WEBVTT Kind: captions Language: en-US 00:00:00.440 --> 00:00:03.300 [inaudible background conversations] 00:00:03.300 --> 00:00:06.680 Good morning, everyone. I think we’re going to get started now. 00:00:06.680 --> 00:00:09.500 Thanks for coming to the Earthquake Science Center weekly seminar. 00:00:09.500 --> 00:00:14.650 Today our speaker is Anne Socquet. She is a professor at the Université 00:00:14.650 --> 00:00:20.130 Grenoble Alpes in ISTerre, or the Institut des Sciences de la Terre. 00:00:20.130 --> 00:00:24.570 And she’s also currently a Miller invited professor and Fulbright Fellow 00:00:24.570 --> 00:00:28.500 at UC-Berkeley Department of Earth and Planetary Sciences. 00:00:28.500 --> 00:00:31.680 And today, she’s going to be talking about intriguing observations 00:00:31.680 --> 00:00:36.060 of the long-term preparation of megathrust earthquakes. 00:00:36.060 --> 00:00:39.340 And with that, I’ll turn it over to Anne. - Thank you very much. 00:00:39.340 --> 00:00:41.440 Thanks for the kind introduction. 00:00:41.440 --> 00:00:45.480 So I’m very happy to be here. Thank you for coming to this seminar. 00:00:45.480 --> 00:00:50.239 So I’m going to try to show you some recent observations that 00:00:50.239 --> 00:00:54.679 have been made of the preparation of megathrust earthquakes. 00:00:54.679 --> 00:00:59.540 So basically, the signal that we can see before the occurrence 00:00:59.540 --> 00:01:02.900 of a large subduction earthquake. 00:01:02.900 --> 00:01:06.600 So, first of all, I would like to acknowledge my co-authors. 00:01:06.600 --> 00:01:13.040 So the work that I’m going to show today has been done in the frame of 00:01:13.040 --> 00:01:17.280 three Ph.D. theses – the Ph.D. thesis of Jorge Jara, 00:01:17.280 --> 00:01:22.080 the Ph.D. thesis of Jesus Peña Valdes, the Ph.D. thesis of Marianne Métois, 00:01:22.080 --> 00:01:26.160 and the former postdoc of Daniel Carrizo. 00:01:26.160 --> 00:01:30.460 And this has been done in collaboration with several people working, 00:01:30.470 --> 00:01:34.730 particularly will thank Christophe Vigny, Michel Bouchon, David Marsan, 00:01:34.730 --> 00:01:39.940 and Fabrice Cotton, who contributed a lot to these observations. 00:01:40.820 --> 00:01:46.280 So a bit of context. So all of you know that most of 00:01:46.280 --> 00:01:50.230 the large – and the largest earthquakes occur at subduction zone. 00:01:50.230 --> 00:01:55.980 And one of the impressive thing is that earthquakes are clustering 00:01:55.980 --> 00:02:01.090 in time but also in space. And we have, like, this clustering 00:02:01.090 --> 00:02:09.890 of a series of megathrust earthquakes that are rupturing one given subduction 00:02:09.890 --> 00:02:15.180 zone on the – so there was the Aleutian Alaska sequence 00:02:15.180 --> 00:02:18.900 that occurred in the ’60s. There was, of course, 00:02:18.900 --> 00:02:21.859 the Sumatra sequence that ruptured the whole Sumatra 00:02:21.859 --> 00:02:26.530 subduction zone over about 10 years. And, at the moment, there is this 00:02:26.530 --> 00:02:32.230 Chile/Peru sequence that is rupturing the whole subduction in Chile. 00:02:32.230 --> 00:02:37.489 And so we can wonder why we have, like, this kind of clustering. 00:02:37.489 --> 00:02:44.069 And if this kind of clustering is somehow related with the preparation 00:02:44.069 --> 00:02:48.090 of the – of those megathrust earthquakes, because those earthquakes 00:02:48.090 --> 00:02:52.379 are occurring on one subduction zone, but they are really far away from each 00:02:52.380 --> 00:02:58.340 other, so it cannot be explained by a single triggering of those earthquakes. 00:02:58.800 --> 00:03:05.500 So, to start with, I’m going to show what kind of data I’m using. 00:03:05.510 --> 00:03:09.799 So I’m using mostly GPS – sorry – GPS data. 00:03:09.799 --> 00:03:19.269 So those GPS data can give you some ideas of how the upper plate is moving. 00:03:19.269 --> 00:03:22.359 Here – you can see here. And this upper plate is moving 00:03:22.359 --> 00:03:26.480 as a response to interseismic loading on the megathrust here. 00:03:26.480 --> 00:03:29.779 And when you have an earthquake, you have the elastic rebound. 00:03:29.779 --> 00:03:36.169 So the message here is, basically, if you can monitor with high precision 00:03:36.169 --> 00:03:43.099 what is – how the upper plate is deforming, then you can get some 00:03:43.100 --> 00:03:48.180 very good insights on what is occurring on the subduction megathrust. 00:03:48.180 --> 00:03:53.760 So, of course, the seismic cycle that is shown here is very simple. 00:03:53.760 --> 00:03:59.069 There is the interseismic loading that is followed by simple rebound. 00:03:59.069 --> 00:04:06.469 And what you measure, in this case, on the – on the GPS time series 00:04:06.469 --> 00:04:09.900 is the following. You have this phase of interseismic 00:04:09.900 --> 00:04:12.900 loading and then the rebound. And this phase of interseismic 00:04:12.900 --> 00:04:17.669 loading and then the rebound. But, of course, the seismic cycle is 00:04:17.669 --> 00:04:23.680 much more complicated than this. And so, first of all, we know that there 00:04:23.680 --> 00:04:27.729 is probably, like, a super cycle. I mean, you have, like, different 00:04:27.729 --> 00:04:33.390 kind of – different magnitude of earthquakes rupturing the megathrust. 00:04:33.390 --> 00:04:37.770 So you can have full ruptures, which are going to generate 00:04:37.770 --> 00:04:42.150 a very large displacement. And partial ruptures. 00:04:42.150 --> 00:04:48.660 Of course, it’s well-known that, after an earthquake, there is a 00:04:48.660 --> 00:04:53.700 postseismic phase where you can have afterslip or viscoelastic relaxation. 00:04:53.700 --> 00:04:59.920 And during this interseismic loading phase, you can have slow-slip events 00:04:59.930 --> 00:05:08.419 or kind of a slow trench that is, like, associated with a variation 00:05:08.419 --> 00:05:12.009 in interseismic loading. And so the subduction I’m going to 00:05:12.009 --> 00:05:17.400 talk today about are subduction zones that are kind of cold. 00:05:17.400 --> 00:05:22.040 So it’s not like Cascadia. It’s more like Chile or Japan, 00:05:22.040 --> 00:05:25.530 offshore Honshu. And those subduction zones 00:05:25.530 --> 00:05:30.340 do not show regular slow-slip events and tremors. 00:05:30.340 --> 00:05:36.780 But instead, they are more associated with recent 00:05:36.780 --> 00:05:41.580 very large megathrust structures. And so there are no tremors, 00:05:41.580 --> 00:05:45.419 and there are no regular slow-slip events, but sometimes you can 00:05:45.419 --> 00:05:51.710 see subtle trench and deformation that occur at these subduction zones. 00:05:51.710 --> 00:05:56.800 And we can wonder whether those subtle trench and deformation 00:05:56.800 --> 00:06:02.420 are related or not with the preparation of earthquakes. 00:06:02.420 --> 00:06:08.740 So there will be four parts in my talk. The first part is going to be quite fast. 00:06:08.740 --> 00:06:12.659 So the first part is going to be, what is the link between interseismic 00:06:12.659 --> 00:06:19.530 loading and the coseismic rupture. Then I will talk about, can we 00:06:19.530 --> 00:06:25.240 observe any precursor to large ruptures from geodetic 00:06:25.240 --> 00:06:27.889 and seismological point of view? 00:06:27.889 --> 00:06:32.030 And then, what is driving the plate interface destabilization? 00:06:32.030 --> 00:06:37.289 So I will go at a larger time and space scale to look at that. 00:06:37.289 --> 00:06:46.979 And then I will try to address weird observations of relationship 00:06:46.979 --> 00:06:52.880 between recent earthquakes during – within the sequence of earthquakes. 00:06:52.880 --> 00:07:00.000 So this is a map of Chile. So I’m going to work a lot about Chile. 00:07:00.000 --> 00:07:03.360 And I’m going to talk a lot about Chile, so, as I told you, 00:07:03.360 --> 00:07:07.539 Chile is a – is a very seismically active area. 00:07:07.539 --> 00:07:12.540 So there was the Maule earthquake that occurred in 2010. 00:07:12.540 --> 00:07:17.930 There was this Iquique earthquake that occurred in 2014. 00:07:17.930 --> 00:07:23.699 And then there was the Illapel earthquake in 2015. 00:07:23.699 --> 00:07:28.960 And here you can see the microseismicity of Chile. 00:07:28.960 --> 00:07:32.330 And this microseismicity was plot before 2010. 00:07:32.330 --> 00:07:37.020 And what you can see is that there are areas where there are gaps of seismicity. 00:07:37.020 --> 00:07:43.719 Notably, here in the area that was later ruptured by the Maule earthquake. 00:07:43.719 --> 00:07:52.740 And here in this area, which is known as the North Chile seismic gap 00:07:52.740 --> 00:07:58.409 that was later partially ruptured by the Iquique earthquake. 