WEBVTT Kind: captions Language: en 00:00:00.080 --> 00:00:04.350 … for today, November 8th. Just a couple of announcements 00:00:04.350 --> 00:00:09.210 about upcoming seminars. Next week, at the same time and place, 00:00:09.210 --> 00:00:14.389 Alan Yong will be – from our Pasadena office will be talking about 00:00:14.389 --> 00:00:18.510 surface wave measurements for site response. 00:00:18.510 --> 00:00:22.380 And we have an added bonus presentation the following day 00:00:22.380 --> 00:00:26.599 at 10:00 a.m. – note the time – 10:00 a.m., not 10:30. 00:00:26.599 --> 00:00:31.439 Art Frankel will be talking about his simulations of strong ground motions 00:00:31.439 --> 00:00:36.490 for magnitude 9 Cascadia events, so that’s – we have two presentations 00:00:36.490 --> 00:00:38.930 next week, Wednesday and Thursday, 00:00:38.930 --> 00:00:42.860 and no presentation on the week of Thanksgiving. 00:00:42.860 --> 00:00:48.400 So I’d like to introduce Ole Kaven who will introduce our speaker. 00:00:49.580 --> 00:00:53.020 Thanks, Tom. It’s my pleasure to introduce Rob Skoumal today. 00:00:53.030 --> 00:00:56.150 He’s Mendenhall in the induced seismicity project 00:00:56.150 --> 00:00:59.490 and started working for us last year in August. 00:00:59.490 --> 00:01:03.420 Rob did his undergraduate work at Eckerd College in Florida and then 00:01:03.420 --> 00:01:08.280 started working with Mike Brudzinski at University of Miami in Ohio. 00:01:08.280 --> 00:01:15.180 And during his master’s and Ph.D., he’s really taken earthquake detection 00:01:15.180 --> 00:01:18.850 from merely template matching to very novel techniques of 00:01:18.850 --> 00:01:23.950 finding events where no templates exist prior. 00:01:23.950 --> 00:01:28.320 And this work has come in really handy for a number of applications, 00:01:28.320 --> 00:01:32.940 mostly in induced seismicity in Ohio during the Ph.D., but also Oklahoma, 00:01:32.950 --> 00:01:39.200 which he’ll talk about today. And here at the Survey, he’s working on 00:01:39.200 --> 00:01:44.950 microseismicity at the Decatur, Illinois, CO2 sequestration site, but has also 00:01:44.950 --> 00:01:48.250 really gone and branched out and worked on seismicity in all sorts of 00:01:48.250 --> 00:01:53.460 different settings, which I think he will get to a little bit today. So, Rob. 00:01:56.260 --> 00:01:58.060 - All right, thanks, Ole. 00:01:58.080 --> 00:02:01.800 So the work I’ll be talking about today benefited greatly from conversations 00:02:01.800 --> 00:02:04.980 with many people here at the Survey, many of – who are in this room. 00:02:04.980 --> 00:02:08.560 But I’d specifically like to acknowledge Mike Brudzinski and Brian Currie at 00:02:08.560 --> 00:02:13.660 Miami University and our very own Ole, Phil, and Steve here at the Survey. 00:02:14.520 --> 00:02:18.700 So I thought I would be talking about kind of a pretty broad set of topics – 00:02:18.709 --> 00:02:20.740 kind of summarize some of the stuff that I’ve been working on over the 00:02:20.740 --> 00:02:24.340 past year here at the Survey. And the first thing I’d like to talk about 00:02:24.340 --> 00:02:28.800 is kind of my efforts to improve our ability to detect repeating earthquakes. 00:02:28.800 --> 00:02:34.300 Then I’ll be applying that type of analysis to – in order to characterize 00:02:34.300 --> 00:02:38.260 the likelihood of injection-induced seismicity across the United States. 00:02:38.260 --> 00:02:41.960 I’ll be trying to scale up those approaches and study induced seismicity 00:02:41.960 --> 00:02:44.840 in Oklahoma. I know many of us know that Oklahoma is the 00:02:44.840 --> 00:02:46.840 most seismically active place in the country right now. 00:02:46.840 --> 00:02:49.960 So it’s kind of a large-scale-type approach. 00:02:49.960 --> 00:02:53.160 Then I’ll be finishing with a pretty interesting case study in Oroville, 00:02:53.160 --> 00:02:58.260 California, looking at some interesting induced seismic events. 00:02:58.270 --> 00:03:02.190 So a lot of my previous work has tried to argue that injection-induced 00:03:02.190 --> 00:03:05.540 seismicity is frequently characterized by swarms of earthquakes. 00:03:05.540 --> 00:03:08.340 And by swarms, I mean that we have – often have hundreds or thousands 00:03:08.340 --> 00:03:11.860 of earthquakes that occur within a relatively short time window. 00:03:11.860 --> 00:03:15.780 Right, and because these events kind of lack a main shock event, 00:03:15.780 --> 00:03:18.720 we kind of classify these as swarms. 00:03:18.720 --> 00:03:22.220 Right, and so my work has tried to say that we could use this 00:03:22.220 --> 00:03:25.280 swarm-like characteristic as a means to distinguish naturally occurring 00:03:25.280 --> 00:03:30.599 seismicity from these induced seismicity throughout the mid-continent. 00:03:30.599 --> 00:03:33.379 So if we look in Ohio, right, I think one of the easiest ways 00:03:33.379 --> 00:03:37.069 to try to quantify the swarm-like behavior is simply to compare the 00:03:37.069 --> 00:03:40.270 number of events in a sequence, along this Y axis, 00:03:40.270 --> 00:03:42.240 to the largest magnitude event in that sequence. 00:03:42.240 --> 00:03:44.790 All right, so if we do that, we see that there’s this kind of separate 00:03:44.790 --> 00:03:48.120 population of induced earthquakes up here in this green box, which is 00:03:48.120 --> 00:03:52.180 very distinct from the naturally occurring kind of background tectonic events. 00:03:52.180 --> 00:03:54.980 If we apply the same kind of analysis in Oklahoma or Texas, 00:03:54.980 --> 00:03:57.270 which, you know, nowadays, virtually all of the earthquakes 00:03:57.270 --> 00:04:02.020 there are induced, we see a similar kind of relationship. 00:04:02.020 --> 00:04:04.810 So this type of analysis, and many others in induced seismicity, 00:04:04.810 --> 00:04:07.480 are limited by our magnitude of completeness. 00:04:07.480 --> 00:04:10.879 Right, so using the existing networks and traditional detection approaches, 00:04:10.879 --> 00:04:13.200 we kind of usually get a magnitude of completeness somewhere in the 00:04:13.200 --> 00:04:18.020 magnitude 2 to 3 range – probably closer to 3 for most of the United States. 00:04:18.020 --> 00:04:20.949 All right, so a lot of my work has been trying to lower this magnitude 00:04:20.949 --> 00:04:24.460 of completeness so that we can better characterize these sequences. 00:04:24.460 --> 00:04:28.280 So one good example of this is a hydraulic fracturing case study in Ohio. 00:04:28.280 --> 00:04:30.900 All right, so in this example, a couple earthquakes were cataloged, 00:04:30.909 --> 00:04:35.229 sort of in the mid-magnitude 2 range. Now, with just this information, I mean, 00:04:35.229 --> 00:04:38.990 it’s virtually impossible to conclusively say whether or not these earthquakes 00:04:38.990 --> 00:04:43.039 were induced, even it occurred right next to a hydraulically fractured well, right? 00:04:43.039 --> 00:04:45.710 But using some correlation approaches, we can turn those two earthquakes 00:04:45.710 --> 00:04:48.550 into thousands of earthquakes. Now, all the sudden, we can look at 00:04:48.550 --> 00:04:50.719 this and say, you know, this does not look natural. 00:04:50.719 --> 00:04:53.930 Right, we see these very tight, temporal clusters of earthquakes 00:04:53.930 --> 00:04:55.509 that are much of this much larger swarm. 00:04:55.509 --> 00:05:00.140 I mean, this is very abnormal for inter-plate naturally occurring seismicity. 00:05:00.780 --> 00:05:03.320 It turns out this is very typical for hydraulically fractured 00:05:03.330 --> 00:05:04.800 induced seismicity. Right, so if you add 00:05:04.800 --> 00:05:07.610 the industry information – all right, so this red line here 00:05:07.610 --> 00:05:10.039 is the time period that was hydraulically fractured. 00:05:10.040 --> 00:05:12.420 We said that there’s no earthquakes prior to hydraulic fracturing. 00:05:12.420 --> 00:05:14.520 We get a bunch of earthquakes when they’re hydraulically fracturing, 00:05:14.520 --> 00:05:16.700 and the earthquakes decay afterwards. 00:05:16.700 --> 00:05:18.909 Right, now these little tiny black bars here, 00:05:18.909 --> 00:05:22.740 those are the individual stimulations they did at this – at this well. 00:05:22.740 --> 00:05:25.240 And so with this information, now we can start to do some real – 00:05:25.240 --> 00:05:26.710 some real meaningful science, right? 