WEBVTT Kind: captions Language: en-US 00:00:01.680 --> 00:00:05.020 [Silence] 00:00:06.160 --> 00:00:09.480 Good morning, everyone. Thank you for coming to our seminar today. 00:00:09.480 --> 00:00:13.680 We just have a few announcements, mostly around the seminars next week. 00:00:13.680 --> 00:00:16.740 We do have a bit of a seminar-a-thon next week. 00:00:16.740 --> 00:00:20.990 Rob Wesson, who is a retired USGS employee, is going to be coming to 00:00:20.990 --> 00:00:25.880 talk about his new book, which is on Darwin’s contribution to Earth sciences. 00:00:25.880 --> 00:00:28.199 So that should be pretty interesting on Monday. 00:00:28.199 --> 00:00:31.640 And then Kang Wang from UC-Berkeley is going to be 00:00:31.640 --> 00:00:36.000 on Wednesday talking about InSAR observation of earthquakes. 00:00:36.000 --> 00:00:38.720 So that’s going to be on Wednesday. But wait, there’s more. 00:00:38.720 --> 00:00:43.350 We have [chuckles] Tom Mitchell on Friday, who is visiting from University 00:00:43.350 --> 00:00:46.870 College London. And he is going to be talking about rock mechanics. 00:00:46.870 --> 00:00:49.410 He’s visiting the rock mechanics lab here. 00:00:49.410 --> 00:00:54.180 So we hope to see you at least one of the seminars next week. 00:00:54.180 --> 00:00:57.860 And Fred Pollitz is going to do the introduction for our speaker today. 00:00:57.860 --> 00:00:59.260 Thank you. 00:01:00.260 --> 00:01:03.520 - Hello. It’s my pleasure to introduce Kathryn Materna. 00:01:03.520 --> 00:01:09.160 She got her undergrad degree at MIT in 2014, advised by Tom Herring. 00:01:09.160 --> 00:01:15.140 She’s a Ph.D. student at UC-Berkeley working with Roland Burgmann. 00:01:15.140 --> 00:01:20.280 And she works on a lot of very interesting things in crustal deformation. 00:01:20.280 --> 00:01:24.300 A lot of wide-ranging interests. One of the things was – 00:01:24.300 --> 00:01:25.860 can you hear me? 00:01:26.780 --> 00:01:27.720 Okay. 00:01:28.660 --> 00:01:31.080 Susan is signaling over there. 00:01:31.080 --> 00:01:35.760 Okay. One of the things is solid Earth deformation from the 00:01:35.760 --> 00:01:38.460 stormwaters of Hurricane Harvey, which is pretty interesting. 00:01:38.460 --> 00:01:41.979 Another thing is studying the source of the deep lower crust 00:01:41.979 --> 00:01:47.020 Botswana earthquake from 2017. And she’s been doing a lot of work 00:01:47.020 --> 00:01:51.620 around the Mendocino Triple Junction, which she’ll talk about today. 00:01:51.620 --> 00:01:54.340 I’ve interacted with Roland’s group for many years, and I’ve known 00:01:54.340 --> 00:01:56.700 Kathryn since she was a first-year student. 00:01:56.700 --> 00:02:01.299 And now, it seems she’s the most senior member of that group, and she’s 00:02:01.300 --> 00:02:04.800 always full of really good ideas. So I look forward to your talk. 00:02:06.660 --> 00:02:08.920 - Okay. Thanks. 00:02:11.980 --> 00:02:15.580 Okay. Good morning, everybody. Thank you for inviting me. 00:02:15.580 --> 00:02:19.440 It’s always really great to come and visit the USGS. 00:02:19.440 --> 00:02:24.400 Can you guys here me? Can the – okay, cool. 00:02:25.360 --> 00:02:28.140 Yeah. It’s great to be here. 00:02:28.140 --> 00:02:30.910 Today I want to talk to you about my research at the 00:02:30.910 --> 00:02:36.319 Mendocino Triple Junction using mostly GPS to investigate 00:02:36.319 --> 00:02:38.840 the styles of crustal deformation there. 00:02:38.840 --> 00:02:45.120 I want to start out by showing a GPS time series to kind of 00:02:45.120 --> 00:02:49.600 motivate the research that we’re doing. 00:02:49.600 --> 00:02:53.720 Here’s a time series of a GPS station in the east component. 00:02:53.720 --> 00:02:56.989 It comes from the Mendocino Triple Junction region. 00:02:56.989 --> 00:03:03.680 And, in particular, the red time series here is one that we’ve corrected for 00:03:03.680 --> 00:03:09.790 earthquake offsets, antenna changes, seasonal, and the normal tectonic trend. 00:03:09.790 --> 00:03:13.360 So basically, we’ve corrected the easy stuff. 00:03:13.360 --> 00:03:17.960 And we expected, at this point, to see a time series that was mostly flat 00:03:17.970 --> 00:03:23.209 with some postseismic – potentially postseismic transients 00:03:23.209 --> 00:03:26.050 because there are – there are earthquakes that 00:03:26.050 --> 00:03:30.439 have happened in this region. And so this is the normal postseismic, 00:03:30.439 --> 00:03:35.680 but after – but beyond that sort of transient, we were expecting to see 00:03:35.680 --> 00:03:40.200 a flat line. And the surprise was that it doesn’t look very flat, and it looks like 00:03:40.200 --> 00:03:45.879 there are systematic variations where sometimes the slope is higher, 00:03:45.880 --> 00:03:47.640 and sometimes the slope is lower. 00:03:47.640 --> 00:03:53.520 And this is the observation that we’re trying to explain. 00:03:53.520 --> 00:03:58.599 This station is one of the stations – one of the GPS stations here in this – 00:03:58.599 --> 00:04:02.569 in the blue arrows at the Mendocino Triple Junction, 00:04:02.569 --> 00:04:08.140 which I think is an important region to study for a few reasons. 00:04:08.140 --> 00:04:11.719 It is our closest subduction zone to here, actually. 00:04:11.719 --> 00:04:15.440 It’s at the southern end of the Cascadia subduction zone and the 00:04:15.440 --> 00:04:19.709 northern end of the San Andreas Fault. There is quite a bit of crustal 00:04:19.709 --> 00:04:26.820 deformation in this region, both on land and out in the ocean. 00:04:26.820 --> 00:04:29.240 And the region hosts more magnitude 6 and 00:04:29.250 --> 00:04:33.660 larger earthquakes than any other part of California. 00:04:33.660 --> 00:04:38.860 The southern end of the Juan de Fuca Plate is called the Gorda Plate, and it is 00:04:38.860 --> 00:04:44.100 internally deforming intensely. As you can – and this results in 00:04:44.100 --> 00:04:48.380 a number of offshore earthquakes that we see these magnitude 6.5 00:04:48.380 --> 00:04:51.510 and larger earthquakes. There’s – they have occurred 00:04:51.510 --> 00:04:56.010 every few years, and there’s two really important ones for this research. 00:04:56.010 --> 00:05:01.000 One is the 2014 earthquake, and one is the 2016 earthquake – 00:05:01.000 --> 00:05:04.160 magnitude 6.8 and magnitude 6.6. 00:05:07.540 --> 00:05:13.470 The picture of this generalized subduction zone in Cascadia – and the 00:05:13.470 --> 00:05:17.480 more specific picture in Cascadia – looks something like this. 00:05:17.480 --> 00:05:25.610 And it’s important to keep this general picture in mind to interpret 00:05:25.610 --> 00:05:30.570 the observations that we make. In the interseismic period, 00:05:30.570 --> 00:05:36.330 the seismogenic portion of the subduction zone is generally locked. 00:05:36.330 --> 00:05:41.690 And it’s – and, as a result of this interseismic locking, there is landward 00:05:41.690 --> 00:05:48.800 motion that we detect in the GPS, basically everywhere on the continent. 00:05:48.800 --> 00:05:54.840 But, as we go deeper into the subduction zone, there’s a transition region that’s 00:05:54.840 --> 00:05:59.520 labeled here as long-term slow-slip events or slow creep, but it’s actually 00:05:59.520 --> 00:06:02.710 pretty unclear what’s going on in this – in this region. 00:06:02.710 --> 00:06:08.310 And then, in Cascadia, as you go deeper, there’s a well-known ETS zone. 00:06:08.310 --> 00:06:10.380 It has episodic tremor and slip. 00:06:10.380 --> 00:06:16.160 So seismically, we record tremor, and geodetically, we observe slip. 00:06:17.130 --> 00:06:24.380 And, in this presentation, all of these zones are important 00:06:24.380 --> 00:06:28.660 for understanding and explaining the GPS data. 00:06:31.340 --> 00:06:35.090 The Mendocino Triple Junction does have some interseismic creep 00:06:35.090 --> 00:06:40.550 on certain interfaces, and this is – I was actually here last year. 00:06:40.550 --> 00:06:46.180 I was lucky enough to give a seminar last year about studying creep on one of 00:06:46.180 --> 00:06:49.610 the interfaces at the Mendocino Triple Junction, specifically the 00:06:49.610 --> 00:06:55.060 Mendocino Fault Zone. But, in this presentation, in terms of 00:06:55.060 --> 00:07:01.960 interfaces to focus on with our data, we’re going to be focusing on this part – 00:07:01.960 --> 00:07:06.540 the lower part of the seismogenic zone and the transition zone of the 00:07:06.540 --> 00:07:10.080 Cascadia subduction zone. And we’re going to look at 00:07:10.080 --> 00:07:13.780 whether there’s time dependence in the coupling there. 