WEBVTT Kind: captions Language: en-US 00:00:01.000 --> 00:01:37.620 [Silence] 00:01:38.120 --> 00:01:41.620 - So I’m going to get started? - Yeah. 00:01:42.240 --> 00:01:46.580 - All right. Hello, everybody. Thanks for coming to seminar 00:01:46.590 --> 00:01:49.620 Just a couple of announcements before getting started. 00:01:49.620 --> 00:01:53.361 I guess we were asked to announce that tomorrow morning there’s going to be 00:01:53.361 --> 00:01:59.080 an all-hands meeting with Mark Sogge. And also then let you know about our 00:01:59.080 --> 00:02:04.540 semester for next week is going to be given by one of our new postdocs here, 00:02:04.540 --> 00:02:09.520 Noha Farghal, discussing the work that she did at Stanford. 00:02:09.530 --> 00:02:11.750 So I’m going to turn the mic over to Ruth Harris 00:02:11.750 --> 00:02:14.700 to introduce today’s speaker. 00:02:20.880 --> 00:02:22.940 - See if I can see with my reading glasses on. 00:02:22.950 --> 00:02:26.590 I’m very happy that Hongfeng has come to visit us. 00:02:26.590 --> 00:02:29.590 He flew from Hong Kong yesterday, and I’m very jealous because 00:02:29.590 --> 00:02:33.220 he flew on Singapore Airlines. We always have to fly on those 00:02:33.220 --> 00:02:37.680 American carriers, but he gets to fly in a comfortable carrier. 00:02:38.470 --> 00:02:43.140 He received his bachelor’s and master’s from USTC – University of Science and 00:02:43.140 --> 00:02:49.680 Technology in China, and then his Ph.D. from St. Louis University in 2010. 00:02:49.680 --> 00:02:55.250 He then went to Woods Hole to be a postdoc from 2010 to 2012, 00:02:55.250 --> 00:03:00.460 and then he was a research scientist at Georgia Tech from 2013 to ’14. 00:03:00.460 --> 00:03:06.040 And since 2014, he’s been a professor at the Chinese University of Hong Kong. 00:03:06.860 --> 00:03:10.760 He serves on the editorial boards of Earth and Planetary Physics and also 00:03:10.760 --> 00:03:15.010 Seismological Research Letters. And he’s received some really cool 00:03:15.010 --> 00:03:19.970 awards – the Editor’s Citation for Excellence in Refereeing from GRL, 00:03:19.970 --> 00:03:25.160 and then also the Young Scientist Award from the Chinese Geophysical Society. 00:03:25.160 --> 00:03:29.640 And this was in recognition of his contributions to integrating earthquake 00:03:29.650 --> 00:03:33.120 rupture dynamics in observational seismology to advance 00:03:33.120 --> 00:03:37.180 our understanding of earthquake physics in seismic hazard evaluation. 00:03:37.180 --> 00:03:40.660 So I’m really happy to present to you Hongfeng. 00:03:41.760 --> 00:03:44.980 - Should I turn it on? - Yep. Turn it on. 00:03:47.600 --> 00:03:49.960 - Okay. And good morning. 00:03:49.960 --> 00:03:55.520 It’s really my great pleasure to visit the USGS and present at seminar today. 00:03:55.520 --> 00:03:59.440 Thanks for the introduction from Ruth. It was really nice. 00:03:59.440 --> 00:04:05.360 So today, I’m going to share with you some recent research we are doing 00:04:05.360 --> 00:04:11.260 on the earthquake magnitude estimation based on interseismic locking. 00:04:11.260 --> 00:04:13.320 - [whispering] 00:04:14.280 --> 00:04:17.000 - Here? All right. Okay. 00:04:18.540 --> 00:04:24.510 So first, before I start – and then I want to thank my students and postdocs and 00:04:24.510 --> 00:04:28.560 collaborators who all contribute to this study in different means. 00:04:28.560 --> 00:04:33.880 And also, of course, I need to thank the sponsor as well. 00:04:33.880 --> 00:04:38.160 And this map I’m showing – I’m showing you basically the purpose, 00:04:38.160 --> 00:04:44.190 the focus, of this talk and also the main geographical region, 00:04:44.190 --> 00:04:50.260 which is in Central – Central America subduction zone Nicoya Peninsula 00:04:50.260 --> 00:04:55.160 and then where I’m going to derive, based on the – on the map is the 00:04:55.160 --> 00:05:00.760 interseismic locking distribution and dynamic rupture scenario. 00:05:00.760 --> 00:05:07.840 But the focus here is to test how this scenario is, you know, 00:05:07.840 --> 00:05:12.000 reliable and then how can we validate such different 00:05:12.000 --> 00:05:16.030 rupture scenarios based on the different locking distributions? 00:05:16.030 --> 00:05:21.790 So I’m interested in this question, or this problem, partly because of the 00:05:21.790 --> 00:05:27.660 question I was always being asked. You know, as a Chinese scientist, I was 00:05:27.660 --> 00:05:32.090 asked often by my relatives or friends, can you predict the earthquake? 00:05:32.090 --> 00:05:36.320 So initially, then, I would just answer, no, we can’t. 00:05:36.320 --> 00:05:40.600 And just – as shown in this Perspective paper more than two decades ago, 00:05:40.600 --> 00:05:45.889 published in Science, and that we think earthquake cannot be predicted. 00:05:45.889 --> 00:05:51.259 And then some of – some of them, we respond rather, you know, 00:05:51.259 --> 00:05:54.259 discouraging after hearing, oh, that we can’t. 00:05:54.259 --> 00:05:58.180 And they would say, okay, you guys suck. [laughter] 00:05:58.180 --> 00:06:02.250 And then some of them would even respond with a rebuttal. 00:06:02.250 --> 00:06:05.140 Can’t earthquake be predicted? 00:06:05.700 --> 00:06:09.860 With a coincidence being identical to this title 00:06:09.870 --> 00:06:13.090 of the Technical Comments on that paper. 00:06:13.090 --> 00:06:16.490 And I don’t think they have read these comments, however, they come up 00:06:16.490 --> 00:06:20.740 with the rebuttal based on different lines of evidence, 00:06:20.740 --> 00:06:24.380 like the animal and cloud patterns that we don’t believe. 00:06:24.380 --> 00:06:28.560 However, this is basically – we, you know, sometimes got challenged. 00:06:28.560 --> 00:06:34.690 And such a simple question seems to be, you know, tricky or difficult to answer. 00:06:34.690 --> 00:06:38.630 And then later, I tried to respond with 00:06:38.630 --> 00:06:43.300 my questions when I was asked again, can you predict the earthquake? 00:06:43.300 --> 00:06:47.870 And then I would tell them, okay, there are three critical elements, which we 00:06:47.870 --> 00:06:53.010 cannot do all of them simultaneously. However, if I ask you to 00:06:53.010 --> 00:06:57.680 pick up your own priority, which one would you care the most? 00:06:57.680 --> 00:07:02.540 Most of them will respond, okay, what about the city I live? 00:07:02.540 --> 00:07:04.590 So they care about locations. 00:07:04.590 --> 00:07:07.900 If I answer, okay, there might be earthquake in your city, 00:07:07.900 --> 00:07:11.460 then they will immediately ask, you know, how big it will be. 00:07:11.460 --> 00:07:15.630 So they care about the magnitude. If I answer, okay, your city is safe. 00:07:15.630 --> 00:07:18.370 Only small earthquakes in magnitude 3, 4, 00:07:18.370 --> 00:07:20.760 and then they don’t bother asking when. 00:07:20.760 --> 00:07:25.620 So, of course, my – you know, limited sampling point, 00:07:25.630 --> 00:07:28.910 and then very small group. But then that made me think, 00:07:28.910 --> 00:07:33.200 okay, how should we put, you know, that critical question 00:07:33.200 --> 00:07:37.720 up in the – in terms of location, magnitude, and time. 00:07:37.720 --> 00:07:42.430 So, in my view, and then, I think our common understanding is that 00:07:42.430 --> 00:07:47.750 we can do location with pretty good confidence as, you know, what we 00:07:47.750 --> 00:07:53.490 can derive based on the historical large earthquake distributions with our – 00:07:53.490 --> 00:07:56.080 all associated with major plate boundaries. 