00:05:26.710 --> 00:05:29.289 We can look at which stages caused earthquakes and which ones did not. 00:05:29.289 --> 00:05:31.469 We could try to locate these earthquakes and identify 00:05:31.469 --> 00:05:33.740 the fault network that’s responsible for them. 00:05:33.740 --> 00:05:36.900 We can look at the distance between the stimulations and the earthquakes. 00:05:36.900 --> 00:05:39.760 We could try to create poroelastic stress models that explain the seismicity. 00:05:39.760 --> 00:05:41.220 You know, there’s a lot of different work that we can do, 00:05:41.229 --> 00:05:43.580 but we need to have these good observations first, right? 00:05:43.580 --> 00:05:46.740 So that’s really what a lot of my work has been dedicated towards. 00:05:47.480 --> 00:05:49.240 So sort of the problem that I’m trying to address 00:05:49.240 --> 00:05:53.379 is that traditional earthquake detection is heavily reliant on relatively 00:05:53.380 --> 00:05:57.300 high signal-to-noise recorded at multiple seismometers. 00:05:57.300 --> 00:06:01.419 Right, now kind of the traditional solution to this is simply go out and 00:06:01.419 --> 00:06:04.389 deploy more seismometers, right? Now, this is obviously the 00:06:04.389 --> 00:06:06.639 best solution, right, but it’s not always the most practical. 00:06:06.639 --> 00:06:08.710 You know, if you’re trying to monitor induced seismicity across 00:06:08.710 --> 00:06:11.480 the United States, we can’t always deploy seismometers where we 00:06:11.480 --> 00:06:15.180 want them. Right, so let’s try to take advantage of this fact 00:06:15.180 --> 00:06:18.189 that induced seismicity occurs as repeating swarms. 00:06:18.189 --> 00:06:22.029 And we can do that by trying to reduce our reliance on signal-to-noise 00:06:22.029 --> 00:06:26.140 by instead shifting the focus for looking for signals that repeat. 00:06:26.140 --> 00:06:30.639 Okay, and so my approach to do this is called the Repeating Signal Detector. 00:06:30.639 --> 00:06:34.379 Right, it’s not a very creative name, but it’s pretty applicable. 00:06:34.379 --> 00:06:37.809 So the – kind of that – this is more of a kind of a general framework, right? 00:06:37.809 --> 00:06:40.409 Kind of a family of characterizations. 00:06:40.409 --> 00:06:42.729 And each of these different kind of steps that I’ll be talking about, you can 00:06:42.729 --> 00:06:46.360 substitute different methods in there depending on what you want to do. 00:06:46.360 --> 00:06:49.040 All right, so the first step is to identify signals of interest, right? 00:06:49.050 --> 00:06:51.389 This is just some sort of input data. 00:06:51.389 --> 00:06:55.440 You could apply a single station, short-term/long-term average to a station. 00:06:55.440 --> 00:06:56.960 You could give it a known earthquake catalog. 00:06:56.960 --> 00:06:59.200 Or you could just give it raw continuous data. 00:06:59.200 --> 00:07:02.760 Like, whatever it is, you have to give it data from at least one seismometer. 00:07:02.760 --> 00:07:05.509 Right, then what it does is it takes the signals of interest 00:07:05.509 --> 00:07:07.330 and converts them to the frequency domain. 00:07:07.330 --> 00:07:10.400 Right, so in this example, we have two different signals of interest. 00:07:10.400 --> 00:07:13.460 I’ve calculated the Fourier transform for the three different channels. 00:07:13.460 --> 00:07:16.120 Okay, what we want to do is then concatenate those 00:07:16.139 --> 00:07:18.669 Fourier transforms across those channels, so that way, 00:07:18.669 --> 00:07:22.440 each signal of interest is now represented by a single array. 00:07:22.440 --> 00:07:24.900 Now, instead of thinking of these as arrays, it might be helpful to 00:07:24.900 --> 00:07:29.180 think of them as some sort of a single point in some higher dimension. 00:07:29.180 --> 00:07:30.979 Right, it might be helpful to think of these as points 00:07:30.979 --> 00:07:35.479 because we want to start clustering these signals of interest, right? 00:07:35.479 --> 00:07:38.499 So this example – right, this is a 2D plot. You can imagine that we have 00:07:38.500 --> 00:07:41.580 these plots – these little points scattered out there in space. 00:07:41.580 --> 00:07:43.360 And what we want to do is we want to start to cluster these 00:07:43.360 --> 00:07:46.339 into different families. Right, so clustering has its own 00:07:46.339 --> 00:07:50.459 scientific discipline, right? It’s a very interesting field. 00:07:50.459 --> 00:07:53.020 So previously I’ve used just kind of simple 00:07:53.020 --> 00:07:55.020 agglomerative clustering to cluster these events. 00:07:55.020 --> 00:07:57.760 But now I’m trying to explore other types of methods, one of them being 00:07:57.770 --> 00:08:03.360 the Ordering Points to Identify the Clustering Structure – called OPTICS. 00:08:03.360 --> 00:08:06.110 But this is kind of relevant, but the main point here is that 00:08:06.110 --> 00:08:08.460 you just kind of pick a clustering algorithm. If you have 00:08:08.460 --> 00:08:12.400 favorite method, you know, use that. That’s great. It’s very adaptable. 00:08:12.400 --> 00:08:15.370 But in this example, let’s say we identified these three different families. 00:08:15.370 --> 00:08:18.110 All right, the next step that we want to do is discard families 00:08:18.110 --> 00:08:20.919 that have few members, the idea here being that we 00:08:20.919 --> 00:08:22.969 want to identify signals that repeat, right? 00:08:22.969 --> 00:08:25.740 So if a signal only has one or two family members, 00:08:25.740 --> 00:08:28.280 maybe it’s not repetitive and we can discard it. 00:08:28.280 --> 00:08:31.420 Right, but if you want to use those families, you can keep them, right? 00:08:31.420 --> 00:08:35.020 You know, you should suit this to your own needs. 00:08:35.020 --> 00:08:38.440 All right, so this kind of procedure so far is what I call the frequency domain 00:08:38.440 --> 00:08:42.020 sorting step, right, because we’re sorting signals in the frequency domain. 00:08:42.020 --> 00:08:45.940 All right, and this is a pretty effective way to try to start to 00:08:45.940 --> 00:08:48.050 initially group signals that are similar. 00:08:48.050 --> 00:08:51.340 So in this example, we have hundreds of waveforms that look fairly similar. 00:08:51.340 --> 00:08:54.240 But the differential phase arrival times vary, right? 00:08:54.240 --> 00:08:57.180 This is just due to slight changes in source location. 00:08:57.180 --> 00:09:00.130 But depending on our end result, we may want to address that issue. 00:09:00.130 --> 00:09:04.640 Right, and we can address that by repeating the sort of sorting step, but this 00:09:04.640 --> 00:09:08.630 time doing it for each individual family, and this time in the time domain. 00:09:08.630 --> 00:09:11.580 All right, so instead of concatenating those Fourier transforms, 00:09:11.580 --> 00:09:14.400 we’re now concatenating the band pass waveforms. 00:09:14.400 --> 00:09:19.270 Right, so the end result of this is, hopefully, a family of events 00:09:19.270 --> 00:09:22.650 that are nicely aligned. And once they’re aligned, 00:09:22.650 --> 00:09:26.030 then we can stack them together, greatly improve the signal-to-noise ratio, 00:09:26.030 --> 00:09:28.720 and this is sort of the characteristic repeating signal. 00:09:28.720 --> 00:09:32.340 Right, this kind of signal is gold in terms of template matching. 00:09:32.340 --> 00:09:34.270 Because now we can take that signal and cross-correlate it 00:09:34.270 --> 00:09:37.640 for years of data and identify even smaller magnitude events. 00:09:37.640 --> 00:09:40.440 All right, so kind of a huge point here is that, even if none of these 00:09:40.450 --> 00:09:42.830 earthquakes were large enough to be cataloged, we can still 00:09:42.830 --> 00:09:46.080 identify those repeating signals and identify them. 00:09:46.760 --> 00:09:51.400 Okay, so one case I’d like to introduce is Harrison County in Ohio. 00:09:51.400 --> 00:09:54.640 So this is kind of eastern Ohio, and this is kind of a zoomed-in 00:09:54.640 --> 00:09:58.220 map of the – of the area. Each of these different-colored lines 00:09:58.220 --> 00:10:02.010 is a different horizontal well that was drilled and fracked. 00:10:02.010 --> 00:10:05.460 And once we apply RSD, we found out that all of the hydraulically fractured 00:10:05.460 --> 00:10:09.800 wells within this red oval – all of those wells have induced seismicity. 