00:07:13.780 --> 00:07:23.340 So the GPS data that we are using in this study comes from the PBO and – 00:07:23.340 --> 00:07:28.540 the Plate Boundary Observatory and University of Nevada-Reno solutions. 00:07:28.540 --> 00:07:35.540 Basically, I used whatever solutions – large-scale solutions I was able to find. 00:07:35.540 --> 00:07:40.890 And, in terms of basic processing, we took the time series, 00:07:40.890 --> 00:07:43.870 de-trended them, and removed the offsets from earthquakes 00:07:43.870 --> 00:07:49.340 and antenna changes and any changes in reference frame. 00:07:50.240 --> 00:07:53.340 And we also removed seasonal components, 00:07:53.340 --> 00:07:56.720 which we can see in this station. The east component here 00:07:56.720 --> 00:08:01.120 has a lot of seasonality. And so, in a few ways, which I’ll 00:08:01.120 --> 00:08:05.150 get into a little bit later, we tried to remove this 00:08:05.150 --> 00:08:09.460 seasonal deformation component. And then, once we had a cleaner set 00:08:09.460 --> 00:08:14.210 of time series for each station, we were able – we are trying to 00:08:14.210 --> 00:08:19.200 investigate the effects of the offshore earthquakes on these time series. 00:08:20.220 --> 00:08:25.940 So here’s about a dozen earthquakes – I mean – sorry – 00:08:25.940 --> 00:08:30.100 about a dozen GPS time series. The east component on the left 00:08:30.100 --> 00:08:33.200 and the north component on the right. 00:08:33.960 --> 00:08:38.880 In this kind of messy plot, we’ve included the earthquakes and the 00:08:38.880 --> 00:08:42.719 seasonal components. And then, after we clean up these time series for 00:08:42.720 --> 00:08:46.420 earthquakes and seasonal components, they look something like this. 00:08:47.940 --> 00:08:54.190 This is kind of the original motivating time series stack that we looked at. 00:08:54.190 --> 00:08:57.760 And we thought, wow, there’s something really strange going on. 00:08:58.460 --> 00:09:03.680 So these dashed lines here represent the times of big magnitude 6 or 7 00:09:03.689 --> 00:09:08.329 offshore earthquakes. And this kind of curvature here 00:09:08.329 --> 00:09:11.880 is a postseismic transient that we would expect. 00:09:11.880 --> 00:09:17.520 So, 2005, 2010, 2014, and 2016. This is where the action is. 00:09:17.520 --> 00:09:21.720 There’s something interesting in the time period between the 2014 and 2016 00:09:21.720 --> 00:09:27.020 earthquakes that actually makes it appear in east component there is an 00:09:27.029 --> 00:09:32.329 increase in slope during this time period, which we call T3. 00:09:32.329 --> 00:09:37.870 And then, after the 2016 earthquake, which was December, so it shows up 00:09:37.870 --> 00:09:46.940 as almost 2017, it looks like the slope returns, or goes more westward. 00:09:47.890 --> 00:09:52.400 So this is interesting, and we see it at many stations, and we want 00:09:52.411 --> 00:09:59.019 to understand why. The next step for us was to 00:09:59.019 --> 00:10:06.380 solve for velocity at T3 and velocity at T2 and plot the differences. 00:10:06.380 --> 00:10:11.380 So here on the left is the velocity change – 00:10:11.380 --> 00:10:14.480 T3 minus T2 – at all of the stations. 00:10:14.480 --> 00:10:21.220 We can easily apply these corrections and the procedure to stations anywhere. 00:10:21.220 --> 00:10:25.440 And this is – this was kind of our big surprise. 00:10:25.440 --> 00:10:30.920 We – so a few important things to point out. 00:10:30.920 --> 00:10:36.470 T3 minus T2 – there is a velocity difference, and it is – and it manifests 00:10:36.470 --> 00:10:41.390 as many stations moving more to the east than they were before. 00:10:41.390 --> 00:10:44.379 The scale bar here is 2 millimeters per year. 00:10:44.379 --> 00:10:50.279 So compared to the normal plate motion that we record at this – in this area, 00:10:50.279 --> 00:10:55.310 about 20 or more millimeters per year relative to North America, these are – 00:10:55.310 --> 00:10:57.930 this is about 10% of that velocity. 00:10:57.930 --> 00:11:02.100 So it’s a smaller perturbation, but it’s spatially coherent. 00:11:02.100 --> 00:11:08.040 And this is the – this is the earthquake that is the dividing line 00:11:08.040 --> 00:11:12.660 between T3 and T2. On the right-hand side, 00:11:12.660 --> 00:11:17.940 we have T4 minus T3, and the dividing line here is this 2016 earthquake. 00:11:17.940 --> 00:11:23.060 Magnitude 6.6 and quite far away – about 200 kilometers away 00:11:23.060 --> 00:11:27.319 from the center of this deformation pattern. 00:11:27.319 --> 00:11:33.970 And, in this velocity difference, the stations are moving more to the west by 00:11:33.970 --> 00:11:38.379 about 2 or 3 millimeters per year more than they – than they were before. 00:11:38.380 --> 00:11:46.380 So, to summarize, we have observed in our time series normal postseismic 00:11:46.380 --> 00:11:51.300 transients following two of the earthquakes in the record in 2005 00:11:51.300 --> 00:11:57.879 and 2010, but a different sort of response following 2014 and 2016, 00:11:57.879 --> 00:12:03.249 where, in these cases, there’s an oppositely signed velocity change. 00:12:03.249 --> 00:12:08.249 And it’s kind of small – about 10% of the normal motion 00:12:08.249 --> 00:12:11.960 at these stations. And it’s directed east-west. 00:12:13.910 --> 00:12:20.080 We – so this was actually very interesting, and the 00:12:20.089 --> 00:12:25.910 spatial coherence makes us really curious about what’s going on. 00:12:25.910 --> 00:12:30.559 And just for background, typically what we would expect 00:12:30.559 --> 00:12:34.490 in taking differences between velocities during the 00:12:34.490 --> 00:12:38.380 interseismic period is we would expect almost nothing. 00:12:40.370 --> 00:12:44.940 So we’ve tried a number of – so our first thought was, 00:12:44.949 --> 00:12:49.249 is this an artifact of something? Seasonal corrections, for example. 00:12:49.249 --> 00:12:55.149 So we’re trying here – we have the east component of a particular station. 00:12:55.149 --> 00:12:58.819 This is the uncorrected east component, and we’re trying three different types 00:12:58.819 --> 00:13:03.930 of filtering techniques to remove seasonals and a loading model 00:13:03.930 --> 00:13:11.190 based on the GRACE gravity field. And what comes out of our analysis 00:13:11.190 --> 00:13:15.360 here is that, in all of them, you can – you can remove the seasonals in 00:13:15.360 --> 00:13:24.120 multiple reasonable ways and still resolve a velocity change. 00:13:24.120 --> 00:13:33.230 So we think that overall it’s not the – it’s not just because we’ve removed 00:13:33.230 --> 00:13:36.360 seasonals in a particular way. We’ve tried actually a few more now – 00:13:36.360 --> 00:13:38.309 so five or six ways of removing seasonals, 00:13:38.309 --> 00:13:43.009 and they all pretty much have consistent outcomes. 00:13:43.009 --> 00:13:49.199 We were also interested in resolving from the data, what’s the time of the – 00:13:49.199 --> 00:13:54.720 of the bottom of this V? Does it coincide with the earthquakes? 00:13:54.720 --> 00:14:00.180 So we tried a filtering technique, passed a long-wavelength filter 00:14:00.180 --> 00:14:07.100 through the data, and we resolved the point of maximum curvature, 00:14:07.100 --> 00:14:12.150 or the place where the slope change should be relative to the time 00:14:12.150 --> 00:14:17.800 of the earthquakes. So on these plots, if the symbol is yellow, it means that 00:14:17.800 --> 00:14:21.699 the slope change happened at the same time as the earthquake. 00:14:21.699 --> 00:14:28.100 And because of the filter wavelength that we’re using, which is a year, 00:14:28.100 --> 00:14:34.309 we wouldn’t expect to see – to resolve changes better than a few months. 