetic slip rates into better 00:27:20.000 --> 00:27:26.000 agreement with the geologic slip rates, and this is also shown in figures in a couple of the SRL 00:27:26.000 --> 00:27:40.000 studies from the paper by Zeng. This also shows the effects of making the ghost transient correction. 00:27:40.000 --> 00:27:51.000 It raises slip rates on a number of the high slip faults and brings them into better agreement with the geologic rates and a similar figure by Shen and Byrd 00:27:51.000 --> 00:27:56.000 shows the same thing. 00:27:56.000 --> 00:28:06.000 This shows the off-fault moment rates contributed by 3 out of the for deformation models, and although there are considerable differences, broadly speaking, they show the same pattern. 00:28:06.000 --> 00:28:14.000 The awful moment rates tend to be concentrated along the San Andreas corridor, 00:28:14.000 --> 00:28:18.000 the Eastern California Shear Zone and Walker Lane, and the Inter-Mountain seismic belt. 00:28:18.000 --> 00:28:34.000 The off-fault moment rates tend to be consistent across the models between 2 and 2 and a half times 10 to the 19 Newton meters per year, and the ghost transient correction tends to have a modest effect 00:28:34.000 --> 00:28:38.000 on raising those overall rates, there's a a bigger variety in the impact of the off- 00:28:38.000 --> 00:28:54.000 Fault moment rates although 3 out of the 4 have similar numbers for the off-fault moment rates and the contribution of the soft moment rates are anywhere from 30 to 60% of the total. 00:28:54.000 --> 00:28:57.000 So this is telling us that slip on faults 00:28:57.000 --> 00:29:03.000 does not even come close to explaining all of the potentials 00:29:03.000 --> 00:29:18.000 seismic moment release in the Western United States. And so we need to consider whether through deformation models or through some other means, we need to consider sources of deformation that are not on 00:29:18.000 --> 00:29:32.000 the faults that we know about. 00:29:32.000 --> 00:29:39.000 Thanks for your attention everyone. Sorry, for the abrupt cut off. 00:29:39.000 --> 00:29:40.000 Thank you very much, Fred, for a great talk. 00:29:40.000 --> 00:29:50.000 Our next speaker is Kai Johnson from Indiana University. Kai. 00:29:50.000 --> 00:30:10.000 Hi there! Kai Johnson, Indiana University and talk about how we implemented fault creep into the 2023 model so work done with Jessica Murray at USGS and Crystal Wespestaqd, who's a former student in Indiana she was very helpful 00:30:10.000 --> 00:30:12.000 In compiling data for us. So this is all summarized in this publication in the 00:30:12.000 --> 00:30:25.000 SRL special focus section on deformation models. So everything I'm talking about is already published. 00:30:25.000 --> 00:30:35.000 So well, I think everybody here knows what fault creep is, but to just be on the same page about terminology, and how we conceptualize this. 00:30:35.000 --> 00:30:40.000 So this is a nice little illustration by Uchida and Burgmann, 2019. 00:30:40.000 --> 00:30:51.000 So we envisioned patches of fully locked areas on faults. Lock patches that are I'll call asparities and then surrounding areas that creep. 00:30:51.000 --> 00:31:06.000 And of course the lock patches are sticks-slip, and rupture and earthquakes the creeping areas we're going to assume are creeps steadily surrounding over time, surrounding the locked areas. 00:31:06.000 --> 00:31:13.000 Okay, so that's the conceptual model. So what the objectives here is to estimate distribution of creep. 00:31:13.000 --> 00:31:20.000 So we first compiled recent surface creep rate observations since the 2013 00:31:20.000 --> 00:31:36.000 UCERF3 model so another 10 years of observations then we used those. Geodetic data in a model, I'll estimatecreep rate at depth on faults. 00:31:36.000 --> 00:31:39.000 And so those depth average creep rates like shown over here. 00:31:39.000 --> 00:31:49.000 This is an actual result of our study. Those depth average creep rates are used to compute a moment reduction for the hazard model. 00:31:49.000 --> 00:31:54.000 So a little history of how this came about and why we're doing this so in UCERF3 00:31:54.000 --> 00:32:02.000 Weldon et. al. came for the method that was very good. 00:32:02.000 --> 00:32:19.000 They took the surface creep rate data. They smoothed it along strike and then took those along strike creep rates. 05:14:38.000 --> 05:14:47.00 Normalized by slip rate, down here at the bottom corner, came up with a curve to relate surface creep rate to how much moment reduction due to depth average creep rate to apply, And on the left, 00:32:28.000 --> 00:32:42.000 I won't go into much detail, but they had a rationalization for this curve, which is in the scenario on the far left that there's only shallow surface creep rate and locked at depth. There's a certain amount of creep rate reduction 00:32:42.000 --> 00:32:47.000 that's computed from an elastic crack model. 00:32:47.000 --> 00:33:08.000 And then they modified that for cases where the faults are not creeping at all, there's creep at all depths, which increases the amount of moment reduction, so they ended up with this kind of curve that was 00:33:08.000 --> 00:33:12.000 applied to creeping faults in UCERF3. 00:33:12.000 --> 00:33:31.000 So what we've done is well, we've tried to improve on that by actually solving at depth average creep rates on faults rather than apply that rule they came up with. Alright so here's the compilation of a surface creep rate so we 00:33:31.000 --> 00:33:34.000 updated; there were 97 measurements in UCEFR3. 00:33:34.000 --> 00:33:40.000 Now we have 400 surface creep rates in this update, most of the new additions are from InSAR studies and some updates 00:33:40.000 --> 00:34:02.000 the alignment data, but largely from InSAR. So on the left summary of where we have new or updated data, those are the red dots, and then the count density, 00:34:02.000 --> 00:34:23.000 these color blue shaded contours just are proportional to how many new observations were added, so you can see, you know where most of the new information from the 2013 model to date, and a histogram on the bottom, 00:34:23.000 --> 00:34:34.000 you can see that most of our data are InSAR alignment array data, and some other cultural offsets, etc., the other kind of data. 00:34:34.000 --> 00:34:36.000 So this is all available to download publicly available here at the Science Base Catalog. 00:34:36.000 --> 00:34:48.000 You can download it. Very quickly 00:34:48.000 --> 00:34:55.000 on the left, not going to spend a lot of time looking at these, but this is a long-strike 00:34:55.000 --> 00:35:07.000 plots of surface creep rate measurements with distance along fault sections. You know the Bartlett Springs, Maacaama, San Andreas, 00:35:07.000 --> 00:35:10.000 and then southern California Southern systems that we modeled. 00:35:10.000 --> 00:35:16.000 And so you can see some of the data and there uncertainties along strike. 00:35:16.000 --> 00:35:24.000 There is quite a bit. I'm not gonna dwell on any of these data right now, but there's quite a bit of spread in some of the InSAR data. 00:35:24.000 --> 00:35:28.000 These are quite a bit of spread in InSAR data Shoe et al. in our data. And so that's the reality of inSAR data. 00:35:28.000 --> 00:35:39.000 But there's a lot of InSAR data which really helps with constraining surface creep rates. 