00:07:58.409 --> 00:08:09.889 So, in this area, it’s interesting also to see that there is deep seismicity. 00:08:09.889 --> 00:08:12.480 And notably, here in the central Andes. 00:08:12.480 --> 00:08:20.419 And this deep seismicity occurs at 80 to 100, 120 kilometers’ depth. 00:08:20.419 --> 00:08:26.340 And this deep seismicity is associated with slab-pull focal mechanisms. 00:08:26.340 --> 00:08:33.240 So we can wonder why there is so much slab-pull seismicity here in this area. 00:08:33.240 --> 00:08:38.640 And what is the potential relationship with what happens above? 00:08:38.640 --> 00:08:48.360 So, after the Maule earthquake occurred, what we have done is – and before, 00:08:48.370 --> 00:08:54.570 we have done a series of GPS measurements all along the coast of 00:08:54.570 --> 00:09:02.660 Chile. And those GPS measurements leaded to this interseismic velocity field. 00:09:02.660 --> 00:09:07.680 And this interseismic velocity field can give us insights on what happens 00:09:07.680 --> 00:09:12.780 on the megathrust. And what we could see is that we have, 00:09:12.780 --> 00:09:17.840 like, areas that are fully coupled before megathrust earthquakes, 00:09:17.850 --> 00:09:22.200 while there are some other areas that are less fully coupled 00:09:22.200 --> 00:09:25.240 in between megathrust earthquakes. 00:09:25.240 --> 00:09:31.000 So notably, what is quite clear from this is that the GPS measurements 00:09:31.000 --> 00:09:38.690 that were done before the Maule earthquake here led to an interseismic 00:09:38.690 --> 00:09:42.230 coupling that was really high. And interestingly, the Maule 00:09:42.230 --> 00:09:44.540 earthquake occurred exactly in this area, 00:09:44.540 --> 00:09:50.500 and it ruptured an area that was heavily locked before this earthquake. 00:09:51.250 --> 00:09:57.040 This was the same for Illapel earthquake that ruptured an area that was heavily 00:09:57.050 --> 00:10:01.880 locked before the earthquake. And Iquique earthquake was also – 00:10:01.880 --> 00:10:05.660 also occurred in an area that was heavily locked. 00:10:06.420 --> 00:10:09.780 Another thing that can be seen is that there are areas that have, 00:10:09.780 --> 00:10:14.579 like, partial coupling. And you can see that, along the 00:10:14.579 --> 00:10:20.029 strike of the subduction zone, we have, like, lateral variations 00:10:20.029 --> 00:10:26.380 of coupling that show lateral segmentation of the subduction zone. 00:10:26.380 --> 00:10:29.959 And this segmentation is probably associated with the 00:10:29.960 --> 00:10:34.800 occurrence of the limitation of future earthquakes. 00:10:35.760 --> 00:10:43.040 So now, can we observe any precursor to large – to large ruptures? 00:10:43.040 --> 00:10:49.529 So there were several studies that showed that there are – 00:10:49.529 --> 00:10:54.350 there is an increased seismicity rate before a large rupture. 00:10:54.350 --> 00:10:59.529 And so this is – this is a figure that was extracted from a paper 00:10:59.529 --> 00:11:05.540 by Bouchon et al. in 2013. And what is quite clear is that – 00:11:05.540 --> 00:11:08.889 so what this plot here is – the normalized stack 00:11:08.889 --> 00:11:14.639 of the cumulative seismic moment of 25 interplate sequence. 00:11:14.639 --> 00:11:19.329 And what is clear is that you have, like, this increase of seismicity 00:11:19.329 --> 00:11:23.220 before the earthquake. And what is puzzling is that, 00:11:23.220 --> 00:11:29.140 here, you have this plot over the 24 hours before the earthquake. 00:11:29.140 --> 00:11:33.260 Here the same plot over the five days before the earthquake. 00:11:33.260 --> 00:11:37.470 And here the same plot over the six months before the earthquake. 00:11:37.470 --> 00:11:41.550 So this paper shows that this increased seismicity seems to 00:11:41.550 --> 00:11:47.399 occur over different time scales. So you have it over a short time scale, 00:11:47.400 --> 00:11:54.720 but you also might have this kind of signal over several months. 00:11:54.720 --> 00:12:03.800 So these are the two end member models of nucleation of earthquakes. 00:12:03.810 --> 00:12:09.610 So the first model is a slow cascade of earthquake where you have one first 00:12:09.610 --> 00:12:13.829 earthquake that triggers the second one, the third one, that triggers the 00:12:13.829 --> 00:12:17.589 fourth one, and eventually it’s going to trigger the main shock. 00:12:17.589 --> 00:12:20.400 And the second model is a pre-slip triggering. 00:12:20.400 --> 00:12:23.930 So it means that you have the fault that is locked. 00:12:23.930 --> 00:12:29.120 And this fault starts to slip slowly. And this slip increases. 00:12:29.120 --> 00:12:35.340 And, as a response to this slow slip, you start to have foreshock seismicity. 00:12:35.340 --> 00:12:38.399 That is a really a response to this slow slip. 00:12:38.399 --> 00:12:43.690 And so, to discriminate between those two models, it’s not completely 00:12:43.690 --> 00:12:51.690 straightforward because most of the observations are done and were done 00:12:51.690 --> 00:12:59.130 using seismicity analysis and stress drop analysis, and it’s really useful 00:12:59.130 --> 00:13:04.780 to try to extract the actual slip, and notably using GPS data, 00:13:04.780 --> 00:13:11.769 this is really useful. Because, if you use GPS data or any kind of 00:13:11.769 --> 00:13:20.009 deformation data, you can access to the actual aseismic part of the slip. 00:13:20.009 --> 00:13:27.000 So – if this exists. So we have looked at that before the Iquique earthquake. 00:13:27.000 --> 00:13:32.410 So we are here in North Chile. So here you see the 00:13:32.410 --> 00:13:37.370 coast of North Chile. Here it’s the boundary with Peru. 00:13:37.370 --> 00:13:43.540 Here it’s the Iquique earthquake – the source of the Iquique earthquake. 00:13:43.540 --> 00:13:50.209 And the different triangles represent the different coastal GPS stations 00:13:50.209 --> 00:13:55.850 that we have. And here you have the cumulative seismicity over the time. 00:13:55.850 --> 00:14:00.139 And you can see that this cumulative seismicity increases with time. 00:14:00.139 --> 00:14:04.670 So it’s starts to increase sometime before the earthquake. 00:14:04.670 --> 00:14:11.480 So there was a swarm, like, eight months before the main shock. 00:14:11.480 --> 00:14:17.260 And then there was a large foreshock that was magnitude 6.7. 00:14:17.260 --> 00:14:20.709 And here you have a zoom of what happens between this large 00:14:20.709 --> 00:14:25.730 foreshock and the main shock itself. And you see that, at that time, 00:14:25.730 --> 00:14:28.910 the seismicity increases dramatically. 00:14:28.910 --> 00:14:31.540 And eventually, you have the main shock. 00:14:31.540 --> 00:14:37.910 So we tried to look – so most of the studies focused on this signal, which 00:14:37.910 --> 00:14:44.040 was this signal that occurred during the two weeks before the main shock. 00:14:44.040 --> 00:14:48.600 And very little is known about potential long-term precursors. 00:14:48.600 --> 00:14:56.610 So I’m going to talk about geodetic observations and seismological 00:14:56.610 --> 00:15:01.740 observations that we have done over longer time period, 00:15:01.740 --> 00:15:09.980 like the monthly scale. And then I will increase even the time window in order 00:15:09.980 --> 00:15:19.220 to look at potential movements that occur over the decadal time scale. 00:15:19.880 --> 00:15:24.410 So here you see a GPS time series of – this is a station 00:15:24.410 --> 00:15:27.779 which is in Iquique – in the city of Iquique. 00:15:27.779 --> 00:15:31.020 So here it’s the northeast placement as a function of time. 00:15:31.020 --> 00:15:35.320 Here it’s the main earthquake. Here it’s the large foreshock. 00:15:35.320 --> 00:15:39.430 And here it’s the swarm that occurred in July 2013. 00:15:39.430 --> 00:15:46.840 On the east displacement, which is the component that is most affected by the 00:15:46.840 --> 00:15:52.300 seismic cycle since the subduction is north-south in this area, 00:15:52.310 --> 00:15:55.079 you can see the interseismic loading here. 00:15:55.079 --> 00:15:58.980 And here you see the coseismic, and here the postseismic tail. 00:15:58.980 --> 00:16:03.209 So if you remove everything that happened after the earthquake, 00:16:03.209 --> 00:16:07.910 and you detrend, here is what you obtain. So here, again, 00:16:07.910 --> 00:16:12.329 the detrended north displacement, detrended east displacement. 00:16:12.329 --> 00:16:19.410 And what is quite clear from this is that there is a strong signal here just before 00:16:19.410 --> 00:16:25.829 the main shock that occurs during the two weeks preceding the earthquake. 00:16:25.829 --> 00:16:29.110 But this is not the only signal that we can see. 00:16:29.110 --> 00:16:39.440 We can also see that here the trend before 2013 and in between July 2013 00:16:39.440 --> 00:16:44.010 and the foreshock is not the same. So it’s a tiny signal. 