00:14:34.309 --> 00:14:38.949 But the results are pretty much consistent with the idea that these – 00:14:38.949 --> 00:14:45.540 that the timing of these velocity changes is coincident with March 10th, 2014 – 00:14:45.540 --> 00:14:46.910 in this case, the time of the earthquake, 00:14:46.910 --> 00:14:52.800 and December 2016, the time of the 2016 earthquake. 00:14:52.800 --> 00:15:00.170 So, to take a step back, we have some strange observations 00:15:00.170 --> 00:15:05.629 of velocity changes – interseismic velocity changes in Mendocino. 00:15:05.629 --> 00:15:12.420 And we wanted to understand why, so we came up with a whole list 00:15:12.420 --> 00:15:18.520 of hypotheses. And we’re trying to chase down which ones are most reasonable. 00:15:18.520 --> 00:15:27.460 So our first hypothesis is that the offshore earthquakes – 00:15:27.460 --> 00:15:32.060 perhaps the offshore earthquakes created static stress changes on 00:15:32.060 --> 00:15:36.959 interfaces that are closer to land, and those changed the state of coupling. 00:15:36.959 --> 00:15:44.860 So we performed Coulomb stress calculations to analyze the effect of the 00:15:44.860 --> 00:15:49.149 static stress changes on the clamping of the onshore faults on the Cascadia 00:15:49.149 --> 00:15:53.079 subduction zone in particular, and we found that this was not a very good 00:15:53.079 --> 00:15:56.879 explanation because the Coulomb stress change from the 2014 event was 00:15:56.879 --> 00:16:02.509 actually in the wrong direction to – compared to the – compared to 00:16:02.509 --> 00:16:06.160 the eastward, you know, increase that we see. 00:16:06.160 --> 00:16:08.120 And, in the case of the 2016 earthquake, 00:16:08.120 --> 00:16:13.040 it’s just too far away, so the static stress changes would be very, very small. 00:16:14.610 --> 00:16:19.760 So we also thought about viscoelastic relaxation or afterslip, but these 00:16:19.779 --> 00:16:26.121 hypotheses struggle to explain the data because they – because the 2014 and 00:16:26.121 --> 00:16:35.350 2016 velocity changes, they look linear with time as opposed to a decaying 00:16:35.350 --> 00:16:42.990 exponential or logarithmic curve, so – and secondly, the velocity changes 00:16:42.990 --> 00:16:46.660 actually extend quite a bit farther inland than we would expect from 00:16:46.660 --> 00:16:54.559 viscoelastic relaxation and afterslip. We thought about potentially reference 00:16:54.559 --> 00:17:01.490 frame issues, although, on intuitive bounds – or, on intuitive means, 00:17:01.490 --> 00:17:05.439 it should not be a reference frame issue because we’re taking differences 00:17:05.439 --> 00:17:08.699 of velocities. So it should be independent of reference frame. 00:17:08.699 --> 00:17:13.250 But anything’s possible, so we wanted to check out the data in a lot of different 00:17:13.250 --> 00:17:19.690 reference frames and see if this was visible across broader scales 00:17:19.690 --> 00:17:24.459 or in any of these frames. But we didn’t really see much of a difference. 00:17:24.459 --> 00:17:30.280 So I’ll show you here five different solutions that we’ve tried. 00:17:30.280 --> 00:17:34.880 And the – and this is the signal of interest up here. 00:17:34.880 --> 00:17:41.040 And it’s different between some of these different fields, and I think that this 00:17:41.040 --> 00:17:46.710 solution in particular is pretty noisy, but there’s a lot of eastward motion 00:17:46.710 --> 00:17:51.840 in the 2014 data in pretty much all the reference frames that we tried. 00:17:53.280 --> 00:18:01.220 So the next really important – the next really important piece of data 00:18:01.220 --> 00:18:07.110 to consider – or, hypothesis to consider has to do with hydrological loading. 00:18:07.110 --> 00:18:12.280 And I want to spend the next 10 or 15 minutes talking about 00:18:12.280 --> 00:18:17.390 hydrological loading because it – there’s reason to believe it could 00:18:17.390 --> 00:18:21.400 be part of what we’re seeing. 00:18:21.400 --> 00:18:26.510 In California, we know there was a massive drought between – starting after 00:18:26.510 --> 00:18:35.290 2011 and ending around 2015 or 2016. And so there’s good reason to suspect 00:18:35.290 --> 00:18:39.040 that, on top of the signal that we’re seeing, there is also a hydrological 00:18:39.040 --> 00:18:46.240 loading component. So you may have noticed on the previous plot, 00:18:46.240 --> 00:18:53.750 we expanded the map and did the same analysis for velocity changes for lots of 00:18:53.750 --> 00:18:59.480 places in the western U.S. So we’ve – this is – this is a broader view. 00:18:59.480 --> 00:19:04.780 In 2010, this is a time period where there was very little going on. 00:19:04.790 --> 00:19:09.420 And these velocity changes are more – are smaller and more random. 00:19:09.420 --> 00:19:12.500 So this is, like, a quiet time. 00:19:12.500 --> 00:19:19.900 And, in this map of velocity changes – T3 minus T2 – this is the signal 00:19:19.900 --> 00:19:23.990 of interest surrounding the 2014 earthquake, but there’s a few other 00:19:23.990 --> 00:19:28.780 things to point out about this map. One is that Oregon has very little 00:19:28.780 --> 00:19:32.310 activity in terms of velocity changes. So here the velocities are constant 00:19:32.310 --> 00:19:35.260 across this time window. But, in the Bay Area, 00:19:35.260 --> 00:19:37.670 there’s something – there’s something going on. 00:19:37.670 --> 00:19:41.880 And it’s consistent in its eastward motion as well with Mendocino. 00:19:41.880 --> 00:19:48.880 And, in the 2016 case – T4 minus T3 – again, nothing in Oregon. 00:19:48.880 --> 00:19:54.390 This response here in the – in Mendocino. 00:19:54.390 --> 00:19:59.240 And something else also – a westward motion, but maybe 00:19:59.240 --> 00:20:02.170 a little more complicated in the southern part of California – 00:20:02.170 --> 00:20:06.481 or, the central part of California. And so we thought it’s really 00:20:06.481 --> 00:20:12.660 important to explain what this is. Because without understanding what 00:20:12.660 --> 00:20:16.910 this is, we can’t really understand what’s going on in Mendocino. 00:20:16.910 --> 00:20:23.640 So we performed a couple of corrections for hydrological loading. 00:20:25.500 --> 00:20:31.210 First – and this is – this is a small detail, we did two kind of manual detailed 00:20:31.210 --> 00:20:35.670 corrections for two stations – a station at – there’s a station 00:20:35.670 --> 00:20:37.720 that’s 1 kilometer away from Lake Shasta. 00:20:37.720 --> 00:20:41.570 And there’s a station that’s 1 kilometer away from Lake Oroville. 00:20:41.570 --> 00:20:47.180 And for those particular stations, we performed a specific correction 00:20:47.180 --> 00:20:51.430 for the lake loading effect of those two lakes. So here’s an example. 00:20:51.430 --> 00:20:58.780 This is the lake level history for Shasta – for Lake Shasta and its modeled loading 00:20:58.780 --> 00:21:05.620 effects in the vertical on this station. So this is – this is the mask that I used 00:21:05.620 --> 00:21:10.820 for Lake Shasta, and that’s the station. And the modeled position – 00:21:10.820 --> 00:21:14.160 I’ve copied it down here in red. So the modeled position is red, 00:21:14.160 --> 00:21:19.300 and the GPS station position is black. And they actually – they track 00:21:19.300 --> 00:21:22.880 very well, which tells us that, at these particular stations, 00:21:22.880 --> 00:21:26.240 when you’re only a kilometer or so away from a major reservoir, 00:21:26.240 --> 00:21:30.860 it’s important to consider the loading effects of that reservoir, 00:21:30.860 --> 00:21:34.820 because they impact the station position quite a bit. 00:21:34.820 --> 00:21:40.030 So we used this model as a correction, and we removed it from the time series, 00:21:40.030 --> 00:21:45.530 and – first we thought, okay, well, we’re going to throw away the station by Lake 00:21:45.530 --> 00:21:48.810 Shasta and the station by Lake Oroville. But when we realized how closely 00:21:48.810 --> 00:21:56.510 these models can fit, we used the lake loading prediction as a correction, 00:21:56.510 --> 00:22:01.400 and we recovered those stations. We kept using those stations. 00:22:02.510 --> 00:22:07.020 On a broader scale, for a discussion of hydrology, 00:22:07.030 --> 00:22:11.810 we thought it made sense to predict the station positions 00:22:11.810 --> 00:22:18.710 using GRACE loading and see how they varied. 