WEBVTT Kind: captions Language: en-US 00:00:00.840 --> 00:00:03.560 [Silence] 00:00:04.740 --> 00:00:10.600 Okay, everybody. Let’s get started. Thank you all so much for coming to 00:00:10.600 --> 00:00:16.600 our joint Earthquake Science Center and Volcano Science Center seminar today. 00:00:16.600 --> 00:00:21.300 Today we have our very own Alicia Hotovec-Ellis giving a talk. 00:00:21.300 --> 00:00:28.870 And for announcements, next week, our regular ESC seminar will be 00:00:28.870 --> 00:00:37.710 by Arjun Kohli from Stanford. So today, Alicia is giving our talk. 00:00:37.710 --> 00:00:43.000 She got her bachelor’s and master’s degrees at Colorado School of Mines 00:00:43.000 --> 00:00:46.560 and her Ph.D. at the University of Washington. 00:00:46.560 --> 00:00:52.980 And she joined us here at the USGS as a Mendenhall postdoctoral fellow in 2016. 00:00:52.980 --> 00:00:55.240 And, take it away. 00:00:57.200 --> 00:00:59.480 - All right. Unmute myself. Awesome. 00:01:00.280 --> 00:01:03.799 All right. Well, good morning, everyone, and thanks for coming out. 00:01:03.799 --> 00:01:06.850 And hello, as well, to the folks watching online. 00:01:06.850 --> 00:01:09.640 I know several of the other volcano observatories were interested in 00:01:09.640 --> 00:01:13.260 listening in, which is actually part of the reason why I’m giving this talk for, 00:01:13.270 --> 00:01:16.759 I think, the third or fourth time, so that everybody that’s interested in listening 00:01:16.760 --> 00:01:22.260 to it can get a chance to do that since this will be recorded for posterity. 00:01:22.260 --> 00:01:27.760 I’m really excited to present to you all the – some of the exciting scientific 00:01:27.760 --> 00:01:33.480 results from this truly incredible and sort of generational kind of eruption, 00:01:33.480 --> 00:01:36.320 which I’m sure that you will be hearing about for years and years 00:01:36.329 --> 00:01:40.520 to come from many different other researchers. 00:01:40.520 --> 00:01:43.549 So last summer, many of us at all of the different volcano 00:01:43.549 --> 00:01:47.400 observatories were consumed by this eruption. 00:01:48.060 --> 00:01:52.060 But, in return, we were given a truly beautiful and complex data set, 00:01:52.060 --> 00:01:54.850 with which we now get to go back through and try to 00:01:54.850 --> 00:01:59.509 piece together and figure out the why of what happened. 00:01:59.509 --> 00:02:03.450 And this work, which I’ve titled Damage and Stress-Induced 00:02:03.450 --> 00:02:07.049 Seismic Velocity Changes During the Progressive Collapse of Kilauea’s 00:02:07.049 --> 00:02:12.260 Summit, is my attempt to figure out, really, just a piece of that puzzle. 00:02:12.260 --> 00:02:16.740 I’d like to take a moment before I jump in to acknowledge my co-authors 00:02:16.740 --> 00:02:22.390 who are seismologists and geodesists at the Hawaiian, Alaskan, Cascades, 00:02:22.390 --> 00:02:27.370 and California Volcano Observatories in sort of this small interdisciplinary 00:02:27.370 --> 00:02:31.030 but truly cross-observatory collaboration. 00:02:31.030 --> 00:02:33.880 So I figured I’d start off the talk with sort of a brief overview of the 00:02:33.880 --> 00:02:38.440 eruption as sort of a jumping-off point so that everybody is on sort of the 00:02:38.440 --> 00:02:41.860 same page, especially for people who only got bits and pieces of it 00:02:41.860 --> 00:02:45.240 through the news or who didn’t really experience it. 00:02:45.240 --> 00:02:48.020 But what I’d really like to emphasize is this really is just 00:02:48.020 --> 00:02:52.040 going to be a highlight reel. So if you’d like a full and more detailed 00:02:52.040 --> 00:02:56.490 overview of the events of last summer, I encourage you to go look up the 00:02:56.490 --> 00:03:01.160 video of our very own Kyle Anderson’s USGS public lecture 00:03:01.160 --> 00:03:07.200 as well as read Neal et al.’s 2019 overview paper in Science. 00:03:07.200 --> 00:03:10.340 All right. So first we have a very quick geography lesson. 00:03:10.350 --> 00:03:13.720 So here we have the Big Island of Hawaii. 00:03:13.720 --> 00:03:19.060 Kilauea Volcano is situated on the southeast coast of that island. 00:03:19.060 --> 00:03:23.140 Unlike volcanoes like Mount St. Helens where it’s basically just one cone, 00:03:23.140 --> 00:03:26.560 and that’s basically the entirety of the volcano, 00:03:26.560 --> 00:03:29.820 Hawaiian volcanoes are very broad and spread out. 00:03:29.830 --> 00:03:33.300 So, for Kilauea, we have the summit caldera. 00:03:33.300 --> 00:03:38.210 And within that, a smaller collapsed crater called Halemaʻumaʻu that has 00:03:38.210 --> 00:03:42.380 a lava lake that has had lava continuously in it since 2008 00:03:42.380 --> 00:03:46.090 and has had lava previous to that as well. 00:03:46.090 --> 00:03:49.960 There is a Southwest Rift Zone that extends down almost to 00:03:49.960 --> 00:03:53.540 the southern tip of the island as well as the East Rift Zone 00:03:53.540 --> 00:03:58.420 which extends out to the eastern tip of the island. 00:03:58.420 --> 00:04:03.730 Since 1983, the main focus of the eruption at Kilauea has been at the 00:04:03.730 --> 00:04:08.680 vent Pu’u’O’o in the middle East Rift Zone here. 00:04:08.680 --> 00:04:13.010 Leading up to the eruption last summer – so the summit 00:04:13.010 --> 00:04:18.709 magma chamber pressurized and caused local inflation. 00:04:18.709 --> 00:04:22.789 And the lava lake level that’s within the crater rose and 00:04:22.789 --> 00:04:26.740 eventually overspilled onto the floor of the crater. 00:04:26.740 --> 00:04:32.430 Then Pu’u’O’o collapsed, magma drained out from the lava lake, 00:04:32.430 --> 00:04:39.460 and eventually withdrew out of sight. And the entire level of sort of magma 00:04:39.460 --> 00:04:44.440 at the chamber went down. And eventually, it worked its way 00:04:44.449 --> 00:04:47.979 from the middle East Rift Zone and along here all the way out to the 00:04:47.979 --> 00:04:52.669 lower East Rift Zone and eventually erupted into the subdivision of Lalani 00:04:52.669 --> 00:04:57.180 Estates, and then lava eventually made its way all the way out to the coast. 00:04:57.180 --> 00:05:01.240 So the lower East Rift Zone produced some truly spectular lava flows and 00:05:01.240 --> 00:05:07.849 garnered some, I think, well-deserved media attention for the truly terrible 00:05:07.849 --> 00:05:12.069 destruction that these lava flows caused and some of the human tragedy 00:05:12.069 --> 00:05:16.719 that occurred last summer that I don’t want to discount. 00:05:16.719 --> 00:05:18.979 But geophysically speaking, the lower East Rift Zone portion 00:05:18.979 --> 00:05:21.949 of the eruption was actually pretty quiet. 00:05:21.949 --> 00:05:26.789 There was a small swarm of earthquakes associated with the intrusion of magma 00:05:26.789 --> 00:05:31.000 into the lower East Rift Zone. There was a magnitude 6.9 earthquake 00:05:31.000 --> 00:05:34.999 that occurred on the décollement between the – sort of the base of 00:05:34.999 --> 00:05:37.710 the island and where it rests on the oceanic lithosphere, 00:05:37.710 --> 00:05:41.939 and the associated aftershocks with that were pretty numerous. 00:05:41.939 --> 00:05:46.219 But that was sort of tangentially associated with the volcanic eruption, 00:05:46.219 --> 00:05:51.150 so I’m going to kind of ignore that earthquake for the most part. 00:05:51.150 --> 00:05:54.160 The story is quite different at the summit, and this is where I’m going 00:05:54.160 --> 00:05:59.490 to be focusing the rest of my talk, obviously, is the response of the summit 00:05:59.490 --> 00:06:04.370 to the evacuation of magma from the summit chamber and the response 00:06:04.370 --> 00:06:08.529 of that system to lava coming out of the East Rift Zone. 00:06:08.529 --> 00:06:13.940 So this is a photograph overlooking this crater, which is – 00:06:13.940 --> 00:06:17.120 oh, and sorry about the horizontal lines here. I’m sorry. 00:06:17.139 --> 00:06:20.229 That’s – yeah, that’s kind of distracting, even for me. 00:06:20.229 --> 00:06:24.230 Anyway, so this is Halemaʻumaʻu crater, and the lava lake would have 00:06:24.230 --> 00:06:27.850 been sitting in the bottom here but now has a very nice ash column 00:06:27.850 --> 00:06:32.580 coming out of it. There were several ash columns that came out of the – 00:06:32.580 --> 00:06:35.680 out of the crater, but those were mostly toward the beginning of the eruption. 