WEBVTT 436caf55-234f-4249-83d8-d9c6c61c7b8f/48-0 00:00:09.697 --> 00:00:10.357 Welcome to the Earthquake Science Center Seminar for Nov. 16. Um. And. 436caf55-234f-4249-83d8-d9c6c61c7b8f/131-0 00:00:11.727 --> 00:00:15.661 Today we have Justin Buckreis from UCLA who's gonna talk about 436caf55-234f-4249-83d8-d9c6c61c7b8f/131-1 00:00:15.661 --> 00:00:19.215 subregional anelastic path effects in California. But 436caf55-234f-4249-83d8-d9c6c61c7b8f/131-2 00:00:19.215 --> 00:00:22.388 before I hand it over to Grace Parker to give the 436caf55-234f-4249-83d8-d9c6c61c7b8f/131-3 00:00:22.388 --> 00:00:26.005 introductions, there are just a couple of announcements. 436caf55-234f-4249-83d8-d9c6c61c7b8f/131-4 00:00:26.005 --> 00:00:29.940 November 18th is the deadline to schedule your user loose and 436caf55-234f-4249-83d8-d9c6c61c7b8f/131-5 00:00:29.940 --> 00:00:33.684 addition. We're soliciting volunteers for the December 7th 436caf55-234f-4249-83d8-d9c6c61c7b8f/131-6 00:00:33.684 --> 00:00:37.364 Earthquake Science Center seminar, which will be probably 436caf55-234f-4249-83d8-d9c6c61c7b8f/131-7 00:00:37.364 --> 00:00:41.299 around 3 practice age. You talk. So if you're going to AG and 436caf55-234f-4249-83d8-d9c6c61c7b8f/131-8 00:00:41.299 --> 00:00:43.457 would like to practice your talk. 436caf55-234f-4249-83d8-d9c6c61c7b8f/166-0 00:00:43.777 --> 00:00:49.582 Please reach out to me and Evan. November 30th is the earthquake 436caf55-234f-4249-83d8-d9c6c61c7b8f/166-1 00:00:49.582 --> 00:00:54.942 Country Alliance meeting at the Aviation Museum in SFO. The 436caf55-234f-4249-83d8-d9c6c61c7b8f/166-2 00:00:54.942 --> 00:01:00.301 airport where there's gonna be a quake break given by yours 436caf55-234f-4249-83d8-d9c6c61c7b8f/166-3 00:01:00.301 --> 00:01:00.837 truly. 436caf55-234f-4249-83d8-d9c6c61c7b8f/243-0 00:01:02.337 --> 00:01:06.914 The 23rd of November at midnight is the sign up deadline for the 436caf55-234f-4249-83d8-d9c6c61c7b8f/243-1 00:01:06.914 --> 00:01:11.279 January 10th to 11th subduction Zone Science Workshop. Please 436caf55-234f-4249-83d8-d9c6c61c7b8f/243-2 00:01:11.279 --> 00:01:15.293 see the e-mail from Keith Nadson for more details. As of 436caf55-234f-4249-83d8-d9c6c61c7b8f/243-3 00:01:15.293 --> 00:01:18.955 yesterday, you should have upgraded all of your iOS 436caf55-234f-4249-83d8-d9c6c61c7b8f/243-4 00:01:18.955 --> 00:01:22.686 devices. And finally, on December 8th there's an all 436caf55-234f-4249-83d8-d9c6c61c7b8f/243-5 00:01:22.686 --> 00:01:26.982 hands virtual from 11:00 to 12:30 where you can meet our new 436caf55-234f-4249-83d8-d9c6c61c7b8f/243-6 00:01:26.982 --> 00:01:30.925 earthquake Science Center director. And without further 436caf55-234f-4249-83d8-d9c6c61c7b8f/243-7 00:01:30.925 --> 00:01:31.207 ado. 436caf55-234f-4249-83d8-d9c6c61c7b8f/271-0 00:01:31.337 --> 00:01:35.262 And please make sure to turn off your cameras and microphones 436caf55-234f-4249-83d8-d9c6c61c7b8f/271-1 00:01:35.262 --> 00:01:39.187 while the speaker is talking. I'll pass it off to Grace to do 436caf55-234f-4249-83d8-d9c6c61c7b8f/271-2 00:01:39.187 --> 00:01:40.327 the introductions. 436caf55-234f-4249-83d8-d9c6c61c7b8f/288-0 00:01:43.067 --> 00:01:47.558 Thanks. It's my pleasure to introduce Tristan Buckreis as 436caf55-234f-4249-83d8-d9c6c61c7b8f/288-1 00:01:47.558 --> 00:01:49.417 today's seminar speaker. 436caf55-234f-4249-83d8-d9c6c61c7b8f/360-0 00:01:50.607 --> 00:01:55.031 Tristan has a bachelor's degree from CSU Long Beach where he 436caf55-234f-4249-83d8-d9c6c61c7b8f/360-1 00:01:55.031 --> 00:01:58.948 completed an honors thesis on a numerical modeling of 436caf55-234f-4249-83d8-d9c6c61c7b8f/360-2 00:01:58.948 --> 00:02:03.300 foundations with Doctor Lisa Star, and he's currently a PhD 436caf55-234f-4249-83d8-d9c6c61c7b8f/360-3 00:02:03.300 --> 00:02:07.725 candidate at UCLA in the Civil and Environmental Engineering 436caf55-234f-4249-83d8-d9c6c61c7b8f/360-4 00:02:07.725 --> 00:02:12.222 Department, where he works on regional ground motion modeling 436caf55-234f-4249-83d8-d9c6c61c7b8f/360-5 00:02:12.222 --> 00:02:16.720 with special attention to soft organic soffits, and you'll be 436caf55-234f-4249-83d8-d9c6c61c7b8f/360-6 00:02:16.720 --> 00:02:20.637 speaking on that topic today. So I will hand it over. 436caf55-234f-4249-83d8-d9c6c61c7b8f/366-0 00:02:21.257 --> 00:02:23.187 Tell him that we can get the seminar started. 436caf55-234f-4249-83d8-d9c6c61c7b8f/394-0 00:02:24.997 --> 00:02:27.960 Thank you, Grace for the introduction. And I'd also like 436caf55-234f-4249-83d8-d9c6c61c7b8f/394-1 00:02:27.960 --> 00:02:31.287 to thank the USGS for inviting me to to give the talk today and 436caf55-234f-4249-83d8-d9c6c61c7b8f/394-2 00:02:31.287 --> 00:02:33.991 for providing me with the opportunity to share this 436caf55-234f-4249-83d8-d9c6c61c7b8f/394-3 00:02:33.991 --> 00:02:35.187 research with everyone. 436caf55-234f-4249-83d8-d9c6c61c7b8f/453-0 00:02:36.047 --> 00:02:39.280 I'd also like to so today I'm going to be discussing 436caf55-234f-4249-83d8-d9c6c61c7b8f/453-1 00:02:39.280 --> 00:02:42.879 subregional analytic path effects in California. This is a 436caf55-234f-4249-83d8-d9c6c61c7b8f/453-2 00:02:42.879 --> 00:02:46.356 work that has been done by myself and my coauthors think 436caf55-234f-4249-83d8-d9c6c61c7b8f/453-3 00:02:46.356 --> 00:02:49.711 they Wong. Who is that Old Dominion University who was 436caf55-234f-4249-83d8-d9c6c61c7b8f/453-4 00:02:49.711 --> 00:02:53.249 formerly a PhD student and postdoc at UCLA, and Professor 436caf55-234f-4249-83d8-d9c6c61c7b8f/453-5 00:02:53.249 --> 00:02:56.177 Scott Brandenburg and Jonathan Stewart at UCLA. 436caf55-234f-4249-83d8-d9c6c61c7b8f/494-0 00:03:00.937 --> 00:03:04.487 Now also like to acknowledge that this research was sponsored 436caf55-234f-4249-83d8-d9c6c61c7b8f/494-1 00:03:04.487 --> 00:03:07.694 by the California Department of Water Resources and any 436caf55-234f-4249-83d8-d9c6c61c7b8f/494-2 00:03:07.694 --> 00:03:11.245 opinions, findings, conclusions and recommendations expressed 436caf55-234f-4249-83d8-d9c6c61c7b8f/494-3 00:03:11.245 --> 00:03:14.739 here are those of myself that do not necessarily reflect the 436caf55-234f-4249-83d8-d9c6c61c7b8f/494-4 00:03:14.739 --> 00:03:15.827 views of the dwarf. 436caf55-234f-4249-83d8-d9c6c61c7b8f/551-0 00:03:18.087 --> 00:03:21.277 First, an outline of what I have prepared for today. I always 436caf55-234f-4249-83d8-d9c6c61c7b8f/551-1 00:03:21.277 --> 00:03:24.468 like to begin presentations with the review of some basics so 436caf55-234f-4249-83d8-d9c6c61c7b8f/551-2 00:03:24.468 --> 00:03:27.350 that we all have a good understanding of the principles 436caf55-234f-4249-83d8-d9c6c61c7b8f/551-3 00:03:27.350 --> 00:03:30.438 that we're gonna be, that I'm going to be discussing today. 436caf55-234f-4249-83d8-d9c6c61c7b8f/551-4 00:03:30.438 --> 00:03:33.269 Then I have a bit more background on what path effects 436caf55-234f-4249-83d8-d9c6c61c7b8f/551-5 00:03:33.269 --> 00:03:34.607 are and how we model them. 436caf55-234f-4249-83d8-d9c6c61c7b8f/593-0 00:03:35.347 --> 00:03:38.689 Which will then lead into a brief discussion of the ground 436caf55-234f-4249-83d8-d9c6c61c7b8f/593-1 00:03:38.689 --> 00:03:42.202 motion data that we used during our model development, how we 436caf55-234f-4249-83d8-d9c6c61c7b8f/593-2 00:03:42.202 --> 00:03:45.488 chose to sub regionalized California and I'll define what 436caf55-234f-4249-83d8-d9c6c61c7b8f/593-3 00:03:45.488 --> 00:03:48.774 sub realization is in the context of this presentation in 436caf55-234f-4249-83d8-d9c6c61c7b8f/593-4 00:03:48.774 --> 00:03:50.077 the background section. 436caf55-234f-4249-83d8-d9c6c61c7b8f/604-0 00:03:50.757 --> 00:03:53.827 And then I will present the model and how it performs. 436caf55-234f-4249-83d8-d9c6c61c7b8f/624-0 00:03:54.487 --> 00:03:57.790 And then I'll end with some conclusions regarding general 436caf55-234f-4249-83d8-d9c6c61c7b8f/624-1 00:03:57.790 --> 00:04:01.151 modeling, a path effects and then also regarding the Model 436caf55-234f-4249-83d8-d9c6c61c7b8f/624-2 00:04:01.151 --> 00:04:02.347 tab that we produced. 436caf55-234f-4249-83d8-d9c6c61c7b8f/670-0 00:04:03.207 --> 00:04:06.593 So beginning first with the review of some basics first, I 436caf55-234f-4249-83d8-d9c6c61c7b8f/670-1 00:04:06.593 --> 00:04:10.209 have some definitions and I'd like to apologize to any purists 436caf55-234f-4249-83d8-d9c6c61c7b8f/670-2 00:04:10.209 --> 00:04:13.710 in the audience today. I use these definitions to I might be 436caf55-234f-4249-83d8-d9c6c61c7b8f/670-3 00:04:13.710 --> 00:04:16.924 a little lenient in how I use some of these definitions 436caf55-234f-4249-83d8-d9c6c61c7b8f/670-4 00:04:16.924 --> 00:04:20.310 because I'm trying to convey messages and important topics 436caf55-234f-4249-83d8-d9c6c61c7b8f/670-5 00:04:20.310 --> 00:04:22.147 rather than textbook definition. 436caf55-234f-4249-83d8-d9c6c61c7b8f/726-0 00:04:22.827 --> 00:04:25.539 That being said, I want to first start with the textbook 436caf55-234f-4249-83d8-d9c6c61c7b8f/726-1 00:04:25.539 --> 00:04:28.537 definition of what ergodic is, so it's relating to or denoting 436caf55-234f-4249-83d8-d9c6c61c7b8f/726-2 00:04:28.537 --> 00:04:31.440 systems or processes with the property that given sufficient 436caf55-234f-4249-83d8-d9c6c61c7b8f/726-3 00:04:31.440 --> 00:04:34.485 time they include our impinge on all points and the given space 436caf55-234f-4249-83d8-d9c6c61c7b8f/726-4 00:04:34.485 --> 00:04:37.007 and can be represented statistically by a reasonably 436caf55-234f-4249-83d8-d9c6c61c7b8f/726-5 00:04:37.007 --> 00:04:38.197 large selection of point. 436caf55-234f-4249-83d8-d9c6c61c7b8f/809-0 00:04:38.807 --> 00:04:42.297 Bit of a mouthful, but it's best easily understood by two other 436caf55-234f-4249-83d8-d9c6c61c7b8f/809-1 00:04:42.297 --> 00:04:45.570 definitions, the first being the time average, which is the 436caf55-234f-4249-83d8-d9c6c61c7b8f/809-2 00:04:45.570 --> 00:04:49.006 average of many observations per particular system over a long 436caf55-234f-4249-83d8-d9c6c61c7b8f/809-3 00:04:49.006 --> 00:04:52.005 period of time, and then ensemble average which is the 436caf55-234f-4249-83d8-d9c6c61c7b8f/809-4 00:04:52.005 --> 00:04:55.387 average of many observations for many systems at a particular 436caf55-234f-4249-83d8-d9c6c61c7b8f/809-5 00:04:55.387 --> 00:04:58.877 time. So essentially, how do we assess or audacity is are these 436caf55-234f-4249-83d8-d9c6c61c7b8f/809-6 00:04:58.877 --> 00:05:01.659 two averages equal to each other? If they are then 436caf55-234f-4249-83d8-d9c6c61c7b8f/809-7 00:05:01.659 --> 00:05:05.095 expectation values are averages can be assumed to be reflected 436caf55-234f-4249-83d8-d9c6c61c7b8f/809-8 00:05:05.095 --> 00:05:06.077 of temporal novel. 436caf55-234f-4249-83d8-d9c6c61c7b8f/827-0 00:05:06.817 --> 00:05:09.841 So to tie this into ground motion modeling, which is 436caf55-234f-4249-83d8-d9c6c61c7b8f/827-1 00:05:09.841 --> 00:05:11.667 subject of today's presentation. 436caf55-234f-4249-83d8-d9c6c61c7b8f/881-0 00:05:12.767 --> 00:05:16.044 We have the ergodic assumption that we use which is that ground 436caf55-234f-4249-83d8-d9c6c61c7b8f/881-1 00:05:16.044 --> 00:05:19.065 motion distribution at a site over time is the same as the 436caf55-234f-4249-83d8-d9c6c61c7b8f/881-2 00:05:19.065 --> 00:05:22.292 ground motion distribution over space. So I will have a little 436caf55-234f-4249-83d8-d9c6c61c7b8f/881-3 00:05:22.292 --> 00:05:25.415 bit more to say on this on our future slide, but essentially 436caf55-234f-4249-83d8-d9c6c61c7b8f/881-4 00:05:25.415 --> 00:05:28.539 under this assumption we can develop models that predict the 436caf55-234f-4249-83d8-d9c6c61c7b8f/881-5 00:05:28.539 --> 00:05:31.817 average at one location by using the average of many locations. 436caf55-234f-4249-83d8-d9c6c61c7b8f/897-0 00:05:32.617 --> 00:05:35.873 So an example to make it clear what innergetic process is 436caf55-234f-4249-83d8-d9c6c61c7b8f/897-1 00:05:35.873 --> 00:05:37.557 before going into nonnarcotic. 436caf55-234f-4249-83d8-d9c6c61c7b8f/915-0 00:05:38.697 --> 00:05:41.540 An example is a coin toss. Assume we have a fair coin, so 436caf55-234f-4249-83d8-d9c6c61c7b8f/915-1 00:05:41.540 --> 00:05:43.747 equal probability of landing on either side. 436caf55-234f-4249-83d8-d9c6c61c7b8f/955-0 00:05:44.417 --> 00:05:47.707 Hold on one side, blue on the other. We have five people. We 436caf55-234f-4249-83d8-d9c6c61c7b8f/955-1 00:05:47.707 --> 00:05:51.159 ask him to flip a coin. We we'd expect a probability of .5, but 436caf55-234f-4249-83d8-d9c6c61c7b8f/955-2 00:05:51.159 --> 00:05:54.611 that realization is impossible given the data we have available 436caf55-234f-4249-83d8-d9c6c61c7b8f/955-3 00:05:54.611 --> 00:05:58.117 to us. We'll get something maybe like .6. Now we can repeat this 436caf55-234f-4249-83d8-d9c6c61c7b8f/955-4 00:05:58.117 --> 00:06:00.167 exercise many times for these people. 436caf55-234f-4249-83d8-d9c6c61c7b8f/1022-0 00:06:00.987 --> 00:06:04.215 And we can see we can get many different probabilities. We can 436caf55-234f-4249-83d8-d9c6c61c7b8f/1022-1 00:06:04.215 --> 00:06:07.391 look at a time average, which is just the probability for one 436caf55-234f-4249-83d8-d9c6c61c7b8f/1022-2 00:06:07.391 --> 00:06:10.209 person or an ensemble average, which is looking at the 436caf55-234f-4249-83d8-d9c6c61c7b8f/1022-3 00:06:10.209 --> 00:06:13.130 probability of all people for one particular trial. Now, 436caf55-234f-4249-83d8-d9c6c61c7b8f/1022-4 00:06:13.130 --> 00:06:16.512 because these are small samples, they might not always equal each 436caf55-234f-4249-83d8-d9c6c61c7b8f/1022-5 00:06:16.512 --> 00:06:19.842 other, but given enough trials, in this case 200, you can see at 436caf55-234f-4249-83d8-d9c6c61c7b8f/1022-6 00:06:19.842 --> 00:06:21.277 the very end, if we look at. 436caf55-234f-4249-83d8-d9c6c61c7b8f/1073-0 00:06:22.327 --> 00:06:26.124 After 200 trials, all the probabilities of each individual 436caf55-234f-4249-83d8-d9c6c61c7b8f/1073-1 00:06:26.124 --> 00:06:30.179 person, which is represented by the lines pretty much converge 436caf55-234f-4249-83d8-d9c6c61c7b8f/1073-2 00:06:30.179 --> 00:06:34.041 to .5 and trials of ensemble sizes of 200 also are around 5 436caf55-234f-4249-83d8-d9c6c61c7b8f/1073-3 00:06:34.041 --> 00:06:38.161 most of the time, so this can be conveyed as an our interpreted 436caf55-234f-4249-83d8-d9c6c61c7b8f/1073-4 00:06:38.161 --> 00:06:39.577 as an ergodic process. 436caf55-234f-4249-83d8-d9c6c61c7b8f/1149-0 00:06:40.487 --> 00:06:43.485 Now at ground motion is an honor. God process. Easiest way 436caf55-234f-4249-83d8-d9c6c61c7b8f/1149-1 00:06:43.485 --> 00:06:46.687 to look at this is if we have five people that do not move and 436caf55-234f-4249-83d8-d9c6c61c7b8f/1149-2 00:06:46.687 --> 00:06:49.126 we have two different earthquakes and different 436caf55-234f-4249-83d8-d9c6c61c7b8f/1149-3 00:06:49.126 --> 00:06:52.226 locations around motions. Are the amplitudes of shaking that 436caf55-234f-4249-83d8-d9c6c61c7b8f/1149-4 00:06:52.226 --> 00:06:55.377 are intensity that they fill for. These earthquakes are going 436caf55-234f-4249-83d8-d9c6c61c7b8f/1149-5 00:06:55.377 --> 00:06:58.527 to be different each time. For each event. We understand that 436caf55-234f-4249-83d8-d9c6c61c7b8f/1149-6 00:06:58.527 --> 00:07:01.678 there's a lot of phenomena that influence ground motions that 436caf55-234f-4249-83d8-d9c6c61c7b8f/1149-7 00:07:01.678 --> 00:07:04.829 could be source effects, path effects and side effects, which 436caf55-234f-4249-83d8-d9c6c61c7b8f/1149-8 00:07:04.829 --> 00:07:07.878 I'll discuss in the next slide. But if we account for those 436caf55-234f-4249-83d8-d9c6c61c7b8f/1149-9 00:07:07.878 --> 00:07:09.657 effects in our model, then we can. 436caf55-234f-4249-83d8-d9c6c61c7b8f/1158-0 00:07:10.417 --> 00:07:12.967 Use their gotta assumption to predict. 436caf55-234f-4249-83d8-d9c6c61c7b8f/1160-0 00:07:13.647 --> 00:07:13.997 Uh. 436caf55-234f-4249-83d8-d9c6c61c7b8f/1166-0 00:07:14.757 --> 00:07:16.397 Intensity at particular sites. 436caf55-234f-4249-83d8-d9c6c61c7b8f/1240-0 00:07:17.827 --> 00:07:21.244 So in a ground motion model Now this is a kind of cartoon that 436caf55-234f-4249-83d8-d9c6c61c7b8f/1240-1 00:07:21.244 --> 00:07:24.389 most people are gonna be familiar with, but we use series 436caf55-234f-4249-83d8-d9c6c61c7b8f/1240-2 00:07:24.389 --> 00:07:27.807 of equations to to model these effects. So assuming we have an 436caf55-234f-4249-83d8-d9c6c61c7b8f/1240-3 00:07:27.807 --> 00:07:31.007 earthquake, it's going to produce some level of shaking at 436caf55-234f-4249-83d8-d9c6c61c7b8f/1240-4 00:07:31.007 --> 00:07:34.207 the site as we can see by the wavelet shown down here. And 436caf55-234f-4249-83d8-d9c6c61c7b8f/1240-5 00:07:34.207 --> 00:07:37.678 these are quantified in source models, Fe, which is shown here, 436caf55-234f-4249-83d8-d9c6c61c7b8f/1240-6 00:07:37.678 --> 00:07:40.986 which use source parameters like magnitude and mechanisms to 436caf55-234f-4249-83d8-d9c6c61c7b8f/1240-7 00:07:40.986 --> 00:07:42.397 predict the level shaking. 