00:07:56.080 --> 00:07:59.470 And then we can also come up with seismic hazard map, 00:07:59.470 --> 00:08:02.910 which has been done as, you know, the leading expert 00:08:02.910 --> 00:08:07.530 in USGS with the probability-based hazard map prediction. 00:08:07.530 --> 00:08:15.170 So we have confidence in where we may face the serious seismic hazard threat. 00:08:15.170 --> 00:08:19.280 However, if we just pick up one particular region, or a smaller scale – 00:08:19.280 --> 00:08:24.030 for example, in the particular segments of the subduction zones 00:08:24.030 --> 00:08:28.800 or in the inter-plate region, it is pretty difficult to say, okay, 00:08:28.800 --> 00:08:31.480 how large the next earthquake will be. 00:08:31.480 --> 00:08:36.340 So that will be something I want to discuss and to share what we want to 00:08:36.340 --> 00:08:40.940 learn, or what we can learn, based on some physical-based models. 00:08:41.940 --> 00:08:45.860 So my focus here will be on subduction zones. 00:08:45.860 --> 00:08:49.940 Because subduction zones is where we may see frequent earthquakes. 00:08:49.940 --> 00:08:55.350 And we all know subduction zones – then in certain subduction zones, 00:08:55.350 --> 00:09:01.940 we may expect to observe the vertical uplift and the horizontal landward 00:09:01.940 --> 00:09:08.000 motion, indicating that some portion on the megathrust is locked. 00:09:08.000 --> 00:09:11.020 Although the locking degree can be calculated if we have 00:09:11.020 --> 00:09:14.870 a very good measurement. For example, here I’m showing, 00:09:14.870 --> 00:09:21.240 in South America subduction zone in Chile. And then all consistent landward 00:09:21.240 --> 00:09:26.310 motion indicates that the megathrust is locked to different degrees. 00:09:26.310 --> 00:09:30.250 And then, if we calculate and then – carefully, then we can derive such a 00:09:30.250 --> 00:09:35.760 colorful map that’s commonly called a locking distribution or locking models. 00:09:35.760 --> 00:09:42.329 And that have been used very effective to qualitatively assess potential – 00:09:42.329 --> 00:09:45.260 earthquake potential in the particular segment. 00:09:45.260 --> 00:09:49.610 And which have been proven true in a lot of places, or occasions, 00:09:49.610 --> 00:09:52.870 when you see a recent megathrust rupture. 00:09:52.870 --> 00:09:57.699 For example, in this map, we can see at least three recent major 00:09:57.699 --> 00:10:06.589 megathrust earthquakes. In the north, 2014, Iquique, magnitude 8.2 00:10:06.589 --> 00:10:13.120 And this part – smaller segment – is the 2015 Illapel, magnitude 8.3 00:10:13.120 --> 00:10:16.860 And this large one is 2010 Maule earthquake. 00:10:16.860 --> 00:10:21.550 So then we feel, okay, kind of a confidence in terms of inferring 00:10:21.550 --> 00:10:24.920 the future earthquake potential based on locking maps. 00:10:24.920 --> 00:10:27.860 Although we all understand earthquakes rupture propagation 00:10:27.860 --> 00:10:31.240 are rather more complicated and not only depending on 00:10:31.240 --> 00:10:34.740 the static imaging of a locking distribution. 00:10:34.740 --> 00:10:40.310 However, if we look – take a closer look – for example, 00:10:40.310 --> 00:10:47.110 if I just take the southern portion of this map, for example here, then we 00:10:47.110 --> 00:10:53.140 see the locking, it seems – the dark color indicating strong locking seems to 00:10:53.140 --> 00:10:56.371 be everywhere, nearly. And then why the rupture 00:10:56.371 --> 00:11:01.779 cannot just rupture all of them once with a much larger event. 00:11:01.779 --> 00:11:07.200 And so the central question is, if we just stare at a locking distribution, 00:11:07.200 --> 00:11:10.480 can we infer the potential rupture segmentation 00:11:10.480 --> 00:11:13.700 and thus the megathrust earthquake magnitude? 00:11:13.700 --> 00:11:16.450 So, if so, then how? 00:11:16.450 --> 00:11:21.089 And of course, after the occurrence of recent megathrust earthquakes, 00:11:21.089 --> 00:11:26.300 then we can attribute all the rupture segmentations are caused by different 00:11:26.300 --> 00:11:30.660 heterogeneities, either subducting, geometrical, heterogeneities, 00:11:30.660 --> 00:11:36.300 or stress or material rheology, or different earthquake histories. 00:11:36.300 --> 00:11:43.550 For the place like in southern Chile, then we know, you know, roughly well, 00:11:43.550 --> 00:11:48.730 where historical rupture occurred. For example, in this case, that made 00:11:48.730 --> 00:11:54.700 me think, okay, the central segment, seemingly locked, but then remain 00:11:54.700 --> 00:12:00.120 unbroken, may be associated with the different rupture histories. 00:12:01.060 --> 00:12:06.380 However, they post a challenge because we know some subduction zones are 00:12:06.380 --> 00:12:11.680 locked, but then they don’t produce recent megathrust earthquakes. 00:12:11.680 --> 00:12:16.790 Not even relative moderate size. That’s the case you guys probably 00:12:16.790 --> 00:12:21.770 know very well, which is in Cascadia. Then here, Cascadia – we believe 00:12:21.770 --> 00:12:25.730 Cascadia has been locked, but the locking degree has been a debate, 00:12:25.730 --> 00:12:28.960 mostly because we are measuring – or, we are 00:12:28.960 --> 00:12:34.220 inferring the offshore locking distribution using inland measurement. 00:12:35.270 --> 00:12:39.760 But the question is, we don’t – you know, it is really challenging 00:12:39.760 --> 00:12:43.135 for us to answer, if next Cascadia earthquake happens, 00:12:43.140 --> 00:12:48.320 it’s going to be rupturing the entire Cascadia [inaudible], or it’s going to 00:12:48.320 --> 00:12:51.800 be segmented, or to certain degrees. I think that question 00:12:51.800 --> 00:12:56.370 is important and also the main focus of today. 00:12:56.370 --> 00:13:01.040 However, I would not put all the examples, or most of the examples, 00:13:01.040 --> 00:13:06.540 along Cascadia because, if we want to derive such, 00:13:06.540 --> 00:13:11.320 you know, spontaneous rupture models, then we first need to validate 00:13:11.320 --> 00:13:16.760 our models in some means. And this is why I chose the Nicoya. 00:13:16.760 --> 00:13:22.740 And then, for Cascadia case, of course, I chose three typical locking models. 00:13:22.740 --> 00:13:26.270 Another question on the locking model is that these models 00:13:26.270 --> 00:13:31.440 may have the different uncertainties. And assuming a different locking degree 00:13:31.440 --> 00:13:35.820 near trench, then we may come up with very different locking distribution, 00:13:35.820 --> 00:13:42.060 even using the exactly same data set. Like, assuming the free variation 00:13:42.060 --> 00:13:45.200 with the near-trench locking or the full locking near trench, 00:13:45.200 --> 00:13:49.040 then we come up with very different locking distributions. 00:13:49.040 --> 00:13:55.860 And the model also depending on the material rheology we are using. 00:13:55.860 --> 00:14:01.810 If we use a viscoelastic upper mantle rheology, and then we may come up 00:14:01.810 --> 00:14:04.120 with such a locking distribution which is 00:14:04.120 --> 00:14:08.080 different than assuming pure elastic material. 00:14:08.080 --> 00:14:13.910 So these all pose the questions, how we can better evaluate the future 00:14:13.910 --> 00:14:19.780 earthquake potential in terms of rupture segmentation and from locking models. 