00:06:35.680 --> 00:06:38.680 And most of what we’re going to be talking about is 00:06:38.680 --> 00:06:42.199 later into the eruption once it really started going. 00:06:42.199 --> 00:06:44.250 The most incredible thing that happened, actually, 00:06:44.250 --> 00:06:48.240 was the morphological change that occurred at the summit. 00:06:48.240 --> 00:06:53.449 So get your oriented here. So this is a radar amplitude image. 00:06:53.449 --> 00:06:56.219 So you kind of have to trick your eyes because the illumination is 00:06:56.219 --> 00:07:01.120 from a slightly different angle. So the look angle is to the southwest, 00:07:01.120 --> 00:07:04.189 I believe, and – or, sorry, the southeast. 00:07:04.189 --> 00:07:08.199 And so this is Kilauea caldera, which is a topographic low. 00:07:08.199 --> 00:07:11.499 And this is Halemaʻumaʻu, which is also a topographic low. 00:07:11.499 --> 00:07:15.259 This is the location of the lava lake that was within that crater as well as 00:07:15.259 --> 00:07:19.729 a few other craters – I believe Kilauea Iki and Keanakāko’i craters. 00:07:19.729 --> 00:07:23.500 And the caldera extends kind of down through here. 00:07:23.500 --> 00:07:30.340 HVO is situated – or, I suppose I should say, used to be situated on the – 00:07:30.340 --> 00:07:34.080 let’s see – here overlooking Kilauea caldera. 00:07:34.080 --> 00:07:38.379 So I’m going to play this animation as we go through with time. 00:07:38.379 --> 00:07:43.289 So what ended up happening is, as magma withdrew from the chamber 00:07:43.289 --> 00:07:47.289 that’s situated kind of underneath here is the entire caldera, or at least 00:07:47.289 --> 00:07:53.240 quite a lot of the caldera actually collapsed. And the floor dropped. 00:07:53.240 --> 00:07:58.640 So you can see here that – yeah, sort of the area to the west 00:07:58.650 --> 00:08:01.870 of Halemaʻumaʻu initially started collapsing, and then eventually 00:08:01.870 --> 00:08:07.339 it moved outboard to the north and east of Halemaʻumaʻu. 00:08:07.339 --> 00:08:11.430 So the topographic change is pretty incredible. 00:08:11.430 --> 00:08:16.900 We’re talking hundreds of meters of change that occurred. 00:08:16.900 --> 00:08:21.700 So this is – I want to say this floor dropped something like a 150 meters, 00:08:21.700 --> 00:08:24.619 and – I don’t remember, but it was quite a – 00:08:24.619 --> 00:08:28.919 in the hundreds of meters of change – the floor of Halemaʻumaʻu. 00:08:28.919 --> 00:08:34.250 So, looking at this – sort of this animation as well as time-lapse 00:08:34.250 --> 00:08:40.829 videos of – one of the videos that was set up overlooking the crater – 00:08:40.829 --> 00:08:43.740 looking at it in time-lapse, it’s almost reminiscent of somebody 00:08:43.740 --> 00:08:47.199 pulling the plug on the bathroom tub, and everything’s just kind of 00:08:47.200 --> 00:08:52.980 draining out and falling. But that’s not quite right because it’s not 00:08:52.980 --> 00:08:58.200 a smooth change that occurred in time. It was actually very episodic. 00:08:58.209 --> 00:09:01.340 And one of the ways of visualizing that is actually in the seismicity. 00:09:01.340 --> 00:09:04.209 So I mentioned that, in the East Rift Zone, there was very little seismicity. 00:09:04.209 --> 00:09:08.700 At the summit, there were hundreds of thousands of earthquakes that occurred. 00:09:08.700 --> 00:09:11.020 Most of them fairly small – in the magnitude – 00:09:11.020 --> 00:09:16.980 maybe 1 or magnitude 2, but as large as magnitude 5.3 00:09:16.980 --> 00:09:21.079 So these two plots I have both show different aspects of the 00:09:21.079 --> 00:09:25.209 seismicity that occurred with time. So this plot up here at the top is 00:09:25.209 --> 00:09:28.510 the number of triggers per hour. So this is roughly the number 00:09:28.510 --> 00:09:30.860 of earthquakes per hour. And at the top, you can see – 00:09:30.860 --> 00:09:35.839 if I have my mouse – that this goes up to 150 earthquakes per hour. 00:09:35.840 --> 00:09:41.949 So almost two earthquakes a minute at its height, which is kind of crazy. 00:09:43.140 --> 00:09:45.959 Let’s see. So we have the magnitude 6.9 earthquake here, 00:09:45.959 --> 00:09:50.880 which did trigger this particular station, and then its aftershocks out here. 00:09:50.880 --> 00:09:55.920 Once we get in – further in, we notice on the RSAM, which is sort of a – 00:09:55.920 --> 00:09:59.450 it’s a representation of the average amplitude of seismicity. 00:09:59.450 --> 00:10:04.000 So it’s a combination of tremor, and it takes into account the size of 00:10:04.000 --> 00:10:08.079 earthquakes as well as their numbers. So basically, more RSAM means more 00:10:08.079 --> 00:10:12.910 and larger earthquakes, and smaller RSAM means very little seismicity. 00:10:12.910 --> 00:10:15.780 So, at the very beginning of the eruption, there was quite a lot 00:10:15.780 --> 00:10:19.720 of tremor, which then died out. And then eventually, it settled into this 00:10:19.720 --> 00:10:26.910 sort of pattern of seismicity that would – if you zoom in here, it ramps up, 00:10:26.910 --> 00:10:31.319 culminates in a magnitude 5.3 – what I’m going to call a collapse event, 00:10:31.319 --> 00:10:32.759 and then the seismicity shuts off. 00:10:32.759 --> 00:10:35.540 And then, after a few hours, it ramps back up again. 00:10:35.540 --> 00:10:40.180 You get more and larger earthquakes. Ramps up to this magnitude 5.3 00:10:40.180 --> 00:10:43.540 and then shuts off. It’s almost like an aftershock sequence in reverse. 00:10:43.540 --> 00:10:45.449 Because there are hardly any aftershocks of these 00:10:45.449 --> 00:10:48.060 magnitude 5 earthquakes. 00:10:48.060 --> 00:10:50.560 I think that’s all I want to say on that. 00:10:50.560 --> 00:10:55.210 So, as part of the monitoring effort for looking at the seismicity and the 00:10:55.210 --> 00:10:58.170 geodetic changes and everything else that was happening as part of the 00:10:58.170 --> 00:11:03.940 24/7 watch cycle for what we had going during the eruption, 00:11:03.940 --> 00:11:08.200 I set up my near-real-time repeating earthquake detector – REDPy – 00:11:08.200 --> 00:11:13.019 shameless plug here – as a way of monitoring the 00:11:13.019 --> 00:11:17.269 repeating earthquakes that were occurring during this eruption. 00:11:17.269 --> 00:11:21.480 So basically, what the program does is, it takes these triggers, or earthquakes, 00:11:21.480 --> 00:11:27.860 and tries to associate them into clusters or groups of similar events based on 00:11:27.870 --> 00:11:30.720 their waveform similarity, which basically just groups them 00:11:30.720 --> 00:11:36.089 roughly by similar location and similar focal mechanism. 00:11:36.089 --> 00:11:39.899 At the top here, I have the number of triggered events, basically as 00:11:39.899 --> 00:11:45.730 a timeline similar to the previous one. Instead, putting a red line for the 00:11:45.730 --> 00:11:51.079 number of earthquakes which were associated into at least one family. 00:11:51.080 --> 00:11:56.020 And the black here – gosh, I still keep losing that – are earthquakes 00:11:56.020 --> 00:11:59.600 that didn’t have any other repeaters in the catalog. 00:11:59.610 --> 00:12:02.980 So it triggered, but it couldn’t associate it with another event. 00:12:02.980 --> 00:12:08.910 And over half of the seismicity was associated with a repeated 00:12:08.910 --> 00:12:15.630 slipping or repeated excitation of a fault or some other seismic source 00:12:15.630 --> 00:12:19.490 near the – near the volcano. On this plot, what I have is 00:12:19.490 --> 00:12:22.379 the occurrence of the largest clusters of events, 00:12:22.379 --> 00:12:26.200 so those with at least 250 earthquake members. 00:12:26.200 --> 00:12:31.389 And each of these horizontal lines is one family, and then the brightness is 00:12:31.389 --> 00:12:34.970 related to when those actually occurred. And so you can see it’s sort of 00:12:34.970 --> 00:12:38.720 clustered in time. You can see sort of – there’s earthquakes, there’s not, 00:12:38.720 --> 00:12:40.351 there’s earthquakes, there’s not, and it just kind of goes 00:12:40.351 --> 00:12:41.720 back and forth and back and forth. 00:12:41.720 --> 00:12:47.750 Each of these vertical lines is one of the 62 collapse events. 00:12:47.750 --> 00:12:52.810 And one of the things that I noticed, especially later on, is that these families 00:12:52.810 --> 00:12:56.060 don’t seem to care that the volcano is collapsing. 