436caf55-234f-4249-83d8-d9c6c61c7b8f/1289-0 00:07:43.687 --> 00:07:46.438 The energy has to get transferred to a site through 436caf55-234f-4249-83d8-d9c6c61c7b8f/1289-1 00:07:46.438 --> 00:07:49.825 wave propagation, which can also be referred to as attenuation, 436caf55-234f-4249-83d8-d9c6c61c7b8f/1289-2 00:07:49.825 --> 00:07:53.106 which will always reduce the energy of the wave, which we can 436caf55-234f-4249-83d8-d9c6c61c7b8f/1289-3 00:07:53.106 --> 00:07:56.175 see is gonna have a lower intensity. This is the focus of 436caf55-234f-4249-83d8-d9c6c61c7b8f/1289-4 00:07:56.175 --> 00:07:59.456 today's presentation, so I'll be discussing this in detail in 436caf55-234f-4249-83d8-d9c6c61c7b8f/1289-5 00:07:59.456 --> 00:08:00.197 future slides. 436caf55-234f-4249-83d8-d9c6c61c7b8f/1308-0 00:08:00.887 --> 00:08:04.717 And then we actually have to get to the site, which is can can be 436caf55-234f-4249-83d8-d9c6c61c7b8f/1308-1 00:08:04.717 --> 00:08:07.677 quite complicated, but there's a lot of work done. 436caf55-234f-4249-83d8-d9c6c61c7b8f/1318-0 00:08:08.367 --> 00:08:12.247 Over the past couple decades or so, looking at site response. 436caf55-234f-4249-83d8-d9c6c61c7b8f/1335-0 00:08:12.887 --> 00:08:16.945 And the factors that influence it and so we have a fairly good 436caf55-234f-4249-83d8-d9c6c61c7b8f/1335-1 00:08:16.945 --> 00:08:18.427 grasp on site response. 436caf55-234f-4249-83d8-d9c6c61c7b8f/1338-0 00:08:19.977 --> 00:08:20.957 Now connecting. 436caf55-234f-4249-83d8-d9c6c61c7b8f/1374-0 00:08:21.577 --> 00:08:24.664 Or Goddess city and ground motion models just recalling or 436caf55-234f-4249-83d8-d9c6c61c7b8f/1374-1 00:08:24.664 --> 00:08:27.804 Gothic is when the time average equals the ensemble average 436caf55-234f-4249-83d8-d9c6c61c7b8f/1374-2 00:08:27.804 --> 00:08:30.943 underlying process of ground motions is non ergodic as seen 436caf55-234f-4249-83d8-d9c6c61c7b8f/1374-3 00:08:30.943 --> 00:08:32.147 by spatial variability. 436caf55-234f-4249-83d8-d9c6c61c7b8f/1391-0 00:08:33.287 --> 00:08:37.237 However, energetic model would be one that's developed under 436caf55-234f-4249-83d8-d9c6c61c7b8f/1391-1 00:08:37.237 --> 00:08:38.727 the ergodic assumption. 436caf55-234f-4249-83d8-d9c6c61c7b8f/1430-0 00:08:39.457 --> 00:08:43.320 And non ergodic model attempts to transform the underlying non 436caf55-234f-4249-83d8-d9c6c61c7b8f/1430-1 00:08:43.320 --> 00:08:46.999 ergodic process into an ergodic process. In other words, we 436caf55-234f-4249-83d8-d9c6c61c7b8f/1430-2 00:08:46.999 --> 00:08:50.739 wanna remove all spatial trends in the residuals but follows 436caf55-234f-4249-83d8-d9c6c61c7b8f/1430-3 00:08:50.739 --> 00:08:52.947 that ground Russian models I would. 436caf55-234f-4249-83d8-d9c6c61c7b8f/1459-0 00:08:53.677 --> 00:08:57.337 Proposed fall on a spectrum of either being or Gothic, which 436caf55-234f-4249-83d8-d9c6c61c7b8f/1459-1 00:08:57.337 --> 00:09:00.817 would be to the left in this case, or none ergodic, which 436caf55-234f-4249-83d8-d9c6c61c7b8f/1459-2 00:09:00.817 --> 00:09:02.977 would be on the right in this case. 436caf55-234f-4249-83d8-d9c6c61c7b8f/1507-0 00:09:03.577 --> 00:09:06.971 And a way to think of that would be nonnarcotic side effects. 436caf55-234f-4249-83d8-d9c6c61c7b8f/1507-1 00:09:06.971 --> 00:09:10.200 This is research that has is fairly mature. We have pretty 436caf55-234f-4249-83d8-d9c6c61c7b8f/1507-2 00:09:10.200 --> 00:09:13.321 good grasp, we understand it where if we have sufficient 436caf55-234f-4249-83d8-d9c6c61c7b8f/1507-3 00:09:13.321 --> 00:09:16.331 number of observations as represented by each of these 436caf55-234f-4249-83d8-d9c6c61c7b8f/1507-4 00:09:16.331 --> 00:09:19.671 dots in residual versus period space, we can pretty much get 436caf55-234f-4249-83d8-d9c6c61c7b8f/1507-5 00:09:19.671 --> 00:09:21.587 the average response at that site. 436caf55-234f-4249-83d8-d9c6c61c7b8f/1579-0 00:09:22.267 --> 00:09:25.002 And then we could compute an amplification that's 436caf55-234f-4249-83d8-d9c6c61c7b8f/1579-1 00:09:25.002 --> 00:09:28.340 representative of this average response and compare it to an 436caf55-234f-4249-83d8-d9c6c61c7b8f/1579-2 00:09:28.340 --> 00:09:31.896 ergodic mob. In this particular case, the amplification is a bit 436caf55-234f-4249-83d8-d9c6c61c7b8f/1579-3 00:09:31.896 --> 00:09:35.014 higher than what an ergodic model would suggest, and the 436caf55-234f-4249-83d8-d9c6c61c7b8f/1579-4 00:09:35.014 --> 00:09:38.352 general shape is also a bit different, and the main benefits 436caf55-234f-4249-83d8-d9c6c61c7b8f/1579-5 00:09:38.352 --> 00:09:41.689 of nonnarcotic models are that we improve the predictions so 436caf55-234f-4249-83d8-d9c6c61c7b8f/1579-6 00:09:41.689 --> 00:09:45.136 we're less biased and we reduced the auditory variability with 436caf55-234f-4249-83d8-d9c6c61c7b8f/1579-7 00:09:45.136 --> 00:09:48.036 respect to side effects by removing the site to site 436caf55-234f-4249-83d8-d9c6c61c7b8f/1579-8 00:09:48.036 --> 00:09:49.787 variability now on the spectrum. 436caf55-234f-4249-83d8-d9c6c61c7b8f/1615-0 00:09:50.487 --> 00:09:53.928 Where what influences how non ergodic insight response model 436caf55-234f-4249-83d8-d9c6c61c7b8f/1615-1 00:09:53.928 --> 00:09:57.088 is? Well fundamentally the number of aspirations if you 436caf55-234f-4249-83d8-d9c6c61c7b8f/1615-2 00:09:57.088 --> 00:10:00.361 have no observations, then you're gonna be fully ergodic. 436caf55-234f-4249-83d8-d9c6c61c7b8f/1615-3 00:10:00.361 --> 00:10:03.746 If you have 100 observations, fully non ergodic or close to 436caf55-234f-4249-83d8-d9c6c61c7b8f/1615-4 00:10:03.746 --> 00:10:05.157 nonnarcotic I should say. 436caf55-234f-4249-83d8-d9c6c61c7b8f/1643-0 00:10:06.097 --> 00:10:09.023 That with path effects it's a bit more complicated because 436caf55-234f-4249-83d8-d9c6c61c7b8f/1643-1 00:10:09.023 --> 00:10:12.049 it's not just a point. Now we're looking at a coupled system 436caf55-234f-4249-83d8-d9c6c61c7b8f/1643-2 00:10:12.049 --> 00:10:15.223 between a source and a site. So how do we what are they and how 436caf55-234f-4249-83d8-d9c6c61c7b8f/1643-3 00:10:15.223 --> 00:10:16.067 do we model them? 436caf55-234f-4249-83d8-d9c6c61c7b8f/1722-0 00:10:17.187 --> 00:10:20.741 So there are two phenomena that kind of control path affects. 436caf55-234f-4249-83d8-d9c6c61c7b8f/1722-1 00:10:20.741 --> 00:10:23.836 The first is geometric spreading, which is related to 436caf55-234f-4249-83d8-d9c6c61c7b8f/1722-2 00:10:23.836 --> 00:10:27.390 energy loss associated to to the way front as it travels over 436caf55-234f-4249-83d8-d9c6c61c7b8f/1722-3 00:10:27.390 --> 00:10:31.002 ever increasing area. Think of it as a water droplet in water. 436caf55-234f-4249-83d8-d9c6c61c7b8f/1722-4 00:10:31.002 --> 00:10:34.441 It's gonna create concentric circles of waves. The waves at 436caf55-234f-4249-83d8-d9c6c61c7b8f/1722-5 00:10:34.441 --> 00:10:37.709 the very close to where the water fell into the pool are 436caf55-234f-4249-83d8-d9c6c61c7b8f/1722-6 00:10:37.709 --> 00:10:41.091 gonna be pretty high. But as they get farther and farther, 436caf55-234f-4249-83d8-d9c6c61c7b8f/1722-7 00:10:41.091 --> 00:10:44.473 they'll eventually disappear because there's the energy is 436caf55-234f-4249-83d8-d9c6c61c7b8f/1722-8 00:10:44.473 --> 00:10:46.537 just spread over such a large area. 436caf55-234f-4249-83d8-d9c6c61c7b8f/1725-0 00:10:47.257 --> 00:10:47.897 The other. 436caf55-234f-4249-83d8-d9c6c61c7b8f/1791-0 00:10:48.677 --> 00:10:52.159 Ohh phenomena which is get the focus of today's presentation is 436caf55-234f-4249-83d8-d9c6c61c7b8f/1791-1 00:10:52.159 --> 00:10:55.478 analyst to continuation which is energy loss associated with 436caf55-234f-4249-83d8-d9c6c61c7b8f/1791-2 00:10:55.478 --> 00:10:59.070 mechanical propagation. A way to think of this is a projectile in 436caf55-234f-4249-83d8-d9c6c61c7b8f/1791-3 00:10:59.070 --> 00:11:02.389 water, it's speed is going to be really fast when it's first 436caf55-234f-4249-83d8-d9c6c61c7b8f/1791-4 00:11:02.389 --> 00:11:05.599 fired but it'll quickly slow down and eventually come to a 436caf55-234f-4249-83d8-d9c6c61c7b8f/1791-5 00:11:05.599 --> 00:11:09.028 stop that will acity is similar to a wave amplitude it's gonna 436caf55-234f-4249-83d8-d9c6c61c7b8f/1791-6 00:11:09.028 --> 00:11:12.129 be pretty high but as it travels farther and farther the 436caf55-234f-4249-83d8-d9c6c61c7b8f/1791-7 00:11:12.129 --> 00:11:14.687 amplitude is going to get smaller and smaller. 436caf55-234f-4249-83d8-d9c6c61c7b8f/1806-0 00:11:16.057 --> 00:11:18.933 A bit of a thought exercise. Assume we have a slight in 436caf55-234f-4249-83d8-d9c6c61c7b8f/1806-1 00:11:18.933 --> 00:11:21.347 California, represented by this gold triangle. 436caf55-234f-4249-83d8-d9c6c61c7b8f/1852-0 00:11:21.947 --> 00:11:24.662 And we're gonna look at some plots and intensity versus 436caf55-234f-4249-83d8-d9c6c61c7b8f/1852-1 00:11:24.662 --> 00:11:27.571 distance, space and within about residual, which is the log 436caf55-234f-4249-83d8-d9c6c61c7b8f/1852-2 00:11:27.571 --> 00:11:30.529 difference of an observation minus a prediction from a model 436caf55-234f-4249-83d8-d9c6c61c7b8f/1852-3 00:11:30.529 --> 00:11:33.293 with any source bias removed. And in this case, if we're 436caf55-234f-4249-83d8-d9c6c61c7b8f/1852-4 00:11:33.293 --> 00:11:36.397 looking at one particular site, the site bias is also the same. 436caf55-234f-4249-83d8-d9c6c61c7b8f/1856-0 00:11:37.497 --> 00:11:38.327 Versus distance. 436caf55-234f-4249-83d8-d9c6c61c7b8f/1867-0 00:11:39.057 --> 00:11:41.957 So we can have an identical event. Now this isn't. 436caf55-234f-4249-83d8-d9c6c61c7b8f/1903-0 00:11:42.717 --> 00:11:45.843 Ever gonna happen, but an identical event that happens at 436caf55-234f-4249-83d8-d9c6c61c7b8f/1903-1 00:11:45.843 --> 00:11:49.023 any azmath from our site with equidistance now zoom that a 436caf55-234f-4249-83d8-d9c6c61c7b8f/1903-2 00:11:49.023 --> 00:11:52.472 path travels in a straight line from our event to our site, and 436caf55-234f-4249-83d8-d9c6c61c7b8f/1903-3 00:11:52.472 --> 00:11:55.814 ergodic model would predict that all of these events have the 436caf55-234f-4249-83d8-d9c6c61c7b8f/1903-4 00:11:55.814 --> 00:11:56.407 same exact. 436caf55-234f-4249-83d8-d9c6c61c7b8f/1969-0 00:11:57.047 --> 00:12:00.097 Behavior. We essentially see this dot, but we interpret the 436caf55-234f-4249-83d8-d9c6c61c7b8f/1969-1 00:12:00.097 --> 00:12:03.453 results to show up as the smooth result and assuming there garlic 436caf55-234f-4249-83d8-d9c6c61c7b8f/1969-2 00:12:03.453 --> 00:12:06.605 model is correct, are within, the residual would be 0 all the 436caf55-234f-4249-83d8-d9c6c61c7b8f/1969-3 00:12:06.605 --> 00:12:09.860 way across. But that's not the case. That's not what we see. We 436caf55-234f-4249-83d8-d9c6c61c7b8f/1969-4 00:12:09.860 --> 00:12:12.910 can have 3 examples of a red event to the northwest of blue 436caf55-234f-4249-83d8-d9c6c61c7b8f/1969-5 00:12:12.910 --> 00:12:15.707 event to the east, and a great event to the southeast. 436caf55-234f-4249-83d8-d9c6c61c7b8f/2019-0 00:12:16.837 --> 00:12:20.352 And the data that we have from these events would just be these 436caf55-234f-4249-83d8-d9c6c61c7b8f/2019-1 00:12:20.352 --> 00:12:23.703 dots and I like to say that Grandma Smallers are essentially 436caf55-234f-4249-83d8-d9c6c61c7b8f/2019-2 00:12:23.703 --> 00:12:26.890 glorified connect the dot people. All we do is we want to 436caf55-234f-4249-83d8-d9c6c61c7b8f/2019-3 00:12:26.890 --> 00:12:30.405 connect these dots to this dot over here in the context of path 436caf55-234f-4249-83d8-d9c6c61c7b8f/2019-4 00:12:30.405 --> 00:12:33.537 model. How we do that is up to us to make our decisions. 436caf55-234f-4249-83d8-d9c6c61c7b8f/2096-0 00:12:34.257 --> 00:12:37.442 So first we can look at this green. Observe the screen event 436caf55-234f-4249-83d8-d9c6c61c7b8f/2096-1 00:12:37.442 --> 00:12:40.627 for example it's path is going to travel kind of pretty much 436caf55-234f-4249-83d8-d9c6c61c7b8f/2096-2 00:12:40.627 --> 00:12:43.604 through the Central Valley. That's all pretty consistent 436caf55-234f-4249-83d8-d9c6c61c7b8f/2096-3 00:12:43.604 --> 00:12:46.632 geology. It might just be that, oh, it tends to attenuate 436caf55-234f-4249-83d8-d9c6c61c7b8f/2096-4 00:12:46.632 --> 00:12:49.922 slightly faster. So it's going to be a smooth transition. Just 436caf55-234f-4249-83d8-d9c6c61c7b8f/2096-5 00:12:49.922 --> 00:12:53.107 blew events has you can see variable terrain that it's gonna 436caf55-234f-4249-83d8-d9c6c61c7b8f/2096-6 00:12:53.107 --> 00:12:56.032 be traveling through, maybe that terrain relates to the 436caf55-234f-4249-83d8-d9c6c61c7b8f/2096-7 00:12:56.032 --> 00:12:59.374 attenuation. So maybe I can Draw Something like this where it's 436caf55-234f-4249-83d8-d9c6c61c7b8f/2096-8 00:12:59.374 --> 00:13:02.298 slow, even slower than fast again. And you can see that 436caf55-234f-4249-83d8-d9c6c61c7b8f/2096-9 00:13:02.298 --> 00:13:04.387 where the effect is always a reduction. 436caf55-234f-4249-83d8-d9c6c61c7b8f/2165-0 00:13:04.477 --> 00:13:07.583 Of intensity. But the rate of reduction is variable as 436caf55-234f-4249-83d8-d9c6c61c7b8f/2165-1 00:13:07.583 --> 00:13:11.028 represented by those Internet residual. Similar case for the 436caf55-234f-4249-83d8-d9c6c61c7b8f/2165-2 00:13:11.028 --> 00:13:14.643 relevent just different results. Now we're just going faster. I 436caf55-234f-4249-83d8-d9c6c61c7b8f/2165-3 00:13:14.643 --> 00:13:18.201 could have just as easily for these great for the blue and the 436caf55-234f-4249-83d8-d9c6c61c7b8f/2165-4 00:13:18.201 --> 00:13:21.477 red events either gone a simpler method and just averaged 436caf55-234f-4249-83d8-d9c6c61c7b8f/2165-5 00:13:21.477 --> 00:13:24.752 everything to make a smooth line, or it could have been a 436caf55-234f-4249-83d8-d9c6c61c7b8f/2165-6 00:13:24.752 --> 00:13:28.254 bit more complicated. And so that's what path modelers do. We 436caf55-234f-4249-83d8-d9c6c61c7b8f/2165-7 00:13:28.254 --> 00:13:30.457 get to decide how we draw these lines. 436caf55-234f-4249-83d8-d9c6c61c7b8f/2242-0 00:13:32.557 --> 00:13:35.462 So now looking at some equations, path mowing is just 436caf55-234f-4249-83d8-d9c6c61c7b8f/2242-1 00:13:35.462 --> 00:13:38.583 the sum of geometric spreading and analytic continuation. 436caf55-234f-4249-83d8-d9c6c61c7b8f/2242-2 00:13:38.583 --> 00:13:41.920 There's also saturation, which is just to say that amplitudes 436caf55-234f-4249-83d8-d9c6c61c7b8f/2242-3 00:13:41.920 --> 00:13:45.094 don't change very much over short distances, but geometric 436caf55-234f-4249-83d8-d9c6c61c7b8f/2242-4 00:13:45.094 --> 00:13:48.377 spreading would be a constant slope down in intensity versus 436caf55-234f-4249-83d8-d9c6c61c7b8f/2242-5 00:13:48.377 --> 00:13:51.390 just in space. Analytic continuation is just additional 436caf55-234f-4249-83d8-d9c6c61c7b8f/2242-6 00:13:51.390 --> 00:13:54.350 energy loss, and for the purposes of this presentation 436caf55-234f-4249-83d8-d9c6c61c7b8f/2242-7 00:13:54.350 --> 00:13:57.686 I'm gonna assume that geometric spreading is constant because 436caf55-234f-4249-83d8-d9c6c61c7b8f/2242-8 00:13:57.686 --> 00:13:58.117 this is. 436caf55-234f-4249-83d8-d9c6c61c7b8f/2277-0 00:13:58.777 --> 00:14:01.629 Part of if we assume that the Earth is a perfect sphere, 436caf55-234f-4249-83d8-d9c6c61c7b8f/2277-1 00:14:01.629 --> 00:14:04.781 geometric spreading will be the same around all the the Earth, 436caf55-234f-4249-83d8-d9c6c61c7b8f/2277-2 00:14:04.781 --> 00:14:07.883 and it's the continuation is related to the properties of the 436caf55-234f-4249-83d8-d9c6c61c7b8f/2277-3 00:14:07.883 --> 00:14:10.635 quest. So it's understandable that this is likely more 436caf55-234f-4249-83d8-d9c6c61c7b8f/2277-4 00:14:10.635 --> 00:14:12.387 variable than geometric spreading. 436caf55-234f-4249-83d8-d9c6c61c7b8f/2286-0 00:14:13.977 --> 00:14:17.657 So the approaches that have been performed for path modeling. 436caf55-234f-4249-83d8-d9c6c61c7b8f/2313-0 00:14:18.037 --> 00:14:21.736 The history lesson we began with just fully global models. There 436caf55-234f-4249-83d8-d9c6c61c7b8f/2313-1 00:14:21.736 --> 00:14:25.095 weren't a lot of data and JW produced just one and elastic 436caf55-234f-4249-83d8-d9c6c61c7b8f/2313-2 00:14:25.095 --> 00:14:27.827 coefficient. For example, for the entire world. 