00:14:20.260 --> 00:14:26.420 I think the critical questions are, first, how we can infer rupture segmentations. 00:14:26.420 --> 00:14:30.600 And then second question is that we have to validate 00:14:30.600 --> 00:14:35.360 any models that we can derive. And based on these two critical 00:14:35.360 --> 00:14:42.380 questions, I choose the geographical focus in Nicoya Peninsula. 00:14:42.380 --> 00:14:48.490 And one of the reasons is that because of the peninsula is protruding to the trench, 00:14:48.490 --> 00:14:52.870 so that we even use only the land measurement, we can construct 00:14:52.870 --> 00:14:57.410 a relatively good locking models in the region. 00:14:57.410 --> 00:15:03.181 And, on the map view shown here, in this color – the red color is 00:15:03.181 --> 00:15:07.860 showing the locking model. And here is the coastline over here. 00:15:07.860 --> 00:15:11.899 And then another important thing in Nicoya Peninsula is that 00:15:11.899 --> 00:15:15.420 there is a megathrust earthquake being well-recorded 00:15:15.420 --> 00:15:19.920 occurring in 2012 with a magnitude 7.6. 00:15:19.920 --> 00:15:26.959 So here – although we may have the – kind of a unique geographical 00:15:26.959 --> 00:15:31.970 environment for Nicoya, but different locking model exist. 00:15:31.970 --> 00:15:35.910 And using the different data set. For example, on the right-hand side, 00:15:35.910 --> 00:15:42.200 this model is mostly derived from trans-parallel GPS measurement. 00:15:42.200 --> 00:15:45.750 And then later, they were – they update the locking model 00:15:45.750 --> 00:15:53.170 using both trans-parallel and trench-normal GPS measurement 00:15:53.170 --> 00:15:57.820 and plus some InSAR constraint. So then they slightly update this. 00:15:57.820 --> 00:16:01.250 So we can see the difference between the two different locking models, 00:16:01.250 --> 00:16:06.130 but the key feature is that they all constrain two major locked patches. 00:16:06.130 --> 00:16:13.450 Which, if ruptured once, then will produce a magnitude 7.8 earthquake. 00:16:13.450 --> 00:16:20.779 So the purpose here is to derive coseismic rupture models 00:16:20.779 --> 00:16:24.690 from the locking model. For example, from the left, 00:16:24.690 --> 00:16:29.980 which is the locking, then I can derive a spontaneous rupture simulation. 00:16:29.980 --> 00:16:35.870 But the key is that I have to validate my model using kinematic model 00:16:35.870 --> 00:16:39.610 in the Nicoya earthquake that occurred in 2012. 00:16:39.610 --> 00:16:43.000 So that is the first step we need to do. 00:16:43.000 --> 00:16:48.200 And how to derive the spontaneous rupture model, and then there are 00:16:48.210 --> 00:16:52.580 some critical ingredients. So first, it’s the geometry. 00:16:52.580 --> 00:16:56.720 And then some – we have to prescribe a friction law as well. 00:16:56.720 --> 00:17:00.950 And the most important thing, we need to specify those parameters 00:17:00.950 --> 00:17:06.289 in the friction law, and then we have to estimate the stress 00:17:06.289 --> 00:17:09.669 on the megathrust from interseismic locking. 00:17:09.669 --> 00:17:14.429 So I will spend some time to explain how we can set up 00:17:14.429 --> 00:17:19.910 the different parameters. And here we use the linear 00:17:19.910 --> 00:17:23.730 slip-weakening law, which is – I know it’s a very simplified version, 00:17:23.730 --> 00:17:28.159 but then it’s commonly used in spontaneous rupture simulations. 00:17:28.159 --> 00:17:30.780 And then, to some extent, it can present some 00:17:30.780 --> 00:17:35.100 coseismic slip behavior in past earthquakes. 00:17:35.100 --> 00:17:39.070 And here – the good thing for this linear slip-weakening law is that 00:17:39.070 --> 00:17:43.670 there are only four parameters. So we need to get the estimate of 00:17:43.670 --> 00:17:48.850 the initial stress. Then we need to also get the yield strength on the fault. 00:17:48.850 --> 00:17:53.049 Then we need to have the critical slip distance, D slope of C, and then 00:17:53.049 --> 00:17:56.779 we need to describe the dynamic stress during the rupture. 00:17:56.779 --> 00:18:01.809 So here, I simplify the problem by looking at the rock-sliding 00:18:01.809 --> 00:18:05.720 experiment with [inaudible] the seismic slip rate in 00:18:05.720 --> 00:18:09.820 meters per second, then also, regardless of the different rock samples 00:18:09.820 --> 00:18:14.399 and experimental conditions. And in the seismic rate, 00:18:14.399 --> 00:18:18.320 nearly all samples are showing very low friction coefficient. 00:18:18.320 --> 00:18:23.700 So that I make the assumption that we assume a constant dynamic stress. 00:18:23.700 --> 00:18:27.410 Then the next question is, how can we get the initial stress? 00:18:27.410 --> 00:18:31.740 And for the yield stress, we prescribe as a uniform yield strength on the 00:18:31.740 --> 00:18:35.580 megathrust, assuming the near-lithostatic pore pressure. 00:18:35.580 --> 00:18:37.750 And for the D slope of C, 00:18:37.750 --> 00:18:41.700 then we can try the different distribution. Later, I will discuss 00:18:41.700 --> 00:18:46.680 some additional constraint we may – we may have, based on observations. 00:18:47.660 --> 00:18:52.580 And then I should say, all the calculation, including static calculation 00:18:52.590 --> 00:18:57.900 and dynamic simulations, are all done using the software [inaudible]. 00:18:57.900 --> 00:19:01.460 So really thank the developer – one of the developers 00:19:01.460 --> 00:19:04.000 sitting in the audience right over here. 00:19:04.760 --> 00:19:10.260 So the key thing for us to think about first is how to convert the locking 00:19:10.269 --> 00:19:14.169 into the estimation of the stress on the megathrust. 00:19:14.169 --> 00:19:20.080 Here I use the strategy is to infer the potential slip deficit on the 00:19:20.080 --> 00:19:25.059 megathrust from locking. So the slip deficit qualitatively can 00:19:25.059 --> 00:19:30.289 be calculated by the locking and times the plate convergence rate in subduction 00:19:30.289 --> 00:19:35.270 zones, and the plate convergence rates are mostly well-constrained. 00:19:35.270 --> 00:19:37.940 And then the years since the last major rupture. 00:19:37.940 --> 00:19:44.070 In Nicoya, the characteristic earthquake rupture history is rather simple. 00:19:44.070 --> 00:19:50.049 And then some earthquake will be generated with magnitude 7.8 00:19:50.049 --> 00:19:56.429 also in roughly 50 years. And with the most recent one, of course, in 2005. 00:19:56.429 --> 00:20:00.110 And then the last most recent one was in 1950. 00:20:00.110 --> 00:20:05.200 So that – I got the years since the last rupture, which is 62 years. 00:20:05.200 --> 00:20:10.330 And then the key is that, by estimating slip deficit over here, 00:20:10.330 --> 00:20:16.830 we implicitly made certain assumptions. For example, the slip deficit 00:20:16.830 --> 00:20:23.669 accumulated since the last major rupture will not be released by other, you know, 00:20:23.669 --> 00:20:29.090 thrust earthquakes on the megathrust. And then there was no other form of 00:20:29.090 --> 00:20:34.580 slip deficit released by, for example, SSEs or any other means. 00:20:34.580 --> 00:20:42.511 So, in Nicoya here – so first, these blue circles – and then they 00:20:42.511 --> 00:20:47.179 indicate the locations of cumulative SSE distributions. 