00:10:09.800 --> 00:10:12.400 All right, so we can look at this in a slightly different way. 00:10:12.400 --> 00:10:14.890 Right, so if we look at the magnitude over time, 00:10:14.890 --> 00:10:18.120 we see there are a bunch of different earthquakes over this time period. 00:10:18.120 --> 00:10:19.990 And we can also look at the hydraulic fracturing period. 00:10:19.990 --> 00:10:22.640 Right, so this is showing the hydraulically fractured wells 00:10:22.640 --> 00:10:23.820 up to 10 kilometers. 00:10:23.820 --> 00:10:27.680 But in this case, we had induced seismicity up to 20 kilometers away. 00:10:28.400 --> 00:10:31.460 If we look at the catalogs – so the ANSS catalog identified 00:10:31.460 --> 00:10:33.562 one earthquake in this sequence. The Ohio Department of Natural 00:10:33.562 --> 00:10:36.490 Resources initially identified two, but after talking to them, 00:10:36.490 --> 00:10:38.630 they were able to increase it up to six. 00:10:38.630 --> 00:10:42.780 But RSD – whoops – RSD identified 14,000 events. 00:10:42.780 --> 00:10:45.440 Right, so kind of the main point here is that template matching would be 00:10:45.450 --> 00:10:49.410 reliant on these cataloged events, but RSD operates completely 00:10:49.410 --> 00:10:51.070 independently from those catalogs, right? 00:10:51.070 --> 00:10:53.480 Even if none of these earthquakes were cataloged, 00:10:53.480 --> 00:10:56.560 we would still identify these 14,000 events. 00:10:56.560 --> 00:10:59.820 All right, so this is kind of an example of kind of more 00:10:59.820 --> 00:11:02.150 local seismicity – you know, out to 20 kilometers. 00:11:02.150 --> 00:11:04.860 But we’ve also applied this up to 50 kilometers away to study 00:11:04.860 --> 00:11:07.080 hydraulic fracturing induced seismicity in Alberta. 00:11:07.080 --> 00:11:10.720 We’ve applied it to subduction-related swarms in Mexico 00:11:10.720 --> 00:11:13.720 and volcanic swarms in – here in California. 00:11:13.720 --> 00:11:15.160 So pretty much the kind of idea is that, 00:11:15.160 --> 00:11:18.130 as long as the signal repeats over time and you have modest 00:11:18.130 --> 00:11:22.040 signal-to-noise ratio, you should be able to identify those events. 00:11:22.860 --> 00:11:24.940 So we’ve tried to kind of scale up this approach and 00:11:24.950 --> 00:11:28.890 apply this across the United States in order to investigate the parameters 00:11:28.890 --> 00:11:32.440 that control the likelihood of inducing earthquakes. 00:11:32.920 --> 00:11:35.080 So, so far, we’ve included analysis in the Appalachian, 00:11:35.080 --> 00:11:38.860 Illinois, and Williston Basins. All right, so in kind of broad terms, 00:11:38.860 --> 00:11:40.980 the Appalachian Basin has had a lot of induced seismicity, 00:11:40.980 --> 00:11:43.700 Illinois has had some, and the Williston Basin has, 00:11:43.700 --> 00:11:47.900 you know, very little evidence for any induced seismicity. 00:11:47.900 --> 00:11:50.670 So initially, we were inspired to look at this kind of idea because 00:11:50.670 --> 00:11:55.590 Cliff Frohlich published a paper in SRL, I think in 2015, and he was looking for 00:11:55.590 --> 00:11:57.330 induced seismicity in the Williston Basin, and he pretty much – 00:11:57.330 --> 00:11:59.400 you know, he found little evidence for it. 00:11:59.400 --> 00:12:02.030 And his main conclusion was, you know – he pretty much asked 00:12:02.030 --> 00:12:04.610 the question, you know, why are there so few 00:12:04.610 --> 00:12:06.460 induced earthquakes in Williston Basin? 00:12:06.460 --> 00:12:11.310 So later work has tried to argue that the injection rate of disposal wells 00:12:11.310 --> 00:12:14.930 is a primary control over the likelihood of induced seismicity. 00:12:14.930 --> 00:12:16.920 But that doesn’t really make much sense in this case, right? 00:12:16.920 --> 00:12:20.330 So the injection wells in the Williston Basin, shown here by the green bars, 00:12:20.330 --> 00:12:24.280 are much larger than anything we have in the Appalachian Basin. 00:12:24.280 --> 00:12:27.680 All right, so this doesn’t really explain the absence of seismicity 00:12:27.682 --> 00:12:31.351 in the Williston Basin. So it’s not simply, you know, 00:12:31.351 --> 00:12:34.600 the more you inject, the greater the likelihood of seismicity. 00:12:34.600 --> 00:12:38.280 You know, there’s some other factor that’s more important. 00:12:38.820 --> 00:12:40.780 All right, so I’d like to kind of step through these different basins 00:12:40.780 --> 00:12:43.820 pretty quickly. But first I’d like to look at the Appalachian Basin. 00:12:43.820 --> 00:12:47.520 All right, so this is a map of the area. We kind of have Ohio here on the west, 00:12:47.520 --> 00:12:50.790 Pennsylvania on the east, West Virginia, Virginia. 00:12:50.790 --> 00:12:53.450 And each of these colored symbols represents all of the hydraulically 00:12:53.450 --> 00:12:56.430 fractured and wastewater disposal wells in this basin. 00:12:56.430 --> 00:12:59.680 Now, the colors of those wells correspond to the proximity 00:12:59.680 --> 00:13:03.090 to the basement. Okay, so the cooler colors are closer to the basement. 00:13:03.090 --> 00:13:05.720 The warmer colors are farther away from the basement. 00:13:05.720 --> 00:13:07.610 So kind of the most obvious trend you might be able to notice 00:13:07.610 --> 00:13:10.980 is that the wells in Ohio tend to be closer to the basement. 00:13:10.980 --> 00:13:16.370 All right, this is largely due to the target formation in Ohio being the Utica Shale. 00:13:16.370 --> 00:13:20.170 But almost as soon as you cross over the political boundary into Pennsylvania, 00:13:20.170 --> 00:13:25.100 West Virginia, the target formation jumps up to the shallower Marcellus. 00:13:25.100 --> 00:13:28.310 All right, so if we look at a cross-section, shown here by the little purple rectangle, 00:13:28.310 --> 00:13:31.370 it might be easier to see. All right, so this is that cross-section. 00:13:31.370 --> 00:13:35.970 All right, so this dark gray portion is our basement model for this area, 00:13:35.970 --> 00:13:40.080 with all the lighter gray sediments being the upper sedimentary strata. 00:13:40.700 --> 00:13:43.720 So here we – and this red line, that’s the Utica-Point Pleasant Shale. 00:13:43.730 --> 00:13:47.140 And this Marcellus is represented there by the green line. 00:13:47.140 --> 00:13:50.010 Now, between the Marcellus and the Utica, we’ve also highlighted 00:13:50.010 --> 00:13:51.800 this area called the Salina Evaporites. 00:13:51.800 --> 00:13:53.610 Right, so this is a very thick layer of evaporites, 00:13:53.610 --> 00:13:56.520 you know, hundreds of meters thick in some areas. 00:13:56.520 --> 00:13:58.800 So throughout our analysis, we also tried to look for geologic 00:13:58.800 --> 00:14:03.240 explanations that might help control the likelihood of induced seismicity. 00:14:03.240 --> 00:14:07.390 The idea here with evaporites is that, if there’s some sort of fluid flow 00:14:07.390 --> 00:14:10.660 or poroelastic stresses across the evaporites, those evaporites 00:14:10.660 --> 00:14:12.490 would seek to inhibit those stress transfers. 00:14:12.490 --> 00:14:15.460 All right, so if there’s a fault immediately below the evaporites, 00:14:15.460 --> 00:14:18.820 the chances of inducing seismicity might be less. 00:14:18.820 --> 00:14:22.460 So across this cross-section, we had four cases of induced seismicity. 00:14:22.460 --> 00:14:26.490 We had the injection – sorry, wastewater disposal in Trumbull County, 00:14:26.490 --> 00:14:30.050 wastewater disposal in Youngtown, and then two cases of hydraulic fracturing 00:14:30.050 --> 00:14:35.180 induced seismicity in the Utica Shale in Poland Township and North Beaver. 00:14:35.180 --> 00:14:39.130 Okay, so all these cases of induced seismicity along this 00:14:39.130 --> 00:14:42.640 transect were within 700 meters of the basement. 00:14:42.640 --> 00:14:46.620 All right, but if we jump over to the Illinois or Williston Basin, 00:14:46.620 --> 00:14:48.779 it looks a lot different. 00:14:48.779 --> 00:14:51.900 So in Illinois, the injection tends to be very shallow, 00:14:51.900 --> 00:14:54.880 with the exception being the Decatur carbon sequestration well, 00:14:54.880 --> 00:14:57.090 which is injecting into the Mount Simon Sandstone, 00:14:57.090 --> 00:15:00.500 which is a basal formation right above the basement. 