00:35:39.000 --> 00:35:42.000 So we have a MoD. We need a model to go from 00:35:42.000 --> 00:35:50.000 data to creep rates at depth. 00:35:50.000 --> 00:36:00.000 And so we adopt this physically constrained creep rate inversion that I've been calling locked so faults are either locked or creep at constant stress. 00:36:00.000 --> 00:36:13.000 So this idea is not new. We implemented it in a couple papers already, and there are a couple of other groups that have been doing very similar type constrained creep inversions. 00:36:13.000 --> 00:36:18.000 Basic idea is that you have the fault, and it's either locked or creeping. 00:36:18.000 --> 00:36:25.000 So we impose a locked spot on the fault, and then we let it creep around that at constant stress. 00:36:25.000 --> 00:36:28.000 So the creeping areas don't accumulate. 00:36:28.000 --> 00:36:39.000 stress is the idea. And so you get this kind of creepy distribution around a locked patch, and we build that model into the forum model and do an inversion. 00:36:39.000 --> 00:36:43.000 A Bayesian version of both surface geodic data. 00:36:43.000 --> 00:36:59.000 So GPS data, GPS derived surface velocity, field and creep rate, surface creep, and the it's implemented in kind of a backslip inversion and 00:36:59.000 --> 00:37:03.000 we generate the long-term deformation field. 00:37:03.000 --> 00:37:09.000 Very similar to the Zeng model, which is a fairly standard method 00:37:09.000 --> 00:37:26.000 now, where faults are extended to infinite depth, and and extended outside the model domain, and slipped at the long term slip rate to generate sort of the far field long-term deformation field. 00:37:26.000 --> 00:37:29.000 Okay. Here's the northern California fault geometry. 00:37:29.000 --> 00:37:34.000 Southern California fault geometry. I didn't mention it, but the kind of greenish faults are faults in the hazard model that we don't impose creep, 00:37:34.000 --> 00:37:43.000 we don't allow to creep. We have no observations, or we don't think they creep. 00:37:43.000 --> 00:37:54.000 And then the meshed surfaces. Here are the faults that are allowed to creep, so more details of that, of course, are in the paper 00:37:54.000 --> 00:37:58.000 if you want to know more about how we actually do all that. Here are some results 00:37:58.000 --> 00:38:04.000 these are for northern California, southern California. The top panel, northern California, showing... oh. 00:38:04.000 --> 00:38:09.000 the left side panels, long term slip rates; middle panels. 00:38:09.000 --> 00:38:11.000 the depth average and seismic creep rate and the right side is a coupling ratio. 00:38:11.000 --> 00:38:14.000 So it's the middle panels that we provide. That's what's given to the hazard model. 00:38:14.000 --> 00:38:32.000 And so you can see, for example, if we look at the coupling ratios right where you see blue, the low coupling lots of creep. 00:38:32.000 --> 00:38:35.000 Unsurprising, that there's lots of creep along the creeping section of the 00:38:35.000 --> 00:38:47.000 San Andreas, Calaveras, Hayward, and there is non-zero, but less creep on many of the other faults throughout the northern California. Okay. 00:38:47.000 --> 00:38:50.000 Here's how we fit the Geodetic datav 00:38:50.000 --> 00:38:54.000 That went very quickly over. 00:38:54.000 --> 00:38:57.000 everything used to constrain the long-term fault slip rates. 00:38:57.000 --> 00:39:01.000 And here's the fit to the northern California model. 00:39:01.000 --> 00:39:12.000 The data and the residuals, and then southern California down here, and as far as these kind of kinematic models go fitting, geodetic data, this is quite a reasonable fit. 00:39:12.000 --> 00:39:24.000 There's largely random residuals left over here's the fit to the surface 00:39:24.000 --> 00:39:27.000 creep rate data on the left. Of the observations on the right is the model and if you just kind of look back and forth, you see that? Yeah? 00:39:27.000 --> 00:39:40.000 Well, the inversion really does fit the general pattern of spatial distribution of surface creep rate. 00:39:40.000 --> 00:39:47.000 really quite well in both northern and southern California. 00:39:47.000 --> 00:39:48.000 Okay, so here's what the creep rates look like. 00:39:48.000 --> 00:39:54.000 I'll show you first creep rates and then locking distribution, intercepts a creep rate. 00:39:54.000 --> 00:39:58.000 This is just the northern California model. Color is millimeters per year. 00:39:58.000 --> 00:40:05.000 Log scale and 2 perspective seared ones looking off to whether the northeast and then on the bottom. 00:40:05.000 --> 00:40:12.000 we're looking off to the southwest. Okay? Just so you orient your brain here. 00:40:12.000 --> 00:40:23.000 So you can see the distribution of creep red being high blue being low, of course, and so that's what the spatial description looks like. 00:40:23.000 --> 00:40:33.000 If we look at the probability of locking that's what we estimate that a patch is either locked or creeping, and a probabilistic inversion. 00:40:33.000 --> 00:40:36.000 so we can get the probability that a patch is locked. 00:40:36.000 --> 00:40:43.000 So red means high probability. Blue means low probability, or being locked. So, blue means creeping. 00:40:43.000 --> 00:41:02.000 So what you'll notice is, there's areas that are quite blue where faults creep when and the you know, the creeping section of the San Andreas is the Hayward fault has a lot of shallow creep, and then throughout northern California there's sort of patches of 00:41:02.000 --> 00:41:08.000 mostly shallow, sometimes a bit deep creep to the north. Okay, we don't resolve very well. 00:41:08.000 --> 00:41:15.000 There's a lot of green on here, because there's a lot of areas of the fault we can't really resolve individual patches whether they're locked or creeping. 00:41:15.000 --> 00:41:16.000 You get sort of a spatial average. And so the spatial average of zeros and ones is 0 point 5. 00:41:16.000 --> 00:41:24.000 And so that's why you see a lot of 0 point 5 that 00:41:24.000 --> 00:41:28.000 means it's mostly locked if it's point 5. 00:41:28.000 --> 00:41:38.000 Okay. So a bit of a whirlwind. But what we ultimately then deliver to the National Seismic Hazard Model is surface creep rates for each of the 5 deformation models. 00:41:38.000 --> 00:41:47.000 So now what we've done is we fix the long-term slip rates 00:41:47.000 --> 00:41:49.000 to the 5 deformation models shown here. 00:41:49.000 --> 00:41:57.000 Geologic model, Evan's model, Paula Schenberg. 00:41:57.000 --> 00:41:58.000 And then this study is shown with this kind of teal color. 00:41:58.000 --> 00:42:08.000 So you see the long-term slip rates with distance along the fault; inner seismic depth average creep rates are the dash line, so it's really the dash lines again. 00:42:08.000 --> 00:42:10.000 that are being delivered to the Hazard Model and the take-home point of this is that there's uncertainty in long-term slip rates. 00:42:10.000 --> 00:42:17.000 That maps into uncertainty in the depth average creep rates, and that's just expected right? 00:42:17.000 --> 00:42:40.000 That's sort of the nature of this. And so the nice thing is that you know the uncertainty and depth average creep rates due to slippery uncertainties being being incorporated. Last point here, if we compare the moment. 