WEBVTT Kind: captions Language: en 00:00:02.360 --> 00:00:07.460 Good morning, everybody. Today’s quick announcement. 00:00:07.460 --> 00:00:14.300 Next week’s seminar, I’ll do a talk. July the 5th, this auditorium. 00:00:14.300 --> 00:00:21.270 And for today, we introduce Martin Schoenball from Stanford University. 00:00:21.270 --> 00:00:27.910 This is – he’s a second postdoc. Previously, he’s done his 00:00:27.910 --> 00:00:31.460 bachelor’s degree in physics at the University of Dresden and 00:00:31.460 --> 00:00:38.340 his master’s and Ph.D. in geophysics at the University of Karlsruhe. 00:00:38.340 --> 00:00:45.340 And then he did his first postdoc here at the USGS called GMAC. 00:00:45.960 --> 00:00:50.900 And his background – he’s done earthquake processes, 00:00:50.900 --> 00:00:55.520 induced seismicity, geothermal systems, borehole geomechanics, 00:00:55.520 --> 00:00:59.820 and as well, hydromechanics. For further ado, Martin. 00:01:03.240 --> 00:01:06.880 - Good morning, everyone. You can hear me? Yeah. 00:01:06.880 --> 00:01:09.340 Yeah, thank you very much for coming here, and thank you for 00:01:09.340 --> 00:01:13.040 giving me the opportunity to present you some of my research 00:01:13.040 --> 00:01:17.060 that I’ve been doing the last year together with Bill Ellsworth. 00:01:17.060 --> 00:01:19.980 So I was – I will talk today about the spatiotemporal evolution 00:01:19.980 --> 00:01:24.320 of induced seismicity clusters in Oklahoma and southern Kansas. 00:01:25.340 --> 00:01:28.960 Fluid-induced seismicity has been observed in a wide range of settings, 00:01:28.979 --> 00:01:33.690 ranging from reservoir impoundment to geothermal systems, hydraulic fracturing, 00:01:33.690 --> 00:01:40.060 and injection of CO2 and wastewater disposal, but also natural causes 00:01:40.060 --> 00:01:44.790 such as rainfall events, snow melt and volcanic swarms. 00:01:44.790 --> 00:01:47.450 And basically, all of these observations of fluid-induced 00:01:47.450 --> 00:01:53.640 seismicity can be explained by one of – both of two mechanisms. 00:01:53.640 --> 00:02:00.620 The first one is the direct effect of the elevated fluid pressure, 00:02:00.620 --> 00:02:05.740 which decreases the effective stress on a fault. 00:02:05.750 --> 00:02:09.960 It unclamps the fault but maintains the shear stress. 00:02:09.960 --> 00:02:13.709 So in this case – [clears throat] sorry – the Mohr’s circle moves to the left, 00:02:13.709 --> 00:02:17.100 touching the failure envelope, and we get an earthquake. 00:02:17.100 --> 00:02:21.610 And the second effect is that of changes to the solid stress, either 00:02:21.610 --> 00:02:25.970 communicated by poor elastic stressing or by changes in the gravitational load. 00:02:25.970 --> 00:02:28.610 And in this case, the Mohr’s circle changes its size 00:02:28.610 --> 00:02:31.440 and may touch the failure envelope this way. 00:02:33.209 --> 00:02:38.360 If we look at the earthquake rate in the United States in the last couple years, 00:02:38.360 --> 00:02:40.860 we’ve noticed this uptick in 2009. 00:02:40.860 --> 00:02:44.920 And in 2013, the seismicity rate basically exploded. 00:02:44.930 --> 00:02:48.099 And most of this seismicity has been observed in the 00:02:48.099 --> 00:02:51.940 Oklahoma and Kansas area. So what’s been going on there? 00:02:51.940 --> 00:02:56.890 You all are very familiar with this. So with the advent of hydraulic 00:02:56.890 --> 00:03:01.280 fracturing or improvement of hydraulic fracturing, formations that 00:03:01.280 --> 00:03:05.989 were previously uneconomical are now being produced economically. 00:03:05.989 --> 00:03:08.550 But these formations have a huge water cut, 00:03:08.550 --> 00:03:13.550 so the produced fluid is mostly water and contains a little bit of oil. 00:03:13.550 --> 00:03:17.580 And this water is considered saltwater, so it has to be disposed of. 00:03:17.580 --> 00:03:21.980 And in Oklahoma and Kansas, it is disposed of in the Arbuckle Formation, 00:03:21.980 --> 00:03:25.840 which is just overlying the crystalline basement. 00:03:25.840 --> 00:03:30.550 The Arbuckle Formation typically lies at depths of about 1 to 2 kilometers. 00:03:30.550 --> 00:03:34.930 And it’s a naturally under-pressured formation, very permeable, so injection 00:03:34.930 --> 00:03:40.810 pressures are actually very small. But the disposal volumes are huge. 00:03:40.810 --> 00:03:45.629 So we get a very large area that is perturbed. 00:03:45.629 --> 00:03:50.300 In this study by Katie Keranen and co-workers, they showed that 00:03:50.300 --> 00:03:54.110 seismicity occurred at distances greater than 20 kilometers from 00:03:54.110 --> 00:03:58.120 injection wells and modeled pressure changes at 00:03:58.120 --> 00:04:03.660 hypocentral locations on the order of a tenth of a megapascal. 00:04:03.660 --> 00:04:07.209 This other study from Nick van der Elst and co-workers showed that the area is 00:04:07.209 --> 00:04:11.500 susceptible to remote dynamic triggering. So stress changes on the order of 00:04:11.500 --> 00:04:20.949 kilopascals from remote earthquake waves trigger earthquakes in this case. 00:04:20.949 --> 00:04:25.580 In this study, I show that earthquakes occur typically about 4 kilometers below 00:04:25.580 --> 00:04:30.020 the injection horizon, so again, this is where the pressure perturbation is 00:04:30.020 --> 00:04:35.369 subsided quite substantially, just because we are far away from the injection. 00:04:35.369 --> 00:04:39.659 And this recent study by Andy Barbour and co-workers showed – they were 00:04:39.659 --> 00:04:44.919 looking at the Pawnee earthquake and made a hydromechanical model for it. 00:04:44.919 --> 00:04:47.939 And they showed that the stress changes at hypocentral depth 00:04:47.939 --> 00:04:50.960 are again on the order of tens of kilopascals. 00:04:50.960 --> 00:04:56.680 So all of these observations tell us that these earthquakes that we see, 00:04:56.680 --> 00:04:59.629 widespread in the region of about 200 kilometers, 00:04:59.629 --> 00:05:03.180 are triggered by very low stress perturbations. 00:05:04.200 --> 00:05:08.229 Although now the earthquake rate is in decline in Oklahoma, 00:05:08.229 --> 00:05:11.050 there is still a lot of challenges ahead to understand 00:05:11.050 --> 00:05:15.710 what’s happening there and to prevent similar cases in the future. 00:05:15.710 --> 00:05:20.089 The big looming question, of course, is why are we seeing this uptick of – 00:05:20.089 --> 00:05:22.809 have we been seeing this uptick in seismicity in Oklahoma 00:05:22.809 --> 00:05:27.500 and not in other areas of significant wastewater disposal? 00:05:27.500 --> 00:05:29.499 And the operators want to know, 00:05:29.499 --> 00:05:34.389 can we mitigate induced seismicity but still dispose of wastewater? 00:05:34.389 --> 00:05:37.589 And why, in particular, the – interested and excited about 00:05:37.589 --> 00:05:41.520 this injection experiment is that it provides us 00:05:41.520 --> 00:05:46.059 great opportunities to learn about the natural system as well. 00:05:46.059 --> 00:05:47.879 So what can we learn about the state of stress 00:05:47.879 --> 00:05:52.919 in these dormant faults that are probably everywhere in the basement? 00:05:52.919 --> 00:05:54.020 How are these basement faults 00:05:54.020 --> 00:05:57.599 hydraulically connected to the sedimentary overburden? 