00:15:00.500 --> 00:15:02.100 So I don’t think it’s a coincidence that that’s the 00:15:02.110 --> 00:15:05.510 one case of induced seismicity [chuckles] in this area. 00:15:05.510 --> 00:15:08.170 In the Williston Basin – so, again, although they’re injecting 00:15:08.170 --> 00:15:11.860 very large volumes, those volumes are very shallow, all right? 00:15:11.860 --> 00:15:14.490 So the Bakken/Three Forks does approach kind of that 00:15:14.490 --> 00:15:18.150 1-kilometer boundary away from the basement. 00:15:18.150 --> 00:15:21.350 But maybe that’s a little bit still too far away. 00:15:21.350 --> 00:15:23.320 Also there’s this little thin layer of evaporites. 00:15:23.320 --> 00:15:27.070 It does pinch out, so it’s not present throughout the basin. 00:15:27.070 --> 00:15:30.339 So this alone does not explain the absence of seismicity in the 00:15:30.339 --> 00:15:34.610 Williston Basin. All right, so I kind of skipped through a lot of the geology in 00:15:34.610 --> 00:15:38.560 these areas, but that’s okay because we’re seismologists, right? [laughter] 00:15:38.560 --> 00:15:42.980 But to try to summarize those results, right – 00:15:42.980 --> 00:15:44.550 so this is an analysis of all those wells. 00:15:44.550 --> 00:15:48.060 We’ve gone in and characterized all the seismicity in these three different basins. 00:15:48.060 --> 00:15:51.710 And this is looking at the percentage of those wells 00:15:51.710 --> 00:15:54.980 at these different distances and the chances that they induce seismicity. 00:15:54.980 --> 00:15:57.970 All right, so if you’re injecting directly into the basement, 00:15:57.970 --> 00:16:02.140 which is not a good idea, you have about a 20% chance of inducing earthquakes. 00:16:02.140 --> 00:16:03.690 But as you get shallower, you know, further away 00:16:03.690 --> 00:16:06.500 from the basement, the likelihood decreases. 00:16:06.500 --> 00:16:10.170 All right, so our kind of conclusion So far is that, while injection volume 00:16:10.170 --> 00:16:13.810 and rate likely contribute to the likelihood of induced seismicity, 00:16:13.810 --> 00:16:14.980 they cannot be the primary control. 00:16:14.980 --> 00:16:18.310 Right, there are other factors that are more important, right? 00:16:18.310 --> 00:16:20.230 So what factors are these, right? [chuckles] 00:16:20.230 --> 00:16:22.839 So I think probably the most important and kind of obvious 00:16:22.839 --> 00:16:25.850 is that you need to have the presence of a critically stressed fault. 00:16:25.850 --> 00:16:28.330 While this sounds obvious, it’s important. 00:16:28.330 --> 00:16:31.540 Right, we should be considering the tectonic histories of these basins. 00:16:31.540 --> 00:16:34.140 So in the Williston Basin, you know, given the tectonic history, 00:16:34.140 --> 00:16:36.730 would we expect the same number of potentially seismogenic 00:16:36.730 --> 00:16:38.760 faults as we would in the Appalachian Basin? 00:16:38.760 --> 00:16:41.320 All right, some people would say that the Williston Basin would have fewer. 00:16:41.320 --> 00:16:43.740 All right, but how do we really know that for sure? 00:16:43.740 --> 00:16:46.130 And that kind of relates to the second point, right? 00:16:46.130 --> 00:16:50.810 I think the next important point here is that the primary control over the induced 00:16:50.810 --> 00:16:55.460 seismicity is the distance to that critically stressed fault from that operation. 00:16:55.460 --> 00:16:58.160 Now, the problem with this statement is that we can’t actually test this 00:16:58.170 --> 00:17:00.440 observationally, right? If we wanted to test this, 00:17:00.440 --> 00:17:02.870 we would need to know where all the faults are located. 00:17:02.870 --> 00:17:05.149 And, you know, almost all induced seismicity 00:17:05.149 --> 00:17:07.470 has occurred along previously unmapped faults. 00:17:07.470 --> 00:17:10.049 It’s totally reasonable to expect that there are other faults that have 00:17:10.049 --> 00:17:12.790 remained aseismic or have at least not released seismicity 00:17:12.790 --> 00:17:15.919 that we have been able to detect. All right, so since we can’t 00:17:15.919 --> 00:17:18.850 observationally test this, I tried to use a proxy. 00:17:18.850 --> 00:17:21.380 All right, and this proxy is that Precambrian basement. 00:17:21.380 --> 00:17:23.760 All right, and this may also still make sense, right? 00:17:23.769 --> 00:17:25.909 Maybe the rock properties in that crystalline rock are more 00:17:25.909 --> 00:17:29.280 conducive to these type of earthquakes. The basement tends to be a little bit 00:17:29.280 --> 00:17:33.460 deeper, so the stress conditions may be more favorable towards earthquakes. 00:17:33.460 --> 00:17:37.380 All right, so there may be explanations for why the basement is still reasonable. 00:17:37.380 --> 00:17:41.740 But once again, you know, this is not a requirement for induced seismicity. 00:17:41.740 --> 00:17:44.600 Even within our case studies in the Appalachian Basin, 00:17:44.610 --> 00:17:48.380 we had induced seismicity within the sedimentary strata, right? 00:17:48.380 --> 00:17:51.360 Now, if you looked at the distance to the fault in the sedimentary strata, 00:17:51.360 --> 00:17:54.020 right, it was still in proximity to the fault, right, 00:17:54.020 --> 00:17:56.740 but it still was far away from the basement. 00:17:57.800 --> 00:18:02.019 And also, I still think that there are geologic influences over 00:18:02.020 --> 00:18:04.500 the chances of inducing seismicity. 00:18:04.500 --> 00:18:09.720 Although, in this case, the halite is not really conclusive in our analysis so far. 00:18:09.720 --> 00:18:11.380 But I think this is kind of just the tip of the iceberg. 00:18:11.380 --> 00:18:13.500 Right, I’m trying to apply this to other basins across 00:18:13.500 --> 00:18:17.179 the United States and look at a little bit more detail. 00:18:17.179 --> 00:18:20.029 And so kind of like to talk about kind of my work in trying to 00:18:20.029 --> 00:18:23.840 characterize the hydraulic fracturing portion in Oklahoma. 00:18:24.360 --> 00:18:27.500 All right, so in Oklahoma, the Oklahoma Geological Survey, 00:18:27.510 --> 00:18:30.889 they identified about 24,000 earthquakes in the state. 00:18:30.889 --> 00:18:32.629 Using those kind of improved methods that I talked about earlier, 00:18:32.629 --> 00:18:35.690 we were able to increase that to over 200,000. 00:18:35.690 --> 00:18:38.749 And this is using the same kind of three-station network 00:18:38.749 --> 00:18:42.149 kind of in the Oklahoma area. Right, so I think we could even 00:18:42.149 --> 00:18:44.710 improve this much further, all right, if we used stations that were even 00:18:44.710 --> 00:18:47.360 closer that have since been deployed. 00:18:47.360 --> 00:18:50.640 But for this analysis, I’ll be using these 200,000 earthquakes. 00:18:50.640 --> 00:18:54.220 All right, so this earthquake catalog that’s now been improved, we now 00:18:54.220 --> 00:18:57.440 need to know, you know, where and when they hydraulically fractured. 00:18:57.440 --> 00:19:02.559 Now, the public records in Oklahoma for hydraulic fracturing are pretty abysmal. 00:19:02.559 --> 00:19:05.749 The only publicly available source of hydraulically fractured dates 00:19:05.749 --> 00:19:09.740 and locations is a chemical repository site called FracFocus. 00:19:09.740 --> 00:19:12.800 All right, so this site is intended to report the chemicals 00:19:12.809 --> 00:19:15.059 that they use in the operation. 00:19:15.059 --> 00:19:17.929 But we’re actually using this in order to investigate induced seismicity. 00:19:17.929 --> 00:19:21.470 So pretty far away from the intended target, but we have to use what we got. 00:19:21.470 --> 00:19:23.860 All right, so in kind of June of 2012, 00:19:23.860 --> 00:19:31.240 Oklahoma began requiring companies to submit the chemical data to this site. 00:19:31.240 --> 00:19:35.730 So post-2012, we would expect this catalog to be complete. 00:19:35.730 --> 00:19:37.830 Okay, so now we have the earthquake catalog. 00:19:37.830 --> 00:19:39.720 We have the hydraulically fractured wells. 00:19:39.720 --> 00:19:41.720 We now want to try to put these together in order to 00:19:41.720 --> 00:19:43.500 look at induced seismicity. 