05:24:59.000 --> 05:25:04.0 reduction, which is proportional to the depth average creep rates to the UCERF3 model 00:42:45.000 --> 00:42:48.000 that curve I showed there, beginning of the talk. This curve right here is what was used in UCERF3. 00:42:48.000 --> 00:42:59.000 The dots are the moment reduction that's computed for all the 5 deformation models for 2023. 00:42:59.000 --> 00:43:05.000 And you can see that the 23 reductions are all systematically higher than what was adopted in UCERF3. 00:43:05.000 --> 00:43:18.000 That's because our model and inversions infer more, creep, more, creep, just sort of almost everywhere, systematically, everywhere more creep, and therefore more moment reductions. 00:43:18.000 --> 00:43:28.000 That's that was really interesting result of this definitely worth some more thought in thinking about these UCERF3. 00:43:28.000 --> 00:43:35.000 And then the 2023 models. I'm going to skip the future for creep modeling. 00:43:35.000 --> 00:43:42.000 I ran out of time, but I'm thinking about some improvements already, so I'll end here because I'm out of time. 00:43:42.000 --> 00:43:46.000 But several conclusions. So we've mapped out the creepy distribution. 00:43:46.000 --> 00:43:53.000 We provide moment reduction for all the faults and our moment reductions are systematically higher than were adopted for UCERF3. 00:43:53.000 --> 00:44:06.000 I've got to stop. Thank you very much. 00:44:07.000 --> 00:44:11.000 Okay, thank you very much, Kaj. Our next speaker will be Kevin Milner from USC. 00:44:11.000 --> 00:44:12.000 [noise] Hi Everyone. My name is Kevin Milner I'm at USC 00:44:12.000 --> 00:44:25.000 in the Southern California Earthquake Center (SCEC) I'm going be talking today about updated fault system inversion solutions and hazard implications with a bit of a northern California slant. 00:44:25.000 --> 00:44:44.000 Of course this is for the 2023 update of the National Seismic Hazard Model. The goal of this study is to improve upon the fault system inversion approach used in UCERF3 by exploring a wider solution space and better fitting available data. We have an updated draft model available, it spans the 00:44:44.000 --> 00:44:49.000 inversion approach to the full Western US. It uses a new multi-fold rupture possibility 00:44:49.000 --> 00:44:59.000 algorithm, it uses a simpler formulation for target magnitude frequency distributions and adds a new segmentation constraint to control the propensity of multi-fault ruptures. 00:44:59.000 --> 00:45:01.000 I don't have a lot of time to get involved with details. 00:45:01.000 --> 00:45:05.000 We're going to be basically drinking from the firehose, 00:45:05.000 --> 00:45:14.000 but hopefully you'll get the gist of the model across, and then I'll end with some hazard comparisons with UCERF3. 00:45:14.000 --> 00:45:23.000 Speaking of hazard, this is a hazard map, computed with the draft model, just a single ground motion model. 00:45:23.000 --> 00:45:30.000 And I'm showing peak ground acceleration with 2% at exceedance and probability in 50 years. 00:45:30.000 --> 00:45:31.000 This is the fault mean map on the left and then here this kind of shows the influence of individual logic 00:45:31.000 --> 00:45:42.000 tree branch choices on the final model. I'm going to be focusing on these fault-based choices in the middle. 00:45:42.000 --> 00:45:43.000 The note that there also is a new grid seismicity model. 00:45:43.000 --> 00:45:44.000 And that group seismicity model can be really important especially outside of California. 00:45:44.000 --> 00:45:55.000 But in California it's the fault-based model that matters the most. I've mentioned UCERF3. 00:45:55.000 --> 00:46:03.000 that's the 3rd Uniform California Earthquake Rupture Forecast it was the California component of the most recent NSHMs 00:46:03.000 --> 00:46:09.000 a key innovation for UCERF3 was that it relaxed fault segmentation assumptions. 00:46:09.000 --> 00:46:26.000 So a prior to UCERF3 ruptures were more or less segmented in that ruptures would start and end at predefined section endpoints and multi-fault ruptures were not allowed, even though a number of them have occurred in nature especially recently. So UCERF3 added 00:46:26.000 --> 00:46:27.000 multi-fault ruptures through a new rupture plausibility m mo del 00:46:27.000 --> 00:46:35.000 and then solve for the rate of each rupture using a newly simulated annealing inversion methodology. 00:46:35.000 --> 00:46:41.000 That's a pretty simple iterative optimization algorithm. 00:46:41.000 --> 00:46:45.000 We have three new logic tree branches this time around. 00:46:45.000 --> 00:46:49.000 So far from what I'm going to be presenting today, everything is evenly weighted. 00:46:49.000 --> 00:46:50.000 That's mostly just to not bias the review process. 00:46:50.000 --> 00:46:59.000 But I would expect that the final model will have some tweaks to these branch weights. 00:46:59.000 --> 00:47:05.000 Let's start with the B value branch and the motivation for this branch is that fault event 00:47:05.000 --> 00:47:08.000 rates are very poorly constrained, you know. On a regional scale 00:47:08.000 --> 00:47:16.000 we think we know something about the rate of earthquakes, the magnitude frequency distribution of earthquakes. 00:47:16.000 --> 00:47:20.000 But we don't know as much for individual faults. 00:47:20.000 --> 00:47:27.000 So, for example, let's consider this fault. Let's say it has a 10 mm per year slip rate 00:47:27.000 --> 00:47:39.000 that's one data constraint we have. That doesn't really tell us how that slip rate is satisfied in fact that's really poorly constrained empirically, but can be highly influential on the final computed hazard. 00:47:39.000 --> 00:47:41.000 So, one viable model would be what we call the minimum rate model. 00:47:41.000 --> 00:48:00.000 So in that model this entire fault is satisfied by a single magnitude 7.5 event. That would have to happen just once every 315 years or so to fully satisfy that slip rate. 00:48:00.000 --> 00:48:08.000 On the other end of the spectrum, the maximum rate model for many smaller or moderately sized ruptures, 6.5 00:48:08.000 --> 00:48:12.000 in this case satisfy the slip rate of this fault. 00:48:12.000 --> 00:48:22.000 Those 6.5 would have to happen on average once if every 15 years along this fault, in order to satisfy its entire slip rate. 00:48:22.000 --> 00:48:26.000 Those are two pretty extreme models. But there's this whole null space of viable models in between these 2 extremes. 00:48:26.000 --> 00:48:48.000 What we are doing this time, I think the answer is probably, for most faults is probably somewhere in between these two models, and what we're doing this time around is trying to explicitly force the inversion to choose a number of solutions that span a viable range between these extreme and member models. This is another way to look 00:48:48.000 --> 00:48:59.000 at that, in terms of a magnitude frequency distribution here is that maximum rate model where the slip rate is cumulative by just the 6.5's a really high event rate 00:48:59.000 --> 00:49:10.000 the minimum rate model here. Between this I've drawn various Gutenberg Richer distributions that also use moderate magnitudes in between in order to satisfy the slip rate. 00:49:10.000 --> 00:49:15.000 So all of these models that I've plotted here satisfy the slip rate equally well. 00:49:15.000 --> 00:49:26.000 They all have a very different cumulative event rate here, and it's that cumulative rate of events that really drives hazard and sites near this fault. 