00:22:18.710 --> 00:22:23.760 So here’s an example of a time series in east, north, and up for a station. 00:22:23.760 --> 00:22:28.190 I’ve actually managed to cut off which station it is, so I’m not quite sure, 00:22:28.190 --> 00:22:31.700 but somewhere in California. You can see the effects of the drought. 00:22:31.700 --> 00:22:39.540 So in the vertical, GRACE is detecting less load due to groundwater, 00:22:39.540 --> 00:22:44.150 soil moistures, snow during the drought. And that’s unloading off the surface 00:22:44.150 --> 00:22:48.200 of the Earth, and the predicted position of the station is going up. 00:22:50.480 --> 00:22:56.080 And similarly, the east and north reflect 00:22:56.080 --> 00:22:59.660 the loading from a distributed mass as well. 00:22:59.660 --> 00:23:08.420 And we take the – we take T3 – you know, T1, T2, and we make 00:23:08.420 --> 00:23:11.880 the subtractions – the same subtractions that we do for the GPS data. 00:23:11.880 --> 00:23:17.950 And the predictions are shown here across California and Oregon, 00:23:17.950 --> 00:23:24.450 so – for the 2014 – T3 minus T2. 00:23:24.450 --> 00:23:27.840 There were a few interesting features. But first I’ll point out is this scale bar 00:23:27.840 --> 00:23:33.570 is very different from the scale bar of the GPS data I was showing you. 00:23:33.570 --> 00:23:36.760 So this is 1 millimeter per year, and these are much smaller 00:23:36.760 --> 00:23:43.100 than 1 millimeter per year. But what we noticed is – 00:23:43.100 --> 00:23:44.720 I’ll start with the Bay Area. 00:23:44.720 --> 00:23:48.140 It’s hard to see, but this velocity change is eastward-directed. 00:23:48.140 --> 00:23:55.840 It’s – which is the same as the data, but it’s smaller by a factor of 10. 00:23:57.240 --> 00:24:02.460 Similarly, a slight eastward and some southward direction here. 00:24:02.460 --> 00:24:07.880 Oregon is different from the data. We saw no velocity changes in the data, 00:24:07.890 --> 00:24:11.070 but certainly velocity change predictions from the hydrological 00:24:11.070 --> 00:24:16.380 loading. And that’s something we don’t entirely understand. 00:24:17.920 --> 00:24:25.280 So the GRACE – the GRACE loading model that we computed, we find that, 00:24:25.280 --> 00:24:30.160 in the area where we’re very interested in understanding hydrology down here, 00:24:30.160 --> 00:24:36.600 it has a similar direction to the data, but it’s too small by a factor of 10. 00:24:38.660 --> 00:24:44.510 There are other loading products that we were considering as well. 00:24:44.510 --> 00:24:51.340 And they are these two loading models called the GLDAS and NLDAS. 00:24:51.340 --> 00:24:58.530 The PBO has a really – a really nice database of loading time series 00:24:58.530 --> 00:25:03.540 that come from these two models. And so we used these as corrections 00:25:03.540 --> 00:25:08.590 as well – an independent set of corrections to see if the slightly 00:25:08.590 --> 00:25:14.611 higher resolution loading product from NLDAS, for example, to see 00:25:14.620 --> 00:25:20.240 if that could explain hydrology in different regions. 00:25:21.730 --> 00:25:28.560 And unfortunately, this slide is a little bit different from the previous one. 00:25:28.560 --> 00:25:34.620 This is not the model from NLDAS. This is the data minus the model. 00:25:34.620 --> 00:25:43.070 So the point is that the systematic eastward and then westward residuals 00:25:43.070 --> 00:25:48.750 that were concerning down here are smaller after correction by NLDAS, 00:25:48.750 --> 00:25:51.340 but not by much. So NLDAS had a very similar – 00:25:51.340 --> 00:25:55.970 has a very similar velocity change pattern compared to GRACE, but – 00:25:55.970 --> 00:26:02.540 a larger amplitude, but it’s still too small to explain all of the data. 00:26:02.540 --> 00:26:07.400 It’s just too small by a factor of 3 instead of a factor of 10. 00:26:10.620 --> 00:26:16.100 And another way to see this is by plotting the GRACE and NLDAS 00:26:16.100 --> 00:26:22.290 and GLDAS models for a particular station against the GPS. 00:26:22.290 --> 00:26:29.260 So the models for a station in Mendocino are these dashed lines here. 00:26:29.260 --> 00:26:34.120 And we can solve for their fit – their best-fitting slopes and do velocity 00:26:34.120 --> 00:26:38.240 differences between these time periods. Their velocity differences 00:26:38.240 --> 00:26:41.500 generally agree in sign with the data but are too small. 00:26:41.500 --> 00:26:45.290 And you can see that because the GPS slopes are here, 00:26:45.290 --> 00:26:50.540 and they’re significantly greater. Like, the variation in actual data 00:26:50.540 --> 00:26:56.580 in the GPS east component is kind of consistent with the – what you’d expect 00:26:56.580 --> 00:27:01.420 from hydrological loading, but is way bigger in Mendocino. 00:27:04.420 --> 00:27:11.200 So we moved on, basically, thinking the hydrological loading is part of the signal 00:27:11.200 --> 00:27:17.740 that we’re seeing, and the drought is certainly something to consider, but we 00:27:17.740 --> 00:27:22.290 can make corrections using NLDAS, GLDAS, or GRACE, and we still 00:27:22.290 --> 00:27:24.580 see a large signal in Mendocino. 00:27:24.580 --> 00:27:32.130 So there’s – so we think that there’s something else going on. 00:27:32.130 --> 00:27:39.710 And the hypothesis that we’re running with right now and considering is that 00:27:39.710 --> 00:27:45.320 the interface – the subduction interface beneath Mendocino, if it changes its 00:27:45.320 --> 00:27:50.190 coupling state, and is sometimes more highly coupled and sometimes more 00:27:50.190 --> 00:27:57.180 weakly coupled, we could explain the eastward motion of the GPS in a – 00:27:57.180 --> 00:28:00.570 at a time when the coupling is higher. And we could explain the westward 00:28:00.570 --> 00:28:04.280 motion of the GPS at a time when the coupling is lower. 00:28:05.650 --> 00:28:10.660 So this is important to consider physically too. 00:28:10.660 --> 00:28:13.380 How could this possibly be happening? 00:28:14.310 --> 00:28:23.560 We took the GPS velocity change data and inverted it for a change in coupling. 00:28:23.560 --> 00:28:29.620 And, in these two plots – this is for the 2014 case, and this is for the 2016 case – 00:28:29.620 --> 00:28:38.340 the colors represent a change in coupling, but it’s important to note 00:28:38.340 --> 00:28:43.290 that the sign is actually different. So, in this case, these coupling 00:28:43.290 --> 00:28:48.540 changes of some number of millimeters per year are increases in coupling. 00:28:48.540 --> 00:28:51.230 They increase the locking on the interface, and that 00:28:51.230 --> 00:28:55.310 makes sense because the velocities are all moving to the east. 00:28:55.310 --> 00:29:01.950 Whereas, in this case, these higher colors represent a decrease in coupling. 00:29:01.950 --> 00:29:06.050 This is similar in concept to a slow-slip event, when we know that, 00:29:06.050 --> 00:29:10.740 when the coupling decreases, the stations move to the west. 00:29:10.740 --> 00:29:16.610 This one is the opposite. It’s kind of like an anti-slow-slip event. 00:29:16.610 --> 00:29:21.480 But in order to explain the data, we have to appeal to both. 00:29:21.480 --> 00:29:25.690 We have to say that in 2014, this is an increase in coupling, 00:29:25.690 --> 00:29:32.050 and this is a decrease in 2016. And so this is – this is a strange result, 00:29:32.050 --> 00:29:37.900 in a sense. And we wanted to understand why. So we looked 00:29:37.900 --> 00:29:40.830 a little bit at the waveforms. I think there’s a lot more work 00:29:40.830 --> 00:29:47.470 that can be done in this – in this area that we haven’t looked into yet. 00:29:47.470 --> 00:29:50.550 But we plotted the waveforms of the different – of these four offshore 00:29:50.550 --> 00:29:56.240 earthquakes, and these are the two that are of interest in the later time period, 00:29:56.240 --> 00:30:01.700 on the same scale, at one of the stations in the Mendocino region. 00:30:01.700 --> 00:30:07.750 And this presents a huge puzzle to us. How could an earthquake offshore, 00:30:07.750 --> 00:30:12.860 at quite a distance, change the coupling on the interface? 