00:19:43.500 --> 00:19:45.399 All right, the problems with this is that there’s so many earthquakes 00:19:45.399 --> 00:19:48.179 and so many wells, and I don’t have a whole lot of free time. 00:19:48.179 --> 00:19:51.580 Right, so I wanted to quantitatively sort these wells in order to determine 00:19:51.580 --> 00:19:55.110 which wells are more likely to be induced and which ones are not. 00:19:55.110 --> 00:19:58.179 Okay, so I looked at other different kind of approaches, and what I decided 00:19:58.180 --> 00:20:01.600 on doing was looking at this kind of delta earthquake rate parameter. 00:20:01.600 --> 00:20:04.320 All right, the idea with this is that, when they’re hydraulically fracturing, 00:20:04.320 --> 00:20:07.889 we would expect the seismicity rate to be high if they induced earthquakes. 00:20:07.889 --> 00:20:09.470 But before and after hydraulic fracturing, 00:20:09.470 --> 00:20:11.230 we’d expect the seismicity rate to be low. 00:20:11.230 --> 00:20:14.580 All right, so we’re looking at changes in the seismicity rate. 00:20:15.200 --> 00:20:18.800 So kind of graphically, right, if this little red dot is a hydraulically fractured well, 00:20:18.809 --> 00:20:21.980 I’d be looking at the seismicity rate at these different distances going away 00:20:21.980 --> 00:20:24.470 from the well, out to 10 kilometers, 00:20:24.470 --> 00:20:26.010 considering these different time windows. 00:20:26.010 --> 00:20:28.720 All right, so now I don’t think hydraulic fractured induced 00:20:28.720 --> 00:20:30.860 seismicity has actually occurred out to 10 kilometers. 00:20:30.860 --> 00:20:34.239 All right, this is just to help deal with any sort of location errors. 00:20:34.239 --> 00:20:38.769 So the location errors, especially early on in the catalog, are quite large. 00:20:38.769 --> 00:20:42.279 So this still allows us to detect those poorly located earthquakes 00:20:42.279 --> 00:20:44.529 while still creating an intrinsic bias towards earthquakes 00:20:44.529 --> 00:20:48.380 that are well-located in proximity to hydraulic fracturing. 00:20:49.280 --> 00:20:51.860 So this is sort of the results of that operation. 00:20:51.860 --> 00:20:54.049 So about 10% of wells had at least one earthquake 00:20:54.049 --> 00:20:55.960 within 10 kilometers of the stimulations. 00:20:55.960 --> 00:20:59.280 All right, so this would mean that 90% of these wells 00:20:59.280 --> 00:21:01.960 didn’t have any cataloged earthquakes, right? 00:21:01.970 --> 00:21:04.470 So this could – you know, it doesn’t mean that they didn’t 00:21:04.470 --> 00:21:05.940 produce any earthquakes. It just means that we did not 00:21:05.940 --> 00:21:09.360 identify any earthquakes for this 90% of the wells. 00:21:10.020 --> 00:21:12.750 If we zoom in on the 10% of wells that did induce the earthquakes, all right, 00:21:12.750 --> 00:21:15.159 we can look at this kind of delta earthquake rate parameter. 00:21:15.159 --> 00:21:18.140 All right, and the idea here being that these wells here at the top with the 00:21:18.140 --> 00:21:21.100 highest rate are more likely to have induced seismicity. 00:21:21.100 --> 00:21:24.299 But as we kind of proceed downwards, they would be less likely, right? 00:21:24.299 --> 00:21:27.179 So I kind of start up here, and I kind of proceed my way down 00:21:27.179 --> 00:21:30.769 until I get bored or busy with something else. All right. 00:21:30.769 --> 00:21:34.620 So an example – so this is showing some earthquakes within 5 kilometers 00:21:34.620 --> 00:21:38.710 of each other. This top plot is showing the magnitude over time. 00:21:38.710 --> 00:21:41.460 This bottom plot is showing the cumulative number of events. 00:21:41.460 --> 00:21:44.059 All right, if we show the periods of hydraulic fracturing, 00:21:44.059 --> 00:21:46.789 shown by the red lines, we see a pretty good correlation. 00:21:46.789 --> 00:21:49.830 All right, we could zoom in, and in this kind of case, all right, we see 00:21:49.830 --> 00:21:53.230 that they start hydraulically fracturing, we see a bunch of earthquakes. 00:21:53.230 --> 00:21:55.220 They stop hydraulic fracturing, the earthquakes stop. 00:21:55.220 --> 00:21:58.100 All right, so this well was successfully identified 00:21:58.100 --> 00:22:00.980 as being potentially induced seismicity. 00:22:00.980 --> 00:22:04.320 But if we look at this first case, no. These earthquakes actually start 00:22:04.320 --> 00:22:07.570 the day before hydraulic fracturing. Now, that’s kind of interesting. 00:22:07.570 --> 00:22:11.149 [chuckles] Now, I’m not a betting man, but I’m willing to bet that the 00:22:11.149 --> 00:22:13.629 person that submitted this report might have had a typo – 00:22:13.629 --> 00:22:15.480 might have meant to include another day or so. 00:22:15.480 --> 00:22:20.600 But because there’s this potential error, right, I did not identify this well, right? 00:22:20.600 --> 00:22:22.080 Now, for the purposes of this research, 00:22:22.090 --> 00:22:25.320 I have to assume that this FracFocus database is accurate. 00:22:25.320 --> 00:22:28.640 Right, so this would not be identified as induced seismicity. 00:22:29.340 --> 00:22:32.500 Right, now this kind of leads into kind of the limitations of this 00:22:32.509 --> 00:22:35.320 type of approach. Right, so I said before, hydraulic fracturing records 00:22:35.320 --> 00:22:40.149 are not complete prior to 2012. All right, so we know of induced cases 00:22:40.149 --> 00:22:44.700 prior to this that we could not detect because we did not have those data. 00:22:44.700 --> 00:22:48.440 Also, as I just mentioned, the hydraulic fracturing reporting 00:22:48.440 --> 00:22:52.360 information may not be completely accurate, but we have to treat it as if it is. 00:22:52.360 --> 00:22:55.119 And finally, chemical reporting is not done in real time, right? 00:22:55.119 --> 00:22:57.499 So if we want to make our research as much – meaningful as possible, 00:22:57.499 --> 00:23:01.179 we want to do this is the most timely manner as possible. 00:23:01.179 --> 00:23:04.190 But if we wait to – if we have to – if we have to wait weeks, months, or even, 00:23:04.190 --> 00:23:09.480 you know, over a year to get this data, it kind of reduces our potential impact. 00:23:09.480 --> 00:23:13.059 Right, so I’m trying to kind of proceed in a new direction. 00:23:13.059 --> 00:23:15.600 You know, can hydraulic fracturing induced seismicity 00:23:15.600 --> 00:23:19.759 be identified confidently using only earthquakes, right? 00:23:19.759 --> 00:23:22.119 So having no industry data, can we still identify earthquakes 00:23:22.119 --> 00:23:24.409 that we believe to be induced by hydraulic fracturing? 00:23:24.409 --> 00:23:27.559 And, you know, these results are certainly early, 00:23:27.559 --> 00:23:31.009 but I would say, so far, yes. You know, we can duplicate our results 00:23:31.009 --> 00:23:34.440 identified by hydraulic fracturing using this delta earthquake rate parameter. 00:23:34.440 --> 00:23:38.409 We can also identify a number of other cases that we think may also be induced. 00:23:38.409 --> 00:23:42.309 Right, so I think it’s pretty interesting. Our results so far indicate that about 00:23:42.309 --> 00:23:46.169 1 to 2% of earthquakes in Oklahoma have been induced by hydraulic fracturing. 00:23:46.169 --> 00:23:48.879 All right, so pretty minor, but I think it has some 00:23:48.879 --> 00:23:51.779 potential implications that I think are pretty interesting. 00:23:51.779 --> 00:23:54.220 All right, so this is kind of one direction that I’m heading, right, 00:23:54.220 --> 00:23:57.139 trying to be completely devoid of industry data. 00:23:57.140 --> 00:24:00.440 But the other branch I’m trying to take is use as much industry data as I can. 00:24:00.440 --> 00:24:04.380 All right, so this top plot is showing kind of the time window from FracFocus. 00:24:04.380 --> 00:24:08.060 All right, we see a start time of hydraulic fracturing, and we see an end time. 00:24:08.060 --> 00:24:12.040 All right, that’s all we have. The science that we can do is sort of limited. 00:24:12.040 --> 00:24:14.739 But this bottom plot shows the stimulation times. 00:24:14.739 --> 00:24:17.750 All right, so every operator, when they stimulate a well, they produce 00:24:17.750 --> 00:24:23.040 a stimulation report, which has a detailed history of what they did at that well. 00:24:23.040 --> 00:24:25.759 All right, the downside in Oklahoma is that the operator is 00:24:25.759 --> 00:24:29.091 not required to submit this information. All right, so if we want this information, 00:24:29.091 --> 00:24:31.100 it has to be voluntarily contributed by the operator. 