00:16:44.010 --> 00:16:48.810 But it’s quite clear from the data. And so, if we try to do the same 00:16:48.810 --> 00:16:58.170 kind of exercise using old GPS stations we have, we obtain this kind of figure, 00:16:58.170 --> 00:17:07.130 where we have here all the GPS time series that have been detrended. 00:17:07.130 --> 00:17:14.480 We have removed all the signal that occurred after the main foreshock. 00:17:14.480 --> 00:17:17.940 So we are just looking at what happens before. 00:17:17.940 --> 00:17:22.170 And the color code represents the velocity anomaly 00:17:22.170 --> 00:17:28.150 that we may see in those GPS time series. 00:17:28.150 --> 00:17:33.929 So it’s – so we are computing this six months’ average velocity 00:17:33.929 --> 00:17:38.440 over the different time series. So we can see that it’s fluctuating a 00:17:38.440 --> 00:17:45.240 little bit, but, in average, it’s zero. And what is popping up very much 00:17:45.250 --> 00:17:53.159 from this figure, it’s this red patch here, where we have, like, consistent velocity 00:17:53.159 --> 00:17:56.440 anomaly, although it’s quite a small velocity anomaly. 00:17:56.440 --> 00:18:00.850 And this velocity anomaly, interestingly, is restricted to the 00:18:00.850 --> 00:18:06.760 stations that are coastal stations close to the epicenter. 00:18:06.760 --> 00:18:11.620 And when you are far away from the epicenter, like here in the Mejillones 00:18:11.620 --> 00:18:15.760 Peninsula, for example, there is not – you cannot see this anomaly. 00:18:15.760 --> 00:18:20.260 Or here, further north, you don’t see this anomaly, either. 00:18:20.260 --> 00:18:25.030 So this strongly suggests that there is something going on. 00:18:25.030 --> 00:18:30.350 So this – so you can compute the displacement associated with 00:18:30.350 --> 00:18:36.250 this velocity anomaly. And the displacement are small, 00:18:36.250 --> 00:18:43.600 but you can – so we tried to invert it. And, by inverting the displacement, 00:18:43.600 --> 00:18:48.420 we obtain the blue patch here – the blue patches. 00:18:48.420 --> 00:18:51.760 And, interestingly, it’s surrounding the main shock. 00:18:51.770 --> 00:18:55.600 So the resolution is not very good, of course, because we are really 00:18:55.600 --> 00:19:01.740 at the limit of detection. But what is interesting is that we 00:19:01.740 --> 00:19:09.280 have an amount of displacement. This is quite well – quite well-resolved, 00:19:09.280 --> 00:19:15.100 and if this amount of displacement corresponds to a slip on the subduction, 00:19:15.100 --> 00:19:18.650 we can compute the magnitude. So the slip distribution is not very 00:19:18.650 --> 00:19:23.169 Well-constrained, but the magnitude is quite well-constrained. 00:19:23.169 --> 00:19:26.390 So we can compute the geodetic magnitude, 00:19:26.390 --> 00:19:33.250 and this geodetic magnitude is 6.5. And, if we compare with the seismicity 00:19:33.250 --> 00:19:38.200 that has been released during the same period, we can see that the 00:19:38.200 --> 00:19:46.900 seismicity represents only 20%. So this slow slip is 80% aseismic. 00:19:46.900 --> 00:19:49.840 It’s mostly an aseismic process. 00:19:49.840 --> 00:19:55.530 Then you can do the same exercise during the two weeks 00:19:55.530 --> 00:20:02.040 before the main shock. And so this is what we obtain. 00:20:02.040 --> 00:20:09.320 Here, the signal is much stronger. We obtain a geodetic magnitude of 7.0. 00:20:09.320 --> 00:20:12.920 And if we compare with the seismicity that is released at the same time, 00:20:12.920 --> 00:20:17.380 and it’s much more seismic, we obtain something which is 00:20:17.380 --> 00:20:25.409 less aseismic – only 35% aseismic. So it’s, like, more a combined process. 00:20:25.409 --> 00:20:32.020 So we have been wondering what happens during the same time period 00:20:32.020 --> 00:20:36.730 and what is the – what are the characteristics of these foreshocks. 00:20:36.730 --> 00:20:42.540 So, to do that, we have looked at the foreshocks we had, and we tried to 00:20:42.540 --> 00:20:47.650 look at the earthquake frequency content of those foreshocks. 00:20:47.650 --> 00:20:57.179 So here, we have performed spectral ratio of different earthquakes. 00:20:57.179 --> 00:21:06.390 So here the spectral ratio are computed between earthquake that all belong 00:21:06.390 --> 00:21:11.780 within the interseismic period. So if we take one earthquake in 00:21:11.780 --> 00:21:15.409 the interseismic period, the other one, we compute the spectral ratio, 00:21:15.409 --> 00:21:21.910 and we make the average, we obtain a spectral ratio, which is – which is 1. 00:21:21.910 --> 00:21:28.520 Then here we computed the average spectral ratio of earthquake that occur 00:21:28.520 --> 00:21:33.790 to the preseismic period with respect to the longer-term interseismic period. 00:21:33.790 --> 00:21:38.450 And what we can see is that, at high frequency, the foreshocks – 00:21:38.450 --> 00:21:44.510 the preseismic earthquakes – seem to be depleted in high frequency. 00:21:44.510 --> 00:21:49.740 And this is the spectral ratio between earthquake that are aftershocks, 00:21:49.740 --> 00:21:54.140 that are belonging to the postseismic period 00:21:54.140 --> 00:21:58.300 with respect to the interseismic period here. 00:21:58.300 --> 00:22:04.280 So it seems that there is a difference between the 00:22:04.280 --> 00:22:10.059 regular background seismicity and the seismicity that occurs 00:22:10.060 --> 00:22:14.640 just before the earthquake or after the earthquake. 00:22:14.640 --> 00:22:20.039 So, in order to do those spectral ratio, we didn’t have a lot of earthquakes. 00:22:20.039 --> 00:22:24.660 Because we need a high magnitude, and to do these spectral ratio, 00:22:24.660 --> 00:22:30.330 we have to actually make sure that we are comparing the source and that 00:22:30.330 --> 00:22:35.820 we are not measuring the distance – the difference in magnitude, so we 00:22:35.820 --> 00:22:39.120 had to take earthquakes that are very close in magnitude, 00:22:39.120 --> 00:22:44.740 very close in distance. So we just had a very small subset of our data set. 00:22:44.740 --> 00:22:52.710 So our idea was to try to use GMPEs – ground motion prediction equations – 00:22:52.710 --> 00:22:58.630 as a backbone to look at the earthquake frequency content. 00:22:58.630 --> 00:23:03.890 So the idea was the following. For each earthquake, we compute 00:23:03.890 --> 00:23:09.590 the residual of this earthquake at each station with respect to 00:23:09.590 --> 00:23:15.279 a GMPE model for the area that has been tested against our data set. 00:23:15.279 --> 00:23:21.170 So, for each earthquake, we will have this distribution of residual, 00:23:21.170 --> 00:23:27.880 and so we can compute, for each residual, the intra-event residual, 00:23:27.880 --> 00:23:31.929 and we can also compute the inter-event residual, which is 00:23:31.929 --> 00:23:39.740 the average residual for one given earthquake with respect to the model. 00:23:39.740 --> 00:23:46.559 So the idea was to try to – and this way, using this inter-event residual, 00:23:46.559 --> 00:23:53.159 you can have a sense of how one given earthquake changes 00:23:53.159 --> 00:23:56.460 with respect to the model. And what we obtained 00:23:56.460 --> 00:24:02.740 was the following. Because then we looked at the 00:24:02.740 --> 00:24:10.320 temporal variability of these residuals, and we obtained the following. 00:24:10.320 --> 00:24:16.130 And so here are the earthquakes within the interseismic period. 00:24:16.130 --> 00:24:19.530 The first eight months preseismic period. 00:24:19.530 --> 00:24:24.279 The 15 days preseismic period. And the postseismic period. 00:24:24.279 --> 00:24:30.169 And this is at frequencies of 0.75 hertz. And you can see that, 00:24:30.169 --> 00:24:32.970 at these frequency, we don’t see a major change. 00:24:32.970 --> 00:24:37.780 All the residuals that we have computed are reddish, right? 00:24:37.780 --> 00:24:44.100 But if you look at higher frequencies – so, like, PGA or 10 hertz, 00:24:44.110 --> 00:24:48.590 what we see is that here, during the interseismic period, it’s reddish. 00:24:48.590 --> 00:24:54.960 While, starting in the preseismic period, we have, like, bluish residuals. 00:24:54.960 --> 00:24:57.820 And it’s the same here for the preseismic – 00:24:57.820 --> 00:25:02.140 the short-term preseismic – and here for the postseismic. 00:25:02.140 --> 00:25:06.840 So this is really puzzling, actually. When we started to look at that, 00:25:06.840 --> 00:25:10.750 we expected to see a change in the earthquake frequency content 00:25:10.750 --> 00:25:14.779 before and after the main shock. Because we thought maybe you have, 00:25:14.779 --> 00:25:19.740 like, a large earthquake. It’s going to change the friction of the interface. 