00:05:57.599 --> 00:06:00.509 And how do fluid-driven earthquake swarms work? 00:06:00.509 --> 00:06:04.909 And this would be also interesting to understand volcanic systems 00:06:04.909 --> 00:06:09.690 where we do see these fluid-driven earthquake swarms but don’t really 00:06:09.690 --> 00:06:13.820 know what the source, or how to characterize the source, of these swarms. 00:06:14.560 --> 00:06:18.479 And then lastly, I’m interested in understanding the role of 00:06:18.480 --> 00:06:22.880 the various triggering processes that go into the earthquake nucleation. 00:06:24.640 --> 00:06:29.319 This is the current earthquake map of the area. 00:06:29.319 --> 00:06:33.610 So the black line here shows the Kansas-Oklahoma state border. 00:06:33.610 --> 00:06:36.099 The thin lines are the counties. 00:06:36.099 --> 00:06:38.879 And dots, of course, earthquakes color-coded by time. 00:06:38.879 --> 00:06:44.119 Early earthquakes – I will just use the same color bar for my talk for time. 00:06:44.120 --> 00:06:49.300 Early earthquakes are in blue, and then later, more recent, earthquakes in yellow. 00:06:51.289 --> 00:06:54.260 And you basically see that we have this carpet of earthquakes 00:06:54.270 --> 00:06:57.419 spanning about 200 kilometers in the region. 00:06:57.419 --> 00:07:00.949 If we look a little bit closer, we see that many of these earthquakes 00:07:00.949 --> 00:07:04.149 are actually occurring in distinct clusters. 00:07:04.149 --> 00:07:07.399 So with the study, I wanted to understand, what are these clusters? 00:07:07.399 --> 00:07:11.249 What are their driving mechanisms? And what is their collective behavior? 00:07:11.249 --> 00:07:16.119 Do they follow some empirical relations that we can use to describe them? 00:07:16.119 --> 00:07:20.530 And also, what is their relation to natural seismicity in the area? 00:07:20.530 --> 00:07:23.599 And by analyzing all of these clusters together, I was – I was hoping 00:07:23.599 --> 00:07:28.919 to use some stacking techniques to inform the temporal evolution 00:07:28.919 --> 00:07:32.149 of these sequences and also the evolution of the largest earthquakes – 00:07:32.149 --> 00:07:36.640 those earthquakes which do drive the seismic hazard in the region. 00:07:37.620 --> 00:07:41.700 So my take-home points for today are the following. 00:07:41.709 --> 00:07:44.720 These earthquake sequences are initiated by anthropogenic stressing, 00:07:44.720 --> 00:07:50.039 but they seem to sustain rupture through earthquake interactions. 00:07:50.039 --> 00:07:55.610 Most of these sequences initiate months before they rise to peak activity. 00:07:55.610 --> 00:07:58.610 And the largest events result from stronger perturbations 00:07:58.610 --> 00:08:02.660 but not from re-activating more fresh asperities. 00:08:03.960 --> 00:08:10.039 But to get to these points, I first had to do an earthquake relocation study. 00:08:10.039 --> 00:08:13.559 Since what I wanted to do – the earthquake catalog 00:08:13.560 --> 00:08:18.060 was just not suitable for what I planned. 00:08:20.100 --> 00:08:27.180 And so what we benefited from is that data from 40 three-component 00:08:27.199 --> 00:08:31.469 stations that were operated by the industry were made available to us. 00:08:31.469 --> 00:08:36.940 Ten of those stations are proprietary, and they are – they were provided 00:08:36.940 --> 00:08:39.490 by sponsors of SCITS – the Stanford Center for 00:08:39.490 --> 00:08:42.169 Induced and Triggered Seismicity. 00:08:42.169 --> 00:08:47.790 And 30 other stations are now available on IRIS for a timespan 00:08:47.790 --> 00:08:52.300 of two and a half years, I think. And it’s the network NX. 00:08:52.300 --> 00:08:57.980 And so these 30 stations are in this area here. Ten other stations are in this area. 00:08:57.990 --> 00:09:00.910 So with these stations, we have a relatively even 00:09:00.910 --> 00:09:04.870 station coverage of stations basing on the order of 25 kilometers 00:09:04.870 --> 00:09:12.089 or better for the whole region, starting in about 2013. 00:09:12.089 --> 00:09:17.720 So for this relocation procedure, I used the ANSS catalog and its phase arrivals. 00:09:17.720 --> 00:09:21.940 I also used the detections of the Oklahoma Geological Survey catalog. 00:09:21.940 --> 00:09:27.880 I did not detect any other earthquakes. I didn’t attempt to. 00:09:27.880 --> 00:09:32.680 And then, using an automatic picker for P and S wave arrivals, I queried those 00:09:32.680 --> 00:09:37.009 40 stations. I used a very simple 1D velocity model, which is closely 00:09:37.009 --> 00:09:41.980 modeled after the one that the Oklahoma Geological Survey is using. 00:09:41.980 --> 00:09:46.360 And then I’m using HypoInverse for absolute locations, and I refined those 00:09:46.360 --> 00:09:50.949 locations using the double-difference method in the software HypoDD. 00:09:50.949 --> 00:09:55.940 And here I’m using differential times from the catalog picks, and as well as 00:09:55.940 --> 00:09:59.829 waveform cross-correlation where I get waveforms to cross-correlate. 00:09:59.829 --> 00:10:04.810 And overall, we covered the period of May 2013, which is the time 00:10:04.810 --> 00:10:11.240 when the 30 stations in this area came online, until November last year. 00:10:11.240 --> 00:10:14.160 So with this, we covered the major uptick after the 00:10:14.160 --> 00:10:18.769 first uptick until the beginning of the decline in last year. 00:10:18.769 --> 00:10:23.180 And, in total, these are 18,600 earthquakes now. 00:10:23.180 --> 00:10:25.960 So, again, this is the catalog that we started with. 00:10:25.960 --> 00:10:30.180 After relocating, all the seismicity collapses quite a bit, 00:10:30.189 --> 00:10:34.060 and we see some structural grain in the area. 00:10:34.060 --> 00:10:37.660 We see that these faults – or these earthquakes tend to 00:10:37.660 --> 00:10:41.240 align in linear features that we can interpret as faults. 00:10:41.240 --> 00:10:43.670 But to really show the quality of the data, we have to zoom in. 00:10:43.670 --> 00:10:46.879 So let’s go into the Kansas part where we do have the best network, 00:10:46.879 --> 00:10:50.139 of course, because we have the closest station spacing there. 00:10:50.139 --> 00:10:54.670 But it’s still sort of representative of what we have in the rest of the area. 00:10:54.670 --> 00:10:59.810 And we see that earthquakes here align in these pretty linear features 00:10:59.810 --> 00:11:03.370 that can be interpreted as the vertical faults. 00:11:03.370 --> 00:11:07.100 But then we still have some cloudy structure in some other areas. 00:11:07.100 --> 00:11:10.740 So we can zoom in again. And if you look at it in 3D, 00:11:10.740 --> 00:11:14.470 we see that there are – in this case, we have two beautifully resolved 00:11:14.470 --> 00:11:18.860 normal faults in an abutting relationship. 00:11:18.860 --> 00:11:23.449 Here I show also the comparison of the focal mechanisms from Bob Hermann 00:11:23.449 --> 00:11:27.310 from St. Louis University, and they are in very good agreement 00:11:27.310 --> 00:11:31.440 with strike and dip of what we resolved just from the hypocenters here. 00:11:31.440 --> 00:11:34.340 So that was very encouraging for us. 00:11:35.500 --> 00:11:38.980 The location uncertainties now are on the order of a couple hundred meters 00:11:38.990 --> 00:11:44.029 in the horizontal and typically better than 1 kilometer in the vertical. 00:11:44.029 --> 00:11:49.020 And the relative locations are about one order of magnitude better than that. 00:11:50.450 --> 00:11:58.740 In this slide, I would like to show you a movie of a subset of the earthquakes. 