00:30:13.360 --> 00:30:17.140 And could it accomplish this dynamically? 00:30:18.040 --> 00:30:22.220 This is a puzzle because the waveforms of these – of these two 00:30:22.230 --> 00:30:27.520 recent earthquakes are kind of more or less similar in amplitude. 00:30:27.520 --> 00:30:32.290 And waveforms – and dynamic shaking involves both positive 00:30:32.290 --> 00:30:37.540 and negative changes in the stress over a time period of seconds. 00:30:37.540 --> 00:30:41.660 So how can it be that a similar amount of shaking with both positive lobes 00:30:41.660 --> 00:30:48.020 and negative lobes resulted in a change that was oppositely signed? 00:30:48.020 --> 00:30:52.020 An increase in 2014 and a decrease in 2016. 00:30:52.020 --> 00:30:58.360 So this is a big puzzle that’s been on our minds for many months now. 00:30:58.360 --> 00:31:04.960 We wanted to look at other data sets that exist. 00:31:04.960 --> 00:31:08.280 And maybe that could give us some sense of what’s going on. 00:31:08.280 --> 00:31:16.820 So we spent some time playing around with the tremor catalog in Cascadia. 00:31:16.820 --> 00:31:23.550 So, for those of you who have used tremor data before, we all know that 00:31:23.550 --> 00:31:26.870 basically it comes from Aaron Wech’s website. 00:31:26.870 --> 00:31:32.310 And that that is the – and this is basically the Aaron Wech catalog 00:31:32.310 --> 00:31:35.150 that we have downloaded and plotted, and we’re interested 00:31:35.150 --> 00:31:37.300 in looking at its evolution in time and space. 00:31:37.300 --> 00:31:42.190 And the coupling change patch that I showed on the inversion results 00:31:42.190 --> 00:31:49.320 is located about here at the up-dip limit of the tremor. 00:31:50.380 --> 00:31:54.240 And we think that that’s interesting and related. 00:31:54.240 --> 00:31:58.380 So we went to Aaron Wech’s catalog. We downloaded off the – 00:31:58.380 --> 00:32:05.640 off this very amazing website here. And this is the catalog that we got. 00:32:05.640 --> 00:32:09.261 Just by downloading – you can plot tremor catalogs in a – 00:32:09.261 --> 00:32:13.070 in a variety of ways, but this is time versus latitude. 00:32:13.070 --> 00:32:16.750 So the Oregon-California border is right here in the middle. 00:32:16.750 --> 00:32:18.640 And this is southern Cascadia. 00:32:18.640 --> 00:32:23.180 The Mendocino Triple Junction is right around here, and it’s very active. 00:32:23.180 --> 00:32:27.830 So, over time, it’s kind of constantly rumbling off tremor. 00:32:27.830 --> 00:32:34.050 And I’ve marked the time periods. 00:32:34.050 --> 00:32:40.450 So we noticed a few things that were a little bit strange about this. 00:32:40.450 --> 00:32:44.491 Well, first, obviously, there’s this big gap earlier in the tremor, which 00:32:44.491 --> 00:32:48.270 means we can’t use this time period. And we talked with Aaron, and this is 00:32:48.270 --> 00:32:51.900 very clearly the data just was not ingested for this region 00:32:51.900 --> 00:32:55.360 of space and time. So that’s not that big of a deal. 00:32:55.360 --> 00:32:59.390 The more concerning question that we were interested in is we noticed that 00:32:59.390 --> 00:33:04.680 there’s a change in the character of the tremor at this – from here. 00:33:04.680 --> 00:33:07.100 And then, at some point in 2016, it looks like something changes. 00:33:07.100 --> 00:33:10.430 And then it’s very quiet after that. But the rumbling that’s constantly 00:33:10.430 --> 00:33:14.050 happening doesn’t happen very much anymore. 00:33:14.050 --> 00:33:21.310 So we asked Aaron to look into it and see what might be going on. 00:33:21.310 --> 00:33:31.730 And he discovered that an instrumental problem right around the middle of 2016 00:33:31.730 --> 00:33:36.960 reduced the detection capability. So there was tremor going on here, 00:33:36.960 --> 00:33:41.610 it just was not detected. And the instrumental problem 00:33:41.610 --> 00:33:45.430 had to do with five – about five stations here. 00:33:45.430 --> 00:33:49.900 They’re the ones that are clear – that don’t have any data in them. 00:33:50.980 --> 00:33:56.440 Sending more or less flat data instead of actual data. 00:33:56.440 --> 00:34:00.540 And they were being sent into the detection scheme. 00:34:00.540 --> 00:34:06.520 And, at the same time, many of these stations had telemetry problems as well 00:34:06.520 --> 00:34:12.169 due to the fact that the stations will all – they send multiple stations’ worth of 00:34:12.169 --> 00:34:18.769 data to a single analog telemetry device, as far as I understand it. 00:34:18.769 --> 00:34:24.060 And occasionally, spikes that affect the analog telemetry device end up 00:34:24.060 --> 00:34:27.180 at all of these four stations. If they’re correlated spikes at four 00:34:27.180 --> 00:34:31.180 stations, they sometimes look like tremor, but it’s not. 00:34:31.180 --> 00:34:39.619 So, in a – so Aaron helped us address these two things by re-running a period 00:34:39.619 --> 00:34:46.429 of time in his catalog, where his detection of 00:34:46.429 --> 00:34:51.659 analog telemetry issues is a little bit better, and where 00:34:51.659 --> 00:34:57.380 he’s more aware of the stations that might be going offline. 00:34:57.380 --> 00:35:00.180 And so he produced a new catalog for us, which was actually 00:35:00.180 --> 00:35:05.069 very, very helpful. Here is the old catalog with the new catalog 00:35:05.069 --> 00:35:09.249 pasted on top, basically. We have removed what was in this time 00:35:09.249 --> 00:35:13.900 period and put in Aaron’s new catalog. And the rumbling tremor that we see 00:35:13.900 --> 00:35:22.220 in the Mendocino region, it may get quieter in a way that’s actually – 00:35:22.220 --> 00:35:27.910 that is, in fact, a real effect. But it doesn’t get quieter by quite 00:35:28.000 --> 00:35:32.780 as much because some of this tremor was not detected before. 00:35:34.100 --> 00:35:37.480 So we take this newer – this new tremor catalog – 00:35:37.480 --> 00:35:41.599 a combined catalog – and we plot it here. 00:35:41.599 --> 00:35:46.549 And we’re particularly interested in the way the tremor evolves in space and 00:35:46.549 --> 00:35:52.960 time and how that might result – how that might relate to our coupling change. 00:35:53.869 --> 00:36:01.760 So I have taken the tremor epicenters and associated each epicenter, 00:36:01.760 --> 00:36:08.500 which is just latitude and longitude, with a depth on the McCrory 2012 00:36:08.500 --> 00:36:14.100 plate interface – the nearest depth. And so now, in addition to latitude 00:36:14.109 --> 00:36:19.060 and longitude boxes, we can study the tremor based on depth ranges too. 00:36:19.060 --> 00:36:25.160 And, in this plot, I’ve decided on depth ranges, such as here – 00:36:25.160 --> 00:36:28.880 24 to 35 kilometers. That’s this purple zone. 00:36:28.880 --> 00:36:32.999 And then 35 kilometers and deeper is this orange zone. 00:36:32.999 --> 00:36:37.400 And I can filter out tremor just in those boxes 00:36:37.400 --> 00:36:42.750 and analyze how it occurs over time. 00:36:42.750 --> 00:36:51.510 The reason I focused on this deeper – on the 24 to 35 kilometers is because 00:36:51.510 --> 00:36:57.089 I wanted to isolate the region of the interface that is active 00:36:57.089 --> 00:37:02.470 only during ETS events and not very active in the background. 00:37:02.470 --> 00:37:13.220 So one of the features of the catalog that’s true of tremor regions around 00:37:13.220 --> 00:37:18.569 Cascadia is that the deeper – the deeper region is more continuous, 00:37:18.569 --> 00:37:20.940 and the shallower region is more episodic. 00:37:20.940 --> 00:37:25.920 And we see that here too, where the deeper portion 00:37:25.920 --> 00:37:28.880 of the interface is active at many times. 00:37:28.880 --> 00:37:33.099 And the shallower portion of the interface is more stair-like. 00:37:33.099 --> 00:37:40.549 The tremor rate per day in the purple zone is given here. 00:37:40.549 --> 00:37:46.039 And this is – this is kind of an illustrative plot to show 00:37:46.039 --> 00:37:49.940 the timing intervals. We can see that a major ETS event 00:37:49.940 --> 00:37:52.890 happens in this region about every seven months. 