00:12:56.060 --> 00:12:59.199 The collapse mechanism seems to be largely non-destructive 00:12:59.199 --> 00:13:02.899 to these earthquake sources. Certainly it may change the waveforms 00:13:02.899 --> 00:13:04.680 for those earthquakes, but it’s doing it 00:13:04.680 --> 00:13:07.930 in a smooth enough way that those sources are 00:13:07.930 --> 00:13:13.389 not being destroyed each time. Which is really great for me because 00:13:13.389 --> 00:13:18.410 one of the techniques that I used in my dissertation requires a catalog 00:13:18.410 --> 00:13:21.019 of repeating earthquakes, and it works best if you have 00:13:21.020 --> 00:13:24.160 repeating earthquakes that last a really long time. 00:13:24.160 --> 00:13:27.860 So, once I noticed this, kind of on a whim, I decided, 00:13:27.879 --> 00:13:31.230 hey, what would happen if I just threw this catalog that I have 00:13:31.230 --> 00:13:35.930 already made into the codes for my dissertation and just see what’s there? 00:13:35.930 --> 00:13:38.029 And that’s what I did. And I wasn’t expecting what 00:13:38.029 --> 00:13:43.320 I got out of it, but that’s going to be the subject of the rest of this talk. 00:13:43.320 --> 00:13:49.740 So the subject of the last two chapters of my dissertation were on using repeating 00:13:49.740 --> 00:13:53.839 earthquakes to look at changes in seismic velocity structure. 00:13:53.839 --> 00:13:56.450 So seismic velocity is simply just the speed at which waves 00:13:56.450 --> 00:14:00.480 travel through the subsurface. And obviously that velocity – 00:14:00.480 --> 00:14:06.880 that sort of property of the subsurface varies depending on where you are. 00:14:06.880 --> 00:14:12.290 So either from the geology or at depth. But it turns out that it’s actually 00:14:12.290 --> 00:14:16.879 not perfectly constant in time. And we think, to first order, 00:14:16.879 --> 00:14:21.029 that you can very slightly change the seismic velocity by opening and 00:14:21.029 --> 00:14:26.940 closing cracks in the subsurface. So we think that, to first order, 00:14:26.940 --> 00:14:31.529 that it’s a function of volumetric strain through third-order elastic constants. 00:14:31.529 --> 00:14:35.829 And it is probably entirely likely and possible that there are other factors 00:14:35.829 --> 00:14:39.889 other than volumetric strain that play a role in changing the seismic velocity, 00:14:39.889 --> 00:14:44.420 but I want to emphasize this is really the lens that I see seismic velocity change 00:14:44.420 --> 00:14:47.399 through, and that really colors sort of the discussion that we’re going to have. 00:14:47.399 --> 00:14:53.009 So I want to be up front with that. So, in this view, seismic velocity 00:14:53.009 --> 00:14:55.759 increases under compression, which would close cracks and 00:14:55.759 --> 00:15:00.420 pore spaces, so you can think – so you can consider rock is fast, fluids are slow, 00:15:00.420 --> 00:15:03.769 and if you close all the cracks, more volumetrically of what 00:15:03.769 --> 00:15:07.100 you’re passing through is fast rock. And then, if you open the cracks 00:15:07.100 --> 00:15:10.250 in dilatation, you introduce more fluids into the system, 00:15:10.250 --> 00:15:13.700 which causes the velocity to be a little bit slower. 00:15:13.700 --> 00:15:17.810 All right. So the main way that most researchers look at seismic 00:15:17.810 --> 00:15:22.170 velocity change is through looking at ambient noise interferometry. 00:15:22.170 --> 00:15:27.260 So basically you take recordings of the ambient noise wavefield at two stations, 00:15:27.260 --> 00:15:30.360 cross-correlate them, and out pops the Green’s function 00:15:30.360 --> 00:15:36.209 of the waves that travel between the source and receiver stations. 00:15:36.209 --> 00:15:41.060 And you get – yeah, sort of this two-sided Green’s function here. 00:15:43.730 --> 00:15:48.320 The path that the waves take – obviously you have the direct path 00:15:48.329 --> 00:15:52.089 that goes between those two stations, but there are also coda waves that are 00:15:52.089 --> 00:15:56.939 the multiply scattered waves that sort of bounce around all of the imperfections 00:15:56.939 --> 00:16:02.260 in the subsurface and sample the subsurface for longer and longer as 00:16:02.260 --> 00:16:05.550 you go later and later into the coda, which makes them extremely sensitive 00:16:05.550 --> 00:16:09.959 to very small differences in the seismic velocity, for example. 00:16:09.959 --> 00:16:13.720 So if you have a seismic velocity decrease, the waves are trapped for 00:16:13.720 --> 00:16:18.129 longer and longer in slower media. So the waves will arrive later and 00:16:18.129 --> 00:16:21.769 later and later as you go into the coda. And so what ends up happening to 00:16:21.769 --> 00:16:26.199 the waveforms is that, if you have a reference waveform from where 00:16:26.199 --> 00:16:30.680 the velocity is faster, compared to a waveform from where the velocity 00:16:30.680 --> 00:16:33.589 is slower, the slower waveform will look stretched out. 00:16:33.589 --> 00:16:37.370 So you can sort of see that here, where basically the red arrivals are 00:16:37.370 --> 00:16:40.829 coming in later and later with time. And you can stretch out – you can 00:16:40.829 --> 00:16:43.800 figure out how much you need to stretch it by and back out what 00:16:43.800 --> 00:16:48.170 the velocity change was by just knowing that stretching factor. 00:16:48.170 --> 00:16:52.649 These techniques have been really popular and show a lot of promise as 00:16:52.649 --> 00:16:56.580 a tool to monitor volcanic eruptions. One of the nice things about ambient 00:16:56.580 --> 00:16:59.529 noise is that, if you have two stations that are recording, and you have 00:16:59.529 --> 00:17:04.060 a continuous source of noise, like the ocean microseism, 00:17:04.060 --> 00:17:06.630 you can basically continuously monitor a volcano. 00:17:06.630 --> 00:17:10.040 And what we’ve seen at several different volcanoes in different 00:17:10.040 --> 00:17:14.700 eruptions is that there seems to be this consistent tendency for the 00:17:14.700 --> 00:17:20.260 velocities to decrease in the days and weeks leading up to volcanic eruptions. 00:17:21.300 --> 00:17:24.959 But we’re still kind of at a point where we understand, okay, we have this 00:17:24.959 --> 00:17:28.600 observation that this is repeatable, but we’re still trying to figure out, 00:17:28.600 --> 00:17:31.840 okay, why? What is it about what’s going on in the volcano 00:17:31.840 --> 00:17:35.480 that’s causing these changes to occur? We know that seismic velocity 00:17:35.480 --> 00:17:38.500 is sensitive to a whole host of different factors. 00:17:38.500 --> 00:17:41.020 So obviously there’s deformation. 00:17:41.020 --> 00:17:44.289 We’ve seen that at different volcanoes, that when the volcano deforms, 00:17:44.289 --> 00:17:47.500 the velocity changes. We also know that it can change 00:17:47.500 --> 00:17:52.720 with the absence or presence of snow, changes in the groundwater level, 00:17:52.720 --> 00:17:56.660 temperature, the occurrence of large earthquakes – there’s a whole 00:17:56.669 --> 00:17:59.200 host of different factors that could come into this. 00:17:59.200 --> 00:18:02.840 And we’re still trying to figure out that, if we have these observations leading 00:18:02.840 --> 00:18:07.630 up to a volcanic eruption, how can we recognize if an eruption is going to 00:18:07.630 --> 00:18:12.180 occur or if it’s some other factor that’s not related to the volcano? 00:18:12.180 --> 00:18:16.440 So, at Kilauea, there has – there has been previous work done on 00:18:16.440 --> 00:18:21.659 looking at seismic velocity change. For this case, they used, actually, 00:18:21.659 --> 00:18:27.330 the spatter and tremor associated with the lava lake itself, rather than 00:18:27.330 --> 00:18:31.600 the ocean microseism, to back out velocity changes there. 00:18:31.600 --> 00:18:36.090 And what they noticed – Donaldson and others noticed is that the seismic 00:18:36.090 --> 00:18:40.780 velocity here in blue seems to change as a function of radial tilt. 00:18:40.780 --> 00:18:44.090 So tilt is a measure of the deformation of the volcano. 