436caf55-234f-4249-83d8-d9c6c61c7b8f/2356-0 00:14:28.667 --> 00:14:32.434 We then got more data and NGOs 2 and we developed broad regional 436caf55-234f-4249-83d8-d9c6c61c7b8f/2356-1 00:14:32.434 --> 00:14:36.260 models for areas like California and Japan where we have a lot of 436caf55-234f-4249-83d8-d9c6c61c7b8f/2356-2 00:14:36.260 --> 00:14:39.680 data to refine those estimates. So we add in this regional 436caf55-234f-4249-83d8-d9c6c61c7b8f/2356-3 00:14:39.680 --> 00:14:41.767 adjustment, Delta, CA in this case. 436caf55-234f-4249-83d8-d9c6c61c7b8f/2401-0 00:14:42.567 --> 00:14:45.777 There's also local methods, local regional methods, Erdem at 436caf55-234f-4249-83d8-d9c6c61c7b8f/2401-1 00:14:45.777 --> 00:14:48.934 all, 2019 proposed a regional adjustment for the Sacramento 436caf55-234f-4249-83d8-d9c6c61c7b8f/2401-2 00:14:48.934 --> 00:14:52.249 San Joaquin Delta, where deep events were found to have faster 436caf55-234f-4249-83d8-d9c6c61c7b8f/2401-3 00:14:52.249 --> 00:14:55.406 attenuation than shallower events, and so those events just 436caf55-234f-4249-83d8-d9c6c61c7b8f/2401-4 00:14:55.406 --> 00:14:55.617 got. 436caf55-234f-4249-83d8-d9c6c61c7b8f/2424-0 00:14:56.727 --> 00:14:59.957 Hadn't additive attenuation term added on to them, which you can 436caf55-234f-4249-83d8-d9c6c61c7b8f/2424-1 00:14:59.957 --> 00:15:03.237 see is represented by the dashed line compared to the solid line. 436caf55-234f-4249-83d8-d9c6c61c7b8f/2502-0 00:15:04.047 --> 00:15:07.985 And lastly, kind of the most sophisticated methods that we 436caf55-234f-4249-83d8-d9c6c61c7b8f/2502-1 00:15:07.985 --> 00:15:11.991 have currently are cell based methods initially proposed by 436caf55-234f-4249-83d8-d9c6c61c7b8f/2502-2 00:15:11.991 --> 00:15:15.929 Dawood and Rodriguez, Merrick 2013 where we discretized an 436caf55-234f-4249-83d8-d9c6c61c7b8f/2502-3 00:15:15.929 --> 00:15:19.534 area, a domain into many different cells and for each 436caf55-234f-4249-83d8-d9c6c61c7b8f/2502-4 00:15:19.534 --> 00:15:23.206 cell we have an analytic coefficient and we can sum up 436caf55-234f-4249-83d8-d9c6c61c7b8f/2502-5 00:15:23.206 --> 00:15:27.211 the product of the analytic coefficient and the path length 436caf55-234f-4249-83d8-d9c6c61c7b8f/2502-6 00:15:27.211 --> 00:15:31.017 through each cell to get our tool analystic attenuation. 436caf55-234f-4249-83d8-d9c6c61c7b8f/2536-0 00:15:31.787 --> 00:15:35.385 Now Queen at all applied this method to California. See they 436caf55-234f-4249-83d8-d9c6c61c7b8f/2536-1 00:15:35.385 --> 00:15:38.925 use the NGA W data set there where there's a lot of data in 436caf55-234f-4249-83d8-d9c6c61c7b8f/2536-2 00:15:38.925 --> 00:15:41.167 the Bay Area and Southern California. 436caf55-234f-4249-83d8-d9c6c61c7b8f/2567-0 00:15:41.797 --> 00:15:45.422 And we can and those data are reflected in where the changes 436caf55-234f-4249-83d8-d9c6c61c7b8f/2567-1 00:15:45.422 --> 00:15:48.869 are observed. We see changes in the Bay Area and Southern 436caf55-234f-4249-83d8-d9c6c61c7b8f/2567-2 00:15:48.869 --> 00:15:52.316 California, good lowered certainty lowered uncertainty in 436caf55-234f-4249-83d8-d9c6c61c7b8f/2567-3 00:15:52.316 --> 00:15:53.327 the Bay Area and. 436caf55-234f-4249-83d8-d9c6c61c7b8f/2651-0 00:15:54.557 --> 00:15:57.822 Uh and Sun, California, but rather large uncertainty and 436caf55-234f-4249-83d8-d9c6c61c7b8f/2651-1 00:15:57.822 --> 00:16:01.489 other areas, but kind of a this is a good modeling approach and 436caf55-234f-4249-83d8-d9c6c61c7b8f/2651-2 00:16:01.489 --> 00:16:04.926 I don't want and I want to emphasize that it it's valid and 436caf55-234f-4249-83d8-d9c6c61c7b8f/2651-3 00:16:04.926 --> 00:16:08.192 it it works well. However a drawback of this approach is 436caf55-234f-4249-83d8-d9c6c61c7b8f/2651-4 00:16:08.192 --> 00:16:11.286 that it's not easily interpreted. For example, we can 436caf55-234f-4249-83d8-d9c6c61c7b8f/2651-5 00:16:11.286 --> 00:16:14.895 see that there's some variable rates up here and some variable 436caf55-234f-4249-83d8-d9c6c61c7b8f/2651-6 00:16:14.895 --> 00:16:18.218 rates down here. But what is producing those rates? Well, 436caf55-234f-4249-83d8-d9c6c61c7b8f/2651-7 00:16:18.218 --> 00:16:21.713 these are just cells. The data suggests what the attenuation 436caf55-234f-4249-83d8-d9c6c61c7b8f/2651-8 00:16:21.713 --> 00:16:22.057 shows. 436caf55-234f-4249-83d8-d9c6c61c7b8f/2685-0 00:16:22.737 --> 00:16:25.660 More so, if we look at smaller features, for example, some of 436caf55-234f-4249-83d8-d9c6c61c7b8f/2685-1 00:16:25.660 --> 00:16:28.347 the cells with the lowest and the OR the slowest and the 436caf55-234f-4249-83d8-d9c6c61c7b8f/2685-2 00:16:28.347 --> 00:16:31.034 fastest attenuation are literally touching each other in 436caf55-234f-4249-83d8-d9c6c61c7b8f/2685-3 00:16:31.034 --> 00:16:31.977 Northern California. 436caf55-234f-4249-83d8-d9c6c61c7b8f/2702-0 00:16:32.887 --> 00:16:36.140 What causes that? What are the reasons why these cells, even 436caf55-234f-4249-83d8-d9c6c61c7b8f/2702-1 00:16:36.140 --> 00:16:39.607 though they're so close to each other, have such different rates 436caf55-234f-4249-83d8-d9c6c61c7b8f/2702-2 00:16:39.607 --> 00:16:40.407 of attenuation? 436caf55-234f-4249-83d8-d9c6c61c7b8f/2704-0 00:16:41.117 --> 00:16:41.557 And. 436caf55-234f-4249-83d8-d9c6c61c7b8f/2734-0 00:16:42.437 --> 00:16:46.291 You can't interpret anything from this model you based off of 436caf55-234f-4249-83d8-d9c6c61c7b8f/2734-1 00:16:46.291 --> 00:16:49.959 this discretization, cause it's just a purely mathematical 436caf55-234f-4249-83d8-d9c6c61c7b8f/2734-2 00:16:49.959 --> 00:16:53.441 model, although it does work well for reducing bias and 436caf55-234f-4249-83d8-d9c6c61c7b8f/2734-3 00:16:53.441 --> 00:16:54.187 variability. 436caf55-234f-4249-83d8-d9c6c61c7b8f/2752-0 00:16:55.387 --> 00:16:59.138 So how these models fall on the spectrum? It's a bit arbitrary 436caf55-234f-4249-83d8-d9c6c61c7b8f/2752-1 00:16:59.138 --> 00:17:01.997 where I place them, but the order is important. 436caf55-234f-4249-83d8-d9c6c61c7b8f/2829-0 00:17:02.747 --> 00:17:06.592 Global models obviously are God. Rod models a little bit better, 436caf55-234f-4249-83d8-d9c6c61c7b8f/2829-1 00:17:06.592 --> 00:17:10.201 but still pretty are God's local models a little bit better? 436caf55-234f-4249-83d8-d9c6c61c7b8f/2829-2 00:17:10.201 --> 00:17:13.988 Probably still are Gothic maybe not ergodic. We're getting into 436caf55-234f-4249-83d8-d9c6c61c7b8f/2829-3 00:17:13.988 --> 00:17:17.833 a fuzzy range and then so based methods kind of span a big range 436caf55-234f-4249-83d8-d9c6c61c7b8f/2829-4 00:17:17.833 --> 00:17:21.324 where we have core sales or large cells or small cells, or 436caf55-234f-4249-83d8-d9c6c61c7b8f/2829-5 00:17:21.324 --> 00:17:25.051 they get closer to nonnarcotic. Obviously regional methods and 436caf55-234f-4249-83d8-d9c6c61c7b8f/2829-6 00:17:25.051 --> 00:17:28.838 cell methods can span different ranges, but they have different 436caf55-234f-4249-83d8-d9c6c61c7b8f/2829-7 00:17:28.838 --> 00:17:30.317 advantages and drawbacks. 436caf55-234f-4249-83d8-d9c6c61c7b8f/2882-0 00:17:31.017 --> 00:17:35.266 Today I'm going to be discussing subregional methods, which take 436caf55-234f-4249-83d8-d9c6c61c7b8f/2882-1 00:17:35.266 --> 00:17:39.384 the advantages of both method and combine them. So we take the 436caf55-234f-4249-83d8-d9c6c61c7b8f/2882-2 00:17:39.384 --> 00:17:43.698 physical basis that local models use, generally defining a domain 436caf55-234f-4249-83d8-d9c6c61c7b8f/2882-3 00:17:43.698 --> 00:17:47.751 and trying to group together areas with similar attributes to 436caf55-234f-4249-83d8-d9c6c61c7b8f/2882-4 00:17:47.751 --> 00:17:51.739 our discretization and then apply the math behind cell based 436caf55-234f-4249-83d8-d9c6c61c7b8f/2882-5 00:17:51.739 --> 00:17:53.177 methods. So hopefully. 436caf55-234f-4249-83d8-d9c6c61c7b8f/2887-0 00:17:53.917 --> 00:17:55.367 Allow us to interpret them all. 436caf55-234f-4249-83d8-d9c6c61c7b8f/2909-0 00:17:56.147 --> 00:17:59.502 Now, as a researcher, I would hope to be able to reach a fully 436caf55-234f-4249-83d8-d9c6c61c7b8f/2909-1 00:17:59.502 --> 00:18:02.911 non organic path model. I don't think that that has ever really 436caf55-234f-4249-83d8-d9c6c61c7b8f/2909-2 00:18:02.911 --> 00:18:03.497 realizable. 436caf55-234f-4249-83d8-d9c6c61c7b8f/2913-0 00:18:04.227 --> 00:18:05.337 But as an engineer. 436caf55-234f-4249-83d8-d9c6c61c7b8f/2950-0 00:18:06.077 --> 00:18:10.536 We had we can achieve something that is practically known about 436caf55-234f-4249-83d8-d9c6c61c7b8f/2950-1 00:18:10.536 --> 00:18:14.995 it, which is to say, what is the least sophisticated model that 436caf55-234f-4249-83d8-d9c6c61c7b8f/2950-2 00:18:14.995 --> 00:18:18.967 gives us the same results as super sophisticated models. 436caf55-234f-4249-83d8-d9c6c61c7b8f/2993-0 00:18:19.777 --> 00:18:23.095 The problem with cell based is some is users may feel 436caf55-234f-4249-83d8-d9c6c61c7b8f/2993-1 00:18:23.095 --> 00:18:26.843 intimidated by the complexity that it looks to be. It's it's 436caf55-234f-4249-83d8-d9c6c61c7b8f/2993-2 00:18:26.843 --> 00:18:30.776 not that complex conceptually, but mathematically there's a lot 436caf55-234f-4249-83d8-d9c6c61c7b8f/2993-3 00:18:30.776 --> 00:18:34.647 going on and users may not want to or may be afraid to use it. 436caf55-234f-4249-83d8-d9c6c61c7b8f/3018-0 00:18:35.757 --> 00:18:39.097 By simplifying the number of cells and making it easy to 436caf55-234f-4249-83d8-d9c6c61c7b8f/3018-1 00:18:39.097 --> 00:18:42.613 interpret, people may be more inclined to use a subregional 436caf55-234f-4249-83d8-d9c6c61c7b8f/3018-2 00:18:42.613 --> 00:18:45.837 method. A cell based method using sub regionalization. 436caf55-234f-4249-83d8-d9c6c61c7b8f/3056-0 00:18:46.537 --> 00:18:50.058 And that's where I think a practical nonnarcotic would fall 436caf55-234f-4249-83d8-d9c6c61c7b8f/3056-1 00:18:50.058 --> 00:18:53.344 is by reducing the number of cells, but still achieving 436caf55-234f-4249-83d8-d9c6c61c7b8f/3056-2 00:18:53.344 --> 00:18:56.924 similar results to a cell based method with a lot of clients 436caf55-234f-4249-83d8-d9c6c61c7b8f/3056-3 00:18:56.924 --> 00:18:59.917 cells. So that's kind of the today's presentation. 436caf55-234f-4249-83d8-d9c6c61c7b8f/3125-0 00:19:01.377 --> 00:19:04.784 So first I want to discuss some ground motion data. A lot of 436caf55-234f-4249-83d8-d9c6c61c7b8f/3125-1 00:19:04.784 --> 00:19:08.079 data has been made available since Ninja W 2 concluded. We 436caf55-234f-4249-83d8-d9c6c61c7b8f/3125-2 00:19:08.079 --> 00:19:11.710 can see the events of NGA events in blue. We have the rich Crest 436caf55-234f-4249-83d8-d9c6c61c7b8f/3125-3 00:19:11.710 --> 00:19:14.894 events for collected a lot of data, pink fade during his 436caf55-234f-4249-83d8-d9c6c61c7b8f/3125-4 00:19:14.894 --> 00:19:18.077 dissertation was investigating site response in Southern 436caf55-234f-4249-83d8-d9c6c61c7b8f/3125-5 00:19:18.077 --> 00:19:21.484 California. So he collected a lot of data. I'm investigating 436caf55-234f-4249-83d8-d9c6c61c7b8f/3125-6 00:19:21.484 --> 00:19:24.780 site response in Northern California. So I collected a lot 436caf55-234f-4249-83d8-d9c6c61c7b8f/3125-7 00:19:24.780 --> 00:19:25.227 of data. 436caf55-234f-4249-83d8-d9c6c61c7b8f/3148-0 00:19:25.947 --> 00:19:28.772 Or I gathered a lot, David rather in Northern California 436caf55-234f-4249-83d8-d9c6c61c7b8f/3148-1 00:19:28.772 --> 00:19:31.797 and you can see there's a lot of sites that we've collected. 436caf55-234f-4249-83d8-d9c6c61c7b8f/3161-0 00:19:32.777 --> 00:19:36.757 Now the data that we use, we only use magnitudes above 4 to 436caf55-234f-4249-83d8-d9c6c61c7b8f/3161-1 00:19:36.757 --> 00:19:37.487 to kind of. 436caf55-234f-4249-83d8-d9c6c61c7b8f/3225-0 00:19:38.967 --> 00:19:42.562 Not include a small magnitude events that sometimes there's 436caf55-234f-4249-83d8-d9c6c61c7b8f/3225-1 00:19:42.562 --> 00:19:46.157 issues with spectral scaling with magnitude and we imply we 436caf55-234f-4249-83d8-d9c6c61c7b8f/3225-2 00:19:46.157 --> 00:19:49.453 enforce magnitude distance criteria. But if we see the 436caf55-234f-4249-83d8-d9c6c61c7b8f/3225-3 00:19:49.453 --> 00:19:52.868 impact of this additional data which is shown by the red 436caf55-234f-4249-83d8-d9c6c61c7b8f/3225-4 00:19:52.868 --> 00:19:56.584 compared to the blue, we have a lot of small magnitude events 436caf55-234f-4249-83d8-d9c6c61c7b8f/3225-5 00:19:56.584 --> 00:20:00.179 and a significant increase in the far distance observations 436caf55-234f-4249-83d8-d9c6c61c7b8f/3225-6 00:20:00.179 --> 00:20:01.917 which is also seen in period. 436caf55-234f-4249-83d8-d9c6c61c7b8f/3233-0 00:20:02.037 --> 00:20:05.647 And period. Umm versus number of record space. 436caf55-234f-4249-83d8-d9c6c61c7b8f/3265-0 00:20:06.977 --> 00:20:10.213 Now for path effects, this has significant implications for NJ 436caf55-234f-4249-83d8-d9c6c61c7b8f/3265-1 00:20:10.213 --> 00:20:13.501 W 2. You can see we have gaps in our data where we can't refine 436caf55-234f-4249-83d8-d9c6c61c7b8f/3265-2 00:20:13.501 --> 00:20:16.223 path effects. Central California, Eastern California 436caf55-234f-4249-83d8-d9c6c61c7b8f/3265-3 00:20:16.223 --> 00:20:18.227 and Northern California predominantly. 436caf55-234f-4249-83d8-d9c6c61c7b8f/3306-0 00:20:18.927 --> 00:20:22.553 With this additional data, we pretty much fill in all of those 436caf55-234f-4249-83d8-d9c6c61c7b8f/3306-1 00:20:22.553 --> 00:20:26.179 gaps. In Central California, we push as Far East as we can get 436caf55-234f-4249-83d8-d9c6c61c7b8f/3306-2 00:20:26.179 --> 00:20:29.633 and we're slowly pushing up north, so we can refine most of 436caf55-234f-4249-83d8-d9c6c61c7b8f/3306-3 00:20:29.633 --> 00:20:32.857 the state with the exception of kind of this NE corner. 436caf55-234f-4249-83d8-d9c6c61c7b8f/3339-0 00:20:33.607 --> 00:20:37.706 Pretty well compared to what previous efforts could do using 436caf55-234f-4249-83d8-d9c6c61c7b8f/3339-1 00:20:37.706 --> 00:20:42.008 just Ng. What's two data? So the hope is that now with all this 436caf55-234f-4249-83d8-d9c6c61c7b8f/3339-2 00:20:42.008 --> 00:20:45.167 data, we can refine these sub regional trends. 436caf55-234f-4249-83d8-d9c6c61c7b8f/3365-0 00:20:46.777 --> 00:20:49.637 So how did I submerge, or how did we rather sub sub 436caf55-234f-4249-83d8-d9c6c61c7b8f/3365-1 00:20:49.637 --> 00:20:53.267 regionalized California? Well, I said one method might just be at 436caf55-234f-4249-83d8-d9c6c61c7b8f/3365-2 00:20:53.267 --> 00:20:54.147 a group similar. 436caf55-234f-4249-83d8-d9c6c61c7b8f/3388-0 00:20:55.057 --> 00:20:59.754 Properties and one property that is kind of readily available is 436caf55-234f-4249-83d8-d9c6c61c7b8f/3388-1 00:20:59.754 --> 00:21:03.945 physiographical provinces. California can be described in 436caf55-234f-4249-83d8-d9c6c61c7b8f/3388-2 00:21:03.945 --> 00:21:05.897 the 13 really 11 provinces. 436caf55-234f-4249-83d8-d9c6c61c7b8f/3444-0 00:21:06.617 --> 00:21:11.012 But some of these provinces span large ranges. For example, this 436caf55-234f-4249-83d8-d9c6c61c7b8f/3444-1 00:21:11.012 --> 00:21:15.339 coastal range all the way from the top of California down to to 436caf55-234f-4249-83d8-d9c6c61c7b8f/3444-2 00:21:15.339 --> 00:21:19.463 central lower end of Central California is all one property, 436caf55-234f-4249-83d8-d9c6c61c7b8f/3444-3 00:21:19.463 --> 00:21:23.385 and it's unlikely that the attenuation within this region 436caf55-234f-4249-83d8-d9c6c61c7b8f/3444-4 00:21:23.385 --> 00:21:27.239 is similar. So this helps to create a starting basis for 436caf55-234f-4249-83d8-d9c6c61c7b8f/3444-5 00:21:27.239 --> 00:21:30.147 suborganization, but we need to refine it. 436caf55-234f-4249-83d8-d9c6c61c7b8f/3515-0 00:21:31.247 --> 00:21:34.850 Umm, one thing to look at that can help that is rock quality 436caf55-234f-4249-83d8-d9c6c61c7b8f/3515-1 00:21:34.850 --> 00:21:38.572 factor which is just the ratio of stored energy over dispersed 436caf55-234f-4249-83d8-d9c6c61c7b8f/3515-2 00:21:38.572 --> 00:21:41.939 energy. It can be thought of as the inverse of anelastic 436caf55-234f-4249-83d8-d9c6c61c7b8f/3515-3 00:21:41.939 --> 00:21:45.543 attenuation. So a high value means slower attenuation. A low 436caf55-234f-4249-83d8-d9c6c61c7b8f/3515-4 00:21:45.543 --> 00:21:49.383 value means faster attenuation. Overheard Phillips 2016 produced 436caf55-234f-4249-83d8-d9c6c61c7b8f/3515-5 00:21:49.383 --> 00:21:53.046 a model. It's depth dependent. Most path models haven't tried 436caf55-234f-4249-83d8-d9c6c61c7b8f/3515-6 00:21:53.046 --> 00:21:56.472 to consider depth dependence yet. So obviously we have to 436caf55-234f-4249-83d8-d9c6c61c7b8f/3515-7 00:21:56.472 --> 00:21:59.367 interpret these results as A2 dimensional plane. 