00:37:52.890 --> 00:37:56.250 And that interval is actually pretty fixed. It doesn’t change very much. 00:37:56.250 --> 00:38:04.599 Six to eight months, maybe, is the range. But it’s pretty well-regulated 00:38:04.599 --> 00:38:10.020 for some reason. So, based on this information, 00:38:10.020 --> 00:38:16.740 we came up with two potential connections, or hypotheses, between 00:38:16.740 --> 00:38:21.630 connecting a tremor catalog with coupling changes on the interface. 00:38:21.630 --> 00:38:29.230 The first hypothesis is that potentially the – if the location 00:38:29.230 --> 00:38:34.180 of tremor, or the dominant location of tremor between one ETS event 00:38:34.180 --> 00:38:39.160 and the next ETS event varies – if it moves up-dip or down-dip, 00:38:39.160 --> 00:38:44.740 potentially that could look like a coupling change to the 00:38:44.740 --> 00:38:47.970 untrained eye of the GPS. It could look like a coupling change. 00:38:47.970 --> 00:38:51.999 So we wanted to investigate whether one ETS event or 00:38:52.000 --> 00:38:56.120 the next had different evolution in space. 00:38:56.120 --> 00:39:01.440 And the second hypothesis that could relate tremor to the occurrence 00:39:01.450 --> 00:39:06.730 of coupling changes is that maybe the physical conditions on the interface 00:39:06.730 --> 00:39:11.250 might change between early in an ETS cycle, during an ETS event, 00:39:11.250 --> 00:39:14.999 or late in the ETS cycle. So perhaps the time since 00:39:14.999 --> 00:39:21.410 the last ETS event is an important piece of information 00:39:21.410 --> 00:39:24.980 to know about and maybe affects the physical conditions. 00:39:24.980 --> 00:39:29.609 So Hypothesis 1 we could rule out pretty fast, actually. 00:39:29.609 --> 00:39:34.210 We separated, say, eight different ETS events, 00:39:34.210 --> 00:39:36.340 plotted the average depth of their tremor. 00:39:36.340 --> 00:39:38.710 This pink one is a – is a weird ETS event 00:39:38.710 --> 00:39:41.269 that I’ll show you on the next slide. 00:39:41.269 --> 00:39:44.539 And all of the other ones were within 5 kilometers. 00:39:44.539 --> 00:39:50.109 And I tried many boxes and definitions of regions, 00:39:50.109 --> 00:39:52.680 and basically there was nothing systematically different 00:39:52.680 --> 00:39:57.960 about one ETS event in terms of its depth compared to the next. 00:39:59.980 --> 00:40:02.630 The eight ETS events that we chose – that we were 00:40:02.630 --> 00:40:07.000 looking at are kind of laid out here. 00:40:07.000 --> 00:40:09.390 We were also thinking about, maybe, what if one ETS event was 00:40:09.390 --> 00:40:11.789 north-propagating and one was south-propagating? 00:40:11.789 --> 00:40:19.900 Could that result in a different state on the interface? But probably not. 00:40:19.900 --> 00:40:23.920 There’s – this is – this is the important one. 00:40:23.930 --> 00:40:30.690 And ETS event that happened only two months – or, for some of the area, 00:40:30.690 --> 00:40:34.990 only two weeks before the 2014 earthquake. 00:40:34.990 --> 00:40:38.249 So we note from this part of the time history that the 00:40:38.249 --> 00:40:43.369 2014 earthquake was early in the ETS cycle. 00:40:43.369 --> 00:40:48.780 And here’s a case – this is the ETS event before the 2016 earthquake. 00:40:48.780 --> 00:40:53.820 The 2016 earthquake was about four months after this. 00:40:53.829 --> 00:40:56.849 So we were about in the middle of an ETS cycle when this 00:40:56.849 --> 00:41:00.940 earthquake happened. However, this ETS was actually quite 00:41:00.940 --> 00:41:06.580 weak in the area that we are researching. It was – it’s a major ETS event 00:41:06.580 --> 00:41:10.710 that went all the way into Oregon in the north, but in the area of – 00:41:10.710 --> 00:41:15.559 in our area of interest, it was actually – it did not produce a lot of tremor. 00:41:15.559 --> 00:41:22.329 So to summarize that – and we can see it from this plot here too. 00:41:22.329 --> 00:41:26.299 The 2014 earthquake occurred early in an ETS cycle, 00:41:26.299 --> 00:41:28.940 while the 2016 earthquake occurred in the middle of 00:41:28.940 --> 00:41:33.980 an ETS cycle, but the previous cycle was actually quite weak. 00:41:37.160 --> 00:41:44.240 We have – so this leads us to think, physically, on the interface, 00:41:44.249 --> 00:41:47.950 what needs to happen in order to achieve a coupling change? 00:41:47.950 --> 00:41:55.470 And we’ve thought of a few ways that – two major ways 00:41:55.470 --> 00:41:59.799 that you could change the coupling on the interface. 00:41:59.799 --> 00:42:03.099 The question of whether these can happen dynamically or not 00:42:03.099 --> 00:42:09.080 is a separate question. But it’s easier to see – to just lay out 00:42:09.080 --> 00:42:12.079 in a schematic way that we could either – we could decrease 00:42:12.079 --> 00:42:15.910 coupling on an interface by either increasing the fluid pressure in that 00:42:15.910 --> 00:42:21.049 fault zone, which would decrease the normal stress, clamping the fault, 00:42:21.049 --> 00:42:27.430 or we could decrease coupling by somehow – if it’s a – if it’s a 00:42:27.430 --> 00:42:33.859 A-minus-B-positive – a fault that likes to slip, to creep at some velocity 00:42:33.859 --> 00:42:37.220 interseismically, somehow we could increase that steady-state velocity, 00:42:37.220 --> 00:42:40.600 and that would also look like a coupling change. 00:42:43.269 --> 00:42:46.499 To achieve the opposite, increasing coupling, we would need to 00:42:46.499 --> 00:42:49.910 decrease fluid pressure in the fault zone, which increases the normal stress and 00:42:49.910 --> 00:42:54.010 clamps the fault, or we would need to decrease the steady-state velocity. 00:42:54.010 --> 00:42:58.210 And somehow, at Mendocino, we actually need to do both. 00:42:58.210 --> 00:43:04.640 Because one of these – one of the earthquakes resulted 00:43:04.640 --> 00:43:08.770 in a coupling increase, and one resulted in a coupling decrease. 00:43:08.770 --> 00:43:16.569 So our schematic interpretation of what’s going on is kind of based on 00:43:16.569 --> 00:43:22.420 an interesting presentation of the – an interesting presentation 00:43:22.420 --> 00:43:27.859 of the ETS – of physical conditions in the ETS zone. 00:43:27.859 --> 00:43:33.769 In this paper here by Pascal Audet and Roland Burgmann, there’s a 00:43:33.769 --> 00:43:38.489 presentation of what’s happening in the ETS zone where, throughout the ETS 00:43:38.489 --> 00:43:44.999 cycle, dehydration reaction of fluids at this part of the interface results in a – 00:43:44.999 --> 00:43:50.569 in a time-dependent change in effective normal stress due to 00:43:50.569 --> 00:43:54.970 an overall increase in pressure in the – in the fault zone through time 00:43:54.970 --> 00:43:58.579 and a release of that pressure during a slip event. 00:43:58.579 --> 00:44:06.420 So we have kind of expanded this schematic idea to interpret what could 00:44:06.420 --> 00:44:13.890 have happened in the 2014 and 2016 earthquakes where our cartoon 00:44:13.890 --> 00:44:16.910 looks something like this. The key elements of the cartoon 00:44:16.910 --> 00:44:23.670 are an ETS zone at a reasonably deep part of the plate interface. 00:44:23.670 --> 00:44:29.800 And just up-dip of an ETS zone is a hydraulically sealed low-permeability 00:44:29.800 --> 00:44:38.330 region, which is thought to exist to seal the high pressures that are likely present 00:44:38.330 --> 00:44:45.760 in the ETS zone from the other, you know, more generic parts of the crust. 00:44:45.760 --> 00:44:54.080 And the two basic ingredients are a time-varying pressure in the ETS zone 00:44:54.080 --> 00:45:02.900 and a hydraulic seal that remains closed during normal times but opens, perhaps 00:45:02.901 --> 00:45:07.630 due to increased fracture permeability, during dynamic shaking. 00:45:07.630 --> 00:45:11.369 And if you have these two ingredients in the – in the fault zone, you can 00:45:11.369 --> 00:45:15.410 make a cartoon picture where, early in the ETS cycle, 00:45:15.410 --> 00:45:22.430 if there’s low pressure here, and it’s low relative to pressure just outside, 00:45:22.430 --> 00:45:28.640 if this hydraulic seal opens, pressure-driven flow could, in fact, 00:45:28.640 --> 00:45:31.960 move high-pressure fluids into the lower ETS zone 00:45:31.960 --> 00:45:34.210 and increase the coupling just up-dip. 00:45:34.210 --> 00:45:43.100 Which is consistent with the picture that we see in the 2014 earthquake. 