00:18:44.090 --> 00:18:49.900 So when the volcano is in inflation, the tilt is out, which is up. 00:18:49.900 --> 00:18:54.890 And when the volcano deflates, the tilt becomes negative or decreases, 00:18:54.890 --> 00:18:56.440 which is down. And, in this case, 00:18:56.440 --> 00:19:00.330 when there is inflation in the volcano, the velocities get faster, 00:19:00.330 --> 00:19:03.990 and when there’s deflation, the velocities get slower. 00:19:03.990 --> 00:19:07.640 And this was noticed pretty consistently, at least on short time scales. 00:19:07.640 --> 00:19:11.480 On the longer time scale, they didn’t agree quite as much, 00:19:11.480 --> 00:19:14.110 which will be a little bit of a theme for later. 00:19:14.110 --> 00:19:18.730 Leading up to the 2018 eruption, other researchers went and did the same thing. 00:19:18.730 --> 00:19:22.710 And for sort of the early portion right before the start of eruption, they noticed 00:19:22.710 --> 00:19:27.720 that the seismic velocity here in blue correlated fairly well with the tilt, 00:19:27.720 --> 00:19:33.940 as it had before here in green, I believe, up until a point here in late April 00:19:33.940 --> 00:19:39.630 where the tilt continues to show inflation, but the velocity decreases. 00:19:39.630 --> 00:19:45.480 And their interpretation for that was that either the velocity change was 00:19:45.480 --> 00:19:49.350 centered in a different location that the deformation wasn’t as sensitive to, 00:19:49.350 --> 00:19:53.770 or that it was some other factor, such as the occurrence of earthquakes that were 00:19:53.770 --> 00:20:00.380 damaging the subsurface and therefore causing the velocities to decrease. 00:20:00.380 --> 00:20:03.250 One of the things you might notice here is that this time series 00:20:03.250 --> 00:20:08.040 of seismic velocity ends at the start of the eruption. 00:20:08.040 --> 00:20:13.049 So ambient noise is great if you have, you know, continuous noise and 00:20:13.049 --> 00:20:15.360 whatnot, but it turns out that, when you have tons of earthquakes 00:20:15.360 --> 00:20:19.110 and lots of other noise emanating from the volcano, the correlation 00:20:19.110 --> 00:20:26.430 functions start to become much messier and less consistent with time. 00:20:26.430 --> 00:20:29.570 So it’s much more difficult to sort of back out what the velocity 00:20:29.570 --> 00:20:32.520 change is during an eruption. I’m sure that there are probably 00:20:32.520 --> 00:20:35.399 ways that you could probably go about getting around those problems, but, 00:20:35.399 --> 00:20:42.830 to this point, I don’t think that I’ve really seen a truly convincing seismic velocity 00:20:42.830 --> 00:20:47.420 change associated with an eruption using traditional ambient noise. 00:20:47.420 --> 00:20:52.260 So, in my view, if you have earthquakes, you may as well use them. 00:20:52.270 --> 00:20:56.330 And it turns out that the – sort of the genesis of these techniques 00:20:56.330 --> 00:20:59.390 was actually in using repeating earthquakes or repeated seismic 00:20:59.390 --> 00:21:05.330 sources or blasts – repeated individual impulsive events. 00:21:05.330 --> 00:21:10.059 So here I have an example of two earthquakes from Kilauea recorded on 00:21:10.059 --> 00:21:17.970 one station in the middle of the eruption in sort of mid-July about a day apart. 00:21:17.970 --> 00:21:22.460 Here we have sort of the body waves. Here are the P wave arrival and 00:21:22.460 --> 00:21:26.340 probably S somewhere in here. And, in the body waves in the early part 00:21:26.340 --> 00:21:30.990 of the waveform – the two waveforms, they line up pretty well. 00:21:30.990 --> 00:21:35.090 I think the cross-correlation coefficient is somewhere between 0.7 and 0.8 – 00:21:35.090 --> 00:21:37.649 something like that. But, as you get further out 00:21:37.649 --> 00:21:40.760 into the coda, or the later part of the waveform, you can see that 00:21:40.760 --> 00:21:47.220 the red – the red waveform is delayed compared to the black one. 00:21:47.220 --> 00:21:50.680 And then, what I’ve done here is, I’ve taken actually the black waveform, 00:21:50.680 --> 00:21:56.910 and I’ve stretched it out by 1-1/2%. And now, the arrivals line up. 00:21:56.910 --> 00:22:01.399 So this implies that there was a 1-1/2% velocity increase between 00:22:01.399 --> 00:22:04.549 the times of these two earthquakes less than a day apart. 00:22:04.549 --> 00:22:08.800 Which, in seismic velocity change terms, is quite a lot. 00:22:09.780 --> 00:22:13.419 I’d like to show you here – this is an example of an entire family of 00:22:13.419 --> 00:22:19.929 earthquakes recorded on a single station. So each horizontal line of pixels is 00:22:19.929 --> 00:22:23.920 an individual recording with the color related to whether the 00:22:23.920 --> 00:22:28.409 motion was up or down. And so the body waves end up showing up as 00:22:28.409 --> 00:22:32.679 these vertical bars of alternating colors. And you can see here in the early 00:22:32.679 --> 00:22:35.260 part of the waveform, again, they line up fairly well. 00:22:35.260 --> 00:22:39.020 There’s a little bit of mess in here. But then, later on, you can kind of 00:22:39.020 --> 00:22:42.360 barely see that there are – they don’t line up vertically. 00:22:42.360 --> 00:22:48.539 There’s actually slope associated with some of the coda waves through here, 00:22:48.539 --> 00:22:52.400 which are consistent from station to station and 00:22:52.400 --> 00:22:55.679 with time from successive events. 00:22:55.679 --> 00:23:00.580 And these horizontal lines are the collapse events that happened, and 00:23:00.580 --> 00:23:06.830 there are breaks in the waveforms across these – across these collapse events. 00:23:06.830 --> 00:23:11.910 So there was some sort of velocity change associated with this event, 00:23:11.910 --> 00:23:15.080 for example, and then there is changes that occurring between the events as 00:23:15.080 --> 00:23:20.580 well, that you can visually see on just the raw waveforms themselves. 00:23:21.690 --> 00:23:25.880 So what I do is I take all of the possible pairs of repeating earthquakes within 00:23:25.880 --> 00:23:29.570 a family, repeat that for all of the families, and then repeat that 00:23:29.570 --> 00:23:35.000 for all of the stations. This is a map showing all of the stations around the 00:23:35.000 --> 00:23:39.539 caldera. So this is the outline of the caldera here. Halemaʻumaʻu here. 00:23:39.539 --> 00:23:44.270 So each of the large triangles is a permanent seismic sensor 00:23:44.270 --> 00:23:49.690 for which I’ve done these seismic velocity change measurements for. 00:23:49.690 --> 00:23:54.529 And these smaller triangles are a subset of the temporary nodal deployment 00:23:54.529 --> 00:23:58.309 that was out for a few weeks in June and July. 00:23:58.309 --> 00:24:04.000 I have a large red triangle here, which is a tiltmeter, which I’m 00:24:04.000 --> 00:24:09.010 going to be showing you data from. And the black square here is 00:24:09.010 --> 00:24:14.390 a GPS sensor that was installed on the crater floor. 00:24:14.390 --> 00:24:17.500 So where are the earthquakes? Because that’s kind of important 00:24:17.500 --> 00:24:19.780 because they’re the sources that I’m going to be using for this. 00:24:19.780 --> 00:24:24.990 So, thanks to Dave Shelly, we have some very nice HypoDD high-precision 00:24:24.990 --> 00:24:29.269 relative relocations for the seismicity that occurred during the eruption. 00:24:29.269 --> 00:24:35.289 And basically all of it was centered here within the crater, just mostly 00:24:35.289 --> 00:24:39.190 to the east of Halemaʻumaʻu. There’s a – there are a few little 00:24:39.190 --> 00:24:42.940 pockets of deeper seismicity out here, but, by and large, 00:24:42.940 --> 00:24:46.429 like, 90-plus percent of the seismicity is in this area here. 00:24:46.429 --> 00:24:51.649 I’ve colored this by depth, so the shallowest seismicity is darkest. 00:24:51.649 --> 00:24:54.309 And then, as you go deeper, it becomes lighter. 