00:49:26.000 --> 00:49:42.000 So what we're gonna do is constrain the inversion to satisfy a range of cumulative event rates drawn from these different Gutenberg Richter distributions ranging from b=0 to b=1, so it's not full range between the 00:49:42.000 --> 00:49:49.000 minimum rate and maximum rate models. But we think it's a pretty viable sub-range in the middle. On a regional scale 00:49:49.000 --> 00:49:58.000 we sum up all those individual single fault magnitude frequency distributions to get a regional target. 00:49:58.000 --> 00:50:05.000 And even though we only constrain it on individual faults to match the total rate implied by the Gutenberg, Richter MFD 00:50:05.000 --> 00:50:09.000 on a regional scale, we constrain it to match the MFD 00:50:09.000 --> 00:50:18.000 as a whole. This is what it looks like, for b=1 reference branch in this case even though each individual MFD is b=1. 00:50:18.000 --> 00:50:25.000 They all have different Mmin and Mmax and that leads to this kind of double taper effect. 00:50:25.000 --> 00:50:29.000 After running it through the inversion this is the result. 00:50:29.000 --> 00:50:35.000 Each individual color here is for different b-value and the mean model is in black here. 00:50:35.000 --> 00:50:52.000 So you can see that we're spending a pretty wide range of total event rates in the model as well as really controlling the rate of both moderate size events, which is higher for the b=1 branch and larger events which is highest for the b=0 but we've also added a 00:50:52.000 --> 00:50:56.000 new segmentation constraint motivation for this is 00:50:56.000 --> 00:51:00.000 some really lively debate on the propensity of multi-fault ruptures. 00:51:00.000 --> 00:51:04.000 Both turn development of the UCERF3 model and subsequently in the literature. 00:51:04.000 --> 00:51:17.000 On one side in 2018, Schwartz argued that we went too far in UCERF3 with multi-fault ruptures, and then Page published rebuttal in 2020 saying that we didn't go far enough. 00:51:17.000 --> 00:51:24.000 We're now attempting to span out a wide range of models that includes both viewpoints and some in between as well. 00:51:24.000 --> 00:51:28.000 Another motivation for this constraint is that we're now using a more permissive multi-fault rupture plausibility model. 00:51:28.000 --> 00:51:41.000 This model includes some really long jumps up to 15 kilometers, partially acknowledging that we just don't know what faults we're missing in our fault system. 00:51:41.000 --> 00:51:47.000 So a 15 kilometer jump in our fault system might actually be a short jump that uses an unknown fault, 00:51:47.000 --> 00:51:55.000 but nonetheless we want those jumps to be taken much less often than some of the more likely shorter jumps. 00:51:55.000 --> 00:52:10.000 And also we lacked an explicit control on ruptures through the creeping section in UCERF3 I didn't end up mattering very much for typical single site hazard, but it can matter a lot if you're doing a portfolio loss calculation where the 00:52:10.000 --> 00:52:21.000 correlation of risk really matters. We define these segmentation constraints in terms of their pass through rate, or the fraction of ruptures that pass through a segmentation boundary. 00:52:21.000 --> 00:52:26.000 So if you think about it as a ruptures going along, the pass through rate, be the fraction of times 00:52:26.000 --> 00:52:34.000 it takes a jump as opposed to not taking that jump which can be of stopping or continuing on the same fa=ult without taking the jump. 00:52:34.000 --> 00:52:37.000 If we don't constrain segmentation, as I'm showing on the left. 00:52:37.000 --> 00:52:45.000 There are many places in the model that have very high rates, very high pastor rates. 00:52:45.000 --> 00:52:46.000 This is what it looks like when we do apply the segmentation model. 00:52:46.000 --> 00:52:49.000 And see if we're really bringing down the rates, especially these really large, large jumps. 00:52:49.000 --> 00:53:09.000 Although there's still a lot of connectivity especially along faults like the San Andreas and in some really closely connected areas. Our primary segmentation model is a distance dependent model from Shaw Dieterich (2007) they used dynamic rupture simulations 00:53:09.000 --> 00:53:16.000 and also some comparisons to the Wenowski data site to come up with a pretty simple exponential model. 00:53:16.000 --> 00:53:20.000 This plot here shows the pass through rate for each individual potential jumping point as a function of jump distance. 00:53:20.000 --> 00:53:26.000 This is without the segmentation constraint. In the model without the segmentation 00:53:26.000 --> 00:53:35.000 and strain inversion uses some really long jumps just as often as it will use for the short jumps. 00:53:35.000 --> 00:53:42.000 When we applied the constraint we end up really controlling the rate of those long jumps. 00:53:42.000 --> 00:53:48.000 Note that this constraint is applied as an inequality constraint, so we're not forcing the model to take jumps, 00:53:48.000 --> 00:53:49.000 we're just limiting it to not exceed the given segmentation 00:53:49.000 --> 00:53:55.000 pass through rate. 00:53:55.000 --> 00:53:59.000 So we have a none branch that applies no segmentation 00:53:59.000 --> 00:54:09.000 penalty, three distance dependent branches, and also a classic model that takes the high segmentation branch and adds some further restrictions. 00:54:09.000 --> 00:54:13.000 The classic model is most like the prior models in California 00:54:13.000 --> 00:54:23.000 before UCERF3. So the UCERF2 model, for example, that allowed multi-segment ruptures within faults such as the San Andreas or the San Jacinto, but not between them. 00:54:23.000 --> 00:54:29.000 And we've also added multiple ruptures in this classic branch that have previously existed in nature. 00:54:29.000 --> 00:54:38.000 So, for example, the Landers, the faults involved in the Lander's rupture, allowed to rupture together on this classic branch. 00:54:38.000 --> 00:54:47.000 Outside of that all faults are treated as completely isolated, and they're solved using the prescriptive approach rather than the inversion. 00:54:47.000 --> 00:54:50.000 We also have a new constraint for ruptures on the creeping section. 00:54:50.000 --> 00:55:04.000 Here we constrain the rate at which ruptures can continue on into the creeping section and on some logic tree branches completely prohibit ruptures from entering and then going through and exiting the other side of the creeping section. 00:55:04.000 --> 00:55:17.000 What this does is really decreases the rate of those throughgoing ruptures, and note that a throughgoing rupture here could be one that just starts here and ends here it's not necessarily the big magnitude 8 wall-to-wall rupture. An aggregate, the segmentation 00:55:17.000 --> 00:55:30.000 models have a pretty large effect on the final model. In this case, the classic model shown in pink really brings up the rate of moderate size earthquakes and brings down the rate of the 00:55:30.000 --> 00:55:37.000 largest ruptures. We also have a new logic tree branch in our model relating to paleoseismic data fits. 00:55:37.000 --> 00:55:45.000 The thinking here is that there are sometimes places in the model where the slip rate and the paleoseismic event rate are seemingly at odds. 00:55:45.000 --> 00:55:52.000 We want to honor both data equally well, and we attempt to do so on this even fit branch. 