00:11:58.740 --> 00:12:03.339 So in yellow, I show injection wells, and the symbol size scales 00:12:03.339 --> 00:12:07.649 with the injection volume. And it will change with time. 00:12:07.649 --> 00:12:15.860 And then earthquakes will start to appear in red and then fade to – fade to gray. 00:12:18.120 --> 00:12:19.500 All right. 00:12:19.509 --> 00:12:23.010 So now, in about summer of 2013, you see that earthquakes start to 00:12:23.010 --> 00:12:25.860 appear basically everywhere in the region. 00:12:25.860 --> 00:12:32.010 We see that some wells are changing their injection schedule a little bit. 00:12:32.010 --> 00:12:35.249 We have some sequences that evolve gradually in all kinds of 00:12:35.249 --> 00:12:38.340 directions with several faults being activated here. 00:12:38.340 --> 00:12:44.089 We have some sequences that are pretty slow, developing over a long time. 00:12:44.089 --> 00:12:47.180 And we have this really nice, intriguing alignment here 00:12:47.180 --> 00:12:50.680 that develops into the Fairview sequence. 00:12:50.680 --> 00:12:52.940 And now, in 2016, we have daily injection data. 00:12:52.940 --> 00:12:56.100 And you see just how these wells over here – 00:12:56.100 --> 00:13:01.220 you don’t see this – how they are pulsating. 00:13:04.080 --> 00:13:06.580 Let’s try this again. 00:13:06.600 --> 00:13:08.780 Doesn’t work. 00:13:11.300 --> 00:13:14.540 So just play it again because there’s so much to see here. 00:13:17.020 --> 00:13:26.640 [ Silence ] 00:13:27.440 --> 00:13:32.900 So now we start to see this alignment over here which turns into Fairview. 00:13:35.080 --> 00:13:44.220 [ Silence ] 00:13:45.020 --> 00:13:47.279 And then notice the pulsation of the injections. 00:13:47.279 --> 00:13:51.029 So this injection schedule is actually varying on a daily basis, or probably 00:13:51.029 --> 00:13:58.089 even on an hourly basis by orders of magnitude from zero to full injection. 00:13:58.089 --> 00:14:04.230 And we don’t really understand yet what this pulsation, or these rapid 00:14:04.230 --> 00:14:11.559 changes of injection rate, actually mean for the mechanical processes there. 00:14:11.560 --> 00:14:14.480 If we look at depth of these earthquakes – 00:14:14.480 --> 00:14:17.009 I showed this slide – this figure before. 00:14:17.009 --> 00:14:20.410 Here I plot depth relative to the top of the basement that 00:14:20.410 --> 00:14:23.379 we get from well bore locks. 00:14:23.380 --> 00:14:28.520 And just as a reminder, the Arbuckle Formation is the one that – 00:14:28.520 --> 00:14:31.760 where the fluid injection is occurring. And most of these earthquakes 00:14:31.769 --> 00:14:37.709 occurred about 4 kilometers below this top of the basement. 00:14:37.709 --> 00:14:42.370 And this is still kind of a mystery to me why this is occurring. 00:14:42.370 --> 00:14:49.389 And since you are going to ask anyway, my speculation for this is that it 00:14:49.389 --> 00:14:51.740 seems to be kind of a sweet spot between the 00:14:51.740 --> 00:14:57.080 brittleness of the crust and how much stress is stored at greater depth. 00:14:57.080 --> 00:15:00.499 But then, at shallower depth, we also have higher pressure changes. 00:15:00.499 --> 00:15:03.110 So my guess is that, at 4-kilometer depth, 00:15:03.110 --> 00:15:07.379 we are just somehow in a sweet spot, and I think this is also supported by 00:15:07.380 --> 00:15:13.809 laboratory studies from Oklahoma University – Brett Carpenter. 00:15:15.440 --> 00:15:17.999 All right, so now we have earthquakes relocated. 00:15:18.000 --> 00:15:22.620 Now we want to study the sequences. So the first step is to 00:15:22.620 --> 00:15:26.840 identify those sequences here. For that, I’m using the 00:15:26.840 --> 00:15:30.779 DBSCAN algorithm, which recursively identifies clusters. 00:15:30.779 --> 00:15:35.100 It only requires two input parameters, which are the minimum number of points 00:15:35.100 --> 00:15:38.750 that are required to call it a cluster – 10 in my case – 00:15:38.750 --> 00:15:42.470 and then a search radius – 700 meters. These 700 meters are 00:15:42.470 --> 00:15:47.160 kind of a magic number. But it seems to work quite nicely to pick 00:15:47.160 --> 00:15:57.310 out those clusters that we see readily by eye and divide – pick up these clusters. 00:15:57.310 --> 00:16:02.519 So on the left-hand side, this is one of those cluster candidates. 00:16:02.519 --> 00:16:05.800 Then in the second step, I manually review those clusters 00:16:05.800 --> 00:16:12.220 in a 3D view, and I interpret faults here. 00:16:12.220 --> 00:16:15.329 So in this case I showed you before, we have these two beautifully resolved 00:16:15.329 --> 00:16:18.010 faults and a couple scattered earthquakes 00:16:18.010 --> 00:16:21.430 around that we cannot attribute to a fault. 00:16:21.430 --> 00:16:25.850 And then, in the second step, we can take hypocenters from each fault and 00:16:25.850 --> 00:16:30.129 compute the covariance matrix, and the eigenvectors give us a strike and dip. 00:16:30.129 --> 00:16:35.420 And we did here in a study from Fred on the Pawnee earthquake. 00:16:36.280 --> 00:16:38.959 Okay, to summarize the relocation procedure, 00:16:38.959 --> 00:16:40.749 this is what we started out with. 00:16:40.749 --> 00:16:46.119 By adding much more data, we have much better control of the 00:16:46.119 --> 00:16:52.630 absolute and relative locations now. We identified sequences and clusters. 00:16:52.630 --> 00:16:56.240 We can now compare these alignments also of the Oklahoma fault map, 00:16:56.240 --> 00:17:02.019 and we basically see no agreement. Only does – I only know one case 00:17:02.020 --> 00:17:06.300 where we actually have earthquakes on a previously known fault strand. 00:17:07.300 --> 00:17:11.980 Importantly, also the default trends that we see already just by looking at 00:17:11.980 --> 00:17:16.020 these lineaments here are not apparent in the Oklahoma fault map. 00:17:16.020 --> 00:17:21.631 So we do have probably a biased view from both ends as to the 00:17:21.640 --> 00:17:26.840 orientation or preferential orientation of faults that we have in the basement. 00:17:26.840 --> 00:17:30.340 We do see a pretty good agreement with these focal mechanisms that 00:17:30.340 --> 00:17:35.399 are provided by Bob Herrmann. So that was very encouraging for us. 00:17:35.399 --> 00:17:37.250 And the catalog will be available very soon. 00:17:37.250 --> 00:17:41.070 We have a paper in press in SRL. And if you’re interested in 00:17:41.070 --> 00:17:45.520 working with this data set, I can also just send you the data set right now. 00:17:46.780 --> 00:17:51.600 Overall, we identified 300 faults ranging from about a couple hundred 00:17:51.600 --> 00:17:59.700 meters to more than 10 kilometers in length for – 00:17:59.700 --> 00:18:03.659 and with strikes that are clearly preferentially aligned. 00:18:03.659 --> 00:18:10.360 I just recently looked at even smaller clusters with less than 10 earthquakes, 00:18:10.360 --> 00:18:14.500 and it seems to be even a couple hundred – maybe 200 more 00:18:14.500 --> 00:18:22.350 that are convincingly elongated to be called a fault, but I don’t show this here. 00:18:22.350 --> 00:18:24.380 So we can compare the fault strikes now 00:18:24.380 --> 00:18:26.640 with what we know about the stress field. 00:18:26.640 --> 00:18:30.240 So on the left-hand side, I show a figure by Alt and Zoback, 00:18:30.240 --> 00:18:34.510 who looked at the stress field in the area, and their main conclusion is that 00:18:34.510 --> 00:18:38.590 the stress field is pretty constant throughout the area, and the maximum 00:18:38.590 --> 00:18:42.680 horizontal stress is oriented at 85 degrees east of north. 