00:24:54.309 --> 00:24:58.179 So this is kind of nice for me because one of – one of the 00:24:58.179 --> 00:25:02.830 assumptions is that I’m sensitive to where the earthquakes are. 00:25:02.830 --> 00:25:06.140 So if the velocity changed in the area where the earthquakes are, 00:25:06.140 --> 00:25:10.240 I should be able to see it. And although it’s sort of spread out, 00:25:10.240 --> 00:25:12.480 you know, I can kind of assume that most of the change 00:25:12.480 --> 00:25:15.559 that I’m seeing is probably occurring here. 00:25:15.559 --> 00:25:20.380 One of the other interesting things is that these earthquakes, if you look at 00:25:20.380 --> 00:25:25.730 where the edge of the collapsed – the new collapsed crater formed, is that 00:25:25.730 --> 00:25:28.630 they’re almost all inside of that as well. So most of these earthquakes seem to 00:25:28.630 --> 00:25:34.680 be occurring likely above or sort of in a similar depth as the magma 00:25:34.680 --> 00:25:39.120 chamber and more or less above it, and maybe to the – off to the sides of it. 00:25:39.120 --> 00:25:45.580 All right. So I’m going to show you the results of the seismic velocity 00:25:45.580 --> 00:25:50.299 change time series from these four stations sort of scattered around 00:25:50.299 --> 00:25:52.820 the caldera, and then this guy just a little bit further away 00:25:52.820 --> 00:25:56.460 so that I can show you how this is a little bit different with distance. 00:25:56.460 --> 00:26:00.840 I basically just use a fairly straightforward linear inversion 00:26:00.840 --> 00:26:06.289 to go from each of these thousands of pairwise measurements to 00:26:06.289 --> 00:26:09.180 a continuous time series that best fits them. 00:26:09.180 --> 00:26:11.140 And so I’m going to show that to you here. 00:26:11.140 --> 00:26:15.520 All right. So we’ve got time at this axis starting at the beginning of June, 00:26:15.539 --> 00:26:17.880 which is really when my sensitivity starts. 00:26:17.880 --> 00:26:20.480 Because I require that there have to be earthquakes occurring. 00:26:20.480 --> 00:26:23.590 It doesn’t really work with tremor. And the beginning of June is really 00:26:23.590 --> 00:26:27.000 when there are enough earthquakes for this technique to work. 00:26:27.000 --> 00:26:33.370 On this axis is relative velocity change, spanning 9% of change. 00:26:33.370 --> 00:26:37.919 The thing to think about for this is that the vertical position of each of 00:26:37.919 --> 00:26:40.929 these time series doesn’t really matter. You can kind of think of this almost 00:26:40.929 --> 00:26:43.790 as a scale bar on the left here. 00:26:43.790 --> 00:26:48.429 So it’s not that this station is necessarily faster than this one. 00:26:48.429 --> 00:26:52.669 It’s just I’ve spaced them out so that they’re a little bit easier to see. 00:26:52.669 --> 00:26:55.669 So what I’d like you to notice is that, to first order, they have a lot of 00:26:55.669 --> 00:26:58.159 similar features, which gives me confidence that what we’re seeing is 00:26:58.159 --> 00:27:02.669 real and probably located, as I said, probably very near the source of the 00:27:02.669 --> 00:27:06.000 earthquakes because that’s what all of these stations have in common. 00:27:06.000 --> 00:27:10.800 So we have sort of this initial decrease in velocity that flattens out. 00:27:10.800 --> 00:27:15.390 And then, on top of that, we have this squiggly cyclic pattern. 00:27:15.390 --> 00:27:17.860 As you go further away with distance – so this station is 00:27:17.860 --> 00:27:21.880 about 4 kilometers away – the amplitude is lower. 00:27:21.880 --> 00:27:23.980 And it gets lower as you go further and further away. 00:27:23.980 --> 00:27:29.970 So, if we have a velocity change that’s centered very near the center of 00:27:29.970 --> 00:27:34.500 the caldera, and we don’t have change further away, when you go to a station 00:27:34.500 --> 00:27:37.940 that’s further away, it’s sampling relatively more of the unchanged 00:27:37.940 --> 00:27:41.190 volume compared to closer stations that are probably 00:27:41.190 --> 00:27:43.960 sampling more of the changed volume. 00:27:43.960 --> 00:27:48.070 This is a little bit weird to look at, so what I’m going to be showing mostly 00:27:48.070 --> 00:27:52.590 for the rest of the talk is an average of the velocity change at six different 00:27:52.590 --> 00:27:56.350 stations around the caldera to really highlight what’s similar between 00:27:56.350 --> 00:27:59.570 the stations rather than focusing on what’s different between them. 00:27:59.570 --> 00:28:04.870 And, again, the main features are this long-term change here that has 00:28:04.870 --> 00:28:09.200 sort of a break in slope, and then this wiggly cyclicity, 00:28:09.200 --> 00:28:12.760 which is not random, as you might expect. 00:28:12.760 --> 00:28:16.710 We can also add several other geophysical time series to compare at 00:28:16.710 --> 00:28:20.409 the same time and see – and get sort of a larger picture of what’s going on. 00:28:20.409 --> 00:28:24.539 So here I’ve got the RSAM again – our seismicity buddy. 00:28:24.539 --> 00:28:28.049 We’ve got the radial tilt here. 00:28:28.049 --> 00:28:32.510 And then the vertical position of the GPS that was 00:28:32.510 --> 00:28:35.120 sitting on the floor of the crater here. 00:28:35.120 --> 00:28:41.250 So I’m going to break this talk into discussions about sort of the long term 00:28:41.250 --> 00:28:45.200 and the short term. And I’m going to start with the short term. 00:28:45.200 --> 00:28:50.289 So I’m going to zoom in so that we can see a little bit better and more clearly 00:28:50.289 --> 00:28:53.880 what’s going on on the short term. All right, so seismicity here. 00:28:53.880 --> 00:28:57.840 Again, we’ve got the ramp-up, culminate in a magnitude 5.3, 00:28:57.840 --> 00:29:00.010 and then shut off. Ramp up. Shut off. 00:29:00.010 --> 00:29:02.830 I’m going to jump down here to the GPS. 00:29:02.830 --> 00:29:07.210 So the position of this sensor is constantly moving down. 00:29:07.210 --> 00:29:14.269 I’ve got these horizontal bars here to illustrate that it’s pretty much constantly 00:29:14.269 --> 00:29:18.139 dropping in elevation as we’re going. But it starts off sort of slow. 00:29:18.139 --> 00:29:22.360 It gets a little bit faster, and then, when the collapse event happens, it drops. 00:29:22.360 --> 00:29:26.230 And then it’s slow, goes a little bit faster, and drops. 00:29:26.230 --> 00:29:30.899 The tilt is kind of interesting. So you would think that, 00:29:30.899 --> 00:29:35.690 as the volcano is collapsing that – it’s deflating the entire time, right? 00:29:35.690 --> 00:29:38.940 And so you would think that the tilt would just be constantly inward. 00:29:38.940 --> 00:29:44.340 But it turns out that the tilt is actually highly sensitive to pressure 00:29:44.340 --> 00:29:47.580 in the magmatic – in the shallow magmatic system. 00:29:47.580 --> 00:29:51.649 So, as magma is draining away and the pressure is decreasing in the 00:29:51.649 --> 00:29:55.980 magma chamber, the roof on top of it collapses into the magma – 00:29:55.980 --> 00:29:59.230 or, on top of the magma chamber and re-pressurizes it. 00:29:59.230 --> 00:30:02.429 And that increase in pressure causes an inflation – 00:30:02.429 --> 00:30:06.390 an inflationary signal in tilt, which is what we see here. 00:30:06.390 --> 00:30:12.340 So it’s deflating, quick inflation, deflates, inflates. 00:30:12.340 --> 00:30:17.320 So the seismic velocity starts off high, decreases – and in some places, 00:30:17.320 --> 00:30:23.399 very quickly – sort of goes to a – to a low level, and then, at the time of the – 00:30:23.399 --> 00:30:27.190 at the time of the collapse events, increases almost instantaneously, 00:30:27.190 --> 00:30:30.309 at least in the limit of my resolution – 00:30:30.309 --> 00:30:32.310 increases at the times of these collapse events. 00:30:32.310 --> 00:30:37.559 So it gets – goes fast, and then it gets slow, and then increases in velocity. 00:30:37.559 --> 00:30:42.799 So we have an increase in velocity at the time of each of these collapse events. 00:30:42.799 --> 00:30:47.580 So this is a little bit weird for me initially. 00:30:47.580 --> 00:30:52.740 So you would – you would think that, when you get a pressure increase or an 00:30:52.740 --> 00:30:58.460 inflation of a magma body at depth, that it should cause dilatation at the surface. 00:30:58.460 --> 00:31:01.440 So you inflate it, and it – and it pops up the surface, 00:31:01.440 --> 00:31:04.410 and it causes dilatation there, which you would expect to cause 00:31:04.410 --> 00:31:08.360 a decrease in velocity, which is opposite of what we see. 00:31:08.360 --> 00:31:13.009 So this is some strain modeling that was done to illustrate – so if you have 00:31:13.009 --> 00:31:18.980 an increase in the pressure at depth, it causes – you would expect to see 00:31:18.980 --> 00:31:22.140 that the velocities decrease at the surface. 