00:20:47.179 --> 00:20:53.519 So they happen to be surrounding the locked patch, with the shallow one 00:20:53.519 --> 00:20:57.570 overlapping with the locked patch. Over here – and then I want to remind 00:20:57.570 --> 00:21:03.690 us that in both the offshore constraint on the SSE location and the locking are 00:21:03.690 --> 00:21:08.799 not very well constrained, so that small overlap may not play a major role 00:21:08.799 --> 00:21:15.379 in terms of releasing the slip deficit. Rather, if the SSEs really occur 00:21:15.379 --> 00:21:19.929 in the surrounding regions, they may likely load the locked patch. 00:21:19.929 --> 00:21:26.019 And second, what about the potential release by the interface earthquakes, 00:21:26.019 --> 00:21:31.190 or smaller size? So, in this region, which I outline 00:21:31.190 --> 00:21:36.520 with a yellow box over here, if we just look at the thrust earthquakes – 00:21:36.520 --> 00:21:42.500 and I can – I took all the moment tensor solutions from GCMT catalog. 00:21:42.500 --> 00:21:46.600 And then we found there were not so many thrust earthquakes 00:21:46.600 --> 00:21:50.799 occurring on the interface. And then the magnitude are much 00:21:50.799 --> 00:21:58.789 smaller comparing to the 7.6 magnitude. And over here, of course, now, after the 00:21:58.789 --> 00:22:03.659 2012 earthquake, only got seven years. But then we don’t observe any 00:22:03.659 --> 00:22:09.590 thrust earthquake larger than 5.5. So maybe in this Nicoya Peninsula, 00:22:09.590 --> 00:22:15.580 this is sort of the characteristic earthquake cycle where you may – 00:22:15.580 --> 00:22:22.220 you know, you may not have a lot of smaller-size thrust earthquake releasing 00:22:22.220 --> 00:22:27.880 the slip deficit. So then here I will say, okay, that’s assumption. 00:22:27.880 --> 00:22:33.460 With very little slip deficit released by other – by other forms may 00:22:33.460 --> 00:22:39.890 hold true in the Nicoya Peninsula. So next part is that I assume the 00:22:39.890 --> 00:22:45.299 entire slip deficit will be released by this single large rupture. 00:22:45.299 --> 00:22:52.259 So here, then I just use that locking and then got the slip deficit. 00:22:52.259 --> 00:22:57.200 If I assume the entire slip deficit will be released by one earthquake, 00:22:57.200 --> 00:23:00.830 then I can basically calculate the static stress drop. 00:23:00.830 --> 00:23:04.809 Then, if we recall that linear slip-weakening, then by knowing 00:23:04.809 --> 00:23:10.250 the constant dynamic stress, and then using the static stress drop calculated 00:23:10.250 --> 00:23:15.590 by the slip deficit, then we can estimate the initial stress distribution. 00:23:15.590 --> 00:23:20.649 So this is how we got the initial stress, which is shown in this map. 00:23:20.649 --> 00:23:24.239 So after we got the initial stress distribution, then we can nucleate 00:23:24.239 --> 00:23:28.090 the rupture. For a spontaneous rupture simulation, we have to 00:23:28.090 --> 00:23:32.639 artificially nucleate the rupture. Then I nucleate the rupture from 00:23:32.639 --> 00:23:38.789 here because this is the hypocenter inferred by the kinematic model – 00:23:38.789 --> 00:23:42.389 kinematic rupture model. Because I need to validate my dynamic 00:23:42.389 --> 00:23:47.380 rupture model by the kinematic model, so I first force the rupture start 00:23:47.380 --> 00:23:52.520 from here, which seems to be in a relatively low stress region. 00:23:53.580 --> 00:23:58.679 So after we nucleate the rupture over here, because the stress 00:23:58.679 --> 00:24:04.399 is heterogeneously distributed, so the stress will be dominant factor 00:24:04.399 --> 00:24:07.799 for the rupture propagation. So the rupture started from 00:24:07.799 --> 00:24:11.830 the hypocenter, and then first propagate up-dip and down-dip, 00:24:11.830 --> 00:24:15.690 and then propagate along strike, and then eventually stops, 00:24:15.690 --> 00:24:21.749 forming moment magnitude 7.6 earthquake in this scenario. 00:24:21.749 --> 00:24:25.490 So at this point, at least the moment magnitude seems to 00:24:25.490 --> 00:24:31.150 be consistent with observations. But then we also need to validate to 00:24:31.150 --> 00:24:36.809 see whether our slip distribution will be consistent with the kinematic model. 00:24:36.809 --> 00:24:40.730 And then, for earthquake, one important factor is the moment rate, which is a 00:24:40.730 --> 00:24:48.249 [inaudible] function. So next thing is to compare, first the slip distribution. 00:24:48.249 --> 00:24:54.960 On the left-hand side, I plot our dynamic model using the black color. 00:24:54.960 --> 00:25:00.379 And the red showing one kinematic model, and blue is 00:25:00.379 --> 00:25:04.779 showing another kinematic model. The blue one is mostly derived from 00:25:04.779 --> 00:25:09.679 GPS, and the red is from a combination with GPS and seismic data. 00:25:09.679 --> 00:25:13.200 So, by looking at the 1-meter slip contour and then this seems to 00:25:13.200 --> 00:25:17.690 be spatially well-correlated. And, although the peak slip region 00:25:17.690 --> 00:25:21.869 may not be identical, but then at least, you know, 00:25:21.869 --> 00:25:25.860 the major slip patch are well-constrained. 00:25:25.860 --> 00:25:30.379 In terms of moment rate, the red and blue are the two different moment rate 00:25:30.379 --> 00:25:36.100 derived from different kinematic models, and the black is our dynamic 00:25:36.100 --> 00:25:40.639 model, which falls in between the two different kinematic models. 00:25:40.639 --> 00:25:45.830 So then I will say, okay, this dynamic rupture models, in terms of a final 00:25:45.830 --> 00:25:52.309 slip distribution and moment rate, are both consistent with observations. 00:25:52.309 --> 00:25:58.559 Then we will build up our confidence in our spontaneous rupture scenarios. 00:25:58.560 --> 00:26:02.760 So here, then, we just basically demonstrate using the interseismic 00:26:02.760 --> 00:26:08.020 locking distribution, we may be able to derive a rupture scenario. 00:26:08.029 --> 00:26:12.730 And then, if we choose the – you know, careful parameters, then we may derive 00:26:12.730 --> 00:26:18.070 some reliable rupture scenarios that are consistent with observations. 00:26:18.070 --> 00:26:22.080 But one challenge is that, if we really want to do, you know, 00:26:22.080 --> 00:26:27.600 the magnitude estimation, or future rupture scenarios, another thing is that 00:26:27.600 --> 00:26:31.100 we don’t really know where the rupture may nucleate before the earthquake 00:26:31.100 --> 00:26:34.779 occurred. And even after a megathrust earthquake occurred, 00:26:34.779 --> 00:26:39.970 in subduction zones, it is typical, we always have the station from the land. 00:26:39.970 --> 00:26:43.859 So we may not be able to constrain the epicenter very well. 00:26:43.859 --> 00:26:48.399 And that’s the case for the Nicoya earthquake shown on the map on 00:26:48.399 --> 00:26:52.919 the right and where we may see the different green stars over here. 00:26:52.919 --> 00:26:58.149 And these are the different hypocenters suggested by different groups. 00:26:58.149 --> 00:27:00.340 And of course, this yellow star, I think, 00:27:00.340 --> 00:27:05.340 is a automatic network location – it’s a USGS location. 00:27:05.340 --> 00:27:10.090 So over here, then, we have to consider this factor because, 00:27:10.090 --> 00:27:15.509 in the heterogeneous stress distribution, different initiation point may 00:27:15.509 --> 00:27:20.470 play different roles in terms of a final slip distribution 00:27:20.470 --> 00:27:22.539 and even moment magnitude. 