436caf55-234f-4249-83d8-d9c6c61c7b8f/3549-0 00:22:00.047 --> 00:22:03.587 But what we can see is that especially in this top row, just 436caf55-234f-4249-83d8-d9c6c61c7b8f/3549-1 00:22:03.587 --> 00:22:07.244 focusing here, there are some areas currently this years where 436caf55-234f-4249-83d8-d9c6c61c7b8f/3549-2 00:22:07.244 --> 00:22:10.437 have which have pretty high QS, so slower attenuation. 436caf55-234f-4249-83d8-d9c6c61c7b8f/3633-0 00:22:11.697 --> 00:22:15.267 There are other areas, for example, the northern coastline 436caf55-234f-4249-83d8-d9c6c61c7b8f/3633-1 00:22:15.267 --> 00:22:18.897 in California, which is pretty red, so would suggest slower 436caf55-234f-4249-83d8-d9c6c61c7b8f/3633-2 00:22:18.897 --> 00:22:22.770 attenuation. And if we look at like Z = 8 kilometer depth, this 436caf55-234f-4249-83d8-d9c6c61c7b8f/3633-3 00:22:22.770 --> 00:22:26.340 coastal range, which I suspect it might have some internal 436caf55-234f-4249-83d8-d9c6c61c7b8f/3633-4 00:22:26.340 --> 00:22:29.547 variation, we have faster attenuation here. A mix of 436caf55-234f-4249-83d8-d9c6c61c7b8f/3633-5 00:22:29.547 --> 00:22:33.239 attenuation as we get to the middle parts around the Bay and 436caf55-234f-4249-83d8-d9c6c61c7b8f/3633-6 00:22:33.239 --> 00:22:36.506 then back to some faster attenuation as we get to the 436caf55-234f-4249-83d8-d9c6c61c7b8f/3633-7 00:22:36.506 --> 00:22:39.713 South. So obviously we need to do some refinement of 436caf55-234f-4249-83d8-d9c6c61c7b8f/3633-8 00:22:39.713 --> 00:22:41.287 physiographical provinces. 436caf55-234f-4249-83d8-d9c6c61c7b8f/3693-0 00:22:42.147 --> 00:22:45.546 And this is one way to do it. Another way would be to look at 436caf55-234f-4249-83d8-d9c6c61c7b8f/3693-1 00:22:45.546 --> 00:22:49.000 if that specific attenuation. Here we just fit this functional 436caf55-234f-4249-83d8-d9c6c61c7b8f/3693-2 00:22:49.000 --> 00:22:52.290 form where we have an event attenuation times distance at a 436caf55-234f-4249-83d8-d9c6c61c7b8f/3693-3 00:22:52.290 --> 00:22:55.690 vertical shift up and down to capture any near source trends, 436caf55-234f-4249-83d8-d9c6c61c7b8f/3693-4 00:22:55.690 --> 00:22:58.651 and this particular event attenuated faster than what 436caf55-234f-4249-83d8-d9c6c61c7b8f/3693-5 00:22:58.651 --> 00:23:01.337 their Gothic model would suggest, which would be 436caf55-234f-4249-83d8-d9c6c61c7b8f/3693-6 00:23:01.337 --> 00:23:04.737 representative in this plot here by the pink colored circles. 436caf55-234f-4249-83d8-d9c6c61c7b8f/3714-0 00:23:05.567 --> 00:23:08.121 The fence, which have slower than equation would be cyan and 436caf55-234f-4249-83d8-d9c6c61c7b8f/3714-1 00:23:08.121 --> 00:23:10.466 events which have average compared to the ergodic would 436caf55-234f-4249-83d8-d9c6c61c7b8f/3714-2 00:23:10.466 --> 00:23:11.597 would pretty much be white. 436caf55-234f-4249-83d8-d9c6c61c7b8f/3783-0 00:23:12.387 --> 00:23:15.985 The problem with this is that these events use observations 436caf55-234f-4249-83d8-d9c6c61c7b8f/3783-1 00:23:15.985 --> 00:23:19.764 from all sub regions. It's not just one particular sub region. 436caf55-234f-4249-83d8-d9c6c61c7b8f/3783-2 00:23:19.764 --> 00:23:23.482 So again we have to interpret the results, but we see similar 436caf55-234f-4249-83d8-d9c6c61c7b8f/3783-3 00:23:23.482 --> 00:23:27.201 results as the QS. For example, in the coastal range where we 436caf55-234f-4249-83d8-d9c6c61c7b8f/3783-4 00:23:27.201 --> 00:23:30.920 have high negatives or large negatives in the north, a mix of 436caf55-234f-4249-83d8-d9c6c61c7b8f/3783-5 00:23:30.920 --> 00:23:34.638 results and then that slowly transitions to positive to maybe 436caf55-234f-4249-83d8-d9c6c61c7b8f/3783-6 00:23:34.638 --> 00:23:38.477 to back again being negative. So there's variation within those 436caf55-234f-4249-83d8-d9c6c61c7b8f/3783-7 00:23:38.477 --> 00:23:39.077 provinces. 436caf55-234f-4249-83d8-d9c6c61c7b8f/3842-0 00:23:40.347 --> 00:23:44.506 We ultimately decided to create 9 broad subregions, and there's 436caf55-234f-4249-83d8-d9c6c61c7b8f/3842-1 00:23:44.506 --> 00:23:48.795 probably many of you that may be offended by this liberalization. 436caf55-234f-4249-83d8-d9c6c61c7b8f/3842-2 00:23:48.795 --> 00:23:52.564 Many Earth scientists that may be offended by the sub sub 436caf55-234f-4249-83d8-d9c6c61c7b8f/3842-3 00:23:52.564 --> 00:23:56.528 regionalization, but I want to emphasize that our purpose of 436caf55-234f-4249-83d8-d9c6c61c7b8f/3842-4 00:23:56.528 --> 00:24:00.103 this liberalization was to capture the broad analystic 436caf55-234f-4249-83d8-d9c6c61c7b8f/3842-5 00:24:00.103 --> 00:24:03.742 attenuation effects. We want to create a model that has 436caf55-234f-4249-83d8-d9c6c61c7b8f/3842-6 00:24:03.742 --> 00:24:04.717 relatively few. 436caf55-234f-4249-83d8-d9c6c61c7b8f/3936-0 00:24:06.607 --> 00:24:10.645 Regions to compare to a model like the Queen Model Tune 2019, 436caf55-234f-4249-83d8-d9c6c61c7b8f/3936-1 00:24:10.645 --> 00:24:14.618 which has about 500 or so cells to see if we can get similar 436caf55-234f-4249-83d8-d9c6c61c7b8f/3936-2 00:24:14.618 --> 00:24:18.526 results. So obviously in the future we can come back and we 436caf55-234f-4249-83d8-d9c6c61c7b8f/3936-3 00:24:18.526 --> 00:24:22.629 can refine issues if we see any within region differences. But 436caf55-234f-4249-83d8-d9c6c61c7b8f/3936-4 00:24:22.629 --> 00:24:26.537 for now these are the 9 sub regions that I use. I just want 436caf55-234f-4249-83d8-d9c6c61c7b8f/3936-5 00:24:26.537 --> 00:24:30.119 to point out the North Coast region because it I'll be 436caf55-234f-4249-83d8-d9c6c61c7b8f/3936-6 00:24:30.119 --> 00:24:34.092 talking a lot about it and I also want to mention the Sierra 436caf55-234f-4249-83d8-d9c6c61c7b8f/3936-7 00:24:34.092 --> 00:24:36.437 Nevada region which is to the east. 436caf55-234f-4249-83d8-d9c6c61c7b8f/3945-0 00:24:36.787 --> 00:24:39.304 Of California and I guess I can also mention Southern 436caf55-234f-4249-83d8-d9c6c61c7b8f/3945-1 00:24:39.304 --> 00:24:39.817 California. 436caf55-234f-4249-83d8-d9c6c61c7b8f/3974-0 00:24:41.607 --> 00:24:44.989 Now for our model development. Now here's where I'm gonna 436caf55-234f-4249-83d8-d9c6c61c7b8f/3974-1 00:24:44.989 --> 00:24:48.663 present some a bunch of words and some equations. A bit of the 436caf55-234f-4249-83d8-d9c6c61c7b8f/3974-2 00:24:48.663 --> 00:24:52.220 boring stuff that we don't like to see, but we have to to at 436caf55-234f-4249-83d8-d9c6c61c7b8f/3974-3 00:24:52.220 --> 00:24:53.037 least present. 436caf55-234f-4249-83d8-d9c6c61c7b8f/4002-0 00:24:54.247 --> 00:24:57.896 We perform our analysis using residuals typical residuals 436caf55-234f-4249-83d8-d9c6c61c7b8f/4002-1 00:24:57.896 --> 00:25:01.294 analysis where we compute residual as the natural log 436caf55-234f-4249-83d8-d9c6c61c7b8f/4002-2 00:25:01.294 --> 00:25:05.259 difference of the observation minus the prediction of a ground 436caf55-234f-4249-83d8-d9c6c61c7b8f/4002-3 00:25:05.259 --> 00:25:06.077 motion model. 436caf55-234f-4249-83d8-d9c6c61c7b8f/4017-0 00:25:06.797 --> 00:25:09.904 Which would be BSA, 14, and our particular case what I'm 436caf55-234f-4249-83d8-d9c6c61c7b8f/4017-1 00:25:09.904 --> 00:25:10.777 presenting here. 436caf55-234f-4249-83d8-d9c6c61c7b8f/4091-0 00:25:11.717 --> 00:25:15.380 We can then partition these total residuals using mixed 436caf55-234f-4249-83d8-d9c6c61c7b8f/4091-1 00:25:15.380 --> 00:25:19.435 effects methods to estimate the event bias or the event term, 436caf55-234f-4249-83d8-d9c6c61c7b8f/4091-2 00:25:19.435 --> 00:25:23.359 but we're careful to only use path unbiased within or total 436caf55-234f-4249-83d8-d9c6c61c7b8f/4091-3 00:25:23.359 --> 00:25:27.480 residuals. In other words, we only use observations within 100 436caf55-234f-4249-83d8-d9c6c61c7b8f/4091-4 00:25:27.480 --> 00:25:31.208 kilometers because they are unlikely to be significantly 436caf55-234f-4249-83d8-d9c6c61c7b8f/4091-5 00:25:31.208 --> 00:25:35.067 affected by analystic variations between subregions, so it 436caf55-234f-4249-83d8-d9c6c61c7b8f/4091-6 00:25:35.067 --> 00:25:38.861 follows that it's actually 2 rounds of mixed effects. The 436caf55-234f-4249-83d8-d9c6c61c7b8f/4091-7 00:25:38.861 --> 00:25:41.347 first round separates a bias and the. 436caf55-234f-4249-83d8-d9c6c61c7b8f/4119-0 00:25:41.707 --> 00:25:45.036 The random effect for the event or the source bias, the second 436caf55-234f-4249-83d8-d9c6c61c7b8f/4119-1 00:25:45.036 --> 00:25:48.101 one is a mixed effects on the within event residual which 436caf55-234f-4249-83d8-d9c6c61c7b8f/4119-2 00:25:48.101 --> 00:25:50.797 separates side effects and the remaining residual. 436caf55-234f-4249-83d8-d9c6c61c7b8f/4149-0 00:25:51.537 --> 00:25:54.827 I have more to say about why we're interested in the side 436caf55-234f-4249-83d8-d9c6c61c7b8f/4149-1 00:25:54.827 --> 00:25:58.514 effects during the Salis. When I talk about the performance. But 436caf55-234f-4249-83d8-d9c6c61c7b8f/4149-2 00:25:58.514 --> 00:26:02.031 just for now, one thing to keep in mind is that within event, 436caf55-234f-4249-83d8-d9c6c61c7b8f/4149-3 00:26:02.031 --> 00:26:03.847 residuals contained within them. 436caf55-234f-4249-83d8-d9c6c61c7b8f/4165-0 00:26:04.617 --> 00:26:08.388 Sort of site bias and path bias, so we're interested in 436caf55-234f-4249-83d8-d9c6c61c7b8f/4165-1 00:26:08.388 --> 00:26:10.207 decoupling the two of them. 436caf55-234f-4249-83d8-d9c6c61c7b8f/4200-0 00:26:11.537 --> 00:26:14.692 The third step is to inspect the trends of the data within 436caf55-234f-4249-83d8-d9c6c61c7b8f/4200-1 00:26:14.692 --> 00:26:17.741 residuals versus distance. If we see for trends, then we 436caf55-234f-4249-83d8-d9c6c61c7b8f/4200-2 00:26:17.741 --> 00:26:20.950 formulate a path model little spoiler because we're talking 436caf55-234f-4249-83d8-d9c6c61c7b8f/4200-3 00:26:20.950 --> 00:26:23.197 about this. Obviously, we've seen trends. 436caf55-234f-4249-83d8-d9c6c61c7b8f/4210-0 00:26:23.817 --> 00:26:27.837 And we develop an adjustment for the ANALYSTIC model which? 436caf55-234f-4249-83d8-d9c6c61c7b8f/4237-0 00:26:28.677 --> 00:26:32.555 Can be added to the source model and the site model in the USA 14 436caf55-234f-4249-83d8-d9c6c61c7b8f/4237-1 00:26:32.555 --> 00:26:35.317 for a a new ground motion model called GM One. 436caf55-234f-4249-83d8-d9c6c61c7b8f/4258-0 00:26:36.507 --> 00:26:39.658 Then we recompute the residuals using this new model and observe 436caf55-234f-4249-83d8-d9c6c61c7b8f/4258-1 00:26:39.658 --> 00:26:42.663 new estimates for the event term within the residuals and the 436caf55-234f-4249-83d8-d9c6c61c7b8f/4258-2 00:26:42.663 --> 00:26:43.197 site terms. 436caf55-234f-4249-83d8-d9c6c61c7b8f/4294-0 00:26:44.847 --> 00:26:47.862 He will repeat steps three and four until the model 436caf55-234f-4249-83d8-d9c6c61c7b8f/4294-1 00:26:47.862 --> 00:26:51.110 coefficients in random effects event terms inside terms 436caf55-234f-4249-83d8-d9c6c61c7b8f/4294-2 00:26:51.110 --> 00:26:54.763 converge and the reason why they will change is because you'll 436caf55-234f-4249-83d8-d9c6c61c7b8f/4294-3 00:26:54.763 --> 00:26:56.387 see in the subsequent steps. 436caf55-234f-4249-83d8-d9c6c61c7b8f/4341-0 00:26:57.807 --> 00:27:01.810 We after we've removed the first couple or the the major path 436caf55-234f-4249-83d8-d9c6c61c7b8f/4341-1 00:27:01.810 --> 00:27:05.490 effects now we can estimate event terms using the entire 436caf55-234f-4249-83d8-d9c6c61c7b8f/4341-2 00:27:05.490 --> 00:27:09.235 data set. So all observations pass 100 kilometers and now 436caf55-234f-4249-83d8-d9c6c61c7b8f/4341-3 00:27:09.235 --> 00:27:13.109 anelastic effects may have an impact on those random effect 436caf55-234f-4249-83d8-d9c6c61c7b8f/4341-4 00:27:13.109 --> 00:27:13.497 terms. 436caf55-234f-4249-83d8-d9c6c61c7b8f/4354-0 00:27:14.517 --> 00:27:17.140 Lastly, we evaluate the performance through residuals 436caf55-234f-4249-83d8-d9c6c61c7b8f/4354-1 00:27:17.140 --> 00:27:18.987 analysis and dispersion calculations. 436caf55-234f-4249-83d8-d9c6c61c7b8f/4416-0 00:27:20.247 --> 00:27:24.033 So what trends did we initially see? So if we look at the data 436caf55-234f-4249-83d8-d9c6c61c7b8f/4416-1 00:27:24.033 --> 00:27:27.939 set as a whole, we can have for one particular intensity measure 436caf55-234f-4249-83d8-d9c6c61c7b8f/4416-2 00:27:27.939 --> 00:27:31.665 .1 seconds .1 seconds versus distance. You can't see that the 436caf55-234f-4249-83d8-d9c6c61c7b8f/4416-3 00:27:31.665 --> 00:27:34.850 bin means are relatively flat. There's quite a large 436caf55-234f-4249-83d8-d9c6c61c7b8f/4416-4 00:27:34.850 --> 00:27:37.976 dispersion, maybe slightly trending downward, large 436caf55-234f-4249-83d8-d9c6c61c7b8f/4416-5 00:27:37.976 --> 00:27:41.942 magnitude events where we didn't add a significant amount of data 436caf55-234f-4249-83d8-d9c6c61c7b8f/4416-6 00:27:41.942 --> 00:27:44.767 because there's not that many available to us. 436caf55-234f-4249-83d8-d9c6c61c7b8f/4479-0 00:27:45.557 --> 00:27:49.292 Do have a slight bend down, but for the most part I'm pretty 436caf55-234f-4249-83d8-d9c6c61c7b8f/4479-1 00:27:49.292 --> 00:27:52.721 unbiased, but for small magnitude events where we added 436caf55-234f-4249-83d8-d9c6c61c7b8f/4479-2 00:27:52.721 --> 00:27:56.640 a substantial amount of data, we essentially doubled the amount 436caf55-234f-4249-83d8-d9c6c61c7b8f/4479-3 00:27:56.640 --> 00:28:00.620 of data available to us. We can see that there's a pretty strong 436caf55-234f-4249-83d8-d9c6c61c7b8f/4479-4 00:28:00.620 --> 00:28:04.417 bent down to about -, .05, so faster attenuation as a kind of 436caf55-234f-4249-83d8-d9c6c61c7b8f/4479-5 00:28:04.417 --> 00:28:05.887 observed for California. 436caf55-234f-4249-83d8-d9c6c61c7b8f/4562-0 00:28:06.807 --> 00:28:10.726 When we include data from more regions that weren't currently 436caf55-234f-4249-83d8-d9c6c61c7b8f/4562-1 00:28:10.726 --> 00:28:14.393 included in Ng W 2, so we see these variations across the 436caf55-234f-4249-83d8-d9c6c61c7b8f/4562-2 00:28:14.393 --> 00:28:17.934 state that we need to capture. One way to look at these 436caf55-234f-4249-83d8-d9c6c61c7b8f/4562-3 00:28:17.934 --> 00:28:21.537 variable rates are to bin observations based off of what 436caf55-234f-4249-83d8-d9c6c61c7b8f/4562-4 00:28:21.537 --> 00:28:25.457 subregion the event originates in. So for example, all events 436caf55-234f-4249-83d8-d9c6c61c7b8f/4562-5 00:28:25.457 --> 00:28:29.250 that originate in the North Coast would be bent here and we 436caf55-234f-4249-83d8-d9c6c61c7b8f/4562-6 00:28:29.250 --> 00:28:33.043 can see that, yeah, the North Coast has faster attenuation, 436caf55-234f-4249-83d8-d9c6c61c7b8f/4562-7 00:28:33.043 --> 00:28:36.837 which aligns well with what the the QS model would suggest. 436caf55-234f-4249-83d8-d9c6c61c7b8f/4573-0 00:28:37.147 --> 00:28:40.176 Sierra Nevada shows slower attenuation. It's bending 436caf55-234f-4249-83d8-d9c6c61c7b8f/4573-1 00:28:40.176 --> 00:28:40.577 upward. 436caf55-234f-4249-83d8-d9c6c61c7b8f/4602-0 00:28:41.327 --> 00:28:44.816 So. So we're over predicting the attenuation and therapeutic 436caf55-234f-4249-83d8-d9c6c61c7b8f/4602-1 00:28:44.816 --> 00:28:48.363 model and and a range of other results. But I'm just going to 436caf55-234f-4249-83d8-d9c6c61c7b8f/4602-2 00:28:48.363 --> 00:28:50.937 focus on these two for the purposes of time. 436caf55-234f-4249-83d8-d9c6c61c7b8f/4624-0 00:28:51.647 --> 00:28:55.849 But in these plots you can say I have also colored each data 436caf55-234f-4249-83d8-d9c6c61c7b8f/4624-1 00:28:55.849 --> 00:28:59.707 point based off of the path weight within a region war. 436caf55-234f-4249-83d8-d9c6c61c7b8f/4641-0 00:29:00.417 --> 00:29:03.632 Now that weight is pretty much the same as how paths are 436caf55-234f-4249-83d8-d9c6c61c7b8f/4641-1 00:29:03.632 --> 00:29:05.607 computed in the cell based method. 436caf55-234f-4249-83d8-d9c6c61c7b8f/4679-0 00:29:06.547 --> 00:29:10.064 If we assume the simplest case where the source and the site 436caf55-234f-4249-83d8-d9c6c61c7b8f/4679-1 00:29:10.064 --> 00:29:13.755 are in the same region are the same sub region, rather the wait 436caf55-234f-4249-83d8-d9c6c61c7b8f/4679-2 00:29:13.755 --> 00:29:17.388 for that sub region is gonna be 1, all other weights are gonna 436caf55-234f-4249-83d8-d9c6c61c7b8f/4679-3 00:29:17.388 --> 00:29:17.677 be 0. 