00:31:22.140 --> 00:31:26.269 At Kilauea, the argument is that the magma chamber is shallow enough 00:31:26.269 --> 00:31:30.769 that actually it pushes sort of this inflationary lobe very near to the 00:31:30.769 --> 00:31:35.180 surface, and, in the depth range that they’re – the techniques that were 00:31:35.180 --> 00:31:39.750 used for this study, that, in those depth ranges, you would actually expect 00:31:39.750 --> 00:31:44.399 to see an increase in velocity. Because, at depth around the sides 00:31:44.399 --> 00:31:47.380 of the magma chamber, it actually is in compression. 00:31:47.380 --> 00:31:52.330 It’s not in dilatation everywhere. It’s actually in compression to the sides. 00:31:52.330 --> 00:31:55.409 So they are argue that, where they have sensitivity, that you would expect to 00:31:55.409 --> 00:32:00.820 see a velocity increase given a pressure increase in the magma chamber. 00:32:00.820 --> 00:32:05.840 So, for my techniques, I’m using a slightly higher frequency range, which 00:32:05.840 --> 00:32:10.900 affects sort of the depth sensitivity. So I would – I had originally expected 00:32:10.900 --> 00:32:17.390 that most of what I would see was sort of this lobe here – the dilatational lobe. 00:32:17.390 --> 00:32:23.050 But, in going back through, I had to reconsider where these earthquakes 00:32:23.050 --> 00:32:28.289 were actually occurring in relation to where the inflation 00:32:28.289 --> 00:32:30.350 in the magma chamber was happening. 00:32:30.350 --> 00:32:34.110 So, as a reminder here, we’ve got the earthquakes here. 00:32:34.110 --> 00:32:37.750 The best estimate that we have at the moment of where the 00:32:37.750 --> 00:32:42.730 magma chamber is based on deformation is here. 00:32:42.730 --> 00:32:46.789 Sort of just slightly east of Halemaʻumaʻu. 00:32:46.789 --> 00:32:48.679 And most of the earthquakes are still occurring sort of on 00:32:48.679 --> 00:32:51.990 the eastern sides of it. And this is the – sort of 00:32:51.990 --> 00:32:56.460 the best approximation that I have access to of what the size 00:32:56.460 --> 00:32:59.169 of the magma chamber might be at depth. 00:32:59.169 --> 00:33:05.840 I redid, with Kyle’s help, the strain modeling with depth. 00:33:05.840 --> 00:33:10.500 So this is a depth cross-section here, but it’s not a true cross-section. 00:33:10.500 --> 00:33:15.019 This is depth and then radial distance because it’s got radial symmetry. 00:33:15.019 --> 00:33:18.720 And you can see here we’ve got the lobe above the magma chamber where you 00:33:18.720 --> 00:33:22.529 would expect the velocities to decrease. Because cracks are – 00:33:22.529 --> 00:33:26.350 cracks should be opening in dilatation. And then down here we have the lobe 00:33:26.350 --> 00:33:29.550 where you would expect the velocities to increase because 00:33:29.550 --> 00:33:33.780 they’re being compressed. And it turns out that about 2/3 to 3/4 00:33:33.780 --> 00:33:38.700 of the earthquakes that I’m using are in this compressional lobe. 00:33:38.700 --> 00:33:41.880 So where you would expect to see a velocity increase, given a pressure 00:33:41.880 --> 00:33:44.950 increase in the magma chamber. But there are still several earthquakes 00:33:44.950 --> 00:33:48.809 shallower that might fall into this place where you would expect 00:33:48.809 --> 00:33:54.710 the velocities to be decreasing. So let’s investigate that a little bit. 00:33:54.710 --> 00:33:58.610 What I can do is, I can take subsets of these earthquakes based on their 00:33:58.610 --> 00:34:02.250 location and just say, okay, if I have a location associated with the 00:34:02.250 --> 00:34:06.899 earthquakes, let’s say – let’s only take earthquakes above a certain depth 00:34:06.899 --> 00:34:09.470 and re-run the inversion just using those earthquakes. 00:34:09.470 --> 00:34:12.490 Let’s take the earthquakes that are only below a certain depth and re-run the 00:34:12.490 --> 00:34:16.829 inversion and see if we can see if there’s any difference between those. 00:34:16.829 --> 00:34:20.909 So here, this is, once again, the inversion using all depths. 00:34:20.909 --> 00:34:25.760 And then the other four lines are for using different subsets above 00:34:25.760 --> 00:34:32.099 two different depth levels, just to kind of demonstrate that it 00:34:32.100 --> 00:34:35.440 doesn’t really matter exactly which depth level you use. 00:34:35.440 --> 00:34:39.339 And I’m going to take the whole thing off so that you can see. 00:34:39.339 --> 00:34:42.129 So you can definitely see that, on the long term, there are changes 00:34:42.129 --> 00:34:46.359 that are very different with depth, which we’ll come back to here in a bit. 00:34:46.359 --> 00:34:50.080 But we really need to zoom in in order to see, sort of on an individual cycle 00:34:50.080 --> 00:34:55.149 scale, if there are differences. And what we see is that the shallower 00:34:55.149 --> 00:35:00.291 earthquakes generally tend to have slightly less amplitude than if you 00:35:00.291 --> 00:35:06.000 use just the deeper earthquakes. This seems to hold sort of well 00:35:06.000 --> 00:35:12.420 if you look at individual stations. So this is outward by distance. 00:35:12.420 --> 00:35:16.590 So these are close-in stations. These are far-away stations. 00:35:16.590 --> 00:35:19.500 This is white in the middle because we’re referencing it to using the 00:35:19.500 --> 00:35:24.560 inversion with all of the earthquakes. And then red colors here indicate that 00:35:24.560 --> 00:35:29.880 there is relatively less amplitude of change when you use shallower 00:35:29.880 --> 00:35:33.740 earthquakes, and relatively more amplitude if you use deeper earthquakes, 00:35:33.740 --> 00:35:37.570 which indicates that more of the change on the short term 00:35:37.570 --> 00:35:41.920 is occurring slightly deeper. So in the range of around 00:35:41.920 --> 00:35:44.910 1-1/2, probably 2, kilometers’ depth. 00:35:44.910 --> 00:35:49.170 What’s interesting here is that this relationship seems to potentially switch 00:35:49.170 --> 00:35:54.770 as we get further out, and you see slightly more amplitude as you go 00:35:54.770 --> 00:35:59.260 further away depth – at depth, and slightly less as you go further away 00:35:59.260 --> 00:36:01.900 if you use the shallow ones, which I haven’t quite figured out yet, 00:36:01.900 --> 00:36:08.420 but – so one of the things here is that the prediction based on sort of this 00:36:08.430 --> 00:36:12.800 very simple strain modeling is that you would actually expect to see, not just 00:36:12.800 --> 00:36:17.240 a decrease in the amount of change if you use the shallower earthquakes 00:36:17.240 --> 00:36:20.540 in that red zone – you would actually expect to see a complete flip 00:36:20.540 --> 00:36:24.211 of the amplitude. So instead of getting a velocity increase, you would expect 00:36:24.211 --> 00:36:27.050 to get a velocity decrease. And that’s not what we see. 00:36:27.050 --> 00:36:32.830 So it could be that the basic assumptions of our model for deformation are wrong. 00:36:32.830 --> 00:36:36.560 Maybe it doesn’t capture the complexity in the near-field. 00:36:36.560 --> 00:36:41.109 It could be that the location is wrong. The shape of the magma chamber is off. 00:36:41.109 --> 00:36:46.450 A whole bunch of different things. Or it could be that part – that this – 00:36:46.450 --> 00:36:50.359 that deformation associated with pressure changes in the 00:36:50.359 --> 00:36:53.839 magmatic chamber is not the whole story. 00:36:53.839 --> 00:36:58.720 So one of the things that I noticed when I did my dissertation was that, 00:36:58.720 --> 00:37:03.960 at Mount St. Helens, at the beginning of its eruption, that the seismic velocity 00:37:03.960 --> 00:37:08.640 change changed [chuckles] in concert with the RSAM – 00:37:08.640 --> 00:37:11.560 the seismic amplitude – the amount of seismicity that occurred. 00:37:11.560 --> 00:37:15.660 And actually, they were anti-correlated. So, for this, I have seismic amplitude, 00:37:15.660 --> 00:37:18.530 but I flipped it upside-down so that it’s the same sense. 00:37:18.530 --> 00:37:22.260 So here what we have is low seismicity and then the seismicity turns on. 00:37:22.260 --> 00:37:28.710 So this is actually high seismicity and low seismic – low, slow wave speeds. 00:37:28.710 --> 00:37:32.160 These were explosions where the seismicity basically turned off 00:37:32.160 --> 00:37:35.380 and the velocity was allowed to recover a little bit. 00:37:35.380 --> 00:37:40.260 And up through, you know, about the point where the first magma 00:37:40.260 --> 00:37:44.250 eventually reached the surface as lava is where these two time series 00:37:44.250 --> 00:37:48.280 seem to change at the same time and in the same sense. 