00:55:52.000 --> 00:55:57.000 But we also want there to be a place in the model where we perfectly honor the paleoseismic data 00:55:57.000 --> 00:56:02.000 that is our only real constraint on the actual event rate on a fault. 00:56:02.000 --> 00:56:07.000 We also have a branch from the slip rate data higher than the paleoseismic event rate 00:56:07.000 --> 00:56:11.000 data to span the whole range of possible behaviors. Okay. 00:56:11.000 --> 00:56:14.000 Now for some hazard comparisons. On the right 00:56:14.000 --> 00:56:19.000 I'm showing a hazard comparison between the new model and UCERF3. 00:56:19.000 --> 00:56:26.000 Here, I'm using using UCERF3 ingredients and only changing the methodology of the inversion 00:56:26.000 --> 00:56:33.000 and our constraint implementations, so we're still using the user free deformation models, fault models and scaling relationships. 00:56:33.000 --> 00:56:36.000 Notice that most of the changes due to methodology are in southern California there's really not much going on in northern California. 00:56:36.000 --> 00:56:52.000 These dumbbell shapes are a pretty small artifact, it actually turns out in UCERF3, it would fill in the ends of faults sometimes with smaller earthquakes in order to satisfy the slip rate, and that's not allowed now due to our more 00:56:52.000 --> 00:56:58.000 explicit event rate constraints. But in southern California it turned out we weren't satisfying the slip rates 00:56:58.000 --> 00:57:12.000 very well. This was due to a really tight regional magnitude frequency distribution for southern California that meant that we couldn't fit the regional MFD. 00:57:12.000 --> 00:57:21.000 The paleo data and the slip rates all simultaneously, and as a result our slip rate fits suffered. Now we're fitting the slip rates much better 00:57:21.000 --> 00:57:24.000 that leads to increased hazard at a number of these sites. 00:57:24.000 --> 00:57:39.000 These areas of hazard increase are also often areas where UCERF3 had very low or even negative b-values on average on each fault using really only large multi-fault ruptures to satisfy the moment rate. 00:57:39.000 --> 00:57:44.000 And now we're forcing them to be more Richter and satisfy more of their slip rate with moderate size 00:57:44.000 --> 00:57:47.000 earthquakes that also causes hazard to increase. 00:57:47.000 --> 00:57:53.000 Now I compare that methodological hazard change map with an ingredient hazard 00:57:53.000 --> 00:58:00.000 change map. Here we're holding the methodology constant and showing just the impact of ingredient changes. 00:58:00.000 --> 00:58:09.000 For example, updated deformation models, added faults, updated scaling relationships and revised paleoseismic recurrence intervals. 00:58:09.000 --> 00:58:17.000 This is the primary driver of hazard changes in northern California. For example, they added some faults and change slip rates on existing faults. 00:58:17.000 --> 00:58:21.000 Now on the left. I'm showing the total hazard change between the draft and SHM 23 00:58:21.000 --> 00:58:30.000 model and UCERF3. On the right, I am highlighting areas where the hazard has changed by more than 10% on coloring them 00:58:30.000 --> 00:58:35.000 by the source of that change. Areas colored in violet are those for ingredient changes or the primary driver of the hazard change 00:58:35.000 --> 00:58:52.000 specifically fault and deformation model updates. Areas in green are those where methodological changes dominate I don't think I left much time for questions but I'm happy to discuss this further. 00:58:52.000 --> 00:58:56.000 After the next talk in the general discussion session. Speaking in the next talk, all of these comparisons just use a single ground motion model. 00:58:56.000 --> 00:59:08.000 I think Mark is going to talk next about how ground motion model changes affect these sorts of hazard comparisons so stay tuned. 00:59:08.000 --> 00:59:10.000 Thank you. 00:59:10.000 --> 00:59:13.000 Thank you very much, Kevin. That was an excellent talk. 00:59:13.000 --> 00:59:19.000 Our final speaker is Mark Peterson from the USGS. 00:59:19.000 --> 00:59:25.000 Hello, everyone. Today, I want to discuss with you our 2,023 national seismic hazard model update. 00:59:25.000 --> 00:59:31.000 I want to thank all the national sismic hazard model update participants from Ese and Ghsc. 00:59:31.000 --> 00:59:38.000 And other places as well. You heard previously from the Erf group discussing some of the model components. 00:59:38.000 --> 00:59:47.000 I want to talk today about our Cascadia update, which is, I don't think the previous talks discussed and the ground motion model 00:59:47.000 --> 00:59:49.000 Let's start with the Cascadia abduction zone. 00:59:49.000 --> 01:00:08.000 We have interpreted the rupture history of magnitude 8 to 9 earthquakes from the 10,000 year record of turbidites mostly from Chris Goldfingers, shown here from his professional paper there are 19 to 20 records in the course showing full Cascadia ruptures of about 01:00:08.000 --> 01:00:16.000 magnitude 9. These are sandy turbidites, and then there are a series of smaller earthquakes that are probably magnitude 8 to 9 that are, and they're about 20. 01:00:16.000 --> 01:00:20.000 3 of these, with segment boundaries located across there. 01:00:20.000 --> 01:00:30.000 So we have developed a rate model based on this data. The logic trees for the Cascadia subduction zone are very complex, and we published these back in 2,014. 01:00:30.000 --> 01:00:38.000 We're still using something very similar. The Gmps have been changed to updated with a Nj. 01:00:38.000 --> 01:00:42.000 Subduction zone models. We have new magnitude, scaling equations that were developed by Bruce Shaw. 01:00:42.000 --> 01:00:56.000 And we're considering a number of these. We have new onshore and offshore geologic data that we're considering to update the models for the Cascadia abductions on rates. 01:00:56.000 --> 01:01:11.000 Hart Frankel, on his colleagues, put together this new proposed 2,023 Cascadia seductions on model, and they used the 2,014 model, but added additional logic tree branches for a time dependent probability ranging from 9% for 01:01:11.000 --> 01:01:14.000 Poisson to 13% for a Brownian passage. 01:01:14.000 --> 01:01:17.000 Time addition, they added a cluster model that accounted for different earthquakes that could occur in a series of magnitude 8. 01:01:17.000 --> 01:01:27.000 That rupture in a short period of time across the Cascadius induction zone for the magnitude 8 partial rupture, logic tree. 01:01:27.000 --> 01:01:37.000 They added information from Chris Goldfinger, from 2,017, and updated his 2012 model. 01:01:37.000 --> 01:01:40.000 They added paleocytic event. Count data from Bradley Lake and Sixes River onshore sites and a maximum rupture model of Nelson. 01:01:40.000 --> 01:01:46.000 At all that included periods of 1,000 607,000 years. 01:01:46.000 --> 01:01:56.000 This didn't really make major changes to the to the overall rates for the Cascadia subduction zone earthquakes. 