00:27:22.539 --> 00:27:26.179 So I will firstly show you a few simple case. 00:27:26.179 --> 00:27:32.399 Over here, the message is clear. If we start the rupture from different 00:27:32.399 --> 00:27:36.059 nucleation point, then the rupture may end up in different moment 00:27:36.059 --> 00:27:41.950 magnitude and slip distribution. For example, if I start with a rupture 00:27:41.950 --> 00:27:47.019 from Point A, then that is the final slip distribution, and that moment 00:27:47.019 --> 00:27:52.399 magnitude is 7.4. And if I start a rupture from this point, which is 00:27:52.399 --> 00:27:58.149 bounded with nearly identical initial stress distribution, but the rupture will 00:27:58.149 --> 00:28:03.720 only rupture one locked patch down-dip and forming a magnitude 7.2. 00:28:03.720 --> 00:28:07.700 So, although the magnitude difference is only 0.2, but the moment is 00:28:07.700 --> 00:28:11.800 actually different by – you know, doubled – twice. 00:28:11.800 --> 00:28:16.440 And so you may think, okay, this is one up-dip or one down-dip. 00:28:16.440 --> 00:28:21.979 And then now, if we try – really the two points just next to each other with 00:28:21.979 --> 00:28:27.340 nearly identical – or, actually, just call it identical stress distribution. 00:28:27.340 --> 00:28:33.299 But at the point from the blue star, then that rupture will only rupture 00:28:33.299 --> 00:28:37.479 the up-dip locked patch, forming a magnitude 7.2. 00:28:37.479 --> 00:28:42.080 And then the case next to it, starting from the red star, 00:28:42.080 --> 00:28:48.039 then we’ll break the entire locked patch from magnitude 7.4. So, again, 00:28:48.039 --> 00:28:54.960 the epicenter – different nucleation points really play critical roles over here. 00:28:54.960 --> 00:29:01.730 So, if we conduct spontaneous rupture simulations at any potential point in the 00:29:01.730 --> 00:29:06.460 locking map, and this is what we have done, and this is what we get. 00:29:06.460 --> 00:29:10.870 So over here – and then I just tried different potential locations 00:29:10.870 --> 00:29:16.779 in the locking. And then I prescribed the nucleation point, the size. 00:29:16.779 --> 00:29:20.249 And then I observe the final moment magnitude. 00:29:20.249 --> 00:29:23.419 So the color is showing here the final moment magnitude. 00:29:23.419 --> 00:29:27.450 And with the black line showing here is the locking degree with 00:29:27.450 --> 00:29:33.320 50% locking and the dashed line showing 75% of the locking. 00:29:33.320 --> 00:29:40.520 So if we consider 75% is high locking, and then yes, in the Nicoya region, 00:29:40.529 --> 00:29:44.139 most of the nucleation points in the high locking will run into 00:29:44.139 --> 00:29:52.179 a magnitude 7.2 or 7.4 earthquake, which is – which is large in this region. 00:29:52.179 --> 00:29:58.349 But, if we start the rupture from intermediate locking regions, then we 00:29:58.349 --> 00:30:05.970 only observe with magnitude, you know, 5 or 6 – you know, that size earthquake. 00:30:05.970 --> 00:30:10.539 And here is a very simple statistic to show the different percent 00:30:10.539 --> 00:30:13.929 of the nucleation point running to different magnitude. 00:30:13.929 --> 00:30:19.739 And, of course, this is – you know, based on how many points we have 00:30:19.740 --> 00:30:24.760 tried. And over here is less than 40% of the rupturing 00:30:24.760 --> 00:30:28.340 or nucleation point that will develop into a magnitude 00:30:28.349 --> 00:30:34.850 earthquake that are – that will be larger than 7 in the Nicoya region. 00:30:34.850 --> 00:30:41.750 So if we change this plot into another view, by evaluating the locking models, 00:30:41.750 --> 00:30:45.549 and then I can plot the moment magnitude, which is showing in 00:30:45.549 --> 00:30:50.999 the Y axis, versus the locking degree. So each point over here is indicating 00:30:50.999 --> 00:30:56.389 the rupture starting from the point with the associated locking degree, 00:30:56.389 --> 00:31:00.690 and then with the final moment magnitude marked over here. 00:31:00.690 --> 00:31:05.809 So we tried a different nucleation size, but then, regardless of the nucleation 00:31:05.809 --> 00:31:12.269 size, as we can see, most nucleation from relatively high locking 00:31:12.269 --> 00:31:19.009 will run into large earthquakes. However, there are significant points 00:31:19.009 --> 00:31:24.149 where the ruptures start from intermediate to moderate locking 00:31:24.149 --> 00:31:29.100 degrees but then will end up being a relatively small earthquake. 00:31:29.100 --> 00:31:35.039 And a interesting feature over here is that, based on the locking distribution, 00:31:35.039 --> 00:31:40.570 then in our spontaneous rupture model predictions, we don’t see, you know, 00:31:40.570 --> 00:31:45.759 magnitude 6 to 7 earthquakes over here. Either we will have larger than 7 00:31:45.759 --> 00:31:52.980 or either we have a smaller – or, smaller than 6. So why there is a gap? 00:31:52.980 --> 00:31:57.029 So, based on the – from the modeling point of view, we can understand it. 00:31:57.029 --> 00:32:00.630 Because of the heterogeneous stress distribution, then the rupture will 00:32:00.630 --> 00:32:04.190 be dominated by the heterogeneity – by the stress. 00:32:04.190 --> 00:32:09.210 So that different locations, then you may have the different size. 00:32:09.210 --> 00:32:12.239 But then why there is a gap? It seems – you know, the magnitude 00:32:12.239 --> 00:32:17.220 seems not to be continuous. So first, we need to check whether 00:32:17.220 --> 00:32:24.249 this is artificial from the locking model or it is somehow, to some extent, real, 00:32:24.249 --> 00:32:27.389 representing something real in the Nicoya Peninsula. 00:32:27.389 --> 00:32:31.889 So first, we test another locking model. And then the epicenter or 00:32:31.889 --> 00:32:37.450 hypocenter-dependent moment magnitude feature seems to be true 00:32:37.450 --> 00:32:41.600 regardless of the locking model. This is a earlier locking model. 00:32:41.600 --> 00:32:45.769 Then also, based on this locking model, we also see that gap. 00:32:45.769 --> 00:32:48.770 And then, of course, there will be a different locking model so that 00:32:48.770 --> 00:32:53.520 the different initial stress distribution, but then we just don’t see that 00:32:53.520 --> 00:32:59.620 magnitude 6 to 7 earthquakes in our model predictions. 00:32:59.620 --> 00:33:04.129 And then we go back to see the observations. 00:33:04.129 --> 00:33:09.590 And if we look in the other interface earthquakes again, 00:33:09.590 --> 00:33:13.779 for the Nicoya Peninsula. And over here, here is showing, 00:33:13.779 --> 00:33:17.989 in the yellow box, and where we extract all the moment tensor 00:33:17.989 --> 00:33:23.460 from GCMT catalog, and indeed, I would call between – you know, 00:33:23.460 --> 00:33:25.979 I draw these two lines kind of artificially. 00:33:25.979 --> 00:33:29.179 I didn’t do – run any statistical means. 00:33:29.179 --> 00:33:34.409 But then we don’t see many earthquakes having magnitude larger than 5.5 00:33:34.409 --> 00:33:39.139 Before 2005, there were not many larger earthquakes anyway, but then there 00:33:39.139 --> 00:33:45.100 are only two potential earthquakes, which this one has large uncertainties 00:33:45.100 --> 00:33:50.179 in terms of locations. Even the fault within the locking region, 00:33:50.179 --> 00:33:55.219 but still, there are only two earthquakes within that range. 