00:37:48.280 --> 00:37:52.170 So the interpretation here was that both the seismicity and the seismic 00:37:52.170 --> 00:37:56.400 velocity were changing because they were sensing similar things. 00:37:56.400 --> 00:38:00.880 Presumably, this was shallow pressure changes or shallow stress 00:38:00.880 --> 00:38:06.440 as the plug of magma was working its way to the surface. 00:38:06.440 --> 00:38:09.329 So maybe something like that is happening here. 00:38:09.329 --> 00:38:14.109 So what we see at Kilauea is that, when there is high seismicity, 00:38:14.109 --> 00:38:17.880 the velocities are low, which is the same as what we see here. 00:38:17.880 --> 00:38:23.840 Another thought raised along that trail of thinking is looking at 00:38:23.840 --> 00:38:30.030 rock mechanics, actually. So back in the ’70s, it was observed that, 00:38:30.030 --> 00:38:35.660 when you put a rock sample under stress and compressed it and forced it to fail, 00:38:35.660 --> 00:38:38.930 that, even under compression, microcracks opened. 00:38:38.930 --> 00:38:43.340 That there was dilatation in these samples that opened cracks and reduced 00:38:43.340 --> 00:38:48.010 the seismic velocity, leading up to the eventual failure of the sample. 00:38:48.010 --> 00:38:51.069 And this was consistent through several experiments. 00:38:51.069 --> 00:38:56.760 So this is a change in the – this V here is not velocity. 00:38:56.760 --> 00:39:00.540 This is – this is – oh, wait, no, maybe it is velocity. Yes, this is velocity. 00:39:00.540 --> 00:39:02.640 And this is P and S wave velocities. 00:39:02.640 --> 00:39:07.450 That initially it sort of increased, but then decreases by quite a bit. 00:39:07.450 --> 00:39:11.640 And here, this is travel time variations as well that shows a initial increase 00:39:11.640 --> 00:39:15.410 in velocity and then decrease. And it decreases at around the 00:39:15.410 --> 00:39:18.369 time where there begin to be more acoustic events as well 00:39:18.369 --> 00:39:21.900 leading up to the eventual failure of the sample. 00:39:21.900 --> 00:39:26.589 So this was originally thought to be a way to predict earthquakes, 00:39:26.589 --> 00:39:31.090 and that didn’t really pan out so much in real life. 00:39:31.090 --> 00:39:36.480 But maybe this might be applicable to our situation here at Kilauea. 00:39:36.480 --> 00:39:42.190 So basically that we’re stressing the roof of the magmatic system where 00:39:42.190 --> 00:39:45.859 all of these earthquakes are happening, and that’s opening cracks at the same 00:39:45.859 --> 00:39:47.530 time that the earthquakes are happening. 00:39:47.530 --> 00:39:50.720 And then, when the collapse event happens, it resets the system. 00:39:50.720 --> 00:39:54.920 The velocities are allowed to recover before decreasing again. 00:39:55.860 --> 00:39:57.720 All right, so that’s the short-term story. 00:39:57.720 --> 00:40:01.069 So let’s try to figure out what’s going on on the long term. 00:40:01.069 --> 00:40:06.319 So, as I mentioned and pointed out several times before, there’s this sort of 00:40:06.319 --> 00:40:10.900 linear decrease in velocity up until about this point in late June, and then 00:40:10.900 --> 00:40:16.960 the velocity kind of stays roughly constant through the rest of the eruption. 00:40:16.960 --> 00:40:20.710 There are some changes that you can sort of maybe trick your eyes into 00:40:20.710 --> 00:40:25.809 seeing for some of the other time series, but it really isn’t terribly obvious. 00:40:25.809 --> 00:40:29.500 There’s sort of a peak in the RSAM about this time. 00:40:29.500 --> 00:40:33.300 Yeah, there’s some changes in the GPS as well, which we’ll get into. 00:40:33.319 --> 00:40:39.400 We can look back at the – at doing these depth subsets and look at events 00:40:39.400 --> 00:40:41.660 that are shallow versus events that are deep. 00:40:41.660 --> 00:40:45.700 And, as I’m sure you noticed before, that there is a difference in this 00:40:45.710 --> 00:40:49.940 long-term slope if you use shallow events versus using deeper events. 00:40:49.940 --> 00:40:54.369 So, unlike before, where we noticed more amplitude at depth 00:40:54.369 --> 00:40:59.820 for the short term, we actually see more amplitude shallower for the long term. 00:40:59.820 --> 00:41:03.760 So this slope is steeper using only shallow earthquakes. 00:41:03.760 --> 00:41:06.900 So I have to come up with some other mechanism to change the 00:41:06.900 --> 00:41:10.900 seismic velocity that isn’t related, at least directly – or indirectly – 00:41:10.900 --> 00:41:13.589 to the state of the magmatic system at depth. 00:41:13.589 --> 00:41:19.540 And this relationship holds robustly with all of the stations, with, you know, 00:41:19.540 --> 00:41:23.770 maybe one exception, kind of here, which is nice. 00:41:23.770 --> 00:41:26.700 You can actually see this if you plot up individual stations. 00:41:26.700 --> 00:41:31.280 There’s a very obvious difference in the slopes here, which is cool. 00:41:31.280 --> 00:41:34.260 So these changes in the long term are probably much shallower 00:41:34.260 --> 00:41:36.630 than changes in the short term. 00:41:36.630 --> 00:41:42.530 So I think, actually, that the main thing that can tell us about what’s going on 00:41:42.530 --> 00:41:46.030 is actually less where the changes are occurring, though that is important, 00:41:46.030 --> 00:41:50.380 and more the timing of when they – when the behavior changed. 00:41:50.380 --> 00:41:55.490 So, at this point here in late – in late June is when the 00:41:55.490 --> 00:41:59.079 magnitude 5.3 collapse events began to become more widely felt. 00:41:59.079 --> 00:42:02.530 So the magnitudes of these earthquakes stayed roughly constant through the 00:42:02.530 --> 00:42:06.740 62 times or something like that that they happened. 00:42:06.740 --> 00:42:09.700 But they began to become more enriched in higher frequencies and 00:42:09.710 --> 00:42:14.350 therefore were felt by humans that were hanging out either in the towns 00:42:14.350 --> 00:42:21.470 nearby or at – or scientists that were at HVO monitoring things. 00:42:21.470 --> 00:42:24.700 So something about the mechanism by which these 00:42:24.700 --> 00:42:28.460 collapse events changed around this time. 00:42:29.180 --> 00:42:35.820 This event here is where the – a GPS that was sitting on the – 00:42:35.829 --> 00:42:41.640 on the old floor of the caldera that soon began to down-drop, this event is where 00:42:41.640 --> 00:42:46.380 that GPS station drops more than 1 meter in each of these events. 00:42:46.380 --> 00:42:51.390 So here – prior to that, it’s kind of, you know, smoother, slightly 00:42:51.390 --> 00:42:55.620 less extreme changes. This is the first event where it really drops. 00:42:56.260 --> 00:42:59.020 And here, I think, is the most telling one. 00:42:59.020 --> 00:43:07.710 So if you look at frames of photography of the – of the caldera, the two bounding 00:43:07.710 --> 00:43:13.430 this sort of kink is when there is the formation of the ring fault that 00:43:13.430 --> 00:43:18.010 eventually became sort of the outline of that down-dropped block. 00:43:18.010 --> 00:43:20.559 And I think that that’s actually much clearer if you look at this animation 00:43:20.559 --> 00:43:25.450 again, which I’ve added the seismic velocity change on top of it 00:43:25.450 --> 00:43:28.460 in white so that you can sort of track it at the same time. 00:43:28.460 --> 00:43:32.000 So I’m going to narrate here. So leading up to it, we have 00:43:32.000 --> 00:43:35.660 sort of collapse to the west, and velocities are decreasing. 00:43:35.660 --> 00:43:40.510 And, bam, they sort of flatten out, and it’s between sort of these 00:43:40.510 --> 00:43:46.210 two frames that it goes from no – or very little – maybe some hints 00:43:46.210 --> 00:43:50.630 of a ring fault at the surface to the ring fault being at the surface here. 00:43:50.630 --> 00:43:57.059 So my interpretation for what’s causing this initial decrease in velocity is the 00:43:57.059 --> 00:44:01.910 formation of this fault coming up from depth and sort of cracking and 00:44:01.910 --> 00:44:08.200 opening new cracks in the subsurface, forming the outline of this fault. 