00:25:19.740 --> 00:25:23.049 And then maybe you will have, like, a very big change. 00:25:23.049 --> 00:25:28.440 You don’t expect the aftershock to work the same way as foreshock, 00:25:28.440 --> 00:25:33.149 most probably. And we were really surprised to see that this change 00:25:33.149 --> 00:25:41.110 actually occurs before the main shock. And it seems to occur at the same time 00:25:41.110 --> 00:25:49.899 as this slow trench and that we have identified with the GPS otherwise. 00:25:49.899 --> 00:25:57.679 So, interestingly, we have done a similar study in Japan before Tohoku, 00:25:57.679 --> 00:26:03.679 and we also see this kind of effect before Tohoku earthquake. 00:26:03.679 --> 00:26:08.880 So it’s not completely straightforward to interpret, but it seems that those 00:26:08.880 --> 00:26:12.310 earthquake, if there is a drop in high frequency, probably the 00:26:12.310 --> 00:26:22.419 foreshocks are – have a lower stress drop and maybe – so they have, like, 00:26:22.419 --> 00:26:27.809 lower slip, or they are rupturing a larger area for the same magnitude. 00:26:27.809 --> 00:26:33.559 So here is a tentative interpretation. So if we mention that the megathrust 00:26:33.559 --> 00:26:41.590 is an interfingering of seismic and aseismic areas, then you have slow slip 00:26:41.590 --> 00:26:48.020 that is occurring on the megathrust on a steady or unsteady manner. 00:26:48.020 --> 00:26:52.890 And, as a response to this slow slip, you have background seismicity that 00:26:52.890 --> 00:26:58.860 is triggered. So these are the red dots. Then – sorry. 00:26:58.860 --> 00:27:03.400 Then, eight months before the megathrust, you start to have 00:27:03.400 --> 00:27:08.620 an increase of this slow slip. And, as a response, you have an 00:27:08.620 --> 00:27:16.880 increase of the foreshock seismicity. And maybe you have also this 00:27:16.880 --> 00:27:22.870 lower stress drop in the – of the foreshock seismicity. 00:27:22.870 --> 00:27:27.400 So maybe the seismic rupture are starting to expand. 00:27:27.400 --> 00:27:34.620 And then, two weeks before the main shock, you have the 00:27:34.630 --> 00:27:41.860 magnitude 6.7 foreshock. And, after that, it’s going to trigger 00:27:41.860 --> 00:27:47.720 aftershocks of the foreshock itself, so you have a cascade – an aftershock 00:27:47.720 --> 00:27:52.830 cascade that is superimposed on the pre-existing slow slip. 00:27:52.830 --> 00:27:59.049 And eventually, the main shock occurs and ruptures the whole area. 00:27:59.049 --> 00:28:07.230 So now, I would like to look at a larger time and space scale and 00:28:07.230 --> 00:28:12.480 maybe ask the question of, what is the – what is driving the 00:28:12.480 --> 00:28:18.890 plate interface destabilization and where the force come from. 00:28:18.890 --> 00:28:22.019 So this is, again, a paper by Michel Bouchon. 00:28:22.019 --> 00:28:30.929 And so this is – so he showed that, before Tohoku and Iquique earthquakes, 00:28:30.929 --> 00:28:36.669 there was a foreshock seismicity that was occurring on the megathrust, 00:28:36.669 --> 00:28:42.480 but there was also a significant amount of foreshocks that were 00:28:42.480 --> 00:28:47.469 occurring at intermediate-depth – at intermediate depth. 00:28:47.469 --> 00:28:55.419 So 80 to 120 kilometers. And interestingly, those earthquakes 00:28:55.419 --> 00:29:02.330 that are occurring at depth are occurring about at the same time 00:29:02.330 --> 00:29:05.779 as earthquakes that are occurring on the megathrust itself. 00:29:05.780 --> 00:29:11.080 So there are doublets like this. So here it’s the time before the 00:29:11.080 --> 00:29:15.380 earthquake. Here it’s the normalized cumulative seismic moment. 00:29:15.380 --> 00:29:21.320 And you see that you have a deep earthquake that is associated 00:29:21.330 --> 00:29:25.600 with shallow earthquakes. Then a period of quiescence. 00:29:25.600 --> 00:29:28.820 Then you have, again, a deep earthquake that is associated 00:29:28.820 --> 00:29:32.150 with shallow earthquakes. Then, again, the quiescence. 00:29:32.150 --> 00:29:37.170 And then, again, this kind of doublet before the megathrust. 00:29:37.170 --> 00:29:40.770 So there seem to be some kind of relationship between this deep 00:29:40.770 --> 00:29:45.140 seismicity and what happens on the megathrust itself. 00:29:45.140 --> 00:29:53.140 So we tried to look in further details to this potential interaction between 00:29:53.140 --> 00:29:59.029 deep and shallow seismicity. And so, North Chile is a good area for this. 00:29:59.029 --> 00:30:01.409 So here you have a context in further detail. 00:30:01.409 --> 00:30:06.950 So here, it’s the Iquique earthquake. And what is remarkable is that 00:30:06.950 --> 00:30:11.830 this Iquique earthquake occurred at the same latitude as an earthquake 00:30:11.830 --> 00:30:17.299 that occurred in 2005, and it was an intraslab earthquake. 00:30:17.299 --> 00:30:22.830 And this intraslab earthquake was a magnitude 7.8, and it ruptured – 00:30:22.830 --> 00:30:27.440 so here you have a cross-section. So here you have six months 00:30:27.440 --> 00:30:32.390 of seismicity after Iquique. Six months of seismicity after 00:30:32.390 --> 00:30:36.250 this intraslab earthquake, which is called Tarapaca earthquake. 00:30:36.250 --> 00:30:44.340 And this earthquake ruptured the slab on a sub-horizontal plane. All right. 00:30:44.340 --> 00:30:50.400 So we were wondering, okay, so if there actually are interactions 00:30:50.400 --> 00:30:52.910 between deep and shallow seismicity, 00:30:52.910 --> 00:30:58.780 probably this is the right place to look at those kind of interactions. 00:30:58.780 --> 00:31:03.210 So we looked again at GPS stations here. 00:31:03.210 --> 00:31:07.059 So, again, here you have, like, the east displacement as a function 00:31:07.059 --> 00:31:12.100 of time, but the time – the time span is larger. It starts in 2000 here. 00:31:12.100 --> 00:31:14.820 We have here the Tarapaca slab-pull earthquake. 00:31:14.820 --> 00:31:17.580 We have here the Iquique earthquake. 00:31:17.580 --> 00:31:23.280 And so you see the postseismic. And here you see the 00:31:23.289 --> 00:31:28.110 interseismic loading. And what you can see is that 00:31:28.110 --> 00:31:31.460 the interseismic loading seems to be changing a little bit. 00:31:31.460 --> 00:31:34.799 So, when you don’t detrend it, it’s not very clear. 00:31:34.799 --> 00:31:37.630 But if you detrend it, it’s quite clear. 00:31:37.630 --> 00:31:44.750 And we – you can compute the difference in the loading for this GPS 00:31:44.750 --> 00:31:49.490 time series, and it’s 4 millimeters of difference, which is quite big. 00:31:49.490 --> 00:31:53.100 So it seems that, before Tarapaca earthquake, we had 00:31:53.100 --> 00:31:57.610 something which was kind of linear. And after that, we also have something 00:31:57.610 --> 00:32:03.160 which has – which is kind of linear. And so we don’t see any, like, 00:32:03.160 --> 00:32:13.300 postseismic tail or very clear – very clear postseismic 00:32:13.309 --> 00:32:19.870 viscoelastic relaxation in those – in those GPS time series. 00:32:19.870 --> 00:32:25.679 So we can wonder what the – what the mechanism is. 00:32:25.679 --> 00:32:31.900 So we could compute this on several GPS stations. 00:32:31.900 --> 00:32:38.279 So unfortunately, Chile doesn’t have a very, very stable GPS array. 00:32:38.279 --> 00:32:43.950 So now it’s been much improved, but back in – at the beginning of 2000, 00:32:43.950 --> 00:32:49.019 it was not the case. So to compute the movement 00:32:49.019 --> 00:32:53.529 before and after 2005 earthquake, we needed to have, like, 00:32:53.529 --> 00:32:57.200 good station at the time, so there are not a lot. 00:32:57.200 --> 00:33:06.700 And here we have the displacement before and after those – 00:33:06.700 --> 00:33:09.850 before and after Tarapaca earthquake. And what we see is that there is 00:33:09.850 --> 00:33:17.580 a decrease in velocity for all the stations that are here around Tarapaca 00:33:17.580 --> 00:33:21.600 and Iquique earthquake, while here, this is no significant difference. 00:33:21.600 --> 00:33:25.380 And here is there is no significant difference, either. 00:33:25.380 --> 00:33:34.880 So interestingly, if we look at the seismicity in this area, 00:33:34.880 --> 00:33:38.820 we see that this seismicity is increasing with time. 00:33:38.820 --> 00:33:42.260 So, of course, we took care of the completeness, magnitude, 00:33:42.260 --> 00:33:47.059 effect, et cetera. And so this goes in the same direction. 