436caf55-234f-4249-83d8-d9c6c61c7b8f/4682-0 00:29:18.617 --> 00:29:19.527 So each. 436caf55-234f-4249-83d8-d9c6c61c7b8f/4720-0 00:29:20.577 --> 00:29:23.804 Observation essentially has a vector of weights with the 436caf55-234f-4249-83d8-d9c6c61c7b8f/4720-1 00:29:23.804 --> 00:29:27.200 number of elements in that vector is equal to the number of 436caf55-234f-4249-83d8-d9c6c61c7b8f/4720-2 00:29:27.200 --> 00:29:30.767 subregions and the sum of all those weeds is always one. So if 436caf55-234f-4249-83d8-d9c6c61c7b8f/4720-3 00:29:30.767 --> 00:29:34.164 we have a path that travels 2 sub regions where you can see 436caf55-234f-4249-83d8-d9c6c61c7b8f/4720-4 00:29:34.164 --> 00:29:34.447 this. 436caf55-234f-4249-83d8-d9c6c61c7b8f/4748-0 00:29:35.327 --> 00:29:39.022 For site two, we have about 70% and 30% in these respective 436caf55-234f-4249-83d8-d9c6c61c7b8f/4748-1 00:29:39.022 --> 00:29:42.901 regions and it's the same case whether you have something more 436caf55-234f-4249-83d8-d9c6c61c7b8f/4748-2 00:29:42.901 --> 00:29:46.227 complicated traveling through five or so regions and. 436caf55-234f-4249-83d8-d9c6c61c7b8f/4805-0 00:29:47.137 --> 00:29:50.833 And that is how we're gonna kind of smooth out the effects to 436caf55-234f-4249-83d8-d9c6c61c7b8f/4805-1 00:29:50.833 --> 00:29:54.529 connect the two dots, the dots that we have here. And if this 436caf55-234f-4249-83d8-d9c6c61c7b8f/4805-2 00:29:54.529 --> 00:29:58.047 was intensity versus just in space, we'd have a dot at the 436caf55-234f-4249-83d8-d9c6c61c7b8f/4805-3 00:29:58.047 --> 00:30:01.803 intensity of the site and that's how we're gonna just create a 436caf55-234f-4249-83d8-d9c6c61c7b8f/4805-4 00:30:01.803 --> 00:30:04.307 smooth transition between those two dots. 436caf55-234f-4249-83d8-d9c6c61c7b8f/4807-0 00:30:05.977 --> 00:30:06.507 So. 436caf55-234f-4249-83d8-d9c6c61c7b8f/4822-0 00:30:07.827 --> 00:30:11.492 What model do we use? So here's the modified ground motion model 436caf55-234f-4249-83d8-d9c6c61c7b8f/4822-1 00:30:11.492 --> 00:30:12.677 that we're proposing. 436caf55-234f-4249-83d8-d9c6c61c7b8f/4890-0 00:30:13.307 --> 00:30:16.551 Here we have constant adjustments that are usually 436caf55-234f-4249-83d8-d9c6c61c7b8f/4890-1 00:30:16.551 --> 00:30:20.432 contained within the source term of the ground motion model, 436caf55-234f-4249-83d8-d9c6c61c7b8f/4890-2 00:30:20.432 --> 00:30:24.186 which C not would be one of those coefficients we add on 2 436caf55-234f-4249-83d8-d9c6c61c7b8f/4890-3 00:30:24.186 --> 00:30:27.812 new terms. This first term is related to a side study of 436caf55-234f-4249-83d8-d9c6c61c7b8f/4890-4 00:30:27.812 --> 00:30:31.693 induced events in the Geysers region in Northern California. 436caf55-234f-4249-83d8-d9c6c61c7b8f/4890-5 00:30:31.693 --> 00:30:35.701 We found that they on average have source effects that produce 436caf55-234f-4249-83d8-d9c6c61c7b8f/4890-6 00:30:35.701 --> 00:30:39.518 weaker ground motions than what they're gonna ground motion 436caf55-234f-4249-83d8-d9c6c61c7b8f/4890-7 00:30:39.518 --> 00:30:40.727 model will predict. 436caf55-234f-4249-83d8-d9c6c61c7b8f/4951-0 00:30:41.537 --> 00:30:45.179 So, but they do not trend with any predictor variables, so we 436caf55-234f-4249-83d8-d9c6c61c7b8f/4951-1 00:30:45.179 --> 00:30:48.586 just add a constant. We did a rigorous study to to assess 436caf55-234f-4249-83d8-d9c6c61c7b8f/4951-2 00:30:48.586 --> 00:30:52.462 whether path effects for induced events were the same as tectonic 436caf55-234f-4249-83d8-d9c6c61c7b8f/4951-3 00:30:52.462 --> 00:30:56.046 events in the same sub region and which ultimately concluded 436caf55-234f-4249-83d8-d9c6c61c7b8f/4951-4 00:30:56.046 --> 00:30:59.746 in that the path effects are the same, which makes sense. It's 436caf55-234f-4249-83d8-d9c6c61c7b8f/4951-5 00:30:59.746 --> 00:31:03.329 just the source effects that change the paths have to travel 436caf55-234f-4249-83d8-d9c6c61c7b8f/4951-6 00:31:03.329 --> 00:31:04.857 through the same material. 436caf55-234f-4249-83d8-d9c6c61c7b8f/4959-0 00:31:05.837 --> 00:31:07.867 But we also add a regional adjustment. 436caf55-234f-4249-83d8-d9c6c61c7b8f/4987-0 00:31:08.767 --> 00:31:12.903 And we modify the path term. The other terms are taken exactly 436caf55-234f-4249-83d8-d9c6c61c7b8f/4987-1 00:31:12.903 --> 00:31:16.449 from BSA 14. We don't adjust these because we're just 436caf55-234f-4249-83d8-d9c6c61c7b8f/4987-2 00:31:16.449 --> 00:31:20.257 interested in adjusting the regional subregional effects. 436caf55-234f-4249-83d8-d9c6c61c7b8f/5030-0 00:31:21.597 --> 00:31:24.447 Before we know that path model is the sum of geometric 436caf55-234f-4249-83d8-d9c6c61c7b8f/5030-1 00:31:24.447 --> 00:31:27.245 spreading in aspiration, I'm only going to modify the 436caf55-234f-4249-83d8-d9c6c61c7b8f/5030-2 00:31:27.245 --> 00:31:30.614 analytic continuation. Assuming geometric spreading is constant, 436caf55-234f-4249-83d8-d9c6c61c7b8f/5030-3 00:31:30.614 --> 00:31:33.464 it's likely not constant, but it's probably a lot less 436caf55-234f-4249-83d8-d9c6c61c7b8f/5030-4 00:31:33.464 --> 00:31:36.314 variable than analytic continuation, so this works for 436caf55-234f-4249-83d8-d9c6c61c7b8f/5030-5 00:31:36.314 --> 00:31:37.247 the current study. 436caf55-234f-4249-83d8-d9c6c61c7b8f/5074-0 00:31:38.727 --> 00:31:42.915 We just take the functional form proposed by BSA 14, which is a 436caf55-234f-4249-83d8-d9c6c61c7b8f/5074-1 00:31:42.915 --> 00:31:46.908 global analytic continuation plus an adjustment for regional 436caf55-234f-4249-83d8-d9c6c61c7b8f/5074-2 00:31:46.908 --> 00:31:51.162 effects times. Distance distance is just for for basic 14 square 436caf55-234f-4249-83d8-d9c6c61c7b8f/5074-3 00:31:51.162 --> 00:31:54.827 root of the joint aboard distance plus the coefficient. 436caf55-234f-4249-83d8-d9c6c61c7b8f/5078-0 00:31:56.867 --> 00:31:57.917 However this. 436caf55-234f-4249-83d8-d9c6c61c7b8f/5089-0 00:31:58.867 --> 00:32:03.597 Delta C3 star is pretty much the cell based method. 436caf55-234f-4249-83d8-d9c6c61c7b8f/5130-0 00:32:04.227 --> 00:32:07.637 But we're no longer computing individual attenuation for each 436caf55-234f-4249-83d8-d9c6c61c7b8f/5130-1 00:32:07.637 --> 00:32:10.772 path. We're just getting one average attenuation for the 436caf55-234f-4249-83d8-d9c6c61c7b8f/5130-2 00:32:10.772 --> 00:32:14.127 entire path. By combining these weights mathematically, it's 436caf55-234f-4249-83d8-d9c6c61c7b8f/5130-3 00:32:14.127 --> 00:32:17.702 pretty much the same. You end up at the same result. So the same 436caf55-234f-4249-83d8-d9c6c61c7b8f/5130-4 00:32:17.702 --> 00:32:21.167 point, but the path that gets there is a little bit different. 436caf55-234f-4249-83d8-d9c6c61c7b8f/5215-0 00:32:22.417 --> 00:32:26.192 What the impact has on our model, if we look at intensity 436caf55-234f-4249-83d8-d9c6c61c7b8f/5215-1 00:32:26.192 --> 00:32:30.487 versus distance space, the black line is the ergodic model of the 436caf55-234f-4249-83d8-d9c6c61c7b8f/5215-2 00:32:30.487 --> 00:32:34.718 essay 14. If we have a positive delta C3 star, so travels mostly 436caf55-234f-4249-83d8-d9c6c61c7b8f/5215-3 00:32:34.718 --> 00:32:38.688 through regions with relatively slower rates of attenuation, 436caf55-234f-4249-83d8-d9c6c61c7b8f/5215-4 00:32:38.688 --> 00:32:42.723 we're gonna reduce our we're gonna subtract less energy as we 436caf55-234f-4249-83d8-d9c6c61c7b8f/5215-5 00:32:42.723 --> 00:32:46.498 travel farther away. The opposite was true, as if if this 436caf55-234f-4249-83d8-d9c6c61c7b8f/5215-6 00:32:46.498 --> 00:32:50.338 path travels mostly through sites, through subregions with 436caf55-234f-4249-83d8-d9c6c61c7b8f/5215-7 00:32:50.338 --> 00:32:53.007 faster generation, where we'll actually. 436caf55-234f-4249-83d8-d9c6c61c7b8f/5241-0 00:32:53.127 --> 00:32:57.126 Subtract more energy. The constant region subregional 436caf55-234f-4249-83d8-d9c6c61c7b8f/5241-1 00:32:57.126 --> 00:33:01.867 adjustment for source effects is kind of self-explanatory. It's 436caf55-234f-4249-83d8-d9c6c61c7b8f/5241-2 00:33:01.867 --> 00:33:05.867 just a constant shift vertically along to the Y axis. 436caf55-234f-4249-83d8-d9c6c61c7b8f/5288-0 00:33:07.437 --> 00:33:11.638 I did mention before that we use iterations in our method and 436caf55-234f-4249-83d8-d9c6c61c7b8f/5288-1 00:33:11.638 --> 00:33:15.840 this kind of highlights the fact that things will change even 436caf55-234f-4249-83d8-d9c6c61c7b8f/5288-2 00:33:15.840 --> 00:33:20.312 after the first iteration. We do see the the largest changes with 436caf55-234f-4249-83d8-d9c6c61c7b8f/5288-3 00:33:20.312 --> 00:33:24.378 respect to the reasonable and elastic coefficient, which is 436caf55-234f-4249-83d8-d9c6c61c7b8f/5288-4 00:33:24.378 --> 00:33:27.767 the top row, and even the regional chorus effect. 436caf55-234f-4249-83d8-d9c6c61c7b8f/5293-0 00:33:28.497 --> 00:33:29.547 Which is the bottom row. 436caf55-234f-4249-83d8-d9c6c61c7b8f/5316-0 00:33:32.737 --> 00:33:36.084 For these particular subregions shown here in North Coast Sierra 436caf55-234f-4249-83d8-d9c6c61c7b8f/5316-1 00:33:36.084 --> 00:33:39.226 Nevada and in California, but they eventually converge after 436caf55-234f-4249-83d8-d9c6c61c7b8f/5316-2 00:33:39.226 --> 00:33:40.977 iterations and and become stable. 436caf55-234f-4249-83d8-d9c6c61c7b8f/5352-0 00:33:42.277 --> 00:33:46.065 Now I also mentioned that we're interested in this because of 436caf55-234f-4249-83d8-d9c6c61c7b8f/5352-1 00:33:46.065 --> 00:33:49.669 the implications path effects can have on non ergodic site 436caf55-234f-4249-83d8-d9c6c61c7b8f/5352-2 00:33:49.669 --> 00:33:53.335 response. Remember us within residual contains site effects 436caf55-234f-4249-83d8-d9c6c61c7b8f/5352-3 00:33:53.335 --> 00:33:55.657 which are contained in the site term. 436caf55-234f-4249-83d8-d9c6c61c7b8f/5399-0 00:33:56.507 --> 00:33:59.773 And also path the path variability our path bias which 436caf55-234f-4249-83d8-d9c6c61c7b8f/5399-1 00:33:59.773 --> 00:34:03.039 can be in the remaining residual, we can see that site 436caf55-234f-4249-83d8-d9c6c61c7b8f/5399-2 00:34:03.039 --> 00:34:06.365 turns similarly will converge while this model is being 436caf55-234f-4249-83d8-d9c6c61c7b8f/5399-3 00:34:06.365 --> 00:34:10.225 developed and if we were to look at the ergodic prediction shown 436caf55-234f-4249-83d8-d9c6c61c7b8f/5399-4 00:34:10.225 --> 00:34:13.966 in black at the very beginning without any adjustments and the 436caf55-234f-4249-83d8-d9c6c61c7b8f/5399-5 00:34:13.966 --> 00:34:15.867 proposed subregional adjustment. 436caf55-234f-4249-83d8-d9c6c61c7b8f/5462-0 00:34:16.887 --> 00:34:20.716 Umm. And read at the very end of our adjustments, we can see that 436caf55-234f-4249-83d8-d9c6c61c7b8f/5462-1 00:34:20.716 --> 00:34:24.196 there's a pretty significant effect on our non ergodic site 436caf55-234f-4249-83d8-d9c6c61c7b8f/5462-2 00:34:24.196 --> 00:34:27.561 response. In this case it's a site in Northern California 436caf55-234f-4249-83d8-d9c6c61c7b8f/5462-3 00:34:27.561 --> 00:34:31.042 where it recorded many events that travel through the North 436caf55-234f-4249-83d8-d9c6c61c7b8f/5462-4 00:34:31.042 --> 00:34:34.465 Coast region and you can see those that strong North Coast 436caf55-234f-4249-83d8-d9c6c61c7b8f/5462-5 00:34:34.465 --> 00:34:37.830 bias that we saw a couple slides ago was causing our site 436caf55-234f-4249-83d8-d9c6c61c7b8f/5462-6 00:34:37.830 --> 00:34:41.137 response to be predicted lower than what it actually is. 436caf55-234f-4249-83d8-d9c6c61c7b8f/5526-0 00:34:41.807 --> 00:34:45.694 Another thing that isn't too easy to see in this spot is that 436caf55-234f-4249-83d8-d9c6c61c7b8f/5526-1 00:34:45.694 --> 00:34:49.455 we actually do decrease the dispersion around our site term 436caf55-234f-4249-83d8-d9c6c61c7b8f/5526-2 00:34:49.455 --> 00:34:53.405 or the uncertainty by quite a bit the the ergodic or the black 436caf55-234f-4249-83d8-d9c6c61c7b8f/5526-3 00:34:53.405 --> 00:34:57.166 results. The the minimum is where my cursor is, the maximum 436caf55-234f-4249-83d8-d9c6c61c7b8f/5526-4 00:34:57.166 --> 00:35:01.053 is up here. But guess what, the maximum red point is here and 436caf55-234f-4249-83d8-d9c6c61c7b8f/5526-5 00:35:01.053 --> 00:35:03.937 the minimum red point is down here, so we're. 436caf55-234f-4249-83d8-d9c6c61c7b8f/5552-0 00:35:05.617 --> 00:35:09.296 Being more certain of a of our site response by accounting for 436caf55-234f-4249-83d8-d9c6c61c7b8f/5552-1 00:35:09.296 --> 00:35:12.917 systematic path effects, if we were to to do a much finer sub 436caf55-234f-4249-83d8-d9c6c61c7b8f/5552-2 00:35:12.917 --> 00:35:16.479 realization, it'd be likely that these results would be even 436caf55-234f-4249-83d8-d9c6c61c7b8f/5552-3 00:35:16.479 --> 00:35:16.947 tighter. 436caf55-234f-4249-83d8-d9c6c61c7b8f/5577-0 00:35:17.977 --> 00:35:21.572 I I would not know whether or not this the the Nonagon site 436caf55-234f-4249-83d8-d9c6c61c7b8f/5577-1 00:35:21.572 --> 00:35:25.348 response that I've shown here would be the same, though that's 436caf55-234f-4249-83d8-d9c6c61c7b8f/5577-2 00:35:25.348 --> 00:35:26.487 for a future study. 436caf55-234f-4249-83d8-d9c6c61c7b8f/5656-0 00:35:27.887 --> 00:35:31.812 So our final model coefficients, I won't spend too much time on 436caf55-234f-4249-83d8-d9c6c61c7b8f/5656-1 00:35:31.812 --> 00:35:35.370 this slide because it's just showing coefficients not too 436caf55-234f-4249-83d8-d9c6c61c7b8f/5656-2 00:35:35.370 --> 00:35:39.295 much to interpret here. The main takeaways is if we compare our 436caf55-234f-4249-83d8-d9c6c61c7b8f/5656-3 00:35:39.295 --> 00:35:42.730 sub regional and elastic coefficients shown on the left 436caf55-234f-4249-83d8-d9c6c61c7b8f/5656-4 00:35:42.730 --> 00:35:46.288 versus period to the regional attenuation coefficients of 436caf55-234f-4249-83d8-d9c6c61c7b8f/5656-5 00:35:46.288 --> 00:35:49.907 China and Turkey and BSA 14 shown by the solid black line, 436caf55-234f-4249-83d8-d9c6c61c7b8f/5656-6 00:35:49.907 --> 00:35:53.955 and Japan and Italy shown by the dashed line, we see that there's 436caf55-234f-4249-83d8-d9c6c61c7b8f/5656-7 00:35:53.955 --> 00:35:57.697 quite a bit of variation within the state which some of the. 436caf55-234f-4249-83d8-d9c6c61c7b8f/5720-0 00:35:58.197 --> 00:36:02.698 The the magnitudes of these anelastic effects are similar to 436caf55-234f-4249-83d8-d9c6c61c7b8f/5720-1 00:36:02.698 --> 00:36:07.051 regional effects that we see globally. The exception being 436caf55-234f-4249-83d8-d9c6c61c7b8f/5720-2 00:36:07.051 --> 00:36:11.331 the North Coast, which has significant faster attenuation 436caf55-234f-4249-83d8-d9c6c61c7b8f/5720-3 00:36:11.331 --> 00:36:15.611 or more negative delta C3 then the other subregions. This 436caf55-234f-4249-83d8-d9c6c61c7b8f/5720-4 00:36:15.611 --> 00:36:20.333 northeastern sub region. There's not a lot of data, but it also 436caf55-234f-4249-83d8-d9c6c61c7b8f/5720-5 00:36:20.333 --> 00:36:24.613 is suggesting that there's faster attenuation in Northern 436caf55-234f-4249-83d8-d9c6c61c7b8f/5720-6 00:36:24.613 --> 00:36:26.237 California in general. 436caf55-234f-4249-83d8-d9c6c61c7b8f/5792-0 00:36:26.987 --> 00:36:30.640 On the right I'm showing the constant adjustment for source 436caf55-234f-4249-83d8-d9c6c61c7b8f/5792-1 00:36:30.640 --> 00:36:34.597 effects and you might ask why I didn't scale the vertical axis a 436caf55-234f-4249-83d8-d9c6c61c7b8f/5792-2 00:36:34.597 --> 00:36:38.007 bit differently, and it's because this vertical axis is 436caf55-234f-4249-83d8-d9c6c61c7b8f/5792-3 00:36:38.007 --> 00:36:41.903 scaled to reflect the magnitude of the ergodic constants in the 436caf55-234f-4249-83d8-d9c6c61c7b8f/5792-4 00:36:41.903 --> 00:36:45.678 graph motion model. So you can see that these are pretty much 436caf55-234f-4249-83d8-d9c6c61c7b8f/5792-5 00:36:45.678 --> 00:36:49.636 negligible when we compare them to the constant adjustments that 436caf55-234f-4249-83d8-d9c6c61c7b8f/5792-6 00:36:49.636 --> 00:36:53.228 are in a ground motion model. There is some effect at long 436caf55-234f-4249-83d8-d9c6c61c7b8f/5792-7 00:36:53.228 --> 00:36:55.907 periods, but for the most part there isn't. 436caf55-234f-4249-83d8-d9c6c61c7b8f/5799-0 00:36:56.627 --> 00:37:00.157 Significant impact of this constant adjustment, which is. 436caf55-234f-4249-83d8-d9c6c61c7b8f/5824-0 00:37:00.857 --> 00:37:05.492 Implying that the model that we've formulated mainly affects 436caf55-234f-4249-83d8-d9c6c61c7b8f/5824-1 00:37:05.492 --> 00:37:10.279 the analytic continuation, which is and which implies that the 436caf55-234f-4249-83d8-d9c6c61c7b8f/5824-2 00:37:10.279 --> 00:37:12.407 model is doing its job well. 436caf55-234f-4249-83d8-d9c6c61c7b8f/5833-0 00:37:14.317 --> 00:37:16.547 So now I wanted to discuss the model performance. 436caf55-234f-4249-83d8-d9c6c61c7b8f/5843-0 00:37:17.677 --> 00:37:21.574 So first just some overall performance metrics. If we look 436caf55-234f-4249-83d8-d9c6c61c7b8f/5843-1 00:37:21.574 --> 00:37:22.367 at the bias. 436caf55-234f-4249-83d8-d9c6c61c7b8f/5886-0 00:37:23.057 --> 00:37:27.048 Umm, that's taken from the mixed effects methods that ergodic 436caf55-234f-4249-83d8-d9c6c61c7b8f/5886-1 00:37:27.048 --> 00:37:30.976 model BSA 14 shown in black proposed model shown in blue. We 436caf55-234f-4249-83d8-d9c6c61c7b8f/5886-2 00:37:30.976 --> 00:37:35.096 can see that we are reducing the bias we're closer to zero when 436caf55-234f-4249-83d8-d9c6c61c7b8f/5886-3 00:37:35.096 --> 00:37:38.702 compared to the ergodic model for short to intermediate 436caf55-234f-4249-83d8-d9c6c61c7b8f/5886-4 00:37:38.702 --> 00:37:39.217 periods. 