01:01:56.000 --> 01:02:10.000 These maps show comparisons of the new 2023 proposed model compared to the 2,018 National hazard model for the Cascadia subduction zone at 0 point 2 one and 3 s periods and a vs. 01:02:10.000 --> 01:02:13.000 30 of 600 meter per second, 2% 50 year return period. 01:02:13.000 --> 01:02:28.000 These were made by Jason Alto, crew. What you see in these plots are differences which are quite small, and ratios, which are very small, typically about 1% changes to the hazard caused by these new models that we just discussed. 01:02:28.000 --> 01:02:35.000 Let's talk for a minute about the Cascadia seduction ground motion model waves that are proposed for the 2,023 updates. 01:02:35.000 --> 01:02:36.000 So we have 2 different types of models. The subduction interface, ground motion models. 01:02:36.000 --> 01:02:53.000 And the subduction, interest, lab, ground, motion. The case of the interface models we've used the Nga subduction models of Abrahamson and Gouletars, coonadal and Parker at all. 01:02:53.000 --> 01:03:01.000 We've applied all the additional epistemic uncertainty to these models for the 80, fifth and fifteenth and and Median percentiles. 01:03:01.000 --> 01:03:09.000 In addition, we've also added the Atkinson, Emasius, and Zow models that were used in the previous versions of the national maps for subduction interest. 01:03:09.000 --> 01:03:25.000 Lab, we decided that we would also include the unadjusted and the adjusted version of the Abrahamson and Guler's model, for, to increase the epistemic uncertainty 01:03:25.000 --> 01:03:37.000 This plot shows the spectral accelerations as a function of spectral period for magnitude 9 at a 100 client distance, with a 760 meter per second shear wave velocity. 01:03:37.000 --> 01:03:44.000 This is for interface ground motion models, and we show the range of the different models. 01:03:44.000 --> 01:03:54.000 You can see that generally the 3 Nga. Subduction models span a factor of about 3 to 5, maybe. 01:03:54.000 --> 01:04:02.000 In short periods, as we get to longer periods, they tend to cluster a little more, and we've added these older models. 01:04:02.000 --> 01:04:19.000 That account for additional epistemic uncertainty that some of the models decrease faster than than previous models, such as the Parker at all model, and that will be seen in later in the hazard models that we will show you 01:04:19.000 --> 01:04:20.000 One important contributor to the seismic hazard is the alliatory variability in this case. 01:04:20.000 --> 01:04:32.000 I'm showing that for interface abduction, earthquakes you can see in this plot the standard deviation against spectral period. 01:04:32.000 --> 01:04:40.000 You can see that the Parker and Coon models typically are higher than the other models for periods less than about 5 s. 01:04:40.000 --> 01:04:59.000 For longer periods. The older models seem to be higher than the current models, so there has been a suggestion made to consider the Median ground motion models in the alliatory Sigmas separately and weighted logic tree branches that means that we would mix and match and various median ground 01:04:59.000 --> 01:05:06.000 Muslim models with alternative alliatory Sigmas 01:05:06.000 --> 01:05:10.000 Let's examine now the interest lab ground motion models in this figure. 01:05:10.000 --> 01:05:19.000 I'm showing the Median spectral accelerations against spectral period for an interest lab scenario of magnitude 7 distance, 50 kilometers, and vs. 01:05:19.000 --> 01:05:38.000 30 of 760 meters per second. You'll notice in the plot that the Nga subduction models are typically about a factor of 3 or 4 for periods up to about 1 s, but they decrease significantly for longer periods we've added the Abrahamson and 01:05:38.000 --> 01:05:47.000 Guler's model for the Cascadia slab and the adjusted slab to account for some additional epistemic uncertainty in this long period region 01:05:47.000 --> 01:06:03.000 These, maps show a sensitivity study for interface earthquakes at 0 point 2 1 2 and 5 s spectral accelerations and 2% 50 year ground motions we're looking at ratios of the new proposed interface earthquakes compared to the 2,018 model and we see that 01:06:03.000 --> 01:06:08.000 ground motions have decreased compared to the previous models. 01:06:08.000 --> 01:06:14.000 We look at the the point 2 s, we can see that they decrease by up to 50%. 01:06:14.000 --> 01:06:21.000 If we look at the longer periods, typically it's 20 or 30% differences. 01:06:21.000 --> 01:06:30.000 This map from Brad a card shows the basin models that we considered in 2,018, 2,023 proposed in 23. 01:06:30.000 --> 01:06:34.000 We would like to look at the central valley of California in more detail. 01:06:34.000 --> 01:06:37.000 We want to apply Brad Eggard's new velocity model. 01:06:37.000 --> 01:06:54.000 So Brad and Sean Adi spend some effort looking at amplifications in the central valley to make sure that these amplification models were working correctly, and they showed that in general they do seem to predict how amplifications in the deep basin like they might expect these mats show 01:06:54.000 --> 01:06:57.000 Ratios of the 2,023 proposed national model. 01:06:57.000 --> 01:07:07.000 Do, divided by the 2,018 national seismic hazard model for 5 s spectral acceleration and a 2% and 50 year ground motion level. 01:07:07.000 --> 01:07:11.000 So we've just applied the model that we discussed previously from Brad. 01:07:11.000 --> 01:07:20.000 A guard, also studied by Sean Oddi, and we see in the central valley of California minor changes across the region. 01:07:20.000 --> 01:07:24.000 Bakersfield down by a percent reading up by 2%. 01:07:24.000 --> 01:07:32.000 Sacramento up by 1%. So these changes are not significant compared to the previous model. 01:07:32.000 --> 01:07:37.000 Let's discuss for a second the areas outside of the basins in 2,018. 01:07:37.000 --> 01:07:43.000 We asked the modelers whether we should amplify only in the deep basin, and not deemplify at the basin edge. 01:07:43.000 --> 01:07:53.000 We are concerned about bass net sites, especially Los Angeles, being underestimated by including these new models. 01:07:53.000 --> 01:08:02.000 The modelers recommended that we use the Nj. West 2 models, but for 2018 we do decided to only deemplify or only amplify in the deepest portion of the basins. 01:08:02.000 --> 01:08:06.000 This made a nice, smooth model. We didn't have any edge effects. 01:08:06.000 --> 01:08:09.000 We didn't really want to deemplify. 01:08:09.000 --> 01:08:16.000 At the edges of basins, because we saw in our data high uncertainties, high levels of ground motions. 01:08:16.000 --> 01:08:19.000 We, knew that there were high damage areas. And they're also high standard deviations in this area that we were not considering. 01:08:19.000 --> 01:08:32.000 We had some new work by Buka Nuke and his co-authors that studied these basin edge sediments as well as other terrains, and found that there were a lot of complications in these models. 01:08:32.000 --> 01:08:42.000 So we decided to make a new study to try and understand these models, and how they differ between San Francisco and Los Angeles. 01:08:42.000 --> 01:08:47.000 For this amplification study we took a region shown in the Blue Polygon. 01:08:47.000 --> 01:08:53.000 We separated the stations that had recorded strong motion into 3 different categories. 