00:45:43.100 --> 00:45:47.210 In the 2016 case, the schematic is slightly different because the pressure 00:45:47.210 --> 00:45:52.700 was higher, potentially, at the time of the – of the earthquake, which means 00:45:52.700 --> 00:45:58.780 that, if that seal is broken, pressure-driven flow could, in fact, 00:45:58.780 --> 00:46:04.650 go in the opposite direction, increasing the fluid pressure, 00:46:04.650 --> 00:46:09.279 decreasing the normal stress, and allowing for more sliding 00:46:09.279 --> 00:46:12.400 just up-dip of the ETS zone, which is, again, consistent with 00:46:12.400 --> 00:46:16.980 the GPS – with the direction of the GPS motion. 00:46:18.160 --> 00:46:32.200 But this is – this is a highly – this is speculative – highly speculative. 00:46:32.200 --> 00:46:36.009 But it is important to think about these things because there are examples – 00:46:36.009 --> 00:46:39.579 they’re rare, but there are a few examples that we’ve found, 00:46:39.579 --> 00:46:43.880 and we’d love to learn about more, elsewhere in the world where coupling 00:46:43.880 --> 00:46:48.109 changed, perhaps in a dynamically triggered way, 00:46:48.109 --> 00:46:53.910 and we don’t really understand why. So I have a few examples that are – 00:46:53.910 --> 00:46:57.839 that I think are interesting. One is a coupling decrease that 00:46:57.840 --> 00:47:01.940 was apparently observed in Chile. And the – following the 00:47:01.940 --> 00:47:07.599 Tocopilla earthquake in 2005. This is – this is the most important line. 00:47:07.599 --> 00:47:11.859 This is the de-trended GPS at a east position of a station. 00:47:11.860 --> 00:47:16.000 And after the earthquake, it obtained a new velocity. 00:47:16.989 --> 00:47:19.840 And this is interpreted to be a decrease in coupling on the 00:47:19.840 --> 00:47:25.930 interface that occurred as a – following – immediately following 00:47:25.930 --> 00:47:28.940 the 2005 earthquake. 00:47:29.740 --> 00:47:35.880 Chile continues to present more challenges for us, actually. 00:47:35.880 --> 00:47:39.599 Because the second example I have is also in Chile, but in a – 00:47:39.599 --> 00:47:46.329 in a few degrees farther south. We actually have recent observations 00:47:46.329 --> 00:47:54.369 that are very similar to Mendocino and also are very difficult to explain. 00:47:54.369 --> 00:47:59.930 There are GPS velocities inland here that moved up to 10 millimeters a year 00:47:59.930 --> 00:48:05.289 faster towards the east than they did before the Iquique earthquake, 00:48:05.289 --> 00:48:09.079 which was actually up here. So the Iquique earthquake was 00:48:09.079 --> 00:48:12.970 several hundred kilometers away, and to the south, even after correcting 00:48:12.970 --> 00:48:19.119 for any small postseismic that they had identified, it looks like the coupling 00:48:19.120 --> 00:48:24.280 somehow increased on this part of the interface, driving everything to the east. 00:48:25.460 --> 00:48:34.539 And the final example is actually in Sumatra, where, in the – where geodetic 00:48:34.539 --> 00:48:41.119 data is often somewhat limited. But there is enough data from some 00:48:41.119 --> 00:48:48.089 islands – forearc islands that suggest that before 2000, there was a certain 00:48:48.089 --> 00:48:54.390 high coupling state in the southern part of this figure here – Enggano Island. 00:48:54.390 --> 00:49:01.740 And in – and after 2001, the coupling state was – it decreased coupling. 00:49:01.740 --> 00:49:06.400 And that’s the interpretation of the change in velocity on this – 00:49:06.400 --> 00:49:08.119 in this station. 00:49:08.119 --> 00:49:13.410 So there are – and it may have been dynamically triggered as well. 00:49:13.410 --> 00:49:19.849 So there are a few examples – and I think the more we dig, 00:49:19.849 --> 00:49:25.450 the more we may find, that the – that the coupling state 00:49:25.450 --> 00:49:31.280 of faults may be more time-dependent than we realize. 00:49:32.630 --> 00:49:38.180 So this – my implication slide is actually mostly just a series 00:49:38.180 --> 00:49:47.250 of open questions that these data kind of bring to bear. 00:49:47.250 --> 00:49:50.589 One is, under what conditions can coupling actually change 00:49:50.589 --> 00:49:54.079 due to dynamic shaking? There’s even – there are examples 00:49:54.079 --> 00:49:58.600 in the Bōsō Peninsula of Japan, where coupling changed in an 00:49:58.600 --> 00:50:00.599 even more mystifying way. 00:50:00.599 --> 00:50:03.999 It changed after slow-slip events where there’s no dynamic shaking. 00:50:03.999 --> 00:50:07.799 So under what conditions can coupling change due to 00:50:07.800 --> 00:50:11.380 dynamic shaking, or even independent of dynamic shaking? 00:50:12.300 --> 00:50:15.160 And does Cascadia, in particular, experience long-term 00:50:15.170 --> 00:50:18.599 coupling variations? We’re very familiar with the ETS 00:50:18.599 --> 00:50:22.859 coupling variations that happen on the time scale of several weeks. 00:50:22.859 --> 00:50:26.200 And they happen months apart. But, on a longer term, are there also 00:50:26.200 --> 00:50:33.059 variations that we need to be aware of? And outside of Cascadia, what other 00:50:33.059 --> 00:50:36.359 regions can experience these types of changes? 00:50:36.359 --> 00:50:41.930 And the implications include – obvious implications include 00:50:41.930 --> 00:50:47.279 time-dependent hazard assessments. Over time, can an interface 00:50:47.279 --> 00:50:51.739 accumulate more strain or less strain than we expected? 00:50:51.739 --> 00:50:58.869 And so, to summarize, I’ll leave this picture up and just lay it out that we 00:50:58.869 --> 00:51:03.540 observe, in Mendocino, velocity changes of about 10% of the regular 00:51:03.540 --> 00:51:06.509 velocity following offshore earthquakes that were 00:51:06.509 --> 00:51:14.799 distant enough that it’s not a static stress type of interaction. 00:51:14.799 --> 00:51:19.869 We think that these velocity changes can be most completely explained 00:51:19.869 --> 00:51:24.290 by a coupling change on the subduction interface. 00:51:24.290 --> 00:51:30.029 And we present a possible mechanism involving the flow of – the pressure- 00:51:30.029 --> 00:51:37.720 driven flow of fluids near the ETS zone to move coupling up-dip and down-dip. 00:51:37.720 --> 00:51:40.840 And the results suggest that perhaps fault coupling in southern 00:51:40.840 --> 00:51:45.020 Cascadia may be time-dependent. Thank you. 00:51:45.020 --> 00:51:51.620 [Applause] 00:51:51.620 --> 00:51:55.280 - All right. Does anyone have any questions for Kathryn? 00:51:57.280 --> 00:52:05.940 [Silence] 00:52:05.940 --> 00:52:10.900 - Really interesting presentation. I guess I want to sort of ask you 00:52:10.900 --> 00:52:14.540 to speculate a little bit about your first question. 00:52:14.540 --> 00:52:20.650 And I guess my broader question is, is why don’t those other two earthquakes 00:52:20.650 --> 00:52:25.140 in the area cause this sort of behavior? Because I would expect that they 00:52:25.140 --> 00:52:27.450 occurred sometime during the ETS cycle, and so they 00:52:27.450 --> 00:52:32.229 should have some sort of effect. And a more broad question would be, 00:52:32.229 --> 00:52:34.720 why isn’t there any other dynamic triggering effect? 00:52:34.720 --> 00:52:38.930 I’m sure that other large earthquakes that are, you know, much more distant, 00:52:38.930 --> 00:52:43.849 but should have similar amplitudes in the same frequency range, 00:52:43.849 --> 00:52:48.690 why wouldn’t that have an effect? - That’s a really good question that 00:52:48.690 --> 00:52:51.400 we’re still kind of working out. We’ve been looking at a lot of the – 00:52:51.400 --> 00:52:54.970 we’ve been looking at the catalog somewhat to see, maybe the 00:52:54.970 --> 00:52:59.249 2014 earthquake – maybe it wasn’t the local one. 00:52:59.249 --> 00:53:03.979 Maybe it was Iquique, which happened within the – within a few weeks. 