00:24:31.100 --> 00:24:35.799 In this case, the operator did provide us the stimulations for this well. 00:24:35.799 --> 00:24:37.409 And so this kind of another direction that I’m heading. 00:24:37.409 --> 00:24:40.859 Right, I’d like to start to try to relate these improved seismology 00:24:40.859 --> 00:24:45.159 observations in order to, you know, kind of understand these poroelastic 00:24:45.159 --> 00:24:49.139 stress models and explain why these earthquakes occur. 00:24:49.139 --> 00:24:53.720 Okay, to kind of summarize these earthquakes, so the top plot is showing 00:24:53.720 --> 00:24:55.799 the earthquakes throughout the state. All right, so most of the earthquakes 00:24:55.800 --> 00:24:59.240 are kind of centered in this more northern portion of the state, or this is – 00:24:59.240 --> 00:25:02.380 this is really where a lot of the wastewater disposal is occurring. 00:25:02.389 --> 00:25:04.889 So my initial concern was that, because the seismicity rate 00:25:04.889 --> 00:25:07.940 is this northern part of the state is so high, we may not be able to 00:25:07.940 --> 00:25:10.660 identify hydraulic fracturing induced seismicity as well. 00:25:10.660 --> 00:25:12.640 But fortunately, this turns out not to be the case. 00:25:12.659 --> 00:25:14.840 You know, we see hydraulic fracturing seismicity pretty much 00:25:14.840 --> 00:25:18.169 throughout the state. All right, so it’s not anticorrelated 00:25:18.169 --> 00:25:21.990 to the seismicity rate, but rather it’s more correlated 00:25:21.990 --> 00:25:24.160 to the density of hydraulic fracturing. 00:25:24.160 --> 00:25:28.179 All right, so pretty much areas where they are hydraulic fracturing a lot, 00:25:28.180 --> 00:25:31.609 the chances of one of those wells will induce seismicity is increased. 00:25:32.160 --> 00:25:33.880 Now, this is not the case throughout the state. 00:25:33.889 --> 00:25:36.260 All right, if you look in kind of the western portion of the state, 00:25:36.260 --> 00:25:40.149 they’re hydraulic fracturing quite a bit. But there are no induced earthquakes. 00:25:40.149 --> 00:25:43.799 So kind of the question there is, you know, why no earthquakes? 00:25:43.800 --> 00:25:46.980 And the answer is the Precambrian basement, right? 00:25:46.980 --> 00:25:50.240 You thought you were done hearing about – me ramble about this, but nope. 00:25:50.250 --> 00:25:53.410 So the Precambrian basement in Oklahoma – my model is 00:25:53.410 --> 00:25:56.690 not completely well-resolved in the southern portion of the state. 00:25:56.690 --> 00:25:59.749 There’s some geological complexity I need to address. 00:25:59.749 --> 00:26:02.620 But so far, it seems to indicate that the cases of hydraulic fracturing 00:26:02.620 --> 00:26:05.440 all occur within about a kilometer of the basement. 00:26:05.440 --> 00:26:07.490 All right, so it seems to be kind of in agreement 00:26:07.490 --> 00:26:11.650 with the previous idea that I suggested that the closer you are to the basement, 00:26:11.650 --> 00:26:14.940 the more likely you are to have induced seismicity. 00:26:15.740 --> 00:26:19.460 All right, so the final concept or topic that I’d like to talk about is a interesting 00:26:19.460 --> 00:26:24.179 case study at the Oroville Dam. So Oroville Dam is located in kind of 00:26:24.179 --> 00:26:28.470 the northern Sierra Nevada foothills. It’s the tallest dam in the United States. 00:26:28.470 --> 00:26:31.399 The Oroville Lake behind it is the second-largest manmade lake in 00:26:31.399 --> 00:26:34.760 California. All right, so it’s a pretty serious operation, and this is all managed 00:26:34.760 --> 00:26:38.880 by our friends over at the California Department of Water Resources. 00:26:39.540 --> 00:26:42.779 So the Oroville Dam made some news back in the ’70s. 00:26:42.779 --> 00:26:46.619 So shortly after completion of the dam in 1975, 00:26:46.619 --> 00:26:49.570 there was a magnitude 5.7 earthquake. All right, and at this time, 00:26:49.570 --> 00:26:52.489 a whole flurry of papers were published that associated 00:26:52.489 --> 00:26:56.760 this earthquake with rapid changes in the lake level of the dam. 00:26:56.760 --> 00:26:59.240 All right, so if we look at the current catalog, you know, earthquakes 00:26:59.249 --> 00:27:01.859 are still occurring in this area. So we would – if we see 00:27:01.860 --> 00:27:04.989 some earthquakes, it would not be super abnormal. 00:27:05.740 --> 00:27:11.200 Now, unfortunately, the Oroville Dam made the news again this past February. 00:27:11.200 --> 00:27:13.830 As all of us here in California know, we had a very wet winter – 00:27:13.830 --> 00:27:19.440 broke some records there. And as a result, the California 00:27:19.440 --> 00:27:23.609 Department of Water Resources had to use the spillway quite a lot. 00:27:23.609 --> 00:27:27.179 And in February, they discovered that part of the spillway had failed and that 00:27:27.179 --> 00:27:31.039 erosion was becoming a significant issue and making the situation much worse. 00:27:31.039 --> 00:27:34.419 Okay, now, over to the west here, all right, they had the – 00:27:34.419 --> 00:27:37.489 what’s called the emergency spillway. I don’t think “emergency spillway” 00:27:37.489 --> 00:27:40.220 is the proper term here because it’s not a really a spillway. 00:27:40.220 --> 00:27:41.770 Right, it’s just unimproved earth. 00:27:41.770 --> 00:27:46.669 All right, so when they started to use this spillway, erosion became a pretty 00:27:46.669 --> 00:27:50.850 significant issue almost immediately. So within hours, they had some pretty 00:27:50.850 --> 00:27:54.840 serious erosion, and they realized that, you know, this is not the solution. 00:27:54.840 --> 00:27:59.160 Right, now, over concerns of the dam failing – you know, 00:27:59.160 --> 00:28:04.020 there’s the town of Oroville right below it – [clears throat] sorry. 00:28:07.160 --> 00:28:13.539 So over concerns of the entire dam failing, 180,000 people were evacuated. 00:28:13.540 --> 00:28:17.040 But fortunately, kind of the situation was resolved. 00:28:17.040 --> 00:28:20.480 They went back to using the main spillway exclusively. 00:28:20.480 --> 00:28:25.440 And sort of the erosion was stabilized. And by stabilized, I mean that, 00:28:25.450 --> 00:28:27.919 although they formed this massive erosional channel, 00:28:27.920 --> 00:28:30.760 the erosion did not propagate back up towards the crest. 00:28:30.760 --> 00:28:33.080 All right, so they were not concerned about the dam failing, 00:28:33.080 --> 00:28:35.549 all right, which is fortunate. 00:28:35.549 --> 00:28:38.090 But we got involved in this because, on Valentine’s Day, 00:28:38.090 --> 00:28:41.370 there was a magnitude 0.8 followed closely by a magnitude 1 00:28:41.370 --> 00:28:43.499 that was identified by Lind Gee’s group. 00:28:43.499 --> 00:28:46.070 All right, so we were kind of interested in this. 00:28:46.070 --> 00:28:49.200 We wanted to see, is this natural seismicity? 00:28:49.200 --> 00:28:53.029 Could this be some sort of precursor events towards another magnitude 5? 00:28:53.029 --> 00:28:57.580 Or does this has some sort of relationship with the operations here at the dam? 00:28:57.580 --> 00:28:59.620 All right, so we went up to investigate this. 00:28:59.639 --> 00:29:02.549 All right, so the seismic network in the area is actually pretty good. 00:29:02.549 --> 00:29:06.700 This is a lot better than what I have to deal with in Oklahoma. 00:29:06.700 --> 00:29:09.899 But we have about five seismometers within kind of 20 kilometers of the – 00:29:09.899 --> 00:29:11.769 of the dam. And I also considered 00:29:11.769 --> 00:29:13.960 a couple seismometers that were 40 kilometers away. 00:29:13.960 --> 00:29:16.440 All right, this was mainly just for testing. 00:29:16.440 --> 00:29:18.840 All right, and I wanted to include these other stations because I 00:29:18.840 --> 00:29:22.620 wanted to try to optimize this sort of correlation approach. 00:29:22.620 --> 00:29:25.600 All right, so for this approach, I took some sort of sub-catalog 00:29:25.600 --> 00:29:28.269 of events, and I correlated the catalog against each other. 00:29:28.269 --> 00:29:31.380 All right, and when I was cataloging – or, when I was correlating them against 00:29:31.380 --> 00:29:34.289 each other, I was varying the different parameters that I was considering. 