436caf55-234f-4249-83d8-d9c6c61c7b8f/5920-0 00:37:39.967 --> 00:37:43.700 Analyst effects pretty much go away at long periods, so that's 436caf55-234f-4249-83d8-d9c6c61c7b8f/5920-1 00:37:43.700 --> 00:37:47.374 why they the results are pretty much identical for those long 436caf55-234f-4249-83d8-d9c6c61c7b8f/5920-2 00:37:47.374 --> 00:37:50.989 periods. But the model does a good job. Maybe we're slightly 436caf55-234f-4249-83d8-d9c6c61c7b8f/5920-3 00:37:50.989 --> 00:37:51.997 under predicting. 436caf55-234f-4249-83d8-d9c6c61c7b8f/5937-0 00:37:52.677 --> 00:37:57.805 The the intensities, but we're less biased than what the SSA 14 436caf55-234f-4249-83d8-d9c6c61c7b8f/5937-1 00:37:57.805 --> 00:37:59.007 was on average. 436caf55-234f-4249-83d8-d9c6c61c7b8f/5969-0 00:37:59.767 --> 00:38:03.776 When we look at similar plots to what I showed before, looking at 436caf55-234f-4249-83d8-d9c6c61c7b8f/5969-1 00:38:03.776 --> 00:38:07.482 all events, large events and small events, we see that these 436caf55-234f-4249-83d8-d9c6c61c7b8f/5969-2 00:38:07.482 --> 00:38:10.337 bin means are a lot more centered about the 0. 436caf55-234f-4249-83d8-d9c6c61c7b8f/6037-0 00:38:11.167 --> 00:38:14.439 Especially for the large magnitude events you can see 436caf55-234f-4249-83d8-d9c6c61c7b8f/6037-1 00:38:14.439 --> 00:38:17.954 they've pretty much are are pretty perfect right at zero. 436caf55-234f-4249-83d8-d9c6c61c7b8f/6037-2 00:38:17.954 --> 00:38:21.348 The small events are a lot better than before. There is 436caf55-234f-4249-83d8-d9c6c61c7b8f/6037-3 00:38:21.348 --> 00:38:25.166 some slight downward dipping at the very far distance, but for 436caf55-234f-4249-83d8-d9c6c61c7b8f/6037-4 00:38:25.166 --> 00:38:28.742 the most part, the systematic path bias has been corrected 436caf55-234f-4249-83d8-d9c6c61c7b8f/6037-5 00:38:28.742 --> 00:38:31.954 with some refinement. We can probably get this to be 436caf55-234f-4249-83d8-d9c6c61c7b8f/6037-6 00:38:31.954 --> 00:38:35.227 perfectly flat or nearly perfectly flat, I would say. 436caf55-234f-4249-83d8-d9c6c61c7b8f/6093-0 00:38:37.817 --> 00:38:40.829 Like I said, we see a pretty good reduction of biostat far 436caf55-234f-4249-83d8-d9c6c61c7b8f/6093-1 00:38:40.829 --> 00:38:43.891 distances if we look at those sub regions that we looked at 436caf55-234f-4249-83d8-d9c6c61c7b8f/6093-2 00:38:43.891 --> 00:38:46.954 before, we can see that OK, the North Coast is pretty flat. 436caf55-234f-4249-83d8-d9c6c61c7b8f/6093-3 00:38:46.954 --> 00:38:49.915 There is definitely some variation still because we can't 436caf55-234f-4249-83d8-d9c6c61c7b8f/6093-4 00:38:49.915 --> 00:38:52.876 correct all distances. We can only really correct the far 436caf55-234f-4249-83d8-d9c6c61c7b8f/6093-5 00:38:52.876 --> 00:38:53.387 distances. 436caf55-234f-4249-83d8-d9c6c61c7b8f/6116-0 00:38:53.987 --> 00:38:58.904 But those far distances for all regions are pretty centered 436caf55-234f-4249-83d8-d9c6c61c7b8f/6116-1 00:38:58.904 --> 00:39:03.412 about 0 and the the the colored dots are a little less 436caf55-234f-4249-83d8-d9c6c61c7b8f/6116-2 00:39:03.412 --> 00:39:05.297 systematic than before. 436caf55-234f-4249-83d8-d9c6c61c7b8f/6121-0 00:39:05.967 --> 00:39:07.127 But there's still some. 436caf55-234f-4249-83d8-d9c6c61c7b8f/6156-0 00:39:07.557 --> 00:39:11.134 Umm systematics, like there's still cluster of a lot of events 436caf55-234f-4249-83d8-d9c6c61c7b8f/6156-1 00:39:11.134 --> 00:39:14.598 which have weight one for North Coast here. So maybe there's 436caf55-234f-4249-83d8-d9c6c61c7b8f/6156-2 00:39:14.598 --> 00:39:18.005 some sub regionalization for this N sub sub regionalization 436caf55-234f-4249-83d8-d9c6c61c7b8f/6156-3 00:39:18.005 --> 00:39:20.447 that needs to be done for the North Coast. 436caf55-234f-4249-83d8-d9c6c61c7b8f/6205-0 00:39:21.157 --> 00:39:24.724 Uh region and if we were to look at this in intensity versus 436caf55-234f-4249-83d8-d9c6c61c7b8f/6205-1 00:39:24.724 --> 00:39:27.999 distance, space or Gothic tradition on the top, the sub 436caf55-234f-4249-83d8-d9c6c61c7b8f/6205-2 00:39:27.999 --> 00:39:31.157 regionalization prediction on the bottom, you can see 436caf55-234f-4249-83d8-d9c6c61c7b8f/6205-3 00:39:31.157 --> 00:39:34.959 obviously overpredicted for the range of magnitudes that's shown 436caf55-234f-4249-83d8-d9c6c61c7b8f/6205-4 00:39:34.959 --> 00:39:38.292 here for the ergodic. But we capture the response or the 436caf55-234f-4249-83d8-d9c6c61c7b8f/6205-5 00:39:38.292 --> 00:39:40.047 observed behavior pretty well. 436caf55-234f-4249-83d8-d9c6c61c7b8f/6212-0 00:39:40.837 --> 00:39:43.097 When we include this sub regionalization. 436caf55-234f-4249-83d8-d9c6c61c7b8f/6293-0 00:39:45.127 --> 00:39:48.972 So I also want to discuss some dispersion of variability. This 436caf55-234f-4249-83d8-d9c6c61c7b8f/6293-1 00:39:48.972 --> 00:39:52.695 is an effect of our which are computed through mixed effects 436caf55-234f-4249-83d8-d9c6c61c7b8f/6293-2 00:39:52.695 --> 00:39:56.601 results. Photo variability just the standard deviation of total 436caf55-234f-4249-83d8-d9c6c61c7b8f/6293-3 00:39:56.601 --> 00:39:59.958 residuals which can be partitioned into the between of 436caf55-234f-4249-83d8-d9c6c61c7b8f/6293-4 00:39:59.958 --> 00:40:03.497 that variability which is standby action of the pet terms 436caf55-234f-4249-83d8-d9c6c61c7b8f/6293-5 00:40:03.497 --> 00:40:07.403 and the within a bit viability standard deviation of the within 436caf55-234f-4249-83d8-d9c6c61c7b8f/6293-6 00:40:07.403 --> 00:40:11.004 a bit residuals which can be then further partitioned into 436caf55-234f-4249-83d8-d9c6c61c7b8f/6293-7 00:40:11.004 --> 00:40:14.239 site to site variability VS2S which is important for 436caf55-234f-4249-83d8-d9c6c61c7b8f/6293-8 00:40:14.239 --> 00:40:15.277 nonnarcotic site. 436caf55-234f-4249-83d8-d9c6c61c7b8f/6296-0 00:40:15.367 --> 00:40:15.707 What is? 436caf55-234f-4249-83d8-d9c6c61c7b8f/6308-0 00:40:16.347 --> 00:40:20.077 And single station variability which contains within it the. 436caf55-234f-4249-83d8-d9c6c61c7b8f/6328-0 00:40:20.897 --> 00:40:25.411 A path to path variability. So the main thing we're focused on 436caf55-234f-4249-83d8-d9c6c61c7b8f/6328-1 00:40:25.411 --> 00:40:29.997 is what is the the impact of fee as S of suburbanization on PS. 436caf55-234f-4249-83d8-d9c6c61c7b8f/6392-0 00:40:30.867 --> 00:40:35.310 Before I get into the results of our model, we did add quite a 436caf55-234f-4249-83d8-d9c6c61c7b8f/6392-1 00:40:35.310 --> 00:40:39.118 lot of data, so we were interested to see what is the 436caf55-234f-4249-83d8-d9c6c61c7b8f/6392-2 00:40:39.118 --> 00:40:43.421 general effect of adding all this data on BSA 14 dispersion. 436caf55-234f-4249-83d8-d9c6c61c7b8f/6392-3 00:40:43.421 --> 00:40:47.088 You can see the black observations are the complete 436caf55-234f-4249-83d8-d9c6c61c7b8f/6392-4 00:40:47.088 --> 00:40:51.531 data set. The Gray are what was is, what's computed when, just 436caf55-234f-4249-83d8-d9c6c61c7b8f/6392-5 00:40:51.531 --> 00:40:55.904 using the NGOs 2 data set that satisfies our criteria for for 436caf55-234f-4249-83d8-d9c6c61c7b8f/6392-6 00:40:55.904 --> 00:40:56.257 data. 436caf55-234f-4249-83d8-d9c6c61c7b8f/6459-0 00:40:56.967 --> 00:41:00.755 You can see in general we increase variability at short to 436caf55-234f-4249-83d8-d9c6c61c7b8f/6459-1 00:41:00.755 --> 00:41:04.801 intermediate periods, which is to be a bit expected because we 436caf55-234f-4249-83d8-d9c6c61c7b8f/6459-2 00:41:04.801 --> 00:41:08.589 have a lot more observations in areas where we didn't have 436caf55-234f-4249-83d8-d9c6c61c7b8f/6459-3 00:41:08.589 --> 00:41:12.699 observations previously. So the parametric range of the data is 436caf55-234f-4249-83d8-d9c6c61c7b8f/6459-4 00:41:12.699 --> 00:41:16.552 a lot wider and so obviously innergetic model can't capture 436caf55-234f-4249-83d8-d9c6c61c7b8f/6459-5 00:41:16.552 --> 00:41:19.891 all of that behavior systematically. So we're gonna 436caf55-234f-4249-83d8-d9c6c61c7b8f/6459-6 00:41:19.891 --> 00:41:23.551 have larger uncertainty at long periods, though we see a 436caf55-234f-4249-83d8-d9c6c61c7b8f/6459-7 00:41:23.551 --> 00:41:26.377 reduction, which is driven entirely by the. 436caf55-234f-4249-83d8-d9c6c61c7b8f/6478-0 00:41:27.117 --> 00:41:31.267 Reduction in the event terms and this we postulate is probably a 436caf55-234f-4249-83d8-d9c6c61c7b8f/6478-1 00:41:31.267 --> 00:41:34.907 result of the fact that we have a lot more observations. 436caf55-234f-4249-83d8-d9c6c61c7b8f/6531-0 00:41:35.647 --> 00:41:39.404 For these recent events and we have a lot more data at long 436caf55-234f-4249-83d8-d9c6c61c7b8f/6531-1 00:41:39.404 --> 00:41:43.223 periods to fall off isn't as as significant As for the NGO W 436caf55-234f-4249-83d8-d9c6c61c7b8f/6531-2 00:41:43.223 --> 00:41:47.168 data. So because we have more data, we can actually accurately 436caf55-234f-4249-83d8-d9c6c61c7b8f/6531-3 00:41:47.168 --> 00:41:50.862 estimate these event terms and just happens to be that the 436caf55-234f-4249-83d8-d9c6c61c7b8f/6531-4 00:41:50.862 --> 00:41:54.682 source model in BSA 14 does a pretty good job on average for 436caf55-234f-4249-83d8-d9c6c61c7b8f/6531-5 00:41:54.682 --> 00:41:55.997 events in California. 436caf55-234f-4249-83d8-d9c6c61c7b8f/6538-0 00:41:56.707 --> 00:41:59.157 Which is why we see the the reduced uncertainty. 436caf55-234f-4249-83d8-d9c6c61c7b8f/6592-0 00:42:00.407 --> 00:42:03.735 Now looking at the effects of the model compared to BSA 14 436caf55-234f-4249-83d8-d9c6c61c7b8f/6592-1 00:42:03.735 --> 00:42:07.008 using the entire data set I have Sigma which is the total 436caf55-234f-4249-83d8-d9c6c61c7b8f/6592-2 00:42:07.008 --> 00:42:10.336 variability by these thick lines. If we're looking at just 436caf55-234f-4249-83d8-d9c6c61c7b8f/6592-3 00:42:10.336 --> 00:42:13.496 all California first, you can see we pretty much reduce 436caf55-234f-4249-83d8-d9c6c61c7b8f/6592-4 00:42:13.496 --> 00:42:16.712 variability and it's pretty significant reduction. Also, 436caf55-234f-4249-83d8-d9c6c61c7b8f/6592-5 00:42:16.712 --> 00:42:20.097 it's not these aren't small, small reductions we're seeing. 436caf55-234f-4249-83d8-d9c6c61c7b8f/6639-0 00:42:21.007 --> 00:42:24.812 The towel is also significantly reduced. This highlights the 436caf55-234f-4249-83d8-d9c6c61c7b8f/6639-1 00:42:24.812 --> 00:42:28.680 impact that path effects can have when estimating event bias. 436caf55-234f-4249-83d8-d9c6c61c7b8f/6639-2 00:42:28.680 --> 00:42:32.486 If they are pretty significant path effects on by correcting 436caf55-234f-4249-83d8-d9c6c61c7b8f/6639-3 00:42:32.486 --> 00:42:36.478 for those path effects, we get a better estimate when using all 436caf55-234f-4249-83d8-d9c6c61c7b8f/6639-4 00:42:36.478 --> 00:42:38.787 of the data to estimate event terms. 436caf55-234f-4249-83d8-d9c6c61c7b8f/6675-0 00:42:40.137 --> 00:42:43.485 P which is shown by the this dotted line also shows a pretty 436caf55-234f-4249-83d8-d9c6c61c7b8f/6675-1 00:42:43.485 --> 00:42:46.560 significant reduction, indicating that the path to path 436caf55-234f-4249-83d8-d9c6c61c7b8f/6675-2 00:42:46.560 --> 00:42:49.908 variability and because the site to site variability doesn't 436caf55-234f-4249-83d8-d9c6c61c7b8f/6675-3 00:42:49.908 --> 00:42:50.787 change too much. 436caf55-234f-4249-83d8-d9c6c61c7b8f/6701-0 00:42:51.517 --> 00:42:54.979 And why it's not shown on these flats? It is also significant 436caf55-234f-4249-83d8-d9c6c61c7b8f/6701-1 00:42:54.979 --> 00:42:58.385 which is shown by the thinner salted lights down here by the 436caf55-234f-4249-83d8-d9c6c61c7b8f/6701-2 00:42:58.385 --> 00:43:00.787 blue line being lower than the black line. 436caf55-234f-4249-83d8-d9c6c61c7b8f/6714-0 00:43:01.617 --> 00:43:05.397 We noticed that the North Coast actually had quite a large. 436caf55-234f-4249-83d8-d9c6c61c7b8f/6733-0 00:43:06.127 --> 00:43:10.216 Contribution to the uncertainty. This might be driven by the fact 436caf55-234f-4249-83d8-d9c6c61c7b8f/6733-1 00:43:10.216 --> 00:43:13.377 that that we don't have a lot of data up here and. 436caf55-234f-4249-83d8-d9c6c61c7b8f/6736-0 00:43:14.477 --> 00:43:14.947 The. 436caf55-234f-4249-83d8-d9c6c61c7b8f/6750-0 00:43:15.997 --> 00:43:18.872 Uncertainty related to the event term because we're also 436caf55-234f-4249-83d8-d9c6c61c7b8f/6750-1 00:43:18.872 --> 00:43:21.597 including those guys are events is also pretty large. 436caf55-234f-4249-83d8-d9c6c61c7b8f/6766-0 00:43:22.597 --> 00:43:25.400 But we see consistent trends where the sub regional model 436caf55-234f-4249-83d8-d9c6c61c7b8f/6766-1 00:43:25.400 --> 00:43:27.527 performs better than a fully ergodic model. 436caf55-234f-4249-83d8-d9c6c61c7b8f/6784-0 00:43:28.317 --> 00:43:31.432 Similarly, if we look at non North Coast events, so now we're 436caf55-234f-4249-83d8-d9c6c61c7b8f/6784-1 00:43:31.432 --> 00:43:34.447 just looking at purely tectonic events which which is what? 436caf55-234f-4249-83d8-d9c6c61c7b8f/6817-0 00:43:35.817 --> 00:43:39.582 The model was the BSA, 14 was designed for. We still see 436caf55-234f-4249-83d8-d9c6c61c7b8f/6817-1 00:43:39.582 --> 00:43:43.480 considerable reductions in uncertainty and when we look at 436caf55-234f-4249-83d8-d9c6c61c7b8f/6817-2 00:43:43.480 --> 00:43:47.577 these reductions, the new model is pretty much equivalent to. 436caf55-234f-4249-83d8-d9c6c61c7b8f/6838-0 00:43:48.737 --> 00:43:54.093 To what has been presented in existing allatori variability 436caf55-234f-4249-83d8-d9c6c61c7b8f/6838-1 00:43:54.093 --> 00:43:58.467 models like ELATIK 2015 or Goulet, all 2018. So. 436caf55-234f-4249-83d8-d9c6c61c7b8f/6870-0 00:43:59.117 --> 00:44:03.303 This the increase of uncertainty by adding the data is offset by 436caf55-234f-4249-83d8-d9c6c61c7b8f/6870-1 00:44:03.303 --> 00:44:07.425 accounting for suborganizations so that we pretty much have the 436caf55-234f-4249-83d8-d9c6c61c7b8f/6870-2 00:44:07.425 --> 00:44:11.547 same aleatory variability models that we previously were using. 436caf55-234f-4249-83d8-d9c6c61c7b8f/6919-0 00:44:12.637 --> 00:44:16.116 Now to to quantify the reduction of variability and to look a 436caf55-234f-4249-83d8-d9c6c61c7b8f/6919-1 00:44:16.116 --> 00:44:19.651 little closer to see where the reduction is actually occurring 436caf55-234f-4249-83d8-d9c6c61c7b8f/6919-2 00:44:19.651 --> 00:44:22.568 most. If we just take the difference between BSA 14 436caf55-234f-4249-83d8-d9c6c61c7b8f/6919-3 00:44:22.568 --> 00:44:25.374 results and the proposed results, we see that the 436caf55-234f-4249-83d8-d9c6c61c7b8f/6919-4 00:44:25.374 --> 00:44:28.292 reductions are pretty much occurring mostly at long 436caf55-234f-4249-83d8-d9c6c61c7b8f/6919-5 00:44:28.292 --> 00:44:29.807 distances greater than 100. 436caf55-234f-4249-83d8-d9c6c61c7b8f/6922-0 00:44:30.347 --> 00:44:31.347 Hello plummeted. 436caf55-234f-4249-83d8-d9c6c61c7b8f/6925-0 00:44:32.827 --> 00:44:33.277 Excuse me. 436caf55-234f-4249-83d8-d9c6c61c7b8f/6987-0 00:44:36.197 --> 00:44:40.172 Which is what we expect because analystic path effects manifest 436caf55-234f-4249-83d8-d9c6c61c7b8f/6987-1 00:44:40.172 --> 00:44:43.837 greatly at far distances, so it makes sense that these far 436caf55-234f-4249-83d8-d9c6c61c7b8f/6987-2 00:44:43.837 --> 00:44:47.316 distances are showing the greatest reduction, and these 436caf55-234f-4249-83d8-d9c6c61c7b8f/6987-3 00:44:47.316 --> 00:44:51.353 reductions are on par with what was achieved in the Q and at all 436caf55-234f-4249-83d8-d9c6c61c7b8f/6987-4 00:44:51.353 --> 00:44:55.329 2019 model. There are a little lower than what Qnet all showed, 436caf55-234f-4249-83d8-d9c6c61c7b8f/6987-5 00:44:55.329 --> 00:44:56.447 but it shows that. 436caf55-234f-4249-83d8-d9c6c61c7b8f/7010-0 00:44:57.237 --> 00:45:00.737 Just with these nine regions, we're getting pretty close to 436caf55-234f-4249-83d8-d9c6c61c7b8f/7010-1 00:45:00.737 --> 00:45:04.470 what Q and it all had were 500 cells. So that's that's a pretty 436caf55-234f-4249-83d8-d9c6c61c7b8f/7010-2 00:45:04.470 --> 00:45:05.637 substantial finding. 436caf55-234f-4249-83d8-d9c6c61c7b8f/7034-0 00:45:07.387 --> 00:45:10.396 Lastly, some conclusions. I have some general conclusions about 436caf55-234f-4249-83d8-d9c6c61c7b8f/7034-1 00:45:10.396 --> 00:45:13.216 path modeling and then some more specific about the model I 436caf55-234f-4249-83d8-d9c6c61c7b8f/7034-2 00:45:13.216 --> 00:45:13.687 presented. 