00:33:55.219 --> 00:33:58.809 So I think, over here, the locking models – although the 00:33:58.809 --> 00:34:03.419 locking model may have a limited spatial resolution, but they somehow 00:34:03.419 --> 00:34:07.440 probably reflect something, you know, to some extent, 00:34:07.440 --> 00:34:11.109 the stress state occurring on the megathrust. 00:34:11.109 --> 00:34:15.750 And so that – in our model predictions, we don’t see that particular magnitude 00:34:15.750 --> 00:34:21.909 earthquakes which is also – seems to be true in the Nicoya Peninsula. 00:34:21.909 --> 00:34:28.429 If we run the comparison by moving a little bit north, and then we see this 00:34:28.429 --> 00:34:34.080 segment, regardless of the occurrence of the 2012 Nicoya earthquake, 00:34:34.080 --> 00:34:39.480 and that earthquake’s magnitude seems to be more consistent with the different 00:34:39.480 --> 00:34:45.589 magnitude distribution on the interface. So I think this is – in the Nicoya 00:34:45.589 --> 00:34:49.600 Peninsula, that locking model, at least to some extent, 00:34:49.600 --> 00:34:53.190 can reflect some earthquake potential. 00:34:53.190 --> 00:35:00.450 And then, based on this hypocenter- dependent magnitude distribution, 00:35:00.450 --> 00:35:04.600 although we may see, you know, some earthquake may end up being 00:35:04.600 --> 00:35:08.380 the same magnitude with different nucleation points, but I think, 00:35:08.390 --> 00:35:10.109 for the subduction zone environment, 00:35:10.109 --> 00:35:14.030 another important thing is potential generation of a tsunami. 00:35:14.030 --> 00:35:19.839 So here I will show you two cases. One case being the rupture – 00:35:19.839 --> 00:35:25.280 the rupture nucleating from the shallow point with magnitude 7.4. 00:35:25.280 --> 00:35:27.710 And then another case is the rupture nucleating 00:35:27.710 --> 00:35:33.300 from the deeper portion with a magnitude 7.4 as well. 00:35:33.300 --> 00:35:38.840 But the final slip distribution may be little bit different in details, 00:35:38.840 --> 00:35:41.410 but the moment magnitude are pretty much the same. 00:35:41.410 --> 00:35:46.010 However, the rupture directivity effect may play 00:35:46.010 --> 00:35:50.760 critical roles in terms of a final ground displacement. 00:35:50.760 --> 00:35:55.780 If we look at the ground displacement produced by the two different case 00:35:55.780 --> 00:36:02.390 scenarios across the – as shown across this profile, the right one is showing 00:36:02.390 --> 00:36:06.470 the rupture nucleating from the deeper portion which is this case. 00:36:06.470 --> 00:36:11.810 And the blue one is showing this case. Actually, the peak – especially in the 00:36:11.810 --> 00:36:18.550 shallow part, in the peak displacement on the ground, maybe twice, or even 00:36:18.550 --> 00:36:24.510 larger, for a deeper-nucleating rupture than a shallower-nucleating rupture. 00:36:24.510 --> 00:36:32.360 So I think, over here, if we incorporate, you know, our ground displacement 00:36:32.360 --> 00:36:37.870 predicted by the different nucleating points, then we may incorporate with 00:36:37.870 --> 00:36:42.670 tsunami models, providing potential constraints for the near-field tsunami 00:36:42.670 --> 00:36:47.400 early warning if we can have, you know, very good locking 00:36:47.400 --> 00:36:51.980 model constraint and then using the pre-computed cases. 00:36:51.980 --> 00:36:56.540 And if we consider all the different nucleation points, and then for the – 00:36:56.540 --> 00:37:04.130 all the deeper portion was – highlight – use the different red points over here. 00:37:04.130 --> 00:37:09.150 And here I am showing you the slip distribution versus the depth 00:37:09.150 --> 00:37:13.780 on the megathrust. And the red one are showing 00:37:13.780 --> 00:37:17.010 all the ruptures nucleating from the deeper portion. 00:37:17.010 --> 00:37:21.050 And then, at the shallow depths, they all tend to have a much larger slip, 00:37:21.050 --> 00:37:26.700 which will, in turn, generate much larger ground displacement at the 00:37:26.700 --> 00:37:34.000 shallower portion versus the rupture nucleating from the shallower depths. 00:37:35.020 --> 00:37:38.980 So, over here then, I tried to demonstrate, we can, you know, 00:37:38.980 --> 00:37:42.060 derive reliable rupture scenarios, and then we also found the 00:37:42.060 --> 00:37:45.980 hypocenter-dependent rupture scenarios. Then you may question it, 00:37:45.980 --> 00:37:49.910 so these are just models. And then I completely agree that 00:37:49.910 --> 00:37:55.550 we got the quote, actually seen often at different occasions – all models 00:37:55.550 --> 00:38:00.230 are wrong. But some are useful. So how would I convince myself and 00:38:00.230 --> 00:38:05.140 then convince, hopefully, some of you to believe, oh, those models are useful. 00:38:05.140 --> 00:38:10.490 I’m thinking, so for example, for the kinematic rupture models, we tend to 00:38:10.490 --> 00:38:14.760 believe, because at least, you know, these models capture some features. 00:38:14.760 --> 00:38:17.320 And then these models can produce synthetics being 00:38:17.320 --> 00:38:21.620 consistent with the data. So can we produce, also from 00:38:21.620 --> 00:38:26.340 our dynamic rupture model, and being consistent with the data? 00:38:27.580 --> 00:38:34.940 So I think – I want to run a little bit quickly, and then, to be short – 00:38:34.940 --> 00:38:39.130 to have a short statement, yes, we can, if we can have well-constrained 00:38:39.130 --> 00:38:42.850 frictional properties on the seismogenic fault. 00:38:42.850 --> 00:38:48.390 Which is not easy, of course. And I’m showing you here is the – 00:38:48.390 --> 00:38:52.360 you know, the difficulty comes from, you know, different means. But one of 00:38:52.360 --> 00:38:58.350 them is the strong tradeoff between the potential strength and critical slip 00:38:58.350 --> 00:39:04.380 distance, even using a very simple friction law, like the linear slip 00:39:04.380 --> 00:39:09.110 friction law. As demonstrated in this paper, if we observe the different 00:39:09.110 --> 00:39:13.620 ground displacement, then even up to 1 hertz, we cannot really 00:39:13.620 --> 00:39:19.790 distinguish whether it’s a high-strength smaller critical slip distance or a 00:39:19.790 --> 00:39:24.790 much lower strength or a much larger critical slip distance. 00:39:24.790 --> 00:39:29.350 If the two models are derived based on the same fracture energy. 00:39:29.350 --> 00:39:35.420 And then that tradeoff is believed to be true, but then recently, 00:39:35.420 --> 00:39:39.820 we come up with another approach, and then we found, yes, there are actually, 00:39:39.820 --> 00:39:45.370 indeed, the strong tradeoff between the strength and the critical slip distance. 00:39:45.370 --> 00:39:48.760 But based on only one single constrained parameter. 00:39:48.760 --> 00:39:53.240 But if we look at the different constraints, for example, the final slip 00:39:53.240 --> 00:40:00.210 or the rupture speed, or the ground velocity versus ground displacement, 00:40:00.210 --> 00:40:04.380 and then the tradeoff trend are different. So for example, as shown here, 00:40:04.380 --> 00:40:09.620 then we see the red line represents the theoretical – the best fit of the model. 00:40:09.620 --> 00:40:14.540 But then this strong tradeoff between the strength and the slip-weakening 00:40:14.540 --> 00:40:18.570 distance. And that is so as well, but the trend are different. 