00:33:47.059 --> 00:33:52.860 It means that, if there is a decrease of velocity of coast – of a coastal station, 00:33:52.860 --> 00:34:00.230 an increase of background seismicity on the megathrust, it means that 00:34:00.230 --> 00:34:04.600 you have a decrease of locking. And this decrease of locking 00:34:04.600 --> 00:34:08.320 is associated with an increase of slow slip. 00:34:08.330 --> 00:34:15.160 And this increase of slow slip is going to trigger background seismicity. 00:34:15.160 --> 00:34:21.740 So those observations are compatible with a change in coupling and 00:34:21.740 --> 00:34:26.780 a reduction in coupling, probably, after Tarapaca earthquake. 00:34:28.560 --> 00:34:37.430 So now we tried to identify pairs of deep and shallow earthquakes, 00:34:37.430 --> 00:34:41.480 as Michel Bouchon has done, but over a longer time period. 00:34:41.480 --> 00:34:47.630 And what we see is that we – so, after Tarapaca earthquake, we saw 00:34:47.630 --> 00:34:53.520 some pairs, like 16. And before 2005 earthquake, we didn’t see any pairs. 00:34:53.520 --> 00:34:58.440 And finally, you can see here the shallow seismicity and 00:34:58.450 --> 00:35:03.570 here the deep seismicity. So this is a sort of – it’s a periodogram. 00:35:03.570 --> 00:35:10.350 So here you have the seismicity rate averaged over different time periods. 00:35:10.350 --> 00:35:15.550 So, at the bottom of the graphs, it’s the yearly seismicity rate. 00:35:15.550 --> 00:35:19.850 And at the top of the graphs, it’s the daily seismicity rate. 00:35:19.850 --> 00:35:24.220 All right. Here you have the time. So this is the Iquique earthquake. 00:35:24.220 --> 00:35:26.210 This is Tarapaca earthquake. 00:35:26.210 --> 00:35:32.910 And the dashed lines represent the other megathrust earthquakes in the area. 00:35:32.910 --> 00:35:37.040 And what is interesting to see is that – so this is the background seismicity, 00:35:37.040 --> 00:35:43.770 which means that all the aftershock seismicity has been removed. 00:35:43.770 --> 00:35:47.370 So we just look at – yeah, the background seismicity, 00:35:47.370 --> 00:35:53.000 which is representative of the loading, probably, of the subduction. 00:35:53.000 --> 00:36:00.010 So this is the shallow – so what we see is that, before Iquique earthquake, 00:36:00.010 --> 00:36:03.740 we have an increased seismicity. We knew this already. 00:36:03.740 --> 00:36:09.930 And it’s associated with deep seismicity as well. We knew this as well. 00:36:09.930 --> 00:36:12.720 And what is interesting is that, after Iquique earthquake, 00:36:12.720 --> 00:36:16.130 there is no seismicity. And there is no deep seismicity. 00:36:16.130 --> 00:36:20.330 So probably we can imagine that, if you have the megathrust, 00:36:20.330 --> 00:36:25.800 and you have – so the driving force – the major driving force, which is 00:36:25.800 --> 00:36:32.470 the slab pull, so you have slab pull. As a response to this slab pull, you have 00:36:32.470 --> 00:36:38.440 deep seismicity, okay, and this deep seismicity somehow is connected with 00:36:38.440 --> 00:36:42.470 what happens on the megathrust. And then, when you have the main 00:36:42.470 --> 00:36:47.560 shock and the megathrust earthquake that occurs, then you have, like, 00:36:47.560 --> 00:36:51.270 some kind of clamping in this area. 00:36:51.270 --> 00:36:56.770 So this might explain why you have a lack of deep seismicity here. 00:36:56.770 --> 00:37:05.940 And interestingly, you can see also this kind of behavior associated with the – 00:37:05.940 --> 00:37:13.280 sorry – 1995 Antofagasta earthquake, where you have seismicity before 00:37:13.290 --> 00:37:17.450 at depths and on the shallow parts. And, after Antofagasta earthquake, 00:37:17.450 --> 00:37:23.400 there was no – not a lot of seismicity on the megathrust. 00:37:23.400 --> 00:37:29.980 And interestingly, here we are showing the seismicity in the area of Iquique. 00:37:29.980 --> 00:37:33.900 So, in this area, while the Antofagasta earthquake was here – 00:37:33.901 --> 00:37:37.130 so it was, like, 500 kilometers away. 00:37:37.130 --> 00:37:43.520 So there seem to be – some large-scale interactions seem to exist. 00:37:43.520 --> 00:37:49.260 So it seems that the occurrence of this Antofagasta earthquake generated 00:37:49.260 --> 00:37:55.240 a change in seismicity in this area. So it seems that you have, like, 00:37:55.240 --> 00:37:59.120 large-scale interactions, and the subduction zone 00:37:59.120 --> 00:38:04.880 probably generates large-scale interactions. 00:38:07.000 --> 00:38:12.480 So if we summarize what we have observed – so before Iquique – 00:38:12.480 --> 00:38:18.280 nine years before Iquique, there was this 2005 slab pull earthquake. 00:38:18.280 --> 00:38:24.330 And this earthquake triggers a decrease of eastward GPS velocities. 00:38:24.330 --> 00:38:28.650 An increase of deep and shallow seismicity, right? 00:38:28.650 --> 00:38:34.380 And those observations are compatible with a decoupling of the interface 00:38:34.380 --> 00:38:41.990 as a response to slab tearing, maybe. And eight months before the 00:38:41.990 --> 00:38:48.070 earthquake, we have, again, a decrease of coastal velocities, 00:38:48.070 --> 00:38:55.060 an increase of seismicity – yeah. And we also see a decrease of 00:38:55.060 --> 00:39:01.650 high-frequency radiations. So we have a magnitude 6.5 00:39:01.650 --> 00:39:06.670 slow-slip event that is mostly aseismic. 00:39:06.670 --> 00:39:12.700 And that is associated with a change in earthquake sources. 00:39:13.630 --> 00:39:18.700 And finally, 15 days before the earthquake, we have a big foreshock, 00:39:18.700 --> 00:39:24.830 which is 6.7. And this big foreshock triggers an abrupt increase of seismicity 00:39:24.830 --> 00:39:29.550 activity and a strong deformation signal that is – that corresponds 00:39:29.550 --> 00:39:38.740 to a magnitude 7 slow slip that is maybe 35% aseismic. 00:39:38.740 --> 00:39:43.370 So now, I would like to take some distance and investigate 00:39:43.370 --> 00:39:48.460 potential large-scale relationship between earthquake within a sequence. 00:39:48.460 --> 00:39:55.110 So we already saw that there are potential relationships between the 00:39:55.110 --> 00:39:57.680 Antofagasta earthquake and the Iquique earthquake, 00:39:57.680 --> 00:40:01.880 or what happens in the area of Iquique earthquake. 00:40:01.880 --> 00:40:08.040 Now I’m going to show some intriguing observations that have 00:40:08.050 --> 00:40:17.150 been done before Illapel earthquake. So this is the – so here, 00:40:17.150 --> 00:40:22.660 the Illapel earthquake occurred in central Chile. So it’s more than 00:40:22.660 --> 00:40:28.690 1,000 kilometers from Iquique. So it’s really far away – very far away. 00:40:28.690 --> 00:40:34.890 And what we can see is that, if we take this box here, and we take the deep 00:40:34.890 --> 00:40:45.190 seismicity – so the seismicity that is below 80 kilometers’ depth on this box, 00:40:45.190 --> 00:40:50.090 we obtain the following terms. So this is the cumulative number 00:40:50.090 --> 00:40:56.180 of earthquakes. And this is the normalized moment in light blue. 00:40:56.180 --> 00:41:02.200 So – I’m sorry. It’s probably too light. So here is the data of Illapel earthquake. 00:41:02.200 --> 00:41:06.960 And here it’s the data of Iquique earthquake that occurred more 00:41:06.960 --> 00:41:10.910 than 1,000 kilometers away. Now, if we take the same box, 00:41:10.910 --> 00:41:18.530 but we restrict the search in depth between 80 and 120 kilometers’ depth, 00:41:18.530 --> 00:41:25.390 this is what we obtain. So we see that the cumulative number of earthquake 00:41:25.390 --> 00:41:33.340 increases at the date of Iquique. And same for the cumulative moment. 00:41:33.340 --> 00:41:42.960 Now, if we restrict the box again by focusing on the epicenter, we see a 00:41:42.960 --> 00:41:49.880 change that is even more abrupt at the – exactly at the date of Iquique. 00:41:49.880 --> 00:41:53.460 And now we restrict the box again, 00:41:53.460 --> 00:41:57.310 and we have a change that is even more abrupt. 00:41:57.310 --> 00:42:03.210 So this is really puzzling. Because we have this earthquake, 00:42:03.210 --> 00:42:09.940 Iquique, that occurs more than 1,000 kilometers away – and, again, 00:42:09.940 --> 00:42:14.950 an earthquake that occurs on the same subduction zone that is really far-field – 00:42:14.950 --> 00:42:18.160 generates a change in the background seismicity. 00:42:18.160 --> 00:42:25.270 And, eventually, maybe, leads to the occurrence of Illapel earthquake. 00:42:25.270 --> 00:42:28.530 So the explanation is not straightforward at all, 00:42:28.530 --> 00:42:32.650 so I’m happy to discuss this with you. 00:42:32.650 --> 00:42:36.130 We may think about large-scale interactions. 00:42:36.130 --> 00:42:41.500 We may think also about dynamic triggering of seismicity. 00:42:42.600 --> 00:42:49.540 There might be, like, several potential explanations to explore. 