01:08:53.000 --> 01:09:05.000 Outside the basin shown in blue the basin edge within 2 kilometers of the basin edge shown in red, and within the basin shown in yellow for our study. 01:09:05.000 --> 01:09:11.000 We examine the intra-event residuals, with and without basin effects for a number of different periods. 01:09:11.000 --> 01:09:15.000 Here I show 5 s in 10 s. Spectral acceleration period. 01:09:15.000 --> 01:09:16.000 The null basin effects are shown as light bars for these 3 different Nj. 01:09:16.000 --> 01:09:24.000 Whists, 2 equations, and the dark bars show with basin effects. 01:09:24.000 --> 01:09:44.000 If you consider the Z one and Z. 2.5 equations, you can see these 3 regions, which in this minus 2, represents areas well outside the basin there is negative 2 to 2 which are within 2 kilometers of the basin edge and greater than 2 are well within the basin. 01:09:44.000 --> 01:09:55.000 We know that in the Basin itself we need to amplify ground motions in order to get closer to the 0 site term values. 01:09:55.000 --> 01:10:11.000 When we're near the basin edge, we basically aren't making change at all when we're well outside the basin, we need to de-amplify the great mostions just like the Nj. West. 01:10:11.000 --> 01:10:16.000 2 modelers predicted. So basically, we come to a similar conclusion to the Nj. 01:10:16.000 --> 01:10:23.000 West molars that we need to d amplify, based on the data in San Francisco Bay Area. 01:10:23.000 --> 01:10:28.000 These maps show comparisons. Difference in ratios of the native Nj. 01:10:28.000 --> 01:10:32.000 West, 2 models that include the Z 1.0 and Z 2.5 values compared to our 2018 model. 01:10:32.000 --> 01:10:46.000 That only amplified in the basins themselves. What we see when we make these ratios is that differences can be quite significant from decreases of point 2 g. 01:10:46.000 --> 01:10:56.000 And ratios show that you can have ground motions that are decreased by almost 45% in this region 01:10:56.000 --> 01:11:05.000 Even though our study did verify what the N. G. West, 2 modelers found, which was that we needed to de-amplify in certain places across the Bay area. 01:11:05.000 --> 01:11:09.000 I still have concerns about doing this, and I wanted to discuss these with you in 2,018 we made a polygon. 01:11:09.000 --> 01:11:15.000 That was developed for deep, basin characterizations and code efficiency. 01:11:15.000 --> 01:11:19.000 It wasn't made for assessing amplifications outside the Basin region. 01:11:19.000 --> 01:11:25.000 So we have a polygon that cuts right through the middle of San Francisco when we run the Nj. 01:11:25.000 --> 01:11:27.000 West 2 hazard models with Z 1.0 and Z. 01:11:27.000 --> 01:11:43.000 2.5 we see these severe edge effects from the current polygons, which can be up to 40% so I'm worried about what plopping in a zone might do to the hazard maps which are very regional in scope we need to develop new polygons and decide where the 01:11:43.000 --> 01:11:47.000 Regional data is best, and we have some new data from bread. 01:11:47.000 --> 01:11:56.000 A guard that we can consider for these think the question that we need to ask is, should we use Z 1.0 Z 2.5 models in urban areas? 01:11:56.000 --> 01:12:00.000 This, the appropriate application for a national hazard map? 01:12:00.000 --> 01:12:04.000 Or is this better handled as a site-specific application? 01:12:04.000 --> 01:12:11.000 Third, we found that regional data for San Francisco seems to be regionally consistent, but we also are worried that maybe we're underrepresenting the epistemic uncertainty. 01:12:11.000 --> 01:12:24.000 And I'll give you some examples of that. We know, for example, that we don't have a real strong database of ground motions of large earthquakes in California. 01:12:24.000 --> 01:12:42.000 We haven't seen big ruptures, a recorded earthquake, strong motion from big ruptures along San Andreas, Heyward ruptures, and so we haven't also looked at the simulations that are made simulations that would help us to have insights into this another consideration is that we didn't 01:12:42.000 --> 01:12:48.000 Consider directivity, which can increase hazard by 15%, which I'll show in a second and the third. 01:12:48.000 --> 01:13:09.000 The final question is, should we be just waiting and considering a non-orgotic application, instead of implementing this right now just mentioned, that directivity could increase the hazard across this region by 1520%, this plot shows that this is a plot made by Kyle withers in his studies of directivity you can 01:13:09.000 --> 01:13:14.000 See, the 2% in the 10% ground motion models. 01:13:14.000 --> 01:13:30.000 These are difference in ratio plots. And in general, when you look at details of the San Francisco Bay area, we see 15% increases compared to the the models that we've used in the past 01:13:30.000 --> 01:13:38.000 So let me conclude by saying that the 2,023 National Map model consider several new models for Northern California, Cascadia, Nga. 01:13:38.000 --> 01:13:44.000 Subduction Central Valley amplification and D. Amplification possibilities outside of the Basin regions. 01:13:44.000 --> 01:13:49.000 The Nj. Subduction models generally bring down the hazard by up to 40%. 01:13:49.000 --> 01:13:52.000 The outside base in regions could be reduced by 30% with Z 1.0 0 point 5 implementations. 01:13:52.000 --> 01:14:04.000 This leads to discontinuity at the Polygon edges, changes in the central valley, hazard are quite small when applying the new Z one and Z. 01:14:04.000 --> 01:14:19.000 2.5 Ngos. 2 models, and also the hazard changes from the New Cascadia subduction model that include the turbidite data my cool finger, also quite small, guess one of the big questions that I have that i'd like you to think about is should we consider Z 01:14:19.000 --> 01:14:32.000 1.0 C. 2.5 in our models across the San Francisco Bay Area 01:14:32.000 --> 01:14:35.000 Okay. Thank you. Mark. 01:14:35.000 --> 01:14:39.000 So now we're going to move on to our discussion section. 01:14:39.000 --> 01:14:56.000 So, if you have questions for any of the speakers in the session, I would encourage you to raise your hand and ask them in person 01:14:56.000 --> 01:15:00.000 And will. Everyone's warming up to ask their question. 01:15:00.000 --> 01:15:04.000 Maybe I'll start with one for Mark, if that's okay. 01:15:04.000 --> 01:15:17.000 And it's about your study of oh, there's plenty of hands, but it's about your study of the face and LED effects. 01:15:17.000 --> 01:15:22.000 I'm glad to see that your results were consistent with the Ngos. 01:15:22.000 --> 01:15:23.000 2 models. Did you look at the standard deviations of your site terms? 01:15:23.000 --> 01:15:30.000 And how did those compare to make existing models 01:15:30.000 --> 01:15:33.000 We did it. We looked at the we looked at the intra event. 01:15:33.000 --> 01:15:51.000 Variability, probably same as you did when you did your study, and when we first looked at it, I have to say we did it 2 times, and the first time we did it we saw huge uncertainties right, at the basin edge that so they were much much stronger second time, we did it we 01:15:51.000 --> 01:15:55.000 didn't see it. So we need to try and understand why the 2 different ways we process the data. 01:15:55.000 --> 01:16:02.000 We're different. But we did see large uncertainties, and I think, etc. periods. 