00:40:18.570 --> 00:40:22.860 So if I combine two of them together, then hopefully I will see the intersection 00:40:22.860 --> 00:40:27.870 between the two theoretically best fit curves so that I can come up with 00:40:27.870 --> 00:40:32.770 a very good constraint with the strength and the critical slip distance. 00:40:32.770 --> 00:40:38.580 So that approach was first derived based on the 2015 Nepal earthquake. 00:40:38.580 --> 00:40:42.100 And here – I will ignore the details, but I would rather show you – 00:40:42.100 --> 00:40:45.900 this is a kinematic model. And also showing that kinematic 00:40:45.900 --> 00:40:50.770 model predictions with the static near-field ground displacement 00:40:50.770 --> 00:40:55.210 and the ground velocity. And here is our best fit dynamic model, 00:40:55.210 --> 00:41:03.410 also showing us the static displacement and then the ground velocity as well. 00:41:03.410 --> 00:41:08.930 So that we can remove the tradeoff between the strength and the critical slip 00:41:08.930 --> 00:41:14.160 distance – come up with a very good or robust ground velocity prediction. 00:41:14.160 --> 00:41:18.770 For the Nicoya, we are conducting the same approach, but deriving the 00:41:18.770 --> 00:41:25.380 best-constrained dynamic model. And here is all the model fits. 00:41:25.380 --> 00:41:29.700 Black is the data observed in the 2005 earthquake. 00:41:29.700 --> 00:41:33.890 And the purple ones are our dynamic model prediction. 00:41:33.890 --> 00:41:38.560 In terms of direction and amplitude, I would call this a very good fit. 00:41:38.560 --> 00:41:42.600 In terms of moment rate function. And the different observed ground 00:41:42.600 --> 00:41:45.540 velocity as well – high-rate GPS record. 00:41:45.550 --> 00:41:49.690 And then the black are data. The red are our models. 00:41:49.690 --> 00:41:54.440 So we can derive some reliable, you know, ground velocity 00:41:54.440 --> 00:41:57.660 prediction up to 1 hertz over here. 00:41:59.570 --> 00:42:06.220 So now we have the location dependence, magnitude distribution. 00:42:06.230 --> 00:42:12.570 We can also put further effort into the ground velocity prediction in Nicoya. 00:42:12.570 --> 00:42:16.920 And then this slide, I wanted to show that such location-dependent 00:42:16.920 --> 00:42:20.830 magnitude distribution, it’s not only limited to Nicoya. 00:42:20.830 --> 00:42:24.160 And then this is a very preliminary model – actually, 00:42:24.160 --> 00:42:28.890 but that was done very long ago. And the work was [inaudible]. 00:42:28.890 --> 00:42:31.780 But here is one of the locking model in Cascadia. 00:42:31.780 --> 00:42:36.470 And then, if we just catch this segment in central Oregon, and then if I – 00:42:36.470 --> 00:42:40.040 the rupture is starting from, you know, the northern part 00:42:40.040 --> 00:42:44.540 in that segment and central and then south, and then we see very 00:42:44.540 --> 00:42:48.650 different final slip distribution and moment magnitude as well. 00:42:48.650 --> 00:42:54.960 So I was – I will – you know, basically, point out and then such feature is 00:42:54.960 --> 00:42:59.920 probably true along every subduction zone given the different heterogeneous 00:42:59.920 --> 00:43:05.500 distribution with the interseismic locking. But then I will highlight the, you know, 00:43:05.500 --> 00:43:10.570 quantitative approach using, you know, spontaneous rupture models. 00:43:10.570 --> 00:43:19.920 And here, for Cascadia again, and then, if we can derive the reliable rupture 00:43:19.920 --> 00:43:25.080 scenarios as well as the ground velocities, and then 00:43:25.090 --> 00:43:29.700 maybe we can think about, in terms of getting a hazard map 00:43:29.700 --> 00:43:32.930 based on those physics-based rupture models. 00:43:32.930 --> 00:43:37.520 And then I’m not sure how easy or reliable that will be for Cascadia 00:43:37.520 --> 00:43:42.050 if we do not have any data to validate. But this is something actually I think 00:43:42.050 --> 00:43:48.110 a lot of experts sitting in the audience who can help me out over here for 00:43:48.110 --> 00:43:52.690 deriving the probability-based, but also physics-based, you know, 00:43:52.690 --> 00:43:58.450 hazard map or potential ground velocities for other subduction zones 00:43:58.450 --> 00:44:03.440 or seismic hazard potential in a seismogenic fault. 00:44:03.440 --> 00:44:10.430 So in the end, I want to summarize that we can derive some rupture scenarios. 00:44:10.430 --> 00:44:15.420 And in the case I demonstrating – the Nicoya case, we can derive a very good 00:44:15.420 --> 00:44:21.560 or reliable rupture scenario that seems to be consistent with the observations. 00:44:21.560 --> 00:44:27.460 And that potential earthquake magnitude actually dependent on 00:44:27.460 --> 00:44:33.260 the locations of the initiation point, which underscore the necessity of 00:44:33.260 --> 00:44:38.510 quantifying the interseismic lockings in terms of hazard distribution. 00:44:38.510 --> 00:44:42.900 And in Nicoya, we also found a magnitude gap in both model 00:44:42.900 --> 00:44:48.550 predictions and in the observations. So the locking models are, at least 00:44:48.550 --> 00:44:53.420 to some extent, reflect potential stress distribution on the megathrust. 00:44:53.420 --> 00:44:57.670 And then eventually, we hope to derive more, you know, 00:44:57.670 --> 00:45:01.380 cases along different locking distributions for different subduction 00:45:01.380 --> 00:45:04.980 zones and then to derive a physics-based hazard map. 00:45:04.980 --> 00:45:07.920 Okay, I will stop there. And thank you very much. 00:45:07.920 --> 00:45:13.480 [Applause] 00:45:16.020 --> 00:45:18.940 - Thank you very much. That was a great talk. 00:45:18.940 --> 00:45:22.740 Lots of – lots of things for us to think about. 00:45:23.600 --> 00:45:25.060 Jess. 00:45:26.500 --> 00:45:29.400 - Hi, thanks. That was a very interesting talk. 00:45:29.400 --> 00:45:32.540 One of the assumptions you talked about that you made was that 00:45:32.540 --> 00:45:36.750 the earthquake would release all the accumulated slip deficit. 00:45:36.750 --> 00:45:39.920 And I’m wondering whether you’ve looked at models in which that isn’t 00:45:39.920 --> 00:45:45.280 the case and what your thinking is on how important it is to consider the 00:45:45.280 --> 00:45:50.030 possibility that any given earthquake is not releasing the full slip deficit. 00:45:50.030 --> 00:45:55.170 - Right. This is a very good question. Here, we want to evaluate the 00:45:55.170 --> 00:46:01.220 potentially extreme scenario, or the worst scenario, so that we assume 00:46:01.220 --> 00:46:06.140 the entire slip deficit will be released by just one single earthquake. 00:46:06.140 --> 00:46:10.810 If only a portion of that will be released, then of course, we may run into much 00:46:10.810 --> 00:46:15.750 smaller magnitude earthquakes. And then, given, you know, that 00:46:15.750 --> 00:46:19.200 locking distribution in Nicoya, then there will probably a magnitude 00:46:19.200 --> 00:46:25.880 of 5 or 6. Yeah. But then we want to evaluate the worst scenario. 00:46:29.680 --> 00:46:33.680 - Early in the talk, you suggested that these might be characteristic 00:46:33.680 --> 00:46:36.520 earthquakes happening here, but how can you reconcile that 00:46:36.580 --> 00:46:41.940 with such a strong sensitivity to nucleation point? 