00:42:49.540 --> 00:42:55.130 But – so, in the case of Illapel, it’s probably a particular case. 00:42:55.130 --> 00:43:00.600 Because Illapel occurs at the boundary between the flat slab and the 00:43:00.600 --> 00:43:04.560 normally dipping slab. So there is a tear in the subduction in this area. 00:43:04.560 --> 00:43:11.680 So it might be a particular setting that generates this particular effect. 00:43:11.680 --> 00:43:19.040 But, still, it was interesting also to see that, in this area, we had a change 00:43:19.040 --> 00:43:24.390 of behavior in the background seismicity generated by the Antofagasta 00:43:24.390 --> 00:43:29.140 earthquake that occurred here. So there are some relationship. 00:43:29.140 --> 00:43:35.440 And Coulomb – I mean, just static stress change cannot explain this. 00:43:35.440 --> 00:43:40.920 All right. So, here I’m going – now show a paper by Sergio Ruiz 00:43:40.920 --> 00:43:43.260 and collaborators. I’m not a co-author of this, 00:43:43.260 --> 00:43:47.960 but I thought it would be interesting for the large-scale picture. 00:43:47.960 --> 00:43:55.400 So these are also observations that have been done during 00:43:55.400 --> 00:43:57.800 Illapel earthquake and before Illapel earthquake. 00:43:57.800 --> 00:44:03.570 So Illapel is here, and we are looking further south. And here – the arrows 00:44:03.570 --> 00:44:09.260 here are at the northern termination of the Maule earthquake. 00:44:09.260 --> 00:44:15.940 All right. So what we see here, those arrows – this GPS station is 00:44:15.940 --> 00:44:20.660 Rocas de Santo Domingo, so you see here, it’s the 2010 Maule earthquake. 00:44:20.660 --> 00:44:26.020 And after that, you have postseismic deformation. 00:44:26.020 --> 00:44:32.500 Here, in Vallenar, you have also postseismic deformation. 00:44:32.500 --> 00:44:40.090 And, for GPS stations which are further north, we can see that there 00:44:40.090 --> 00:44:48.290 is a change in the velocity before and after Maule earthquake. 00:44:48.290 --> 00:44:56.000 So here, what is plot is just the change in velocity before and after Maule. 00:44:56.000 --> 00:45:00.160 All right, so you remove the interseismic movement before Maule, 00:45:00.160 --> 00:45:05.350 and you just compute the change. And what is quite clear is that there 00:45:05.350 --> 00:45:12.450 is a sort of a vortex here – sort of vortex deformation with a significant 00:45:12.450 --> 00:45:17.660 deformation quite far away from Maule itself, right? 00:45:17.660 --> 00:45:22.800 So interestingly, this kind of pattern – so here, it’s the same 00:45:22.800 --> 00:45:27.460 kind of observation. So these are the observations. 00:45:27.460 --> 00:45:34.390 Here it’s the area where Illapel occurred. And, at the first order, those 00:45:34.390 --> 00:45:41.450 observations can be well explained by a viscoelastic relaxation model. 00:45:41.450 --> 00:45:47.710 But, when you start to look at details in this area where Illapel earthquake 00:45:47.710 --> 00:45:52.890 occurred, you have – sorry – you have this pattern. 00:45:52.890 --> 00:45:58.850 So this is a zoom on this area exactly. And here you have the GPS in 00:45:58.850 --> 00:46:02.980 orange and the model in blue. And you can see that there is a 00:46:02.980 --> 00:46:10.510 systematic offset and a systematic residual between 00:46:10.510 --> 00:46:15.960 the model and the GPS data. So it means that the model 00:46:15.960 --> 00:46:22.320 cannot explain the fact that it’s going further east like this. 00:46:22.320 --> 00:46:28.840 And so this observation has been made by another group, which is the group 00:46:28.850 --> 00:46:35.010 of Daniel Melnick and collaborators. So this is a paper that they have 00:46:35.010 --> 00:46:38.420 published on the same kind of observation. 00:46:38.420 --> 00:46:45.480 And so this thing that cannot be explained by the viscoelastic model, 00:46:45.480 --> 00:46:49.840 they interpret it as a change in interseismic coupling. 00:46:50.740 --> 00:46:56.280 So here it’s the interseismic coupling map before the Maule earthquake. 00:46:56.280 --> 00:47:01.650 And here, it’s the interseismic coupling map after the Maule earthquake. 00:47:01.650 --> 00:47:06.100 And what you see is that – well, at the first order, it looks the same. 00:47:06.100 --> 00:47:15.310 But, if you compute the locking as a function of the latitude, you can 00:47:15.310 --> 00:47:22.400 see that, after the – sorry – after the 2010 earthquake, 00:47:22.400 --> 00:47:29.000 there was an increase in coupling in the area of Maule earthquake. 00:47:29.000 --> 00:47:37.720 So it seems that, in addition to the – to the viscoelastic relaxation that 00:47:37.720 --> 00:47:43.420 we can explain, there might be also some changes that are large-scale 00:47:43.431 --> 00:47:48.810 and that are generating a large-scale change in coupling. 00:47:48.810 --> 00:47:53.530 And, actually, this has been seen in Japan also. 00:47:53.530 --> 00:48:00.400 And, instead of interpreting it as a change in coupling, 00:48:00.400 --> 00:48:05.880 Heki and Mitsui have interpreted it as a change of plate velocity. 00:48:05.880 --> 00:48:09.860 Because, if you increase the coupling in one area, 00:48:09.860 --> 00:48:14.000 you will have, like, a change in the upper-plate velocity. If you increase 00:48:14.000 --> 00:48:19.010 the plate velocity, you will also have a change in the upper-plate velocity. 00:48:19.010 --> 00:48:22.980 So if you consider this, you might have – 00:48:22.980 --> 00:48:27.610 okay, so this is before the earthquake. This is after the earthquake. 00:48:27.610 --> 00:48:31.060 So, before the earthquake, you have the slab pull force. 00:48:31.060 --> 00:48:34.250 You have the ridge push force. 00:48:34.250 --> 00:48:40.240 To counterbalance these forces, you have the side resistance effect 00:48:40.240 --> 00:48:44.560 and also the friction and the megathrust that are – and all this is 00:48:44.560 --> 00:48:51.940 resulting in one given plate velocity. And what they say is that, after a large 00:48:51.940 --> 00:48:57.160 earthquake, you might have a different friction of the megathrust – 00:48:57.160 --> 00:49:01.980 a reduced friction on the megathrust. And then, as a result, you can have 00:49:01.980 --> 00:49:06.960 an increased velocity of the down-going plate. 00:49:06.960 --> 00:49:11.761 So I don’t know if this is really the explanation, but you can imagine that, 00:49:11.761 --> 00:49:17.550 if you have one earthquake in this area – a large earthquake that is moving the 00:49:17.550 --> 00:49:23.430 subduction like this, then you have, like, a bend in the slab itself. 00:49:23.430 --> 00:49:29.960 And this bend can propagate very far away from the earthquake itself. 00:49:29.960 --> 00:49:35.090 And this might be something to take into account in the modeling 00:49:35.090 --> 00:49:40.160 in order to explain this kind of large-scale deformation. 00:49:40.160 --> 00:49:48.080 So, yeah, and that’s about it. So I’ve been showing, like, 00:49:48.090 --> 00:49:53.450 some links between interseismic loading and coseismic ruptures. 00:49:53.450 --> 00:49:58.690 I have shown that we could see precursory deformation before 00:49:58.690 --> 00:50:03.290 the Iquique earthquake, and this is something that is probably really 00:50:03.290 --> 00:50:07.660 interesting to look at, and it has been shown in Japan as well. 00:50:07.660 --> 00:50:13.490 We have shown that maybe there is some weird relationship 00:50:13.490 --> 00:50:16.530 between deep seismicity and shallow seismicity that is 00:50:16.530 --> 00:50:22.620 maybe linked to the short-term dynamics of the subduction zone. 00:50:22.620 --> 00:50:30.280 And also, we have observations of a link between earthquakes 00:50:30.280 --> 00:50:33.530 that are within the segments on the same subduction zone – 00:50:33.530 --> 00:50:38.500 large-scale interactions. And I think that all these questions 00:50:38.500 --> 00:50:45.530 are maybe – probably needs to be further explored and 00:50:45.530 --> 00:50:49.330 raise a lot of questions. So I think, at the end of this talk, 00:50:49.330 --> 00:50:52.520 I have more questions than at the beginning. [laughs] 00:50:52.520 --> 00:50:55.500 So that’s all. Thank you for your attention. 00:50:55.500 --> 00:51:01.300 [Applause] 00:51:03.460 --> 00:51:06.440 - … any questions for the speaker? 00:51:09.040 --> 00:51:17.640 [Silence] 00:51:17.640 --> 00:51:21.180 - I had a question about the ground motions that you were showing. 00:51:21.190 --> 00:51:25.570 So you’re arguing that the earthquakes in the main shock 00:51:25.570 --> 00:51:28.390 sequence have a lower stress drop. 00:51:28.390 --> 00:51:33.570 And I’d like you to speculate on why you think that is. 00:51:33.570 --> 00:51:36.820 But a second question – I noticed the locations in general – 00:51:36.820 --> 00:51:41.080 the earthquakes in that sequence do look to be a little bit deeper, 00:51:41.080 --> 00:51:45.480 and could this really be just a function of the depth of these earthquakes? 