436caf55-234f-4249-83d8-d9c6c61c7b8f/7071-0 00:45:14.377 --> 00:45:17.177 Health models exists on antibiotic and endergonic 436caf55-234f-4249-83d8-d9c6c61c7b8f/7071-1 00:45:17.177 --> 00:45:20.705 spectrum. We can never get to be fully or not non ergodic, but 436caf55-234f-4249-83d8-d9c6c61c7b8f/7071-2 00:45:20.705 --> 00:45:23.953 ideally we'll get to a point where things are practically 436caf55-234f-4249-83d8-d9c6c61c7b8f/7071-3 00:45:23.953 --> 00:45:24.457 narcotic. 436caf55-234f-4249-83d8-d9c6c61c7b8f/7085-0 00:45:25.417 --> 00:45:28.959 Several path modeling combines advantages of regional and cell 436caf55-234f-4249-83d8-d9c6c61c7b8f/7085-1 00:45:28.959 --> 00:45:30.477 based methods where we can. 436caf55-234f-4249-83d8-d9c6c61c7b8f/7100-0 00:45:31.477 --> 00:45:34.495 Constrain our discretization based off of physical principles 436caf55-234f-4249-83d8-d9c6c61c7b8f/7100-1 00:45:34.495 --> 00:45:36.297 so that we can interpret the models. 436caf55-234f-4249-83d8-d9c6c61c7b8f/7167-0 00:45:37.187 --> 00:45:40.993 And then apply the cell based math for the Arabic effects. I'd 436caf55-234f-4249-83d8-d9c6c61c7b8f/7167-1 00:45:40.993 --> 00:45:44.437 also argue that cell based methods are kind of only have 436caf55-234f-4249-83d8-d9c6c61c7b8f/7167-2 00:45:44.437 --> 00:45:48.243 applications only for the region which they are developed, but 436caf55-234f-4249-83d8-d9c6c61c7b8f/7167-3 00:45:48.243 --> 00:45:51.928 I've perceived there's potential future application of using 436caf55-234f-4249-83d8-d9c6c61c7b8f/7167-4 00:45:51.928 --> 00:45:55.553 lessons learned through sub regionalization 2 regions where 436caf55-234f-4249-83d8-d9c6c61c7b8f/7167-5 00:45:55.553 --> 00:45:58.876 there's less data, for example where we can't do these 436caf55-234f-4249-83d8-d9c6c61c7b8f/7167-6 00:45:58.876 --> 00:46:00.387 nonnarcotic path studies. 436caf55-234f-4249-83d8-d9c6c61c7b8f/7179-0 00:46:01.257 --> 00:46:04.616 But that's sort of future work that that needs to be flushed 436caf55-234f-4249-83d8-d9c6c61c7b8f/7179-1 00:46:04.616 --> 00:46:04.837 out. 436caf55-234f-4249-83d8-d9c6c61c7b8f/7223-0 00:46:05.967 --> 00:46:09.397 But as kind of what I mentioned there, as with other methods 436caf55-234f-4249-83d8-d9c6c61c7b8f/7223-1 00:46:09.397 --> 00:46:12.884 modeling methods, subregional methods are limited by the data 436caf55-234f-4249-83d8-d9c6c61c7b8f/7223-2 00:46:12.884 --> 00:46:16.314 availability, but I think we tried to optimize the amount of 436caf55-234f-4249-83d8-d9c6c61c7b8f/7223-3 00:46:16.314 --> 00:46:19.857 data within each subregion by creating large sub regions with. 436caf55-234f-4249-83d8-d9c6c61c7b8f/7227-0 00:46:21.007 --> 00:46:22.007 Physical constraints. 436caf55-234f-4249-83d8-d9c6c61c7b8f/7262-0 00:46:22.977 --> 00:46:26.879 Now conclusions about our path model. The biggest conclusion is 436caf55-234f-4249-83d8-d9c6c61c7b8f/7262-1 00:46:26.879 --> 00:46:30.781 that we find variable rates of analytic continuation across the 436caf55-234f-4249-83d8-d9c6c61c7b8f/7262-2 00:46:30.781 --> 00:46:34.196 state, not necessarily new findings, but with new data, 436caf55-234f-4249-83d8-d9c6c61c7b8f/7262-3 00:46:34.196 --> 00:46:36.757 we're able to refine these results a lot. 436caf55-234f-4249-83d8-d9c6c61c7b8f/7288-0 00:46:37.647 --> 00:46:40.808 Better than what previous efforts have have found, which 436caf55-234f-4249-83d8-d9c6c61c7b8f/7288-1 00:46:40.808 --> 00:46:43.970 they've postulated that, oh, there's variable rates, but 436caf55-234f-4249-83d8-d9c6c61c7b8f/7288-2 00:46:43.970 --> 00:46:47.187 they're what's insufficient data to really quantify them. 436caf55-234f-4249-83d8-d9c6c61c7b8f/7380-0 00:46:48.317 --> 00:46:51.140 Mainframe is there is significantly higher than 436caf55-234f-4249-83d8-d9c6c61c7b8f/7380-1 00:46:51.140 --> 00:46:54.374 average Ng North Coast sub region so fast regeneration 436caf55-234f-4249-83d8-d9c6c61c7b8f/7380-2 00:46:54.374 --> 00:46:57.844 there slower than average in the eastern subregions, maybe 436caf55-234f-4249-83d8-d9c6c61c7b8f/7380-3 00:46:57.844 --> 00:47:01.549 related to the High Q West and the as the Waves travel through 436caf55-234f-4249-83d8-d9c6c61c7b8f/7380-4 00:47:01.549 --> 00:47:05.254 the Sierra Nevada coming back to this first point, this faster 436caf55-234f-4249-83d8-d9c6c61c7b8f/7380-5 00:47:05.254 --> 00:47:08.959 attenuation might be related to the fact that the rocks in the 436caf55-234f-4249-83d8-d9c6c61c7b8f/7380-6 00:47:08.959 --> 00:47:12.193 North Coast area have been tormented in their tectonic 436caf55-234f-4249-83d8-d9c6c61c7b8f/7380-7 00:47:12.193 --> 00:47:15.663 history. So the QS is so low and that's what's driving the 436caf55-234f-4249-83d8-d9c6c61c7b8f/7380-8 00:47:15.663 --> 00:47:18.957 attenuation low. I'd be interested to know what anyone. 436caf55-234f-4249-83d8-d9c6c61c7b8f/7407-0 00:47:19.447 --> 00:47:23.211 And the earth scientist would suggest might be causing these 436caf55-234f-4249-83d8-d9c6c61c7b8f/7407-1 00:47:23.211 --> 00:47:26.852 observations and then lastly, intermediate rates for other 436caf55-234f-4249-83d8-d9c6c61c7b8f/7407-2 00:47:26.852 --> 00:47:30.370 subregions, which tend to trend to be slightly faster to 436caf55-234f-4249-83d8-d9c6c61c7b8f/7407-3 00:47:30.370 --> 00:47:30.987 emulation. 436caf55-234f-4249-83d8-d9c6c61c7b8f/7453-0 00:47:32.047 --> 00:47:35.878 Current models of dominated by small magnitude events, but we 436caf55-234f-4249-83d8-d9c6c61c7b8f/7453-1 00:47:35.878 --> 00:47:39.585 assume that the variations are equally applicable for large 436caf55-234f-4249-83d8-d9c6c61c7b8f/7453-2 00:47:39.585 --> 00:47:43.231 magnitude. It wouldn't make sense that the the energy loss 436caf55-234f-4249-83d8-d9c6c61c7b8f/7453-3 00:47:43.231 --> 00:47:46.876 is directly related to the magnitude of the earthquake, so 436caf55-234f-4249-83d8-d9c6c61c7b8f/7453-4 00:47:46.876 --> 00:47:50.769 our unless we start to get into side effects with nonlinearity 436caf55-234f-4249-83d8-d9c6c61c7b8f/7453-5 00:47:50.769 --> 00:47:52.067 and things like that. 436caf55-234f-4249-83d8-d9c6c61c7b8f/7472-0 00:47:53.457 --> 00:47:57.215 See a significant reduction at very far just this, which can be 436caf55-234f-4249-83d8-d9c6c61c7b8f/7472-1 00:47:57.215 --> 00:48:00.387 achieved using a relatively few number of subregions. 436caf55-234f-4249-83d8-d9c6c61c7b8f/7500-0 00:48:01.587 --> 00:48:04.831 And I believe that further refinement of these subregions 436caf55-234f-4249-83d8-d9c6c61c7b8f/7500-1 00:48:04.831 --> 00:48:08.131 holds significant potential to bring us to a practical non 436caf55-234f-4249-83d8-d9c6c61c7b8f/7500-2 00:48:08.131 --> 00:48:11.487 ergodic path model that can be attractive to practitioners. 436caf55-234f-4249-83d8-d9c6c61c7b8f/7523-0 00:48:12.407 --> 00:48:16.386 That's easy to interpret and easy to implement, so that's 436caf55-234f-4249-83d8-d9c6c61c7b8f/7523-1 00:48:16.386 --> 00:48:20.777 sort of the the next steps would be to refine this model to be. 436caf55-234f-4249-83d8-d9c6c61c7b8f/7531-0 00:48:21.987 --> 00:48:24.427 Able to fully capture not necrotic path effects. 436caf55-234f-4249-83d8-d9c6c61c7b8f/7553-0 00:48:25.127 --> 00:48:28.000 With that, I have some references that I've referred to 436caf55-234f-4249-83d8-d9c6c61c7b8f/7553-1 00:48:28.000 --> 00:48:31.078 throughout this presentation, and I'd like to thank you and 436caf55-234f-4249-83d8-d9c6c61c7b8f/7553-2 00:48:31.078 --> 00:48:32.617 would end up to any questions. 436caf55-234f-4249-83d8-d9c6c61c7b8f/7579-0 00:48:43.457 --> 00:48:47.345 Thanks Justin for the awesome talk. Do we have any questions 436caf55-234f-4249-83d8-d9c6c61c7b8f/7579-1 00:48:47.345 --> 00:48:51.042 from the audience? You can either raise your hand or type 436caf55-234f-4249-83d8-d9c6c61c7b8f/7579-2 00:48:51.042 --> 00:48:51.807 in the chat. 436caf55-234f-4249-83d8-d9c6c61c7b8f/7582-0 00:48:56.847 --> 00:48:57.437 Grace. 436caf55-234f-4249-83d8-d9c6c61c7b8f/7596-0 00:49:02.147 --> 00:49:05.060 Hi Kristen, thanks for the really great talk to you. Did a 436caf55-234f-4249-83d8-d9c6c61c7b8f/7596-1 00:49:05.060 --> 00:49:06.097 great job explaining. 436caf55-234f-4249-83d8-d9c6c61c7b8f/7617-0 00:49:07.987 --> 00:49:11.642 All the work that you did, I have a question about the 436caf55-234f-4249-83d8-d9c6c61c7b8f/7617-1 00:49:11.642 --> 00:49:15.297 constant, the regional constant terms that delta cnos. 436caf55-234f-4249-83d8-d9c6c61c7b8f/7653-0 00:49:15.977 --> 00:49:19.165 And the first part of the question is, can you talk a 436caf55-234f-4249-83d8-d9c6c61c7b8f/7653-1 00:49:19.165 --> 00:49:22.885 little bit more about how they were estimated, where they like 436caf55-234f-4249-83d8-d9c6c61c7b8f/7653-2 00:49:22.885 --> 00:49:26.782 a parameter that you fit in your model or were they taken as like 436caf55-234f-4249-83d8-d9c6c61c7b8f/7653-3 00:49:26.782 --> 00:49:28.377 the average of event terms? 436caf55-234f-4249-83d8-d9c6c61c7b8f/7658-0 00:49:29.347 --> 00:49:31.247 And the second part. 436caf55-234f-4249-83d8-d9c6c61c7b8f/7668-0 00:49:32.547 --> 00:49:36.156 Of the question is, I was surprised that they were so 436caf55-234f-4249-83d8-d9c6c61c7b8f/7668-1 00:49:36.156 --> 00:49:36.557 small. 436caf55-234f-4249-83d8-d9c6c61c7b8f/7692-0 00:49:37.597 --> 00:49:41.292 Could you talk a maybe a bit more about why you think that is 436caf55-234f-4249-83d8-d9c6c61c7b8f/7692-1 00:49:41.292 --> 00:49:44.629 and how you thought about addressing trade-offs between 436caf55-234f-4249-83d8-d9c6c61c7b8f/7692-2 00:49:44.629 --> 00:49:46.537 the path and the source effects? 436caf55-234f-4249-83d8-d9c6c61c7b8f/7722-0 00:49:47.587 --> 00:49:51.959 Yes. So to answer the first question, the way that we 436caf55-234f-4249-83d8-d9c6c61c7b8f/7722-1 00:49:51.959 --> 00:49:56.656 estimated the Delta C knots is it's not strictly a source 436caf55-234f-4249-83d8-d9c6c61c7b8f/7722-2 00:49:56.656 --> 00:50:00.057 adjustment original source adjustment as. 436caf55-234f-4249-83d8-d9c6c61c7b8f/7733-0 00:50:00.737 --> 00:50:05.427 Some researchers have tried to start have proposed for. 436caf55-234f-4249-83d8-d9c6c61c7b8f/7741-0 00:50:06.597 --> 00:50:10.867 Smaller regions. The reason why we didn't want to propose a. 436caf55-234f-4249-83d8-d9c6c61c7b8f/7771-0 00:50:11.997 --> 00:50:15.206 A pure source adjustment is because some of these sub 436caf55-234f-4249-83d8-d9c6c61c7b8f/7771-1 00:50:15.206 --> 00:50:19.069 regions don't have many events, especially as we get out to long 436caf55-234f-4249-83d8-d9c6c61c7b8f/7771-2 00:50:19.069 --> 00:50:20.377 periods. So trying to. 436caf55-234f-4249-83d8-d9c6c61c7b8f/7843-0 00:50:21.357 --> 00:50:25.076 Define a a true adjustment based off of a few number of 436caf55-234f-4249-83d8-d9c6c61c7b8f/7843-1 00:50:25.076 --> 00:50:29.061 observations would have such large uncertainty that we felt 436caf55-234f-4249-83d8-d9c6c61c7b8f/7843-2 00:50:29.061 --> 00:50:32.913 that it wasn't necessarily to our advantage and we really 436caf55-234f-4249-83d8-d9c6c61c7b8f/7843-3 00:50:32.913 --> 00:50:36.965 weren't at that point in with the data that we had available 436caf55-234f-4249-83d8-d9c6c61c7b8f/7843-4 00:50:36.965 --> 00:50:40.817 to us to confidently develop a source adjustment. So even 436caf55-234f-4249-83d8-d9c6c61c7b8f/7843-5 00:50:40.817 --> 00:50:44.603 though I refer to these as a constant source adjustment, 436caf55-234f-4249-83d8-d9c6c61c7b8f/7843-6 00:50:44.603 --> 00:50:48.057 really what they are, if I go back a couple slides. 436caf55-234f-4249-83d8-d9c6c61c7b8f/7856-0 00:50:49.697 --> 00:50:52.957 Or an adjustment with respect to the within of that residual. 436caf55-234f-4249-83d8-d9c6c61c7b8f/7858-0 00:50:53.657 --> 00:50:54.267 So. 436caf55-234f-4249-83d8-d9c6c61c7b8f/7870-0 00:50:55.437 --> 00:51:00.627 When when we fit the the total, I guess this this shows it here. 436caf55-234f-4249-83d8-d9c6c61c7b8f/7911-0 00:51:01.387 --> 00:51:05.253 When we fit the events specific attenuation, we can see that 436caf55-234f-4249-83d8-d9c6c61c7b8f/7911-1 00:51:05.253 --> 00:51:09.182 each event is going to have some offset up and down for short 436caf55-234f-4249-83d8-d9c6c61c7b8f/7911-2 00:51:09.182 --> 00:51:12.858 periods, depending on the functional form that we use and 436caf55-234f-4249-83d8-d9c6c61c7b8f/7911-3 00:51:12.858 --> 00:51:16.407 when we went to fit the the groupings for some regions. 436caf55-234f-4249-83d8-d9c6c61c7b8f/7995-0 00:51:19.857 --> 00:51:23.706 Which is shown in this slide here you can see that the the 436caf55-234f-4249-83d8-d9c6c61c7b8f/7995-1 00:51:23.706 --> 00:51:27.621 the region which showed the the largest impact would be the 436caf55-234f-4249-83d8-d9c6c61c7b8f/7995-2 00:51:27.621 --> 00:51:31.405 North Coast where we can see that there's a slight offset 436caf55-234f-4249-83d8-d9c6c61c7b8f/7995-3 00:51:31.405 --> 00:51:35.124 from zero. This offset is actually what that delta C not 436caf55-234f-4249-83d8-d9c6c61c7b8f/7995-4 00:51:35.124 --> 00:51:38.908 term is, it's the offset with respect to the within about 436caf55-234f-4249-83d8-d9c6c61c7b8f/7995-5 00:51:38.908 --> 00:51:42.823 residual for each subregion which was estimated using these 436caf55-234f-4249-83d8-d9c6c61c7b8f/7995-6 00:51:42.823 --> 00:51:47.064 clusters of events that occurred within each sub region. So this 436caf55-234f-4249-83d8-d9c6c61c7b8f/8002-0 00:51:46.977 --> 00:51:50.397 So it's after the event terms have been removed. 436caf55-234f-4249-83d8-d9c6c61c7b8f/7995-7 00:51:47.064 --> 00:51:47.717 data here. 436caf55-234f-4249-83d8-d9c6c61c7b8f/8009-0 00:51:50.707 --> 00:51:52.927 Yes, after the event terms have been removed. 436caf55-234f-4249-83d8-d9c6c61c7b8f/8022-0 00:51:53.257 --> 00:51:56.920 Did you look at your event terms banned by these different 436caf55-234f-4249-83d8-d9c6c61c7b8f/8022-1 00:51:56.920 --> 00:51:57.417 regions? 436caf55-234f-4249-83d8-d9c6c61c7b8f/8058-0 00:51:58.477 --> 00:52:02.337 Yes. So we did that helped us to come up with the sub 436caf55-234f-4249-83d8-d9c6c61c7b8f/8058-1 00:52:02.337 --> 00:52:06.913 regionalization. We spent quite a bit of time we and it kind of 436caf55-234f-4249-83d8-d9c6c61c7b8f/8058-2 00:52:06.913 --> 00:52:10.774 also helped us to identify that the geyser events had 436caf55-234f-4249-83d8-d9c6c61c7b8f/8058-3 00:52:10.774 --> 00:52:12.347 significant bias if I. 436caf55-234f-4249-83d8-d9c6c61c7b8f/8064-0 00:52:13.357 --> 00:52:14.977 I have a hidden slide that. 436caf55-234f-4249-83d8-d9c6c61c7b8f/8070-0 00:52:15.837 --> 00:52:17.767 Kind of show what what we found. 436caf55-234f-4249-83d8-d9c6c61c7b8f/8093-0 00:52:18.617 --> 00:52:23.078 So this particular slide here shows when we looked at the the 436caf55-234f-4249-83d8-d9c6c61c7b8f/8093-1 00:52:23.078 --> 00:52:26.317 binning of event terms, most regions it had. 436caf55-234f-4249-83d8-d9c6c61c7b8f/8166-0 00:52:27.077 --> 00:52:30.646 Similar average event terms relatively well captured by the 436caf55-234f-4249-83d8-d9c6c61c7b8f/8166-1 00:52:30.646 --> 00:52:33.799 ergodic model, except for the North Coast, which had 436caf55-234f-4249-83d8-d9c6c61c7b8f/8166-2 00:52:33.799 --> 00:52:37.667 significantly lower event terms and we found when looking at our 436caf55-234f-4249-83d8-d9c6c61c7b8f/8166-3 00:52:37.667 --> 00:52:40.998 data that it was driven primarily by these Kaiser event 436caf55-234f-4249-83d8-d9c6c61c7b8f/8166-4 00:52:40.998 --> 00:52:44.627 terms which the ergodic model predicts to be a lot lower. So 436caf55-234f-4249-83d8-d9c6c61c7b8f/8166-5 00:52:44.627 --> 00:52:48.019 we're seeing lower amplitudes than what BSA 14 predicts. 436caf55-234f-4249-83d8-d9c6c61c7b8f/8166-6 00:52:48.019 --> 00:52:51.529 That's why we introduced this adjustment to to correct for 436caf55-234f-4249-83d8-d9c6c61c7b8f/8166-7 00:52:51.529 --> 00:52:52.957 this into source effect. 436caf55-234f-4249-83d8-d9c6c61c7b8f/8182-0 00:52:57.237 --> 00:53:00.798 I think because does that answer your questions? I'm not sure if 436caf55-234f-4249-83d8-d9c6c61c7b8f/8171-0 00:52:58.017 --> 00:52:58.447 Thanks. 436caf55-234f-4249-83d8-d9c6c61c7b8f/8182-1 00:53:00.798 --> 00:53:02.277 I answered the second part. 436caf55-234f-4249-83d8-d9c6c61c7b8f/8186-0 00:53:04.457 --> 00:53:05.527 I. 436caf55-234f-4249-83d8-d9c6c61c7b8f/8216-0 00:53:06.237 --> 00:53:11.852 Think so? I was thinking that I think the fact that the Delta C 436caf55-234f-4249-83d8-d9c6c61c7b8f/8216-1 00:53:11.852 --> 00:53:16.765 not are fit to the width in event residual explains why 436caf55-234f-4249-83d8-d9c6c61c7b8f/8216-2 00:53:16.765 --> 00:53:18.257 they're so small. 436caf55-234f-4249-83d8-d9c6c61c7b8f/8218-0 00:53:19.037 --> 00:53:19.337 Yeah. 436caf55-234f-4249-83d8-d9c6c61c7b8f/8221-0 00:53:21.377 --> 00:53:21.697 Thank you. 436caf55-234f-4249-83d8-d9c6c61c7b8f/8226-0 00:53:34.307 --> 00:53:35.087 In Marie. 