00:18:42.680 --> 00:18:46.370 And the fault strikes that we observe are pretty much consistent 00:18:46.370 --> 00:18:49.000 with what we would expect in a strike-slip faulting 00:18:49.000 --> 00:18:52.400 regime with this kind of stress orientation. 00:18:53.640 --> 00:18:57.180 All right. Let’s have a look at how these sequences evolve. 00:18:57.190 --> 00:19:01.920 At first, I was interested in migration patterns. 00:19:01.920 --> 00:19:07.450 So I was looking – I was hoping to find horizontal migration patterns 00:19:07.450 --> 00:19:10.299 that would show up as some coherent domains that 00:19:10.300 --> 00:19:13.540 several clusters migrate in the same direction. 00:19:13.540 --> 00:19:16.320 To cut a long story short, I didn’t find any of these. 00:19:16.320 --> 00:19:21.280 Maybe I have to look a little bit – a little bit harder, but for now, I stop there. 00:19:21.280 --> 00:19:23.960 I did, however, see some systematics in the vertical migration. 00:19:23.970 --> 00:19:26.820 This is what I would like to show you here. 00:19:26.820 --> 00:19:31.570 So this is an example from the Woodward sequence in westernmost – 00:19:31.570 --> 00:19:35.400 the westernmost area of the crisis area. 00:19:35.400 --> 00:19:40.900 And, again, I color-code earthquakes by time beginning in blue. 00:19:40.900 --> 00:19:45.580 The sequence systematically migrates to the left in this case. 00:19:45.580 --> 00:19:50.340 If we look at it in depth view, we see, of course, the same behavior. 00:19:50.340 --> 00:19:54.679 But then there’s a second migration direction overlaying. 00:19:54.679 --> 00:19:58.420 As a particular strike of a fault was activated, later earthquakes 00:19:58.420 --> 00:20:03.490 appear to be at greater depth. And then later, another strike 00:20:03.490 --> 00:20:07.120 was activated, and seismicity migrates downward. 00:20:07.120 --> 00:20:11.950 And in the second half of the sequence, 00:20:11.950 --> 00:20:15.140 we saw reactivation of a shallow part of this fault strand. 00:20:15.140 --> 00:20:20.520 But once this part was activated, again, seismicity migrated downward here. 00:20:21.580 --> 00:20:26.220 So this is already a relatively complex behavior to capture with a simple 00:20:26.230 --> 00:20:30.970 algorithm that tries to quantify diverging migrations either up or down. 00:20:30.970 --> 00:20:34.799 Because we have these shallow earthquakes in the end, 00:20:34.799 --> 00:20:38.350 but still this downward migration, an algorithm that would compare 00:20:38.350 --> 00:20:42.830 just the hypocentral depth of earthquakes of the first and second half 00:20:42.830 --> 00:20:47.090 would call this sequence migrating upward, which it clearly does not. 00:20:47.090 --> 00:20:52.500 So this is why I opted for a manual interpretation of this vertical migration. 00:20:53.560 --> 00:20:58.340 And in this manual interpretation, I also tried to differentiate 00:20:58.340 --> 00:21:01.500 between individual faults in single clusters. 00:21:01.500 --> 00:21:05.940 So some of these clusters have several faults being activated in succession. 00:21:05.940 --> 00:21:09.380 So in this case, I interpret first the vertical migration 00:21:09.380 --> 00:21:13.030 along this fault and then on this fault here. 00:21:13.030 --> 00:21:17.700 And the result of this interpretation is that the first faults that are 00:21:17.700 --> 00:21:22.500 being activated tend to show a preference for downward migration, 00:21:22.500 --> 00:21:27.580 which suggests that their driving source, it’s above these faults. 00:21:27.580 --> 00:21:30.860 But for second faults that are being activated, I do not see 00:21:30.860 --> 00:21:34.769 this preferential migration direction downward. 00:21:34.769 --> 00:21:38.549 So this suggests that, for second faults being activated, 00:21:38.549 --> 00:21:44.149 a different process is coming to play, such as a fault interaction that would 00:21:44.149 --> 00:21:50.700 maybe not have this preferential vertical migration direction imposed. 00:21:52.830 --> 00:21:56.940 If we look at this sequence that I showed you before, 00:21:56.950 --> 00:21:59.450 it ruptured with a couple of magnitude 3 earthquakes 00:21:59.450 --> 00:22:04.820 in the first three months until it hit this second fault here. 00:22:04.820 --> 00:22:11.820 And then, once it hit this fault, seismicity evolved in a burst, 00:22:11.820 --> 00:22:16.070 reaching magnitude 4.3 earthquake and the highest activity. 00:22:16.070 --> 00:22:18.649 So here we have a delay from the initiation – 00:22:18.649 --> 00:22:22.269 the first recorded earthquake – until the maximum activity 00:22:22.269 --> 00:22:26.549 we observed in this sequence of about three months. 00:22:26.549 --> 00:22:29.980 And it turns out that this delay is actually very characteristic 00:22:29.980 --> 00:22:34.470 of earthquakes in the area. Only about a quarter of sequences 00:22:34.470 --> 00:22:40.020 reached their maximum activity within the first month of initiation. 00:22:40.020 --> 00:22:43.909 And many sequences reached their maximum activities, or the maximum 00:22:43.909 --> 00:22:50.279 earthquake rate, delayed by several months or even longer than a year. 00:22:50.279 --> 00:22:54.210 So here we do have some kind of warning time 00:22:54.210 --> 00:23:01.680 even at this relatively coarse completeness monitoring that we had. 00:23:04.200 --> 00:23:07.210 Then we might be interested in looking when the largest earthquakes 00:23:07.210 --> 00:23:15.260 occur and see if there’s a development there or if these – 00:23:15.260 --> 00:23:18.029 if the largest events occur at a random time. 00:23:18.029 --> 00:23:22.700 So they’re just a random Gutenberg-Richter draw. 00:23:23.880 --> 00:23:27.580 And I looked at it in two ways here for all sequences that had 00:23:27.580 --> 00:23:33.120 at least 10 earthquakes, which is the light gray data set here. 00:23:33.120 --> 00:23:35.779 And it looks like, in this cumulative density plot, 00:23:35.779 --> 00:23:39.029 that it follows very closely aligned and would correspond 00:23:39.029 --> 00:23:43.220 to a random occurrence of the largest earthquakes. 00:23:43.220 --> 00:23:46.990 So this would follow a uniform distribution, 00:23:46.990 --> 00:23:53.740 and the Kolmogorov-Smirnov test actually does not reject a null hypothesis 00:23:53.740 --> 00:24:00.500 that the largest events are uniformly distributed 00:24:00.500 --> 00:24:02.990 over time or along their position. 00:24:02.990 --> 00:24:06.789 If we only look at the larger sequences that have at least 30 earthquakes, 00:24:06.789 --> 00:24:10.170 there seems to be some tendency for the largest events 00:24:10.170 --> 00:24:12.769 to occur during the middle of the sequences. 00:24:12.769 --> 00:24:16.809 And in this case, the Kolmogorov- Smirnov test actually does reject 00:24:16.809 --> 00:24:22.200 the null hypothesis that these events are randomly distributed. 00:24:25.150 --> 00:24:29.260 So we can continue analyzing earthquake rates and how they 00:24:29.269 --> 00:24:33.309 develop towards the main shock. So if we consider, for each cluster, 00:24:33.309 --> 00:24:40.970 the largest event to be the main shock and all the events that occurred before, 00:24:40.970 --> 00:24:44.299 we can call them pre-shocks, and then all the events after 00:24:44.299 --> 00:24:46.790 the largest event are aftershocks. 