00:46:41.940 --> 00:46:45.810 - Okay. Yeah. This is a question we have been thinking about for 00:46:45.810 --> 00:46:51.490 a long time after we see that result. So, in Nicoya, based on observations, 00:46:51.490 --> 00:46:57.750 and then we do see magnitude 7.7, 7.8, drop every 50 years. 00:46:57.750 --> 00:47:01.510 But based on the locking distributions, and then after we checked the different 00:47:01.510 --> 00:47:06.170 parameters, different input locking models, and then we think, if this is 00:47:06.170 --> 00:47:12.050 somehow, you know, reflecting something real, then that may help us 00:47:12.050 --> 00:47:17.580 to understand, for some earthquakes, or some sequences, that might occur 00:47:17.580 --> 00:47:21.830 in close spatial locations. But some of them run into – 00:47:21.830 --> 00:47:25.820 we call it the maybe smaller size or some foreshocks. 00:47:25.820 --> 00:47:30.360 And then some of them are – you know, ending up being a large one, 00:47:30.360 --> 00:47:32.580 then we call them main shock. 00:47:32.580 --> 00:47:37.260 So, in the Nicoya case, I think it’s rather sensitive to the location – 00:47:37.260 --> 00:47:41.580 where the earthquake may start. And then for these … 00:47:46.600 --> 00:47:52.920 For these ones – actually, for those black ones, these are the earthquakes – 00:47:52.930 --> 00:47:58.920 interface earthquake before the 2005 7.6 event. 00:47:58.920 --> 00:48:03.200 So I think they just didn’t start in the right position. 00:48:03.200 --> 00:48:07.020 And of course, based on the different time since the last major rupture, 00:48:07.020 --> 00:48:09.830 even though they may start in that particular position, 00:48:09.830 --> 00:48:13.450 they may not be magnitude 7.6. It should be smaller. 00:48:13.450 --> 00:48:18.000 But then, the main reason over here is that likely they didn’t start 00:48:18.000 --> 00:48:21.840 from the right position. That would be my interpretation. 00:48:27.100 --> 00:48:33.960 - Following up on Justin’s question, it seems to me that it’s not necessary 00:48:33.960 --> 00:48:37.700 that – I mean, the implication is that you assume that the previous earthquake 00:48:37.700 --> 00:48:42.490 released all the stress. And so you can do a sort of a slip predictable model. 00:48:42.490 --> 00:48:47.720 Whereas, if the previous one hadn’t released all the stress, then your 00:48:47.720 --> 00:48:53.830 assumption that a slip predictable model only gives a 7.8 is a pretty 00:48:53.830 --> 00:48:57.210 big assumption. And it wouldn’t be the worst case. So which is it? 00:48:57.210 --> 00:49:02.330 Do you have to assume that every single earthquake is a maximum stress drop? 00:49:02.330 --> 00:49:07.560 And this sort of gets at the point where you chose Nicoya for sort of specific 00:49:07.560 --> 00:49:12.260 reasons in terms of the resolution, no slow slip earthquakes, and so 00:49:12.270 --> 00:49:15.620 how well does this methodology work when you go to these sort of 00:49:15.620 --> 00:49:21.480 more complex areas that tend to be sort of maybe more of our subduction 00:49:21.480 --> 00:49:27.570 zones than what the Nicoya region has? - Right. This a very good question. 00:49:28.460 --> 00:49:35.160 Yes. If we – based on this approach, certainly, we assume the last major 00:49:35.160 --> 00:49:38.740 earthquake released most of the – or all the slip deficit 00:49:38.740 --> 00:49:40.800 before that earthquake as well. 00:49:40.800 --> 00:49:46.290 If, indeed, certain slip deficit was not completely released, then that will just 00:49:46.290 --> 00:49:53.080 put into certain perturbations to our – for example, the slip deficit. 00:49:53.080 --> 00:49:59.630 Then that will affect our initial stress distribution over here, to some extent. 00:49:59.630 --> 00:50:03.510 So we may think, in this stress – you know, initial stress distribution, 00:50:03.510 --> 00:50:07.210 then there will be some difference in quantities. 00:50:07.210 --> 00:50:11.280 And then, that will, in turn, affect this distribution. 00:50:11.280 --> 00:50:16.120 But then I don’t think it will play a critical role unless we perturb 00:50:16.120 --> 00:50:20.320 this by a certain significant amount. 00:50:20.320 --> 00:50:24.660 And then with such location-dependent feature. 00:50:24.660 --> 00:50:29.740 So that is something I think, you know, how to consider an answer to these. 00:50:29.740 --> 00:50:34.010 But, in terms of Nicoya, yes, I agree this is rather a unique 00:50:34.010 --> 00:50:37.610 subduction environment. We choose here because we 00:50:37.610 --> 00:50:41.640 want to validate the model. To – you know, we need to have 00:50:41.640 --> 00:50:46.540 a large earthquake that can be used to validate the spontaneous rupture model, 00:50:46.540 --> 00:50:50.510 and then we want to validate from a different perspective, 00:50:50.510 --> 00:50:55.760 including that magnitude gap. And then, if we want to, you know, 00:50:55.760 --> 00:51:00.010 apply this to other subduction zones, then we have to be more careful 00:51:00.010 --> 00:51:03.230 in terms of getting this, you know, rupture history. 00:51:03.230 --> 00:51:10.750 For example, in Chile, then we have to consider the different potential segment 00:51:10.750 --> 00:51:16.510 released in different time, and then calculating the slip deficit 00:51:16.510 --> 00:51:21.130 across the different segment. For Cascadia, then there is pretty much 00:51:21.130 --> 00:51:28.191 no interface earthquakes, and then that is, I think, close to a Nicoya case. 00:51:28.191 --> 00:51:32.520 And then, if we assume the slip deficit is building up and then released by 00:51:32.520 --> 00:51:38.660 some degree, then we can probably use the similar approach. 00:51:41.420 --> 00:51:47.220 [Silence] 00:51:47.580 --> 00:51:51.840 - So, since you have this figure up, there’s a big locked patch, and I think 00:51:51.840 --> 00:51:56.200 it’s the southern end, that doesn’t rupture in any of the scenarios. 00:51:56.200 --> 00:52:00.150 What would you have to change about the mapping between the locking map 00:52:00.150 --> 00:52:06.220 and the slip-weakening parameters to allow it to rupture through to that patch? 00:52:06.880 --> 00:52:12.860 - Right. Here – if we want to rupture this patch, then, over here, then 00:52:12.860 --> 00:52:18.150 we basically have to artificially increase either the stress – 00:52:18.150 --> 00:52:24.210 initial stress or sort of decrease the fracture energy across that gap. 00:52:24.210 --> 00:52:26.430 But here, this patch – we put it over here because 00:52:26.430 --> 00:52:31.020 that is included in every locking model. 00:52:31.020 --> 00:52:38.140 But in every locking model, this patch is not as well-constrained as this one. 00:52:38.140 --> 00:52:42.250 Because this is really sitting on the peninsula, and this is rather 00:52:42.250 --> 00:52:47.230 to the south offshore. So we didn’t really attempt to break, 00:52:47.230 --> 00:52:51.080 you know, that central locking patch to the south. 00:52:52.140 --> 00:52:54.540 But in models, yes, we can. 00:52:54.540 --> 00:52:59.360 But we don’t know whether this is something real or not. 00:53:01.880 --> 00:53:04.900 [Silence] 00:53:05.380 --> 00:53:07.060 - Any more questions? 00:53:09.940 --> 00:53:12.500 - Does anyone have any more questions? 00:53:13.460 --> 00:53:16.520 All right. Well, that’s the end of our seminar for today. 00:53:16.530 --> 00:53:22.920 We will taking Hongfeng to lunch here on campus if anyone wants to join us. 00:53:22.920 --> 00:53:25.800 And let’s thank our speaker once again. - Okay. Thank you very much. 00:53:25.800 --> 00:53:29.960 [Applause]