00:51:45.480 --> 00:51:47.630 Are you certain they’re all interface earthquakes 00:51:47.630 --> 00:51:50.420 as opposed to slab earthquakes? 00:51:50.860 --> 00:51:56.840 - Yeah. So we think they are interface earthquakes because 00:51:56.840 --> 00:52:03.540 these are earthquakes that have a thrust focal mechanism, 00:52:03.540 --> 00:52:08.160 but we are not 100% sure of it, of course. 00:52:08.160 --> 00:52:12.270 There is a definitely a tradeoff between the earthquake frequency 00:52:12.270 --> 00:52:16.120 content and the stress drop and the depth on subduction zones. 00:52:16.120 --> 00:52:19.180 This has been shown by several authors. 00:52:20.340 --> 00:52:25.840 And I agree that the method that we use is probably a bit unconventional, 00:52:25.850 --> 00:52:32.080 but [laughs] – well, I mean, it’s been shown that there is 00:52:32.080 --> 00:52:38.300 a very good tradeoff between even residuals that are computed 00:52:38.300 --> 00:52:43.330 with respect to a GMPE model and stress drop. 00:52:43.330 --> 00:52:49.360 So it’s been shown that it’s working really well. 00:52:49.360 --> 00:52:55.720 So I think that interpreting these residual as a change 00:52:55.720 --> 00:53:00.960 in stress drop is probably reasonable. 00:53:00.960 --> 00:53:08.220 Then the reason for this is probably open to discussion, so … [laughs] 00:53:08.220 --> 00:53:11.480 I don’t know if I fully answered your question. 00:53:12.860 --> 00:53:15.580 - Are there any more questions? 00:53:16.620 --> 00:53:19.620 - Hi. That was a really fascinating talk. 00:53:19.620 --> 00:53:26.200 You described transient processes operating over a variety of time scales. 00:53:26.200 --> 00:53:30.500 And – including over, you know, months and years. 00:53:30.500 --> 00:53:33.500 And it starts to make one wonder whether it’s really possible to 00:53:33.500 --> 00:53:43.080 identify a true interseismic period. And if you can’t so definitively define that, 00:53:43.080 --> 00:53:48.860 then how do you, you know, determine what is anomalous and what is a rate 00:53:48.870 --> 00:53:52.160 change from one period to another? So I’m interested in your thoughts 00:53:52.160 --> 00:53:57.140 on that and how we might address that problem, especially given that, 00:53:57.140 --> 00:54:01.580 you know, thus far, we have GPS time series for a couple decades, 00:54:01.580 --> 00:54:04.359 but it may not be enough yet. 00:54:04.359 --> 00:54:07.420 - Yeah. So if I understand your question, you question is, 00:54:07.420 --> 00:54:11.540 how do you define interseismic? That’s it? Sorry. 00:54:11.540 --> 00:54:15.070 - Yeah. It might be more of a comment than a question, but I’m just curious of 00:54:15.070 --> 00:54:21.020 your thoughts on that general issue of, you know, once you start talking about, 00:54:21.020 --> 00:54:25.540 you know, interactions among earthquakes over several years, 00:54:25.540 --> 00:54:31.180 or 10 years, you know, how do you feel confident in defining, okay, this is a 00:54:31.180 --> 00:54:36.740 rate change versus the earlier behavior. - Yeah. I see. Yeah. 00:54:36.740 --> 00:54:42.620 Well, [laughs] that’s a broad question. I’m not sure I will be able to answer it, 00:54:42.620 --> 00:54:49.040 really, but so, yeah, defining the interseismic is probably very difficult 00:54:49.040 --> 00:54:54.370 because you use – so you need to use the interseismic far enough 00:54:54.370 --> 00:54:59.150 from the previous earthquake and far enough from the next earthquake. 00:54:59.150 --> 00:55:04.680 So all this is really relative, I would say. So you take one reference, and with 00:55:04.680 --> 00:55:08.180 respect to this reference, you look at potential changes, and then 00:55:08.180 --> 00:55:12.970 you interpret it as, you know, being part of the seismic cycle. 00:55:12.970 --> 00:55:19.710 But what is – what seems quite clear is that there are relationships between 00:55:19.710 --> 00:55:25.570 the average interseismic loading map that are done with GPS 00:55:25.570 --> 00:55:31.530 and long-term features. So it seems that those coupling 00:55:31.530 --> 00:55:37.070 asperities are areas that are going to generate very large earthquakes seem to 00:55:37.070 --> 00:55:43.000 be more or less stable over time – over several maybe million years. 00:55:43.000 --> 00:55:47.410 Because there is a relationship between the gravity and the 00:55:47.410 --> 00:55:50.890 interseismic coupling. There is a relationship between 00:55:50.890 --> 00:55:55.560 the topography or the basins or the tectonics and interseismic coupling. 00:55:55.560 --> 00:55:59.440 So the first order, I think the segmentation does not change 00:55:59.440 --> 00:56:03.940 completely, you know. So you – and, on top of this – 00:56:03.940 --> 00:56:08.230 on top of this background, I think you have tiny changes. 00:56:08.230 --> 00:56:11.290 But these tiny changes do not change completely 00:56:11.290 --> 00:56:16.310 the interseismic loading map. So if you take, for example, 00:56:16.310 --> 00:56:21.250 the figure of Melnick and collaborators, they look at 00:56:21.250 --> 00:56:24.430 the interseismic coupling before and after Maule. 00:56:24.430 --> 00:56:27.860 The pattern, at the first order, looks very much the same, right? 00:56:27.860 --> 00:56:36.210 But you have tiny changes that are associated with increase of locking, 00:56:36.210 --> 00:56:42.170 or decrease of locking, associated with an earthquake nearby. 00:56:42.170 --> 00:56:56.720 So I guess it’s – what I’m – what I think is pretty strong is this evidence that the 00:56:56.720 --> 00:57:02.740 segmentation of one given fault or the subduction zone is something that 00:57:02.740 --> 00:57:08.750 seems to be pretty stable in time. And then, on top of that, you have, 00:57:08.750 --> 00:57:11.560 like, trench and deformation that comes and complicates 00:57:11.560 --> 00:57:16.060 a bit the pattern that does not change completely the pattern. 00:57:18.160 --> 00:57:25.060 [Silence] 00:57:25.060 --> 00:57:28.680 - Could we look at your second slide from Bouchon where he showed 00:57:28.680 --> 00:57:34.400 the cumulative moment from 1900 to the present? 00:57:36.360 --> 00:57:45.960 [Silence] 00:57:46.300 --> 00:57:47.760 - Yes. 00:57:50.560 --> 00:57:52.120 This one, you mean? 00:57:52.120 --> 00:57:58.540 - There we are. Yeah. So do you think you have begun to 00:57:58.540 --> 00:58:06.010 understand that clustering – you know, these – the step – is it – are you 00:58:06.010 --> 00:58:13.180 beginning to – you didn’t talk about this slide any further in the talk. 00:58:13.180 --> 00:58:18.400 - Well, I talked about the fact that we have sequences that 00:58:18.410 --> 00:58:21.700 are rupturing one given subduction zone. 00:58:21.700 --> 00:58:28.020 And this slide shows that you have – I’m sorry – so this ongoing sequence 00:58:28.020 --> 00:58:33.590 in South America. And we had other examples of sequences. 00:58:33.590 --> 00:58:37.360 So here in the area of Sumatra, for example, and the very famous 00:58:37.360 --> 00:58:43.450 sequence is the Aleutian and Alaska sequence in the ’60s. 00:58:43.450 --> 00:58:48.490 So it’s interesting to see – well, looking at this slide, 00:58:48.490 --> 00:58:56.360 I think I have – I have the impression that you have sequences – you have 00:58:56.360 --> 00:59:00.851 one subduction zone that is starting to rupture, and eventually, maybe 00:59:00.851 --> 00:59:05.300 everything is going to rupture. And then you have, like, 00:59:05.300 --> 00:59:09.280 a period of quiescence. And maybe this subduction zone 00:59:09.280 --> 00:59:17.600 will reload, and then you will have another sequence of ruptures. 00:59:17.600 --> 00:59:23.820 And then, the question behind this is, what is the relationship between 00:59:23.820 --> 00:59:27.280 the individual earthquakes within one given sequence? 00:59:27.280 --> 00:59:33.890 Is it something that is just, like, large-scale relationship like 00:59:33.890 --> 00:59:42.520 dynamic triggering of earthquakes? But, if so, why would they be located 00:59:42.520 --> 00:59:46.910 within one given subduction zone? Maybe there are large – even 00:59:46.910 --> 00:59:50.760 larger-scale interactions between Sumatra and Chile 00:59:50.760 --> 00:59:52.940 and Japan earthquake. 00:59:52.980 --> 00:59:56.960 Some people think that there are larger-scale interactions, definitely. 00:59:56.960 --> 01:00:02.760 But the question is also, why do you have those sequences of 01:00:02.760 --> 01:00:07.510 earthquakes that are occurring within one given subduction zone? 01:00:07.510 --> 01:00:14.440 And here I showed the example of Chile was 2010, 2014, and 2015. 01:00:14.440 --> 01:00:17.570 And it seems that there are some relationships. 