00:24:46.790 --> 00:24:52.470 So, in this case, I stacked earthquake rates along the 00:24:52.470 --> 00:24:57.580 largest event at zero and any earthquakes at negative time are pre-shocks. 00:24:57.580 --> 00:25:01.680 Any events at positive time are aftershocks. 00:25:01.680 --> 00:25:06.730 If I did this – or, if I do this, we get these kind of plots. 00:25:06.730 --> 00:25:11.190 So in this case, the main shock would be in between the two plots here. 00:25:11.190 --> 00:25:15.900 And let’s look first at the aftershocks. We do see, in this log-log plot 00:25:15.900 --> 00:25:20.529 of the number and time that we have this kind of linear decay, 00:25:20.529 --> 00:25:26.389 which means that we can approximate the earthquake rate 00:25:26.389 --> 00:25:30.160 after the main shock by an Omori-type decay. 00:25:30.160 --> 00:25:32.149 The P value, which describes the 00:25:32.149 --> 00:25:38.549 decay behavior, has a relatively low value of about 0.6. 00:25:38.549 --> 00:25:42.180 So typical single main shock-aftershock sequences show larger values of 00:25:42.180 --> 00:25:46.929 about 0.8 to 1.2 or something. So these are relatively low values. 00:25:46.929 --> 00:25:49.900 But there has been also other studies where this kind of stacking technique 00:25:49.900 --> 00:25:55.640 was used that show similarly small values here. 00:25:55.640 --> 00:25:59.200 If we look at the foreshocks on the left – or, pre-shocks on the left side, 00:25:59.200 --> 00:26:03.559 we first notice that we have a lot of pre-shocks. 00:26:03.559 --> 00:26:09.260 In fact, cutting the catalog in two parts of pre-shocks 00:26:09.260 --> 00:26:12.260 and aftershocks basically cuts the catalog in half. 00:26:12.260 --> 00:26:17.160 We have the same number of pre-shocks as aftershocks. 00:26:17.160 --> 00:26:21.639 We still see kind of this increasing behavior that we can describe with – 00:26:21.639 --> 00:26:24.360 sort of with a reverse Omori law. 00:26:24.360 --> 00:26:28.700 But it doesn’t follow this linear trend as closely. 00:26:28.700 --> 00:26:31.690 Instead, we have this – appear to have this jump of 00:26:31.690 --> 00:26:36.029 activity a month before the main shock is occurring. 00:26:36.029 --> 00:26:43.200 And then pre-shock activity seems to be on a relatively constant high level. 00:26:46.280 --> 00:26:51.380 [ Silence ] 00:26:51.980 --> 00:26:57.340 Okay, now we may be interested in how do – or, why do these 00:26:57.340 --> 00:27:01.720 largest earthquakes occur? Do they occur because we are – 00:27:01.720 --> 00:27:05.620 continue pumping after the sequence has been initiated? 00:27:05.620 --> 00:27:09.820 Or do they occur because we reactivate fresh fault asperities 00:27:09.820 --> 00:27:14.139 that have not been activated before, and we basically just have a larger chance of 00:27:14.139 --> 00:27:20.510 finding a nice asperity that is well-suited for rupturing in a big earthquake? 00:27:20.510 --> 00:27:25.440 And to follow up on this question, I was studying each sequence 00:27:25.440 --> 00:27:28.950 and then looking at when the currently largest earthquake 00:27:28.950 --> 00:27:32.080 occurred and the extent of these sequences. 00:27:32.080 --> 00:27:34.889 So in this case here, for the Woodward sequence, 00:27:34.889 --> 00:27:37.720 the second earthquake occurs 600 meters away from the 00:27:37.720 --> 00:27:42.900 first earthquake, and the larger of the two is a magnitude 2.8. 00:27:42.900 --> 00:27:46.139 And then the next-largest earthquake was a 3.5 00:27:46.140 --> 00:27:49.520 once the sequence spread to 2-1/2 kilometers. 00:27:49.520 --> 00:27:54.220 Then we get a 3.7 once we have 4 kilometers reactivated, 00:27:54.220 --> 00:27:59.000 and a 3.8 once the whole sequence is 8 kilometers long. 00:27:59.800 --> 00:28:03.019 The Fairview sequence evolved quite differently here. 00:28:03.019 --> 00:28:08.470 So here we observe a 3.1 earthquake already 00:28:08.470 --> 00:28:13.210 at a distance of 10 – or, spread out over 10 kilometers. 00:28:13.210 --> 00:28:17.490 With the same extent, we get a 3.4. 00:28:17.490 --> 00:28:21.169 Just growing the sequence a little bit gives us a 4.3. 00:28:21.169 --> 00:28:25.330 Still kind of in the center of what’s been activated before. 00:28:25.330 --> 00:28:28.120 Same extent, we get a 4.4. 00:28:28.120 --> 00:28:32.300 And then almost the same extent, we culminate in a 5.1 here. 00:28:33.120 --> 00:28:35.440 So these are pretty different behaviors. 00:28:35.450 --> 00:28:40.730 And I plot them in this graph here of cluster length that has been 00:28:40.730 --> 00:28:44.560 reactivated and the maximum magnitude that has been observed. 00:28:44.560 --> 00:28:49.000 And for Woodward, we get this upward trajectory here. 00:28:49.000 --> 00:28:51.460 And for Fairview, we get this sort of horizontal – 00:28:51.460 --> 00:28:56.669 or, yeah, constant trajectory there. So the Fairview sequence basically 00:28:56.669 --> 00:29:00.909 increased its effective stress drop throughout the sequence, 00:29:00.909 --> 00:29:03.710 while the Woodward sequence decreased its effective stress drop. 00:29:03.710 --> 00:29:10.789 So it’s getting – well, less energetic, if you want. 00:29:10.789 --> 00:29:14.100 And we can interpret these kind of trajectories. 00:29:14.100 --> 00:29:18.480 The steeper trajectories as resulting from – the largest events result 00:29:18.480 --> 00:29:23.850 from reactivating longer fault strands, while the horizontal trajectories 00:29:23.850 --> 00:29:29.430 would result from stronger perturbation of a continued injection. 00:29:29.430 --> 00:29:33.020 If I do this for all of these studies – all of these sequences, 00:29:33.020 --> 00:29:35.820 we get this pretty messy graph here. 00:29:35.820 --> 00:29:40.039 But on the right-hand side, I show this simplified in a histogram. 00:29:40.039 --> 00:29:43.010 And we see that we have this preference for sequences to occur, 00:29:43.010 --> 00:29:48.080 or to show the largest events because of stronger perturbation. 00:29:48.080 --> 00:29:51.150 So we have these horizontal trajectories here and only 00:29:51.150 --> 00:29:56.620 a smaller set that is actually resulting from longer reactivation. 00:29:56.620 --> 00:30:00.690 If we look at the largest earthquakes, so earthquakes greater than 4, or even 00:30:00.690 --> 00:30:07.429 greater than 4-1/2, this preference for the stronger perturbation increases. 00:30:07.429 --> 00:30:12.559 So there’s, in fact, no earthquake larger than 4.5 that was occurring 00:30:12.560 --> 00:30:19.580 because we activated a new fault strand or new part of a fault. 00:30:20.720 --> 00:30:22.120 All right. 00:30:23.240 --> 00:30:28.000 Then I was interested in quantifying the earthquake interactions. 00:30:28.000 --> 00:30:32.309 And I’m using – I’m trying to quantify this using temporal clustering. 00:30:32.309 --> 00:30:35.380 Temporal clustering is a fundamental property of earthquakes 00:30:35.380 --> 00:30:40.549 as we see in mainshock-aftershock sequences, for instance. 00:30:40.549 --> 00:30:44.409 So if we have a catalog that does not show temporal clustering, 00:30:44.409 --> 00:30:47.960 this can be interpreted as – events can be interpreted as 00:30:47.960 --> 00:30:51.460 independent background events with no interaction. 00:30:51.460 --> 00:30:53.500 But if there is temporal clustering occurring – 00:30:53.500 --> 00:30:57.190 so many events occur at a small section of time, 00:30:57.190 --> 00:31:02.950 and then we have large stretches in time where you don’t. So if an earthquake – 00:31:02.950 --> 00:31:08.039 this means that earthquake interactions are important, 00:31:08.039 --> 00:31:11.070 and we do get this temporal clustering. 