00:29:34.289 --> 00:29:36.110 So things like the network configuration – you know, 00:29:36.110 --> 00:29:39.440 which stations I used, the band pass frequencies, the template start times, 00:29:39.440 --> 00:29:43.340 the template lengths – pretty much anything I could vary, I varied. 00:29:43.340 --> 00:29:45.840 All right, and the idea here was that I wanted to determine 00:29:45.840 --> 00:29:49.360 which set of conditions produced the highest signal-to-noise. 00:29:49.360 --> 00:29:51.110 All right, pretty much the highest cross-correlation 00:29:51.110 --> 00:29:53.680 coefficient relative to the noise level. 00:29:53.680 --> 00:29:57.399 All right, so this example is showing the band pass frequency 00:29:57.400 --> 00:29:59.220 optimizations for two different stations. 00:29:59.220 --> 00:30:02.280 All right, so this is holding all other constant – all other parameters constant 00:30:02.280 --> 00:30:06.420 and just varying the band pass frequencies for these stations. 00:30:06.420 --> 00:30:09.720 So for WR PAM, turns out there’s a lot of high-frequency noise, 00:30:09.729 --> 00:30:11.700 and it pretty much told us, you know, don’t use any of 00:30:11.700 --> 00:30:13.660 this high-frequency energy in your analysis. 00:30:13.660 --> 00:30:17.160 But this other station, NC OGO, pretty much said, you know, 00:30:17.160 --> 00:30:19.980 do a simple, you know, 7 hertz high pass, and you’ll be fine. 00:30:19.980 --> 00:30:23.980 All right, so this result isn’t, you know, super interesting, I would say. 00:30:23.980 --> 00:30:28.289 But it’s really, I think, fascinating to me because I went from having to 00:30:28.289 --> 00:30:31.820 make kind of educated guesses about what the ideal parameters were 00:30:31.820 --> 00:30:33.749 to being able to say, you know, these are the optimal parameters 00:30:33.749 --> 00:30:36.580 that we should be using in order to detect earthquakes. 00:30:36.580 --> 00:30:38.909 All right, this became a very important issue in Oroville 00:30:38.909 --> 00:30:39.989 because the events were so small. 00:30:39.989 --> 00:30:43.700 You know, we were diving down really small into the negative magnitudes here. 00:30:44.220 --> 00:30:48.519 All right, so another kind of concept that I try to play around with 00:30:48.520 --> 00:30:50.920 is the issue of false positives. 00:30:50.920 --> 00:30:53.500 Right, in traditional correlation detection, all right, we usually set 00:30:53.509 --> 00:30:57.580 a threshold high enough so that we would not expect 00:30:57.580 --> 00:31:01.129 false positives to get through. But because these events are so small – 00:31:01.129 --> 00:31:05.160 they’re kind of buried within the noise – false positives were a reality. 00:31:05.660 --> 00:31:09.160 Now, kind of the approach here that I assumed was that, you know, 00:31:09.160 --> 00:31:12.740 correlations with random noise should occur randomly in time. 00:31:12.740 --> 00:31:15.059 Right, so we would expect some kind of low sort of 00:31:15.060 --> 00:31:19.820 correlation rate – detection rate with false positives over time. 00:31:19.820 --> 00:31:23.299 But our real signals, we’d expect these events to cluster in time, right? 00:31:23.300 --> 00:31:26.200 So very different from the false positives. 00:31:26.200 --> 00:31:27.869 But in the real world, right, we have a combination of both 00:31:27.869 --> 00:31:31.540 false positives and the real events. So how do we sort out the false positives? 00:31:31.540 --> 00:31:32.960 Right, so my solution to this was 00:31:32.960 --> 00:31:36.020 to just set some sort of detection threshold, right? 00:31:36.020 --> 00:31:39.509 So if fewer than some number of events were detected in a given day, 00:31:39.509 --> 00:31:41.639 we would just exclude those events from our catalog. 00:31:41.640 --> 00:31:43.860 So that we were left with just the events that were real. 00:31:43.860 --> 00:31:46.580 All right, so this is kind of a elementary cartoon view, all right, 00:31:46.580 --> 00:31:49.009 but I think it kind of demonstrates the point pretty well. 00:31:49.009 --> 00:31:51.840 And it turns out that all this is a pretty simple approach. 00:31:51.840 --> 00:31:54.400 It worked really, really well in Oroville. 00:31:54.400 --> 00:31:57.820 All right, so is the timing of events going back to 1993. 00:31:57.820 --> 00:32:00.200 All right, and you see events, you know, fairly continuously. 00:32:00.200 --> 00:32:03.590 But if you look at those sequences, you know, they look fairly abnormal. 00:32:03.590 --> 00:32:06.660 We get these real tight temporal clusters of earthquakes. 00:32:06.660 --> 00:32:09.180 And initially, we were kind of stumped about, you know, 00:32:09.180 --> 00:32:12.320 [chuckles] why are these earthquakes occurring when they did? 00:32:12.320 --> 00:32:15.500 I was comparing, you know, different lake levels or, you know, river inflow – 00:32:15.500 --> 00:32:17.680 pretty much all the publicly available information, 00:32:17.680 --> 00:32:19.220 and it didn’t really make sense. 00:32:19.220 --> 00:32:22.760 But fortunately, we contacted the Department of Water Resources, 00:32:22.760 --> 00:32:26.100 and they were happy to provide us with the operations at the dam. 00:32:26.100 --> 00:32:29.080 And in the data that they sent us included the spillway usage. 00:32:29.080 --> 00:32:32.640 All right, so as soon as we plotted that, we saw this great correlation, 00:32:32.649 --> 00:32:35.630 and we know – [chuckles] we knew we were pretty happy, right? 00:32:35.630 --> 00:32:39.200 So going back to 1995, we saw earthquakes pretty much 00:32:39.200 --> 00:32:40.660 every time they used the spillway. 00:32:40.660 --> 00:32:43.940 All right, so maybe this occurring also prior to 1993. 00:32:43.940 --> 00:32:46.059 We don’t know because we don’t have data going back that far. 00:32:46.060 --> 00:32:50.320 But it’s definitely a long-lived case of these seismic events. 00:32:51.000 --> 00:32:55.620 All right, so we used NonLinLoc to locate the magnitude 1 event. 00:32:55.629 --> 00:32:58.340 Our maximum likelihood location was right below the spillway. 00:32:58.340 --> 00:33:01.039 If we look at particle motions from the closest station, those particle motions 00:33:01.039 --> 00:33:05.049 are also in agreement with that location originating from that spillway. 00:33:05.049 --> 00:33:07.020 So also further evidence. 00:33:07.020 --> 00:33:10.799 We can also look at the magnitudes of events related to the spillway volume. 00:33:10.799 --> 00:33:13.359 All right, so the magnitude here is along the Y axis. 00:33:13.360 --> 00:33:16.660 Along the X axis is the spillway discharge. 00:33:16.660 --> 00:33:18.780 So although, you know, the largest-magnitude events 00:33:18.789 --> 00:33:21.889 occurred when they’re using the high spillway, we still see this kind of 00:33:21.889 --> 00:33:25.960 correlation even at the lower magnitudes and well into the negative magnitudes. 00:33:26.880 --> 00:33:30.320 So sort of the results from this is that the magnitude 8 – 00:33:30.320 --> 00:33:33.460 or, the 0.8 and magnitude 1 events were part of this larger sequence 00:33:33.460 --> 00:33:35.929 going back to at least 1995. 00:33:35.929 --> 00:33:40.000 Our interpreted source is that there’s some sort of near-vertical joint 00:33:40.000 --> 00:33:42.330 that’s within that spillway. So whenever they use the spillway, 00:33:42.330 --> 00:33:46.149 that fluid would get into this joint, volumetrically expand it, and then it would 00:33:46.149 --> 00:33:49.980 collapse, producing these seismic events. All right, so they’re not real earthquakes, 00:33:49.980 --> 00:33:53.140 but they’re some sort of other induced type of seismic event. 00:33:53.140 --> 00:33:56.330 All right, now, the independent forensic team, they went out and 00:33:56.330 --> 00:33:59.379 investigated the spillway to try to look at the causes for this failure. 00:33:59.379 --> 00:34:01.690 And they did identify that there are these joints 00:34:01.690 --> 00:34:06.639 that we think could potentially explain the seismicity. 00:34:06.639 --> 00:34:10.500 Finally, the source is non-destructive. Because we had – you know, 00:34:10.500 --> 00:34:14.260 these events have been occurring for decades prior to the break-up. 00:34:14.260 --> 00:34:16.390 So they’re not expected to be directly hazardous. 00:34:16.390 --> 00:34:19.840 All right, but this does raise some interesting questions. 