00:53:03.979 --> 00:53:10.900 The 2016 change, that occurred at a time when globally, there weren’t 00:53:10.900 --> 00:53:18.180 very many, like, magnitude 8s or something like that that could cause it. 00:53:18.180 --> 00:53:22.650 But that’s a really good question about the 2005 and 2010. 00:53:22.650 --> 00:53:28.779 I think it’s possible that the 2010 event may have just been smaller – 6.5 – 00:53:28.780 --> 00:53:31.680 and it seemed to not have very much surface wave energy. 00:53:31.680 --> 00:53:39.520 But, in terms of what frequency range is most effective and what 00:53:39.520 --> 00:53:43.260 duration of shaking, for example, we really don’t know yet. 00:53:44.569 --> 00:53:47.339 - But what makes those two so special? - These two? 00:53:47.340 --> 00:53:50.380 - Yeah. - The latest two. 00:53:51.120 --> 00:53:59.320 Perhaps there were velocity changes. So we’re – the 2005 earthquake 00:53:59.329 --> 00:54:04.799 may have had velocity changes too. But there were only about five stations 00:54:04.799 --> 00:54:07.650 installed a few months before that earthquake happened. 00:54:07.650 --> 00:54:11.859 And when we look at their time series, it’s suggested that there may have been. 00:54:11.860 --> 00:54:15.760 And the 2010 earthquake, I think, may have been too small. 00:54:18.520 --> 00:54:22.360 - Two quick questions. Is there episodic tremor and 00:54:22.369 --> 00:54:25.479 slip down in the Chile examples that you were showing? 00:54:25.479 --> 00:54:28.660 And might that – you know, how does relate its characteristics 00:54:28.660 --> 00:54:30.720 to what’s happening in Cascadia? 00:54:30.720 --> 00:54:33.779 And the other question was, what about the Bay Area? 00:54:33.779 --> 00:54:37.769 You showed that hydrologic loading didn’t really, you know, explain the 00:54:37.769 --> 00:54:42.089 magnitude of the arrows up in the Mendocino area in the Bay Area. 00:54:42.089 --> 00:54:47.539 Any speculation on what might make similar-magnitude changes 00:54:47.539 --> 00:54:56.450 in motion down here? - Yeah. Chile is kind of an anomaly 00:54:56.450 --> 00:55:02.740 of subduction zone because it doesn’t have tremor, as best as we know. 00:55:02.740 --> 00:55:09.519 It may have some longer-term slow slip or variations even beyond the two that 00:55:09.519 --> 00:55:15.960 I showed, but we haven’t detected any ETS activity. 00:55:15.960 --> 00:55:18.970 Which means that perhaps what’s going on there is different – 00:55:18.970 --> 00:55:23.720 just, like, qualitatively, a different situation. 00:55:24.860 --> 00:55:28.520 And your section question – oh, in the Bay Area. Yeah. 00:55:28.530 --> 00:55:35.749 So what we’re working on now is taking the hydrologic loading 00:55:35.749 --> 00:55:42.670 models and scaling them up. Because it’s possible that they’re – 00:55:42.670 --> 00:55:45.750 in their overall characteristics, the loading models are correct, 00:55:45.750 --> 00:55:50.440 but in the local site response, they could be off by some factor. 00:55:50.440 --> 00:55:57.190 So our latest results that didn’t end up here were, station by station, 00:55:57.190 --> 00:56:05.339 increasing the amplitude of the loading correction until it matches the amplitude 00:56:05.339 --> 00:56:09.719 of the GPS annual signal and then using that as a correction. 00:56:09.719 --> 00:56:14.690 And what it showed us, just quite recently, is that, if we have that extra 00:56:14.690 --> 00:56:18.489 degree of freedom, we can make the – we can explain 00:56:18.489 --> 00:56:25.240 the Bay Area residuals pretty well. And the Mendocino effect is still there. 00:56:26.140 --> 00:56:28.060 - [inaudible] 00:56:30.240 --> 00:56:32.420 - Any other questions? 00:56:34.080 --> 00:56:38.000 [Silence] 00:56:38.000 --> 00:56:40.900 - Yeah. Thanks again for a really interesting, thorough talk. 00:56:40.910 --> 00:56:45.920 I was curious what the sensitivities are with respect to that loading calculation. 00:56:45.920 --> 00:56:49.440 I think you said you used the PREM velocity model. 00:56:49.440 --> 00:56:52.680 - Mm-hmm. - What happens if you allow for 00:56:52.690 --> 00:56:56.980 shallow structure that deviates from that significantly or something? 00:56:56.980 --> 00:56:59.930 I’m not sure what the key parameters are there. 00:56:59.930 --> 00:57:05.099 - Mm-hmm. Yeah. We haven’t tried forward-calculating 00:57:05.099 --> 00:57:12.220 the load on another structure. But there are groups in Europe who are 00:57:12.220 --> 00:57:16.869 doing this, and they suggest, actually, that it may affect horizontal. 00:57:16.869 --> 00:57:19.390 So it’s something that’s definitely worth looking into. 00:57:19.390 --> 00:57:22.880 It affects horizontals and phase, actually. 00:57:24.780 --> 00:57:28.100 [Silence] 00:57:28.100 --> 00:57:30.200 - Any other questions? 00:57:32.380 --> 00:57:37.840 [Silence] 00:57:37.840 --> 00:57:42.320 - Yeah. The hypothesis with the coupling change due to the – 00:57:42.329 --> 00:57:46.099 you know, this fault valving behavior where the permeability 00:57:46.099 --> 00:57:50.390 changes and the water escapes somehow, I think, is common for 00:57:50.390 --> 00:57:53.190 reverse faults and stuff like this. But these earthquakes are on 00:57:53.190 --> 00:57:57.489 those transform faults offshore. So it – are you saying that maybe 00:57:57.489 --> 00:58:01.390 the permeability change and this dynamic triggering happens just 00:58:01.390 --> 00:58:04.029 due to the shaking from that? And is that – do you think that 00:58:04.029 --> 00:58:08.290 shaking is strong enough at the depth of the ETS zone? 00:58:08.290 --> 00:58:17.359 - We’re not sure. It’s – that’s something we wonder about a lot too. 00:58:17.359 --> 00:58:24.769 The shaking at the surface – it’s maybe not – it’s not easy to 00:58:24.769 --> 00:58:29.269 take the measurements that we make of seismic shaking at the surface 00:58:29.269 --> 00:58:33.789 and just claim that it’s exactly the same amplitude of shaking 00:58:33.789 --> 00:58:36.849 on the interface itself. There could be extra structure 00:58:36.849 --> 00:58:41.309 and a wave guide effect or something that make it 00:58:41.309 --> 00:58:45.500 hard to predict exactly the amount of shaking. 00:58:47.520 --> 00:58:51.680 Something that we want to look into further is going back, I think, 00:58:51.680 --> 00:58:59.160 to Justin’s question that perhaps other seismic events – big earthquakes 00:58:59.160 --> 00:59:02.499 elsewhere in the world may have had a similar amplitude of shaking. 00:59:02.499 --> 00:59:08.359 And it’s definitely an important question to push forward. 00:59:08.360 --> 00:59:11.500 Why didn’t that cause the same thing? 00:59:13.269 --> 00:59:16.980 - Just to follow on Evan’s question, is there any evidence that those 00:59:16.980 --> 00:59:22.160 two earthquakes triggered tremor dynamically at that same time? 00:59:22.820 --> 00:59:28.780 - Not very much, actually. The website catalog does have some 00:59:28.910 --> 00:59:33.180 dynamically triggered tremor on the day of the 2014 earthquake. 00:59:33.180 --> 00:59:37.049 But with Aaron’s help, we looked at it more closely, 00:59:37.049 --> 00:59:41.119 and it seems like that’s false detections from aftershock activity. 00:59:41.119 --> 00:59:45.309 So on the – when we looked at the 24 hours surrounding that earthquake, there 00:59:45.309 --> 00:59:50.630 was lots of earthquake stuff, but nothing that was a systematic tremor detection. 00:59:50.630 --> 00:59:53.529 We got really excited when we saw lots of tremor on that day. 00:59:53.529 --> 00:59:59.920 We thought it – there was movement everywhere, but it seems to be false. 01:00:02.180 --> 01:00:04.040 - Okay. Any last questions? 01:00:07.020 --> 01:00:11.440 Great. So let’s thank Kathryn one more time. Give her a round of applause. 01:00:11.440 --> 01:00:16.180 [Applause] 01:00:16.180 --> 01:00:19.849 If anyone is interested in coming and having lunch with us, 01:00:19.849 --> 01:00:23.869 please come up to the front of the room, and we will arrange a lunch. 01:00:23.869 --> 01:00:26.800 Also, Kathryn still has some spots available in the afternoon. 01:00:26.800 --> 01:00:30.840 If you want to arrange meeting with her, please contact Fred or myself or Jeanne, 01:00:30.849 --> 01:00:32.430 and we will make that happen. Anyway, thank you. 01:00:32.430 --> 01:00:34.880 We’ll see you here next Monday. 01:00:37.300 --> 01:00:39.160 [Silence]