00:31:11.070 --> 00:31:16.059 And to quantify this temporal clustering for these relatively small subsets here, 00:31:16.059 --> 00:31:21.470 I’m using the coefficient of variation, which is – of the inter-event times, 00:31:21.470 --> 00:31:23.160 which is defined as the standard deviation 00:31:23.160 --> 00:31:26.159 of inter-event times divided by the mean. 00:31:26.159 --> 00:31:30.260 And this quantity has a value of about 1 for 00:31:30.260 --> 00:31:34.880 uniform distributed occurrence times, so a Poissonian process. 00:31:34.880 --> 00:31:39.090 And it takes larger values for sequences where we do have temporal clustering, 00:31:39.090 --> 00:31:43.760 such as a mainshock-aftershock sequence that I use as an example here. 00:31:44.860 --> 00:31:49.419 If I do this for the sequences, we basically get the full spectrum 00:31:49.419 --> 00:31:53.720 from sequences that show no temporal clustering – 00:31:53.720 --> 00:31:58.590 so they behave as background events – as independent background events. 00:31:58.590 --> 00:32:00.240 And then many sequences which have 00:32:00.240 --> 00:32:03.190 a pretty strong temporal clustering component. 00:32:03.190 --> 00:32:10.440 So these sequences seem to be strongly influenced by earthquake interactions. 00:32:11.800 --> 00:32:14.140 And I’ll come back to this in just a bit. 00:32:15.560 --> 00:32:19.800 But now I would like to talk briefly about activity fronts. 00:32:19.800 --> 00:32:23.800 So the current paradigm with current understanding for these 00:32:23.809 --> 00:32:27.519 sequences is that they are triggered by pore pressure perturbations 00:32:27.519 --> 00:32:31.720 which propagate by a diffusive process. 00:32:31.720 --> 00:32:39.270 And if so, we can describe the outer envelope of quakes as a distance or of 00:32:39.270 --> 00:32:46.640 a time by this relationship here – square root of 4-pi times diffusivity times time. 00:32:46.640 --> 00:32:51.140 And we can use this to estimate diffusivity in the underground. 00:32:52.560 --> 00:32:55.220 And I did this for the larger sequences here, 00:32:55.230 --> 00:33:00.409 and we get diffusivity values that span about three orders of magnitude. 00:33:00.409 --> 00:33:03.919 And they are sort of in agreement with diffusivity values 00:33:03.919 --> 00:33:08.940 that have been obtained in this kind of manner from reservoir impoundment – 00:33:08.940 --> 00:33:12.100 typically larger values or volcanic swarms. 00:33:12.100 --> 00:33:16.639 In this case, I should compare with Dave Shelley’s work here. 00:33:16.639 --> 00:33:21.019 And then also lower values that we see in southern California, 00:33:21.020 --> 00:33:25.080 tectonic swarms, and even smaller values than this. 00:33:26.970 --> 00:33:33.280 And so here I plot – a cross-plot of the diffusivity on the right-hand side, 00:33:33.300 --> 00:33:36.560 and on the top, the temporal clustering, 00:33:36.560 --> 00:33:39.299 quantified by the coefficient of variation. 00:33:39.299 --> 00:33:41.970 And with this figure, I want to suggest that there might be 00:33:41.970 --> 00:33:45.890 a correlation between the two in the way that sequences that 00:33:45.890 --> 00:33:50.460 have no temporal clustering, they do seem to propagate slower. 00:33:50.460 --> 00:33:56.830 So their seismic diffusivity, as I call it now, is relatively small. 00:33:56.830 --> 00:34:00.159 And then sequences that have strong temporal clustering 00:34:00.159 --> 00:34:02.659 seem to propagate also much faster. 00:34:02.659 --> 00:34:07.400 So in this case, the seismic diffusivity that I get is larger. 00:34:07.400 --> 00:34:10.660 And I would suggest that these sequences are propagating 00:34:10.660 --> 00:34:15.460 through a stress transfer, while the slower sequences are 00:34:15.460 --> 00:34:21.120 propagating through slower processes such as pore pressure diffusion. 00:34:23.580 --> 00:34:26.350 If we look closely, we also see, in some sequences, 00:34:26.350 --> 00:34:31.390 what can be called a cessation front. So we have – from the same origin, 00:34:31.390 --> 00:34:35.410 we have a spreading, so of quiescence, behind which 00:34:35.410 --> 00:34:38.220 almost no earthquakes occur. 00:34:38.220 --> 00:34:42.650 So this corresponds to some kind of relaxation process here. 00:34:42.650 --> 00:34:47.020 This is another example where – see a very well-developed cessation front. 00:34:47.020 --> 00:34:50.270 But in this case, a triggering front is actually not very well-developed, 00:34:50.270 --> 00:34:55.220 where it’s actually kind of interrupted by this burst of activity 00:34:55.220 --> 00:34:59.190 in the middle of the sequence where seismicity ruptured outward 00:34:59.190 --> 00:35:04.030 by a couple kilometers from the origin in a very short time. 00:35:04.030 --> 00:35:07.820 But again, the cessation front was pretty striking to me. 00:35:07.820 --> 00:35:11.890 So I had a closer look onto this example here. 00:35:11.890 --> 00:35:16.020 So here are a couple panels. In this map view, I show the 00:35:16.020 --> 00:35:20.210 earthquakes like dots and the crosses correspond to injection wells 00:35:20.210 --> 00:35:23.200 color-coded by the distance from the first earthquake 00:35:23.200 --> 00:35:25.960 that we record in this sequence. 00:35:27.370 --> 00:35:30.100 So wells that are within 15 kilometers 00:35:30.110 --> 00:35:34.370 of the origin have this kind of injection schedule here. 00:35:34.370 --> 00:35:40.110 In red, I show the injection schedule for wells within 6 kilometers. 00:35:40.110 --> 00:35:43.680 And then, in green, it’s just this one single well which is less than 00:35:43.680 --> 00:35:48.820 a kilometer from what we determined as the origin of this sequence here. 00:35:48.820 --> 00:35:54.400 And if we zoom into this gray area on the right and look at – 00:35:54.400 --> 00:36:01.080 so this is the same timeframe now – and compare the occurrence of 00:36:01.080 --> 00:36:05.060 earthquakes with the injection schedule for this one particular well which sits 00:36:05.060 --> 00:36:10.150 close at the origin, we see a pretty good agreement with the 00:36:10.150 --> 00:36:15.640 ramp-up of this injection well to much higher values than it was used before 00:36:15.640 --> 00:36:18.480 and the occurrence of these earthquakes. 00:36:18.480 --> 00:36:23.110 And in this case, the operator realized that they have a well which is 00:36:23.110 --> 00:36:27.440 very well-correlated with seismicity, so they decided to shut it down. 00:36:27.440 --> 00:36:33.000 And, as an effect, we do have this declining pressure now spreading 00:36:33.000 --> 00:36:41.070 from this injection well. And as a – and we see this cessation front. 00:36:41.070 --> 00:36:46.950 So just to explain it a little further, if we start injecting – and in this cartoon, 00:36:46.950 --> 00:36:49.930 the black line represents the shear stress on a fault, 00:36:49.930 --> 00:36:53.390 and the blue line is the pore pressure perturbation. 00:36:53.390 --> 00:36:56.420 And in cases where the pore pressure perturbation crosses, 00:36:56.420 --> 00:37:02.330 or touches, the residential strength of this fault, we do get earthquakes. 00:37:02.330 --> 00:37:08.270 If you continue injecting, we continue driving seismicity outward, 00:37:08.270 --> 00:37:11.680 but we still get earthquakes behind the triggering front 00:37:11.680 --> 00:37:14.740 because we are still increasing pressure there. 