00:44:08.200 --> 00:44:11.230 And then, once that’s fully formed, it doesn’t need to open any more cracks 00:44:11.230 --> 00:44:17.970 anymore and can just sort of slide down on that formed fault and keep 00:44:17.970 --> 00:44:21.640 the velocity relatively constant through the rest of the eruption. 00:44:21.640 --> 00:44:24.390 So that’s pretty much most of what I wanted to say. 00:44:24.390 --> 00:44:27.410 I’m sure you guys have questions. But I’ll leave you with this photograph 00:44:27.410 --> 00:44:33.750 of the new configuration of Kilauea caldera and the incredible amount 00:44:33.750 --> 00:44:39.270 of cracking and fracturing and just change that occurred last summer. 00:44:39.270 --> 00:44:41.980 And with that, I’d be happy to take questions. 00:44:41.980 --> 00:44:49.000 [Applause] 00:44:50.840 --> 00:44:53.120 - Do you have any questions? 00:44:54.340 --> 00:44:58.120 [Silence] 00:44:58.120 --> 00:45:03.560 - Could you go back to your slide that has A and B and then the path or 00:45:03.560 --> 00:45:08.740 the scattered waves – the coda waves? - Yeah. So Walter’s question is that 00:45:08.740 --> 00:45:13.020 I need to go back to a slide toward the beginning showing sort of the initial … 00:45:13.020 --> 00:45:17.960 - Right. That would be – thank you. - … sort of the illustration of – 00:45:17.960 --> 00:45:20.400 there we go. - Oh, that’s great. 00:45:20.400 --> 00:45:22.089 - Yes. - I have a question. 00:45:22.089 --> 00:45:25.079 - Sure. - How can you tell for sure whether 00:45:25.079 --> 00:45:29.830 the velocity changes or just that those paths change because 00:45:29.830 --> 00:45:34.730 the structure has changed? - So you can tell the difference 00:45:34.730 --> 00:45:38.470 between a change in velocity and a change in scattering. 00:45:38.470 --> 00:45:41.520 So, if you form new cracks, you would expect for the 00:45:41.520 --> 00:45:47.530 scattering properties to change. So if you have a change in scattering, 00:45:47.530 --> 00:45:51.040 it’ll change the cross-correlation coefficient because different arrivals 00:45:51.040 --> 00:45:56.050 are arriving at different times. And they aren’t consistent. 00:45:56.050 --> 00:45:58.589 They aren’t consistently arriving late. 00:45:58.589 --> 00:46:02.130 So, with a change in velocity, they are the same arrivals, 00:46:02.130 --> 00:46:05.790 or very similar arrivals, that are arriving later and later and later. 00:46:05.790 --> 00:46:09.640 So you can tell based on whether or not you can fit it with stretching 00:46:09.640 --> 00:46:14.840 or if it’s just a change in the cross- correlation coefficient of the coda. 00:46:14.840 --> 00:46:20.400 Originally, I had tried to do an add-on project with this looking 00:46:20.400 --> 00:46:23.650 at changes in scattering, but that didn’t end up panning out. 00:46:23.650 --> 00:46:26.750 But you can – you can tell the difference with that. 00:46:26.750 --> 00:46:31.930 - Okay. So if the stretching works, it has to be velocity. 00:46:31.930 --> 00:46:35.119 - That’s one of the fundamental assumptions that’s in this technique, yes. 00:46:35.119 --> 00:46:38.240 - Yeah, okay. Yeah, I’d buy that. - And I have some cutoffs of how 00:46:38.240 --> 00:46:42.180 well it has to fit after you stretch it in order to keep the measurements as well. 00:46:42.180 --> 00:46:45.360 That it has to be a robust measurement. 00:46:48.420 --> 00:46:51.060 - Any additional questions? 00:46:55.140 --> 00:46:58.080 - Nice talk, Alicia. 00:46:58.089 --> 00:47:02.270 One just general suggestion is trying to take a look at a coda Q and see 00:47:02.270 --> 00:47:06.640 how that changes in time. I know people have done that in the past. 00:47:06.640 --> 00:47:08.100 - Yes, absolutely. 00:47:08.100 --> 00:47:10.440 - But can you go back to one of the pictures sort of 00:47:10.440 --> 00:47:13.200 showing the ring fault formation? 00:47:13.200 --> 00:47:16.320 - Sure thing. Toward the end. 00:47:17.800 --> 00:47:21.460 - I’m just – okay, I’m just trying to get a sense of how much – because you 00:47:21.460 --> 00:47:24.190 have some of your stations to the west and some of your stations to the east. 00:47:24.190 --> 00:47:26.069 - Mm-hmm. - And I was just trying to get 00:47:26.069 --> 00:47:28.260 a sense of how much change there really is on the west. 00:47:28.260 --> 00:47:31.020 It looked like there was already a lot of deformation on the west? 00:47:31.020 --> 00:47:34.960 - Yes. So, looking at individual stations and trying to – instead of 00:47:34.960 --> 00:47:38.030 comparing what’s similar between them, comparing what’s different. 00:47:38.030 --> 00:47:42.730 It actually turns out that most of the difference based on azimuth 00:47:42.730 --> 00:47:49.140 is actually – there’s a difference to the north and the south for the – for, I think, 00:47:49.140 --> 00:47:53.109 the short-term velocity changes, that there tends to be slightly 00:47:53.109 --> 00:47:55.460 more amplitude to the north and slightly less than you would 00:47:55.460 --> 00:47:58.190 expect to the south, which is a little bit weird. 00:47:58.190 --> 00:48:03.010 But we – but I actually don’t see that much consistent difference with azimuth 00:48:03.010 --> 00:48:09.490 looking at sort of the slope of this – of this change here. 00:48:09.490 --> 00:48:13.790 Which is a little surprising. - So one other – one other comment is – 00:48:13.790 --> 00:48:16.180 I think this is really interesting, you know, looking at – 00:48:16.180 --> 00:48:20.319 I’ve looked at a lot of these sort of changes for earthquakes. 00:48:20.320 --> 00:48:24.500 And it’s very clear that a lot of the velocity changes you’re seeing in 00:48:24.500 --> 00:48:29.780 the coda are actually local to these individual areas as opposed to 00:48:29.780 --> 00:48:34.020 local to the stations. But if you look at earthquake-induced 00:48:34.020 --> 00:48:36.520 velocity changes, it’s really – the coda is really generated 00:48:36.520 --> 00:48:39.480 more locally to the station. So I think that’s really interesting. 00:48:39.480 --> 00:48:41.640 - Yeah. Yeah. 00:48:43.480 --> 00:48:48.240 [Silence] 00:48:48.240 --> 00:48:51.400 - Any other questions? 00:48:52.320 --> 00:48:56.960 [Silence] 00:48:56.960 --> 00:48:59.779 - Yeah. Could you go to the last slide? 00:48:59.779 --> 00:49:02.010 - The very last slide with the photo? - Yeah. 00:49:02.010 --> 00:49:03.810 Yeah, that’s a very impressive photo there. 00:49:03.810 --> 00:49:09.220 - Here, actually, yeah. - So what is the volume 00:49:09.220 --> 00:49:14.700 of that collapse there? And how well does that compare 00:49:14.700 --> 00:49:19.300 with the volume of extruded lava? - That’s a great question that I don’t 00:49:19.300 --> 00:49:22.080 have the answer to, but there’s someone in the room that I think does. 00:49:22.080 --> 00:49:25.300 And I’ll repeat it – or yeah. Thanks. 00:49:25.300 --> 00:49:27.839 - So, yeah, the question was, what’s the volume of the 00:49:27.839 --> 00:49:30.609 collapsed caldera, and then what was the erupted volume. 00:49:30.609 --> 00:49:35.089 The collapsed caldera was 825 million cubic meters, give or take. 00:49:35.089 --> 00:49:38.319 The erupted volume estimates are still being sorted out, and there’s 00:49:38.319 --> 00:49:41.350 challenges because a lot of the erupted volume actually went into the ocean. 00:49:41.350 --> 00:49:44.700 But generally – and also there’s issues with density, of course, 00:49:44.700 --> 00:49:47.579 in comparing the two. But, to first order, it looks like what 00:49:47.579 --> 00:49:51.150 was erupted seems to be a bit larger than what collapsed at the summit, 00:49:51.150 --> 00:49:56.800 at least those absolute numbers – a bit over a cubic kilometer, but preliminary. 00:49:56.800 --> 00:49:58.300 - Thanks, Kyle. 00:50:01.340 --> 00:50:03.860 - How can they be different? They have to be the same. 00:50:03.860 --> 00:50:06.920 [laughter] 00:50:06.920 --> 00:50:08.340 - You’re wrong again. 00:50:08.340 --> 00:50:10.440 [laughter] 00:50:11.240 --> 00:50:14.480 [multiple inaudible comments] 00:50:15.480 --> 00:50:19.440 - All right. - Okay. Any last questions? 00:50:20.560 --> 00:50:23.100 - Or you can just, you know, talk to me later about it. I don’t care. 00:50:23.100 --> 00:50:25.220 - Okay, well, let’s thank Alicia one more time. 00:50:25.220 --> 00:50:26.180 - Thank you. 00:50:26.180 --> 00:50:31.640 [Applause] 00:50:33.460 --> 00:50:34.520 All right. 00:50:35.160 --> 00:50:45.080 [Silence]