436caf55-234f-4249-83d8-d9c6c61c7b8f/8237-0 00:53:36.567 --> 00:53:39.880 Yeah, and thanks, Tristan. It was really great and a really 436caf55-234f-4249-83d8-d9c6c61c7b8f/8237-1 00:53:39.880 --> 00:53:40.157 nice. 436caf55-234f-4249-83d8-d9c6c61c7b8f/8268-0 00:53:41.647 --> 00:53:45.982 Walk through of all the your work and the non organic models 436caf55-234f-4249-83d8-d9c6c61c7b8f/8268-1 00:53:45.982 --> 00:53:50.246 and maybe following up on on Grace's question, you showed a 436caf55-234f-4249-83d8-d9c6c61c7b8f/8268-2 00:53:50.246 --> 00:53:52.947 little bit later on the overall bias. 436caf55-234f-4249-83d8-d9c6c61c7b8f/8275-0 00:53:54.167 --> 00:53:56.407 UM here? Yeah, so. 436caf55-234f-4249-83d8-d9c6c61c7b8f/8290-0 00:53:57.967 --> 00:54:00.678 We've observed and yes, or maybe this was kind of covered the 436caf55-234f-4249-83d8-d9c6c61c7b8f/8290-1 00:54:00.678 --> 00:54:02.997 discussion and grace are having, but we've observed. 436caf55-234f-4249-83d8-d9c6c61c7b8f/8318-0 00:54:03.317 --> 00:54:06.635 And that there's a difference between Northern California and 436caf55-234f-4249-83d8-d9c6c61c7b8f/8318-1 00:54:06.635 --> 00:54:10.008 Southern California, and Kunal and other people have seen that 436caf55-234f-4249-83d8-d9c6c61c7b8f/8318-2 00:54:10.008 --> 00:54:13.327 as well. So am I understanding that with your proposed model? 436caf55-234f-4249-83d8-d9c6c61c7b8f/8323-0 00:54:14.327 --> 00:54:15.447 There's your. 436caf55-234f-4249-83d8-d9c6c61c7b8f/8335-0 00:54:16.267 --> 00:54:19.907 Are you? I guess this is the bias for all of the data. 436caf55-234f-4249-83d8-d9c6c61c7b8f/8374-0 00:54:21.507 --> 00:54:24.688 Can you? I guess. Can you explain that that account for 436caf55-234f-4249-83d8-d9c6c61c7b8f/8374-1 00:54:24.688 --> 00:54:28.380 any difference between Southern California, Northern California? 436caf55-234f-4249-83d8-d9c6c61c7b8f/8374-2 00:54:28.380 --> 00:54:31.618 Are you inherently then more correctly modeling the path 436caf55-234f-4249-83d8-d9c6c61c7b8f/8374-3 00:54:31.618 --> 00:54:35.196 attenuation and thus removing any bias between the two regions 436caf55-234f-4249-83d8-d9c6c61c7b8f/8374-4 00:54:35.196 --> 00:54:36.787 or how do we interpret this? 436caf55-234f-4249-83d8-d9c6c61c7b8f/8390-0 00:54:38.737 --> 00:54:41.917 Bias going from, you know, pretty significantly negative to 436caf55-234f-4249-83d8-d9c6c61c7b8f/8390-1 00:54:41.917 --> 00:54:43.507 0 for for most of the periods. 436caf55-234f-4249-83d8-d9c6c61c7b8f/8419-0 00:54:44.597 --> 00:54:49.337 Thank you. So I'm aware of the differences in the that North 436caf55-234f-4249-83d8-d9c6c61c7b8f/8419-1 00:54:49.337 --> 00:54:54.466 Coast events are generally found to have bias with respect to the 436caf55-234f-4249-83d8-d9c6c61c7b8f/8419-2 00:54:54.466 --> 00:54:58.507 magnitudes. I think it's related to the catalog of. 436caf55-234f-4249-83d8-d9c6c61c7b8f/8429-0 00:54:59.427 --> 00:55:01.147 Which reports the source parameters. 436caf55-234f-4249-83d8-d9c6c61c7b8f/8444-0 00:55:02.087 --> 00:55:06.948 In our study, though, if I go back up to methodology, we don't 436caf55-234f-4249-83d8-d9c6c61c7b8f/8444-1 00:55:06.948 --> 00:55:08.877 inherently in our method. 436caf55-234f-4249-83d8-d9c6c61c7b8f/8490-0 00:55:09.437 --> 00:55:13.118 Umm try to correct for those differences with respect to 436caf55-234f-4249-83d8-d9c6c61c7b8f/8490-1 00:55:13.118 --> 00:55:16.928 Northern California and Southern California source effects 436caf55-234f-4249-83d8-d9c6c61c7b8f/8490-2 00:55:16.928 --> 00:55:21.190 because we remove the event term and within a residual we're only 436caf55-234f-4249-83d8-d9c6c61c7b8f/8490-3 00:55:21.190 --> 00:55:25.129 looking at within residuals. When we developed the model, so 436caf55-234f-4249-83d8-d9c6c61c7b8f/8490-4 00:55:25.129 --> 00:55:25.517 those. 436caf55-234f-4249-83d8-d9c6c61c7b8f/8507-0 00:55:26.357 --> 00:55:29.485 What are? I guess we operate under the assumption that the 436caf55-234f-4249-83d8-d9c6c61c7b8f/8507-1 00:55:29.485 --> 00:55:31.607 systematic differences with respect to. 436caf55-234f-4249-83d8-d9c6c61c7b8f/8533-0 00:55:31.677 --> 00:55:35.881 Yeah, to event bias, which can be contained within the event 436caf55-234f-4249-83d8-d9c6c61c7b8f/8533-1 00:55:35.881 --> 00:55:40.154 term, is subtracted out of the within repositioning and isn't 436caf55-234f-4249-83d8-d9c6c61c7b8f/8533-2 00:55:40.154 --> 00:55:42.567 mapped into our model development. 436caf55-234f-4249-83d8-d9c6c61c7b8f/8551-0 00:55:45.007 --> 00:55:49.394 But then, didn't you just say that there was no systematic 436caf55-234f-4249-83d8-d9c6c61c7b8f/8551-1 00:55:49.394 --> 00:55:52.517 difference in the event terms regionally? 436caf55-234f-4249-83d8-d9c6c61c7b8f/8633-0 00:55:54.797 --> 00:55:58.783 So there was. So I did state that there wasn't any strong 436caf55-234f-4249-83d8-d9c6c61c7b8f/8633-1 00:55:58.783 --> 00:56:03.182 system. I don't have a slide to show that system as differences 436caf55-234f-4249-83d8-d9c6c61c7b8f/8575-0 00:56:01.527 --> 00:56:02.237 That's fine, yeah. 436caf55-234f-4249-83d8-d9c6c61c7b8f/8633-2 00:56:03.182 --> 00:56:07.374 in the event terms. There were obviously differences between 436caf55-234f-4249-83d8-d9c6c61c7b8f/8633-3 00:56:07.374 --> 00:56:11.842 them. However, when we looked at bins and essentially histograms 436caf55-234f-4249-83d8-d9c6c61c7b8f/8633-4 00:56:11.842 --> 00:56:15.759 of event terms, the statistics of each group were pretty 436caf55-234f-4249-83d8-d9c6c61c7b8f/8633-5 00:56:15.759 --> 00:56:19.883 similar, except for the North Coast, which after correcting 436caf55-234f-4249-83d8-d9c6c61c7b8f/8620-0 00:56:18.177 --> 00:56:18.467 Right. 436caf55-234f-4249-83d8-d9c6c61c7b8f/8633-6 00:56:19.883 --> 00:56:24.007 for the the Geysers were still slightly more negative than. 436caf55-234f-4249-83d8-d9c6c61c7b8f/8642-0 00:56:24.137 --> 00:56:27.127 For example, Southern California, but it wasn't. 436caf55-234f-4249-83d8-d9c6c61c7b8f/8663-0 00:56:27.967 --> 00:56:31.781 Like astonishing me, negative, which we're seeing here. So 436caf55-234f-4249-83d8-d9c6c61c7b8f/8657-0 00:56:31.537 --> 00:56:32.847 Right. So yeah. 436caf55-234f-4249-83d8-d9c6c61c7b8f/8663-1 00:56:31.781 --> 00:56:35.207 there's definitely some, some bias still there, but. 436caf55-234f-4249-83d8-d9c6c61c7b8f/8665-0 00:56:35.617 --> 00:56:35.897 Right. 436caf55-234f-4249-83d8-d9c6c61c7b8f/8746-0 00:56:37.697 --> 00:56:40.997 Yes. So let's talk more about this later maybe, but I guess 436caf55-234f-4249-83d8-d9c6c61c7b8f/8746-1 00:56:40.997 --> 00:56:44.298 I'm curious if you're you're basically you have an improved 436caf55-234f-4249-83d8-d9c6c61c7b8f/8746-2 00:56:44.298 --> 00:56:47.763 attenuation model and if that is sort of inherently correcting 436caf55-234f-4249-83d8-d9c6c61c7b8f/8746-3 00:56:47.763 --> 00:56:51.009 and more correctly accounting for this observed difference 436caf55-234f-4249-83d8-d9c6c61c7b8f/8746-4 00:56:51.009 --> 00:56:54.475 between Northern California and California, right, so it could 436caf55-234f-4249-83d8-d9c6c61c7b8f/8746-5 00:56:54.475 --> 00:56:57.445 be because of the catalog magnitude, it could also be 436caf55-234f-4249-83d8-d9c6c61c7b8f/8746-6 00:56:57.445 --> 00:57:00.966 because the attenuation models are different, right? So I guess 436caf55-234f-4249-83d8-d9c6c61c7b8f/8730-0 00:56:59.497 --> 00:57:00.497 Precisely yes. 436caf55-234f-4249-83d8-d9c6c61c7b8f/8746-7 00:57:00.966 --> 00:57:04.101 it's, yeah, I'm curious if you you're inherently perhaps 436caf55-234f-4249-83d8-d9c6c61c7b8f/8746-8 00:57:04.101 --> 00:57:06.137 accounting for that in a better way. 436caf55-234f-4249-83d8-d9c6c61c7b8f/8819-0 00:57:06.817 --> 00:57:10.020 I haven't thought about that, but I it makes sense that by 436caf55-234f-4249-83d8-d9c6c61c7b8f/8819-1 00:57:10.020 --> 00:57:13.494 correcting for the attenuation we would correct for some of the 436caf55-234f-4249-83d8-d9c6c61c7b8f/8819-2 00:57:13.494 --> 00:57:16.371 bias that we're seeing this event, like I said, is a 436caf55-234f-4249-83d8-d9c6c61c7b8f/8819-3 00:57:16.371 --> 00:57:19.792 Northern California event, and if you were to compute an event 436caf55-234f-4249-83d8-d9c6c61c7b8f/8819-4 00:57:19.792 --> 00:57:23.157 term using all the data would pretty much just be an average. 436caf55-234f-4249-83d8-d9c6c61c7b8f/8819-5 00:57:23.157 --> 00:57:26.469 So we'd get something negative close to -, .9 or so. So that 436caf55-234f-4249-83d8-d9c6c61c7b8f/8796-0 00:57:25.667 --> 00:57:25.867 Right. 436caf55-234f-4249-83d8-d9c6c61c7b8f/8819-6 00:57:26.469 --> 00:57:29.835 would be pretty much strongly influenced by the path effects. 436caf55-234f-4249-83d8-d9c6c61c7b8f/8819-7 00:57:29.835 --> 00:57:33.201 So by correcting for that path effect, we bring these bins up 436caf55-234f-4249-83d8-d9c6c61c7b8f/8819-8 00:57:33.201 --> 00:57:34.287 and get closer to 0. 436caf55-234f-4249-83d8-d9c6c61c7b8f/8826-0 00:57:33.927 --> 00:57:36.457 Right, right. Cool. 436caf55-234f-4249-83d8-d9c6c61c7b8f/8825-0 00:57:35.047 --> 00:57:36.247 So I agree. 436caf55-234f-4249-83d8-d9c6c61c7b8f/8829-0 00:57:37.607 --> 00:57:38.147 Thanks. 436caf55-234f-4249-83d8-d9c6c61c7b8f/8884-0 00:57:40.157 --> 00:57:44.731 OK, I think we have time for one more question for the recorded 436caf55-234f-4249-83d8-d9c6c61c7b8f/8884-1 00:57:44.731 --> 00:57:49.163 part. So art Frankel had his hand up and after that if you're 436caf55-234f-4249-83d8-d9c6c61c7b8f/8884-2 00:57:49.163 --> 00:57:53.452 willing to stay on Tristan, we sometimes have this informal 436caf55-234f-4249-83d8-d9c6c61c7b8f/8884-3 00:57:53.452 --> 00:57:57.455 session where we stop the recording and people get just 436caf55-234f-4249-83d8-d9c6c61c7b8f/8884-4 00:57:57.455 --> 00:58:01.387 still ask stuff. Yeah. So go ahead with your question. 436caf55-234f-4249-83d8-d9c6c61c7b8f/8891-0 00:58:01.567 --> 00:58:04.197 And thanks. Great. Talk really interesting. 436caf55-234f-4249-83d8-d9c6c61c7b8f/8897-0 00:58:05.727 --> 00:58:07.987 A lot of times people see. 436caf55-234f-4249-83d8-d9c6c61c7b8f/8949-0 00:58:09.367 --> 00:58:13.956 Strong attenuation when for past that sort of go across the 436caf55-234f-4249-83d8-d9c6c61c7b8f/8949-1 00:58:13.956 --> 00:58:18.699 structural grain of an area as opposed to pass that sort of a 436caf55-234f-4249-83d8-d9c6c61c7b8f/8949-2 00:58:18.699 --> 00:58:23.594 line along parallel to like the faults and things like that. So 436caf55-234f-4249-83d8-d9c6c61c7b8f/8949-3 00:58:23.594 --> 00:58:28.184 do you see any kind of trends like that where past you know 436caf55-234f-4249-83d8-d9c6c61c7b8f/8949-4 00:58:28.184 --> 00:58:31.397 going across these structural boundaries? 436caf55-234f-4249-83d8-d9c6c61c7b8f/8960-0 00:58:32.127 --> 00:58:36.248 It produces more attenuation and pass along the structural 436caf55-234f-4249-83d8-d9c6c61c7b8f/8960-1 00:58:36.248 --> 00:58:37.017 boundaries. 436caf55-234f-4249-83d8-d9c6c61c7b8f/9018-0 00:58:38.077 --> 00:58:41.255 So we haven't systematically looked at any as muscle 436caf55-234f-4249-83d8-d9c6c61c7b8f/9018-1 00:58:41.255 --> 00:58:44.733 variation is is what I would consider that to be we we've 436caf55-234f-4249-83d8-d9c6c61c7b8f/9018-2 00:58:44.733 --> 00:58:48.091 just kind of lumped all observations together and found 436caf55-234f-4249-83d8-d9c6c61c7b8f/9018-3 00:58:48.091 --> 00:58:51.509 these general differences. I wouldn't be surprised if we 436caf55-234f-4249-83d8-d9c6c61c7b8f/9018-4 00:58:51.509 --> 00:58:55.467 could see that in the data and I think that's an excellent follow 436caf55-234f-4249-83d8-d9c6c61c7b8f/9018-5 00:58:55.467 --> 00:58:56.427 up to look into. 436caf55-234f-4249-83d8-d9c6c61c7b8f/9027-0 00:58:56.207 --> 00:59:00.416 And I'm also big concern that you're using response Spectra 436caf55-234f-4249-83d8-d9c6c61c7b8f/9027-1 00:59:00.416 --> 00:59:00.697 and. 436caf55-234f-4249-83d8-d9c6c61c7b8f/9033-0 00:59:01.357 --> 00:59:02.927 Rather than Fourier Spectra. 436caf55-234f-4249-83d8-d9c6c61c7b8f/9052-0 00:59:04.147 --> 00:59:08.584 I realized response Spectra. The metric that's used in the ground 436caf55-234f-4249-83d8-d9c6c61c7b8f/9052-1 00:59:08.584 --> 00:59:10.937 motion models. But if you want to. 436caf55-234f-4249-83d8-d9c6c61c7b8f/9074-0 00:59:12.097 --> 00:59:16.407 Uh, you know, expand this to different magnitudes. You might. 436caf55-234f-4249-83d8-d9c6c61c7b8f/9074-1 00:59:16.407 --> 00:59:20.370 There might be some discrepancy because you're using the 436caf55-234f-4249-83d8-d9c6c61c7b8f/9074-2 00:59:20.370 --> 00:59:22.247 response Spectra which are. 436caf55-234f-4249-83d8-d9c6c61c7b8f/9102-0 00:59:22.927 --> 00:59:27.199 You know also contain the source Spectra of the earthquakes, so 436caf55-234f-4249-83d8-d9c6c61c7b8f/9102-1 00:59:27.199 --> 00:59:31.205 you might have some discrepancy there of extrapolating this 436caf55-234f-4249-83d8-d9c6c61c7b8f/9102-2 00:59:31.205 --> 00:59:35.077 result to larger magnitudes. Do you have comment on that? 436caf55-234f-4249-83d8-d9c6c61c7b8f/9119-0 00:59:37.167 --> 00:59:41.012 I definitely agree with the fact that a response spectrum 436caf55-234f-4249-83d8-d9c6c61c7b8f/9119-1 00:59:41.012 --> 00:59:42.537 analysis might produce. 436caf55-234f-4249-83d8-d9c6c61c7b8f/9147-0 00:59:43.657 --> 00:59:46.901 Better results. The reason why we pursued this method is 436caf55-234f-4249-83d8-d9c6c61c7b8f/9147-1 00:59:46.901 --> 00:59:50.430 ultimately this project kind of came out as a side project of 436caf55-234f-4249-83d8-d9c6c61c7b8f/9147-2 00:59:50.430 --> 00:59:52.707 another study where we were looking at. 436caf55-234f-4249-83d8-d9c6c61c7b8f/9221-0 00:59:53.167 --> 00:59:57.016 Uh site response in Northern California and we found, for 436caf55-234f-4249-83d8-d9c6c61c7b8f/9221-1 00:59:57.016 --> 01:00:00.866 example, that the Northern California had this this awful 436caf55-234f-4249-83d8-d9c6c61c7b8f/9221-2 01:00:00.866 --> 01:00:04.583 path effect that was just skewing our site response. So 436caf55-234f-4249-83d8-d9c6c61c7b8f/9221-3 01:00:04.583 --> 01:00:08.698 our site response study is done in PSA. The response spectral 436caf55-234f-4249-83d8-d9c6c61c7b8f/9221-4 01:00:08.698 --> 01:00:12.814 domain. So we needed a way to try to cap to correct for these 436caf55-234f-4249-83d8-d9c6c61c7b8f/9221-5 01:00:12.814 --> 01:00:16.796 path effects. Hence why this this work, it broadly captures 436caf55-234f-4249-83d8-d9c6c61c7b8f/9221-6 01:00:16.796 --> 01:00:21.044 the effects. We don't go couple steps farther to try to capture 436caf55-234f-4249-83d8-d9c6c61c7b8f/9221-7 01:00:21.044 --> 01:00:24.297 everything. I think it's definitely future work. 436caf55-234f-4249-83d8-d9c6c61c7b8f/9236-0 01:00:24.737 --> 01:00:28.037 And I agree that a response spectral or sorry for your 436caf55-234f-4249-83d8-d9c6c61c7b8f/9236-1 01:00:28.037 --> 01:00:29.117 spectral analysis. 436caf55-234f-4249-83d8-d9c6c61c7b8f/9247-0 01:00:30.407 --> 01:00:33.277 Would be probably a good way to to pursue it. 436caf55-234f-4249-83d8-d9c6c61c7b8f/9244-0 01:00:31.767 --> 01:00:32.137 I. 436caf55-234f-4249-83d8-d9c6c61c7b8f/9283-0 01:00:34.257 --> 01:00:37.993 Uh, I don't know if I get to make three comments, but you 436caf55-234f-4249-83d8-d9c6c61c7b8f/9283-1 01:00:37.993 --> 01:00:42.116 mentioned anelastic attenuation. This could also be scattering. 436caf55-234f-4249-83d8-d9c6c61c7b8f/9283-2 01:00:42.116 --> 01:00:45.916 Attenuation can also contribute to this effect. It doesn't 436caf55-234f-4249-83d8-d9c6c61c7b8f/9283-3 01:00:45.916 --> 01:00:48.107 change any of that analysis, but. 436caf55-234f-4249-83d8-d9c6c61c7b8f/9294-0 01:00:49.057 --> 01:00:51.957 To say it's just all anelastic attenuation. 436caf55-234f-4249-83d8-d9c6c61c7b8f/9314-0 01:00:52.647 --> 01:00:56.925 I think it's a little simplified because obviously scattering 436caf55-234f-4249-83d8-d9c6c61c7b8f/9314-1 01:00:56.925 --> 01:01:00.997 also can cause a seismic waves to decay with distance. So. 436caf55-234f-4249-83d8-d9c6c61c7b8f/9364-0 01:01:01.457 --> 01:01:04.939 Yeah, I definitely agree. We we went with the simplification 436caf55-234f-4249-83d8-d9c6c61c7b8f/9364-1 01:01:04.939 --> 01:01:08.422 assumption that they analyzed. Continuation was gonna vary a 436caf55-234f-4249-83d8-d9c6c61c7b8f/9364-2 01:01:08.422 --> 01:01:12.018 lot more than the the geometric spit reading or the scattering 436caf55-234f-4249-83d8-d9c6c61c7b8f/9364-3 01:01:12.018 --> 01:01:15.672 attenuation. So obviously for a for a full adjustment to a path 436caf55-234f-4249-83d8-d9c6c61c7b8f/9364-4 01:01:15.672 --> 01:01:18.527 model, we'd have to look at all of these effects. 436caf55-234f-4249-83d8-d9c6c61c7b8f/9369-0 01:01:19.447 --> 01:01:20.797 OK. Thank you very much. 436caf55-234f-4249-83d8-d9c6c61c7b8f/9378-0 01:01:27.077 --> 01:01:30.857 OK, I think that's all the time we have for the. 436caf55-234f-4249-83d8-d9c6c61c7b8f/9398-0 01:01:31.547 --> 01:01:37.329 Formal session. So Susan, if you could stop the recording or I 436caf55-234f-4249-83d8-d9c6c61c7b8f/9398-1 01:01:37.329 --> 01:01:38.247 can do it.