00:37:17.060 --> 00:37:22.160 However, if we stop injecting, we still get the triggering front, 00:37:22.160 --> 00:37:27.070 and then we stop, so we are still propagating the triggering front outward. 00:37:27.070 --> 00:37:30.820 But behind, the pressure is actually declining, so we do not get any more 00:37:30.820 --> 00:37:36.640 earthquakes there. This is why we do observe this cessation front here. 00:37:36.640 --> 00:37:42.190 So these observations together let me conclude that the 00:37:42.190 --> 00:37:46.540 triggering front actually characterizes the seismogenic process. 00:37:46.540 --> 00:37:49.560 So it is a convolution of the pore pressure diffusion 00:37:49.560 --> 00:37:52.290 and other processes that come into play, 00:37:52.290 --> 00:37:58.350 such as static and dynamic stress transfer or other kind of earthquake interactions. 00:37:58.350 --> 00:38:02.360 And the cessation front actually characterizes hydraulic diffusivity since 00:38:02.360 --> 00:38:09.220 it shows us when the pore pressure falls behind its previous maximum value. 00:38:09.220 --> 00:38:12.460 And we see that in this kind of example here 00:38:12.460 --> 00:38:16.160 that it spreads over periods longer than a year. 00:38:16.160 --> 00:38:19.090 What is also interesting to notice that the diffusivity that we get 00:38:19.090 --> 00:38:23.170 from these cessation fronts fall into a pretty narrow range 00:38:23.170 --> 00:38:28.410 of 1 to – 1 to 3, 10 to the minus 3 square meters per second. 00:38:28.410 --> 00:38:30.850 And incidentally, these are the lowest values we also get 00:38:30.850 --> 00:38:35.740 for the triggering front for these very slow sequences 00:38:35.740 --> 00:38:39.540 where we do not see earthquake interactions to occur. 00:38:40.520 --> 00:38:44.500 All right. So just to repeat my key observations, 00:38:44.510 --> 00:38:48.890 seismicity occurs in distinct clusters that may span several faults. 00:38:48.890 --> 00:38:51.770 These sequences typically occur at about 4 kilometers below the top 00:38:51.770 --> 00:38:55.940 of the basement, which means that small stress perturbations 00:38:55.940 --> 00:39:01.600 at depth are sufficient to get widespread earthquake triggering. 00:39:01.600 --> 00:39:04.540 The majority of cluster shows a downward migration, 00:39:04.540 --> 00:39:08.100 which indicates that the driving source is above. 00:39:08.100 --> 00:39:11.980 The position of the largest events appears to be random for the smaller sequences, 00:39:11.980 --> 00:39:16.560 but there seems to be some kind of development for the larger sequences. 00:39:16.560 --> 00:39:18.940 So in order to reduce risk for large earthquakes, 00:39:18.940 --> 00:39:22.580 we have to reduce the overall activity. 00:39:24.440 --> 00:39:29.780 Events are triggered by fluid diffusion and by faster earthquake interactions. 00:39:29.780 --> 00:39:33.150 This means that faults are critically stressed throughout and are 00:39:33.150 --> 00:39:37.800 capable of sustaining rupture once they get initiated. 00:39:38.810 --> 00:39:41.420 No major sequence actually starts with the largest earthquakes, 00:39:41.420 --> 00:39:46.580 and many – and often many months before they rise to peak activity. 00:39:46.580 --> 00:39:50.620 So with an improved seismic monitoring here, we can actually have 00:39:50.620 --> 00:39:55.700 some kind of early warning where we see that there’s something developing. 00:39:56.800 --> 00:40:01.020 And to conclude, earthquake sequences are initiated by anthropogenic stressing 00:40:01.030 --> 00:40:04.930 but sustain rupture through earthquake interactions. 00:40:04.930 --> 00:40:08.570 Sequences initiate months before they rise to peak activity. 00:40:08.570 --> 00:40:10.970 And the largest earthquakes result from stronger perturbations 00:40:10.970 --> 00:40:14.940 but not from reactivating more fresh asperities. Thank you. 00:40:15.200 --> 00:40:22.340 [ Applause ] 00:40:24.100 --> 00:40:26.000 - Questions? 00:40:30.400 --> 00:40:34.840 - Thank you. That was a fantastic talk. Can you show the last slide again? 00:40:34.850 --> 00:40:41.830 Say that the rate peaks months after this first event. 00:40:41.830 --> 00:40:46.960 How – can you tell how much later than when they start the injection? 00:40:46.960 --> 00:40:48.980 Or you don’t often know when they started the injection. 00:40:48.980 --> 00:40:52.400 It could be years after they start the injection, right? 00:40:52.400 --> 00:40:57.360 - So in most cases, it’s just impossible to connect a 00:40:57.360 --> 00:41:00.020 single well with a single sequence. - Too many wells? 00:41:00.020 --> 00:41:04.530 - We have several hundred wells in the area that have been injecting for decades 00:41:04.530 --> 00:41:09.280 before this crisis actually occurred. And then there was just a ramp-up in, 00:41:09.280 --> 00:41:17.180 I don’t know, 2012 or so that caused these many earthquakes. 00:41:17.180 --> 00:41:21.980 So I think that’s just – it’s really difficult. 00:41:25.600 --> 00:41:30.140 - It’s just great to see the insights that are emerging from all this work. 00:41:31.740 --> 00:41:36.300 So there’s been a major decrease in seismicity in the last year or so. 00:41:36.300 --> 00:41:41.350 Is that – does that owe to management practices? Or what’s going on there? 00:41:41.350 --> 00:41:45.260 - So earthquake injection rates have been declining through 00:41:45.260 --> 00:41:51.640 mandated reductions, but also through, I guess, market forces. 00:41:53.580 --> 00:42:00.430 This is probably the major cause. So we are – I’m not quite sure 00:42:00.430 --> 00:42:04.670 how it’s going to develop after now, but it’s definitely reductions 00:42:04.670 --> 00:42:09.220 in injection rate that do cause this decline. 00:42:13.940 --> 00:42:17.120 - Yeah. Really nice body of work there, Martin. 00:42:17.170 --> 00:42:20.100 I was going to ask about the last conclusion, 00:42:20.100 --> 00:42:25.270 which was the larger earthquakes that are resulting from 00:42:25.270 --> 00:42:28.450 stronger perturbations rather than reacting fresh asperities. 00:42:28.450 --> 00:42:32.681 So you showed the kind of – the, I guess, map view of the 00:42:32.681 --> 00:42:38.260 length of the sequence versus time and whether that’s expanding or not. 00:42:38.260 --> 00:42:42.900 How do you account for changes either in depth or in – 00:42:42.900 --> 00:42:46.540 I think in some cases, there’s evidence that, you know, the fluid pressure 00:42:46.540 --> 00:42:52.340 changes are – result from injection from wells in multiple locations. 00:42:52.340 --> 00:42:57.140 So, like in the upper-left plot there, do we know that that event in the 00:42:57.140 --> 00:43:01.310 lower left – sorry, the – yeah, the upper- left plot – the event in the lower left, 00:43:01.310 --> 00:43:04.960 do we know that that one is caused sort of by the same fluid pressure front? 00:43:04.960 --> 00:43:08.430 Or could that be a fluid pressure perturbation coming from a different 00:43:08.430 --> 00:43:12.340 well and they’re sort of meeting in the middle of that fault? 00:43:13.300 --> 00:43:16.640 - So for this sequence, actually, this is the Fairview sequence, 00:43:16.640 --> 00:43:19.720 which is actually very far away from major injection wells. 00:43:19.720 --> 00:43:24.100 So it’s a pretty interesting sequence by itself just because it was triggered 00:43:24.100 --> 00:43:27.860 so far away from injection wells. So I’m pretty confident that 00:43:27.860 --> 00:43:34.830 the earthquake over here is driven by the same source as this one. 00:43:34.830 --> 00:43:38.840 As to the other comment on depth, for now, I don’t look at depth. 