01:00:17.570 --> 01:00:20.580 We see some changes in the background deformation, in the 01:00:20.580 --> 01:00:29.400 background seismicity, that, at least, I don’t completely understand. 01:00:29.400 --> 01:00:35.980 But those observations are – seem to show that, yeah, I mean, 01:00:35.990 --> 01:00:42.420 there is probably further to dig into our understanding of interactions 01:00:42.420 --> 01:00:45.930 between earthquakes. Because you cannot explain these interactions 01:00:45.930 --> 01:00:52.500 by simple static triggering. It’s just impossible. So we need to have a larger- 01:00:52.500 --> 01:00:58.700 scale mechanism. So the mantle can play a role. The slab can play a role. 01:00:58.700 --> 01:01:05.160 Maybe activation of fluids can play a role by dynamic triggering. 01:01:05.700 --> 01:01:11.480 And any suggestions are welcome. [laughs] 01:01:13.660 --> 01:01:16.660 [Silence] 01:01:16.660 --> 01:01:19.880 - I had a question to follow up on Walter’s. 01:01:19.880 --> 01:01:22.200 There was a large earthquake in Kamchatka. 01:01:22.200 --> 01:01:26.180 I think it was approximately a magnitude 9 in 1952. 01:01:26.180 --> 01:01:30.780 Is that taken into account on your graph there? 01:01:31.580 --> 01:01:36.780 Yes. I think it’s probably this one. 01:01:39.080 --> 01:01:45.720 [Silence] 01:01:45.720 --> 01:01:47.420 - Yeah. I … - I mean, it’s … 01:01:47.420 --> 01:01:52.220 - I would have thought it showed a little bit more cumulative moment than that – 01:01:52.220 --> 01:01:56.350 or, moment contribution than … - Yes. But, you know, I mean, 01:01:56.350 --> 01:02:04.100 this graph is very much – when you have, like, the 1960 Chile earthquake 01:02:04.100 --> 01:02:07.750 that is taking the whole moment, basically, or most of the moment, 01:02:07.750 --> 01:02:12.540 you know. So you have, like, those clusters where you have, like, 01:02:12.540 --> 01:02:18.980 Chile, Alaska – you see an acceleration. And you have – you have Chile 01:02:18.980 --> 01:02:22.730 and Alaska that are really about at the same time. 01:02:22.730 --> 01:02:24.540 - Sure. - And then you have this quiescence, 01:02:24.540 --> 01:02:30.880 and then it’s resuming again with Sumatra, Chile, and Japan. 01:02:30.880 --> 01:02:36.050 But, yeah, my point was more like, okay, what is the relationship 01:02:36.050 --> 01:02:38.270 between an individual earthquake within an sequence. 01:02:38.270 --> 01:02:41.960 I’m not addressing, really, the relationship – 01:02:41.960 --> 01:02:46.760 the larger-scale relationship and this clustering. 01:02:46.760 --> 01:02:51.460 This is probably even more – well, it’s a wider problem. 01:02:51.460 --> 01:02:53.520 - All right. Thanks. 01:02:55.700 --> 01:03:00.040 [Silence] 01:03:00.040 --> 01:03:01.360 - Thank you. 01:03:01.740 --> 01:03:04.040 Really interesting talk. 01:03:04.040 --> 01:03:08.840 I have a follow-up related to what Jessica was asking you. 01:03:11.060 --> 01:03:19.330 The – what you show that looked convincing to me was the change in rate 01:03:19.330 --> 01:03:26.960 from the GPS time series before the earthquake at a number of stations. 01:03:28.100 --> 01:03:31.580 And Jessica’s question was essentially, well, lots of stuff 01:03:31.580 --> 01:03:37.400 could be going on interseismically. And some coincidentally occur 01:03:37.400 --> 01:03:42.210 before the earthquake. 01:03:42.210 --> 01:03:46.550 Have you looked at all of the interseismic signals – 01:03:46.550 --> 01:03:53.630 all of the interseismic period from the continuous stations that exist, say, 01:03:53.630 --> 01:04:00.290 along the Chile subduction zone to see what kind of variability there is? 01:04:00.290 --> 01:04:04.180 - The variability in the interseismic loading, you mean, along the 01:04:04.180 --> 01:04:10.320 subduction zone associated with any particular earthquake or … 01:04:10.320 --> 01:04:15.840 - That’s in the interseismic period that isn’t obviously related to an earthquake. 01:04:15.840 --> 01:04:23.580 - Yes. Well, I mean, this is what I have shown a little bit at the end. 01:04:23.580 --> 01:04:28.220 Well, this is something that I would like to do to actually examine in 01:04:28.220 --> 01:04:33.350 a systematic way potential changes along the Chile subduction zone. 01:04:33.350 --> 01:04:40.730 At the moment, I don’t have – I’m not working very much 01:04:40.730 --> 01:04:43.360 on Chile at the moment. I’m more working on Japan. 01:04:43.360 --> 01:04:47.240 So we are doing this on Japan. The good thing with Japanese data 01:04:47.240 --> 01:04:52.260 is that there are – like, it’s just an incredible quality, so … [laughs] 01:04:52.260 --> 01:04:56.680 So we have done this exercise. And so we have obtained and 01:04:56.680 --> 01:05:02.390 reproduced results similar to what had been seen by our colleagues 01:05:02.390 --> 01:05:08.340 of Stanford, for example, [inaudible] and collaborators. 01:05:08.360 --> 01:05:14.740 So we have done an independent GPS processing, so it was interesting to see 01:05:14.740 --> 01:05:22.080 that we see the same – the same signal. I think it’s quite important to use GPS 01:05:22.080 --> 01:05:27.620 data that are not – that do not all come from the same processing center. 01:05:27.620 --> 01:05:36.980 And – because notably, in Japan, there are – sorry – reference frame 01:05:36.980 --> 01:05:41.860 issues because it’s a small island. And so – well, whatever. 01:05:41.860 --> 01:05:47.960 And what we see is that we find an acceleration like this before Tohoku 01:05:47.970 --> 01:05:53.090 but we also find an acceleration in the area of Bōsō, which is this 01:05:53.090 --> 01:06:00.100 peninsula which is close to Tokyo. So this is interesting. 01:06:00.100 --> 01:06:04.960 And a bit scary, I would say. So – and this acceleration that we see 01:06:04.960 --> 01:06:11.560 in the GPS is also seen in the background seismicity, actually. 01:06:12.829 --> 01:06:21.880 - Thank you. I have a – just a comment about the seismicity rate changes 01:06:21.890 --> 01:06:25.740 that you showed and the cumulative moment plots. 01:06:25.740 --> 01:06:34.840 In those – if I understand correctly, in each of those cases, the analyst has 01:06:34.840 --> 01:06:40.210 a number of arbitrary choices to make. The time interval over which 01:06:40.210 --> 01:06:48.580 these changes are occurring. The area that is – that is included. 01:06:48.580 --> 01:06:59.920 And I worry a little bit that that subjectivity can lead to 01:06:59.920 --> 01:07:03.340 misleading results. 01:07:04.240 --> 01:07:05.780 - All right. 01:07:05.780 --> 01:07:07.800 [laughter] 01:07:08.980 --> 01:07:13.540 Well, yeah, definitely. I mean, that’s what I showed. 01:07:13.540 --> 01:07:18.890 I mean, if you take one given window around – which is a bit large, 01:07:18.890 --> 01:07:21.870 around the epicenter, then you don’t see a very big change. 01:07:21.870 --> 01:07:27.410 And once you start to refine the window and make it closer and closer, then you 01:07:27.410 --> 01:07:35.190 see this change that is – that gets really sharp and that is getting really clear. 01:07:35.190 --> 01:07:40.710 So what is – I think, yeah, of course. I mean, probably it was – 01:07:40.710 --> 01:07:46.140 it was by chance. I think – so this is a paper – 01:07:46.140 --> 01:07:50.040 I mean, Michel Bouchon is the one who is leading this work, definitely, 01:07:50.040 --> 01:07:54.190 so – but I think he’s looking, you know, very much at the 01:07:54.190 --> 01:07:57.270 different catalogs and exploring a lot. 01:07:57.270 --> 01:08:02.400 And when he sees an interesting signal, he digs further. 01:08:02.400 --> 01:08:09.220 And, yeah. This is definitely why we have decided to make a systematic 01:08:09.220 --> 01:08:16.480 analysis in the North Chile area with, you know, statistical methods, et cetera, 01:08:16.480 --> 01:08:22.120 exploring, on a systematic basis, the new catalog and to look at, 01:08:22.120 --> 01:08:25.630 you know, probabilities and things like this. 01:08:25.630 --> 01:08:28.569 And we obtain results that are compatible with 01:08:28.569 --> 01:08:32.970 what Michel had found. Because he had been looking, 01:08:32.970 --> 01:08:37.750 you know, manually, in great detail at the catalog. 01:08:37.750 --> 01:08:43.960 And what we obtained was something that was very compatible and that was 01:08:43.960 --> 01:08:51.799 even larger in time and space scale. So probably it would be interesting 01:08:51.799 --> 01:08:56.730 to make a more systematic analysis in the area of Iquique as well. 01:08:56.730 --> 01:09:00.380 And more generally along the subduction zones. 01:09:00.380 --> 01:09:02.080 Yes, I agree with you. 01:09:02.080 --> 01:09:09.560 I mean, it’s really important to automatize this type of exploration. 01:09:11.700 --> 01:09:13.760 - Okay. Let’s thank our speaker again. 01:09:14.200 --> 01:09:16.600 [Applause]