01:16:02.000 --> 01:16:19.000 You can see large uncertainties at the basin edge which concerns me because you can see that some of the highest ground motions we recorded were near the basin edge, and some of the lowest, and it makes sense that you'd have this real complication and differences at that base net so I 01:16:19.000 --> 01:16:35.000 think a lot of people are are in agreement that that might be a place of concern, and I think it might be a natural way for us to consider in the future either that uncertainty in the modeling, or I think maybe nonergotic applications may actually give that to us too. 01:16:35.000 --> 01:16:39.000 Right, my follow-on is going to be. Could you use that uncertainty to maybe guide? 01:16:39.000 --> 01:16:40.000 Write a logic tree approach which sounds like yes. 01:16:40.000 --> 01:16:44.000 Yeah, right? I think it's a good idea. 01:16:44.000 --> 01:16:45.000 Okay. 01:16:45.000 --> 01:16:52.000 Thank you very much, Mark. So next, Don, please go ahead. 01:16:52.000 --> 01:16:55.000 Thanks a question for Alex Alex. It looks like the the West. 01:16:55.000 --> 01:17:02.000 Tracy follows, is included in in the new model, and I think I saw the Midway fault as well. 01:17:02.000 --> 01:17:05.000 I'm wondering what other delta faults may have been included. 01:17:05.000 --> 01:17:14.000 For example, like was is the Midland in the Bronalis, also included in the 2,020 01:17:14.000 --> 01:17:15.000 Yeah. Good. Question, done? So those faults. The West Tracy faults, and Midway are added. 01:17:15.000 --> 01:17:31.000 You're right. We worked with Steve Delong quite a bit, and as well as Judy Zaccharis, and then Tim Dawson, to get those faults in lot of that data for those come from the dreams. 01:17:31.000 --> 01:17:40.000 Report, if I remember correctly, in terms of other Delta faults that are added. I'm sorry. Which faults did you say 01:17:40.000 --> 01:17:43.000 The Midland and Vernalis 01:17:43.000 --> 01:17:49.000 I think Midland was already in. Am I wrong and saying that? 01:17:49.000 --> 01:17:50.000 It might have been 01:17:50.000 --> 01:17:57.000 And Brunell is not in. Okay. I can look quickly and and type to you in the checks 01:17:57.000 --> 01:18:00.000 Thanks. 01:18:00.000 --> 01:18:03.000 Thanks to Don and and Alex Buka, you're next. 01:18:03.000 --> 01:18:07.000 If you want to ask a question 01:18:07.000 --> 01:18:12.000 Yeah, how you doing? Thanks, Grace. This question is for for more game. 01:18:12.000 --> 01:18:16.000 Mark, how are you doing? I want to ask about the the basic edge study is, what is the echo? 01:18:16.000 --> 01:18:28.000 What Grace was saying when you, when you study it and use you essentially went through, and you get had, like a graded level of base and edges where it's like, Okay, you know, you go 1010 kilometers away 20 kilometers away. 01:18:28.000 --> 01:18:45.000 And more. Did you see any difference in those in those different bins as you get got close and closer to, let's say on, quote the send up the base of the main basin, and if you did see any difference well, it's substantial enough to wear I I guess maybe probably was because you said you're concerned 01:18:45.000 --> 01:18:51.000 But wha wha what was it you saw when you looked at the difference in grades between those those different distances 01:18:51.000 --> 01:18:58.000 So excellent question, Buka, and we we we made many different categories. 01:18:58.000 --> 01:19:12.000 I showed you just the 3 category plot that showed, you know, within 2 kilometers of the basin edge, and the more distant distant distances away from the basin and within the basin, and we saw a lot of complexity. 01:19:12.000 --> 01:19:17.000 I think I think when you look, yeah, a lot of different. 01:19:17.000 --> 01:19:26.000 Bins you get a lot of uncertainty because the because you get much you don't have as much data to analyze. 01:19:26.000 --> 01:19:30.000 And so it's hard to to really make a lot of judgments. 01:19:30.000 --> 01:19:36.000 I like just the 3 bidding kind of thing, but what you're saying is, if you look at a lot of detail, can you see that? 01:19:36.000 --> 01:19:41.000 And we could see some more detail than I showed in those plots, and I think it's something to look at. 01:19:41.000 --> 01:19:49.000 I you know there are some times when you see them high and low. And again, I think it has to do with data, quality and data availability. 01:19:49.000 --> 01:19:53.000 And those bin so I'm not sure exactly what to make of it. 01:19:53.000 --> 01:20:00.000 But sometime you and I should take a look at those and see if we see more complexity that we should talk about 01:20:00.000 --> 01:20:04.000 Definitely, I definitely agree. I'll I'll I'll reach out after you know. 01:20:04.000 --> 01:20:05.000 After the the video at the conference, and we'll I'll return thanks. 01:20:05.000 --> 01:20:07.000 Right. Thanks. Thank you. 01:20:07.000 --> 01:20:09.000 Mark. 01:20:09.000 --> 01:20:15.000 Thanks, Mark Keith, please take it away. 01:20:15.000 --> 01:20:24.000 Well, first I'd like to start by congratulating it, congratulating everyone on the you of work that went into this and the and the really great presentations. 01:20:24.000 --> 01:20:33.000 Of course the Board thanks everybody but my question right now is for Mark, and I'm I'm wondering if there are any plans and it's not fair to ask this, because you're still trying to finish. 01:20:33.000 --> 01:20:40.000 But in in thinking about, where should we do research to try to improve things in the future? 01:20:40.000 --> 01:20:52.000 Are you? Are you going to somehow try to figure out what what is driving the uncertainty in different areas like, you know? 01:20:52.000 --> 01:21:01.000 Maybe I of course I have no idea how much work this would take, but create a grid map with tornado diagrams or something like that. 01:21:01.000 --> 01:21:10.000 Is there a way for for people to think about? Where can we get the most improvement by focusing on different aspects 01:21:10.000 --> 01:21:14.000 No, a excellent question. Keys, and we we we obviously have. 01:21:14.000 --> 01:21:21.000 I mean in some ways we can see, for example, characterized Cascadia. 01:21:21.000 --> 01:21:30.000 We don't see a lot of change. Well, in the in the source model, and yet we know certainty and Cascadia Source model. 01:21:30.000 --> 01:21:32.000 So I'm not sure you know, we're just looking at some minor tweaks and things. 01:21:32.000 --> 01:21:43.000 I think obviously, there's we just had a meeting last week on the seduction science, and there are all kinds of things for us to do to improve that. 01:21:43.000 --> 01:21:52.000 I think if we look at the, at, at the shallow basins, I I you can see that the shallow basins. 01:21:52.000 --> 01:21:58.000 I I didn't show the Reno and the Las Vegas and Salt Lake Basins, but some of the models for Z. 01:21:58.000 --> 01:22:07.000 One, and Z. 2.5 for shallower type basins don't show any amplification, and yet we've heard over and over from John Louis and others that you see a lot of amplification in the actual records. 01:22:07.000 --> 01:22:14.000 We're just not seeing them in the the models. 01:22:14.000 --> 01:22:19.000 The Nga. West to vs. 30 amplification models, and and and the Z. 1 point on Z. 01:22:19.000 --> 01:22:30.000 2.5