00:43:38.840 --> 00:43:42.520 And I know this is a problem in this characterization, 00:43:42.520 --> 00:43:46.900 so I might have to think about how to incorporate depth here. 00:43:48.040 --> 00:43:49.740 But I … 00:43:58.020 --> 00:44:03.690 I think for these larger sequences, they don’t – well, so typically, 00:44:03.690 --> 00:44:06.850 the depth extent of the sequences are on the order of 2 kilometers, 00:44:06.850 --> 00:44:10.270 so this would go to magnitude 4. So maybe there is definitely – 00:44:10.270 --> 00:44:14.720 there might be a component of depth in there I should look into, yeah. 00:44:14.720 --> 00:44:20.900 - In this case, is it – it looks like it – the sequence broadens a bit toward the end. 00:44:20.900 --> 00:44:24.320 Is that … - It’s a dipping structure. 00:44:24.320 --> 00:44:27.450 - Okay. - Yep. It’s like a dip of 60, 70 degrees. 00:44:27.450 --> 00:44:32.500 - Okay. So there is – it’s maybe showing an increase in the depth extent there 00:44:32.500 --> 00:44:36.500 that’s reflected by that broadening, or at least filling out that … 00:44:36.500 --> 00:44:38.690 - Potentially, yeah. 00:44:39.400 --> 00:44:42.660 - What’s the time span for that figure? 00:44:44.840 --> 00:44:50.520 - I think it’s something like 1-1/2 years, maybe. I’m not quite sure. 00:44:52.840 --> 00:45:04.820 [ Silence ] 00:45:05.480 --> 00:45:11.060 - So I’m curious about the plot of the earthquake locations versus depth. 00:45:11.060 --> 00:45:18.940 And first question is, so I think you mentioned it, but you’re taking account 00:45:18.940 --> 00:45:24.760 of the change in depth of the Arbuckle or the basement contact there? 00:45:24.760 --> 00:45:26.940 - Yep. - Okay. 00:45:26.940 --> 00:45:31.010 And then, you know, if you put that in together with 00:45:31.010 --> 00:45:39.690 how you’ve actually defined your faults, do you find that some of the seismicity 00:45:39.690 --> 00:45:46.220 in those clusters would reach up to the Arbuckle contact in most of those cases? 00:45:46.220 --> 00:45:49.310 Or are these faults – seem more confined to that 00:45:49.310 --> 00:45:52.640 kind of 4 kilometers below the contact? 00:45:52.640 --> 00:45:55.910 Or could you comment a little bit on what you see there? 00:45:55.910 --> 00:46:00.560 - We do see single sequences that rupture probably to the contact – 00:46:00.560 --> 00:46:10.440 Cushing, for instance. I think most of the sequences, though, do stay deep. Yeah. 00:46:10.440 --> 00:46:14.710 So I guess, for most of sequences, we can say that their vertical extent 00:46:14.710 --> 00:46:21.180 is about 2-ish kilometers except for the very big ones. 00:46:22.580 --> 00:46:25.180 So they would stay at depth. 00:46:26.620 --> 00:46:28.140 - Nice talk, Martin. 00:46:28.140 --> 00:46:33.660 I actually want to return to what Dave and Randy were asking you about. 00:46:33.670 --> 00:46:38.780 I was – I was sort of wondering, when you – when you have these sort of longer 00:46:38.780 --> 00:46:45.660 fault structures, I would expect that you have more earthquakes in general. 00:46:45.660 --> 00:46:48.510 And so couldn’t – just to play devil’s advocate, couldn’t you 00:46:48.510 --> 00:46:53.150 just be seeing the biggest earthquakes in these sort of sequences due to 00:46:53.150 --> 00:46:57.020 a statistical effect, basically? The more earthquakes you have, 00:46:57.020 --> 00:47:00.250 the more likely you are to have big earthquakes, and it’s 00:47:00.250 --> 00:47:07.100 not that it’s a pressure effect. It’s just a statistical result. 00:47:07.100 --> 00:47:09.240 - Certainly. [laughter] 00:47:10.980 --> 00:47:13.160 I mean, yes, if we have more earthquakes, we also 00:47:13.160 --> 00:47:20.370 increase just the likelihood of rolling the dice with a better number. 00:47:20.370 --> 00:47:24.400 But then the question is, why are we seeing more earthquakes? [laughs] 00:47:24.400 --> 00:47:26.200 So, yeah. 00:47:29.060 --> 00:47:34.780 - Yeah, is there a depth dependence of magnitude or a magnitude 00:47:34.790 --> 00:47:38.750 dependence on depth? And in particular, in the next slide, 00:47:38.750 --> 00:47:46.390 what are the depths of those four or five earthquakes that go from 3.1 to 4.4? 00:47:46.390 --> 00:47:48.060 In the next slide? 00:47:48.060 --> 00:47:51.780 - Next – I’m not sure which one is the next slide you meant. 00:47:51.780 --> 00:47:55.160 - Well, the one where … - This one? 00:47:57.080 --> 00:47:58.700 - The Fairview slide. - Oh. 00:47:58.700 --> 00:48:01.610 - The – yeah, that one. 00:48:01.610 --> 00:48:06.160 So what are the depths of – are they all at the same depth, or … 00:48:06.160 --> 00:48:12.830 - I didn’t look into depth too much for magnitudes for now. 00:48:12.830 --> 00:48:18.740 I should look into this depth dependence of magnitude. I still have to do that. 00:48:18.740 --> 00:48:22.500 - And then in the following slide, how did you estimate the stress drop 00:48:22.500 --> 00:48:26.620 for those various earthquakes? - So these lines of stress drop are 00:48:26.620 --> 00:48:34.860 just kind of equivalent stress drop taking the size of the reactivated fault, 00:48:34.860 --> 00:48:38.030 taking this as – diameter of a rupture source. 00:48:38.030 --> 00:48:42.201 And then, in this case, the magnitude gives us the … 00:48:42.201 --> 00:48:46.840 - Okay, so you spread the slip over the whole length of the … 00:48:46.840 --> 00:48:48.910 - Right. Yes. - … fault as it existed at that time. 00:48:48.910 --> 00:48:51.040 - So these have nothing to do with actual stress drops of earthquakes. 00:48:51.040 --> 00:48:55.500 It’s just kind of the – a proxy for the energy release in a single sequence. 00:48:55.500 --> 00:48:57.820 - Okay. Thank you. 00:49:00.660 --> 00:49:08.600 [ Silence ] 00:49:09.500 --> 00:49:13.320 - I’m guessing there’s no observation wells that go 00:49:13.320 --> 00:49:16.840 3 or 4 kilometers into basements. So these changes in conditions 00:49:16.840 --> 00:49:20.240 of seismogenic depths are inferred from models? 00:49:20.240 --> 00:49:23.800 - We wish. [laughs] - Did you have – aren’t there data from – 00:49:23.800 --> 00:49:28.120 are there downhole data from, say, the injection depth 00:49:28.120 --> 00:49:30.600 these days that are available? 00:49:30.600 --> 00:49:35.600 - Well, Andy or Bob, do you want to comment on this? [laughs] 00:49:39.160 --> 00:49:43.840 - The answer is, it’s complicated. We have one monitoring well 00:49:43.850 --> 00:49:47.460 that’s basically outside of where earthquakes are at this point. 00:49:47.460 --> 00:49:53.510 And the OJS has monitoring wells that is inside the seismically active zone. 00:49:53.510 --> 00:49:58.980 So these are old injection wells that have been instrumented. 00:49:58.980 --> 00:50:03.060 Although we haven’t seen any of those data basically, so … 00:50:04.100 --> 00:50:06.480 - No? [laughter] 00:50:06.480 --> 00:50:09.300 - We’re waiting for more. - Yeah. We’re trying for more. 00:50:11.960 --> 00:50:17.560 [ Silence ] 00:50:18.260 --> 00:50:20.680 - Any more – any more questions? 00:50:26.620 --> 00:50:28.980 All right, well, let’s thank our speaker. 00:50:28.980 --> 00:50:33.380 [ Applause ] 00:50:33.380 --> 00:50:35.860 If you guys want to grab lunch with Martin, 00:50:35.860 --> 00:50:41.820 we’ll be outside the pole, like, around 11:40. 00:50:41.820 --> 00:50:43.500 Thank you. 00:50:46.940 --> 00:50:54.160 [ Silence ]