00:34:19.840 --> 00:34:22.080 You know, does this indicate that there’s some sort of problem with 00:34:22.080 --> 00:34:26.860 the spillway construction prior to the eventual break-up of the spillway? 00:34:26.860 --> 00:34:29.160 You know, when they start using the spillway again this November – 00:34:29.160 --> 00:34:31.640 so they’ve been repairing it this summer. 00:34:31.640 --> 00:34:33.440 All right, so when they start using the spillway again, 00:34:33.440 --> 00:34:37.300 will these events continue again? Or are they completely done with? 00:34:37.300 --> 00:34:39.980 And finally, you know, are these events occurring in other spillways? 00:34:39.980 --> 00:34:43.600 Is this a common occurrence? Or is this isolated to this kind of case? 00:34:43.600 --> 00:34:44.860 All right, so those are the kind of questions 00:34:44.860 --> 00:34:47.000 that I’m interested in addressing next. 00:34:47.600 --> 00:34:49.800 So sort of in conclusion, these are kind of the different topics. 00:34:49.800 --> 00:34:52.670 All right, so with repeating earthquake detection, I would say 00:34:52.670 --> 00:34:56.380 that the swarm-like nature can be exploited using this RSD 00:34:56.380 --> 00:34:59.570 and optimized template matching to detect these earthquakes. 00:34:59.570 --> 00:35:02.460 With regards to the likelihood of induced seismicity, all right, 00:35:02.460 --> 00:35:06.600 I think it’s largely a factor of the proximity to the fault, which we can’t 00:35:06.600 --> 00:35:12.440 address observationally, so I’m saying that it’s proximity to the basement. 00:35:12.440 --> 00:35:14.970 In Oklahoma, yes, the hydraulic fracturing 00:35:14.970 --> 00:35:17.500 induced seismicity is happening. All right, it’s pretty small, though – 00:35:17.500 --> 00:35:19.820 only a couple percentage. All right, but, you know, 00:35:19.820 --> 00:35:22.510 a couple percentage of hundreds of thousands of earthquakes, 00:35:22.510 --> 00:35:24.850 you know, that’s still fairly sizable. 00:35:24.850 --> 00:35:26.500 And finally, in the Oroville Dam, you know, it turns out they’re 00:35:26.500 --> 00:35:28.920 not real earthquakes, but I think it’s still pretty darn neat. 00:35:28.920 --> 00:35:31.660 All right, so with that, I’ll be happy to answer any questions. Thank you. 00:35:31.660 --> 00:35:37.640 [ Applause ] 00:35:37.640 --> 00:35:40.000 - Thanks, Rob. Very, very interesting talk. 00:35:40.000 --> 00:35:42.140 Do we have any questions for Rob? 00:35:50.620 --> 00:35:53.120 - I was counting on you, Dave. - Right. 00:35:53.120 --> 00:35:54.860 [laughter] 00:35:54.860 --> 00:35:57.940 Yeah, you covered a lot of ground there, Rob. 00:35:57.950 --> 00:36:01.550 So I’ll ask a question about the earlier part of the talk. 00:36:01.550 --> 00:36:04.220 So I think you made a pretty convincing argument that 00:36:04.220 --> 00:36:10.990 proximity to the basement or to faults is an important criteria. 00:36:10.990 --> 00:36:13.830 And you also mentioned at one point that, of course, 00:36:13.830 --> 00:36:17.050 those faults need to be critically stressed to have induced seismicity. 00:36:17.050 --> 00:36:23.240 So do we see any places where we have injection that’s close to basement 00:36:23.240 --> 00:36:26.000 that we just don’t have the critically stressed faults? 00:36:26.000 --> 00:36:28.060 Is that all those other wells that aren’t generating … 00:36:28.060 --> 00:36:29.980 - Sure. Of course. 00:36:31.640 --> 00:36:33.140 - And is … 00:36:33.820 --> 00:36:36.340 - Yeah, I mean, so even within the Appalachian Basin, right, 00:36:36.340 --> 00:36:39.240 we still have plenty of wells that are within the proximity to basement, 00:36:39.240 --> 00:36:40.320 and they don’t all produce seismicity – 00:36:40.320 --> 00:36:42.500 at least seismicity that we have cataloged, right? 00:36:42.500 --> 00:36:45.030 So by no means in this a requirement. Right, you’re not going to 00:36:45.030 --> 00:36:47.900 always cause earthquakes if you’re close to the basement. 00:36:47.900 --> 00:36:50.460 And you don’t have to be close to the basement to cause earthquakes. 00:36:50.460 --> 00:36:56.560 - Do we have any constraint on the stress state of the faults in North Dakota? 00:36:56.560 --> 00:37:00.200 Is that the area that didn’t have any seismicity? 00:37:00.200 --> 00:37:03.360 - All right, so it has seismicity, all right, but there has – there’s very little 00:37:03.360 --> 00:37:06.730 evidence to suggest that any of the seismicity has been induced. 00:37:06.730 --> 00:37:13.160 Right, so yeah, I’m not familiar with the stress state type of analysis in that area. 00:37:13.160 --> 00:37:16.540 But it’s certainly very important, yeah. - Okay, thanks. 00:37:19.460 --> 00:37:21.540 - Any further questions? 00:37:24.440 --> 00:37:26.160 - Actually … - Oh, all right. 00:37:26.160 --> 00:37:28.220 - Yeah. - Is this on? Thanks, Rob. 00:37:28.220 --> 00:37:30.180 That was a … - [inaudible] 00:37:30.180 --> 00:37:31.760 - That was a really … [laughter] 00:37:31.760 --> 00:37:33.660 That was a really great talk. Can you just go back to the 00:37:33.670 --> 00:37:37.310 very beginning and maybe explain, like, a little bit more your, like, 00:37:37.310 --> 00:37:41.310 first or second slide about your overview of the swarm-like nature … 00:37:41.310 --> 00:37:42.310 - Yeah. 00:37:42.310 --> 00:37:46.400 - Yeah. And, I guess, not being familiar at all with what’s going on 00:37:46.400 --> 00:37:50.700 in Ohio, for example. You – like, just explain a little bit more 00:37:50.710 --> 00:37:54.080 what’s going on in these figures? - Oh, yeah. Yeah. 00:37:54.080 --> 00:37:56.790 So each of these letters – this is a name for a different 00:37:56.790 --> 00:37:58.820 induced seismic sequence. 00:37:58.820 --> 00:38:01.160 All right … - So they’re observed sequences. 00:38:01.160 --> 00:38:02.270 - Yeah. - Okay. 00:38:02.270 --> 00:38:05.670 - So these are all previously published in my earlier work. All right, so it’s … 00:38:05.670 --> 00:38:09.180 - And then the Ohio swarm in the bottom left … 00:38:09.180 --> 00:38:11.640 - So, yeah, these are just natural earthquakes. 00:38:11.640 --> 00:38:14.340 So we found out that all of the induced earthquakes that, you know, 00:38:14.340 --> 00:38:17.710 are up here, these are all within 10 kilometers of a well. 00:38:17.710 --> 00:38:19.490 But all these naturally occurring earthquakes, 00:38:19.490 --> 00:38:22.540 these are all further than 10 kilometers away from a well. Right. 00:38:22.540 --> 00:38:26.800 - And then the red line you’ve drawn in there to sort of distinguish between, like, 00:38:26.800 --> 00:38:28.520 a main shock-aftershock sequence and swarms. 00:38:28.520 --> 00:38:30.660 And then you used the same line for the different … 00:38:30.660 --> 00:38:32.160 - Yep. - … regions. Would there be any 00:38:32.160 --> 00:38:34.640 evidence that, you know, those – 00:38:34.640 --> 00:38:37.840 I don’t know, the slopes or the intercepts of those lines should 00:38:37.840 --> 00:38:42.920 change in different regions or … - Yeah. So this line is pretty arbitrary. 00:38:42.920 --> 00:38:44.960 All right, it’s from Vidale and Shearer. 00:38:44.960 --> 00:38:49.100 Well, they didn’t explicitly draw this line. We actually drew in this line. 00:38:49.100 --> 00:38:51.600 It’s more for comparison for these different zones. 00:38:51.600 --> 00:38:55.790 All right, but it’s an attempt to try to quantify that swarm-like behavior. 00:38:55.790 --> 00:39:01.360 Right, so if there’s a sequence, you know, right close to the line, 00:39:01.360 --> 00:39:04.660 right, it doesn’t really mean it’s a swarm or an aftershock. 00:39:04.660 --> 00:39:08.860 It just more means that it doesn’t have a really strong demonstration 00:39:08.860 --> 00:39:12.760 of a swarm-like characteristic, right? So the – but the further away from 00:39:12.760 --> 00:39:17.400 this line you get, you know, the more likely it is to be, you know, an actual 00:39:17.400 --> 00:39:22.040 swarm. All right, it’s more just to try to quantify this kind of relationship. 00:39:28.780 --> 00:39:30.600 - Hi, Rob. - Hey. 00:39:30.610 --> 00:39:35.800 - The largest induced earthquake caused by fluid injection 00:39:35.800 --> 00:39:40.130 is the Pawnee earthquake in September last year. 00:39:40.130 --> 00:39:46.160 That had very few foreshocks and very few aftershocks. 00:39:46.160 --> 00:39:53.260 And I was just wondering how you would classify it on that plot that you’re 00:39:53.260 --> 00:39:58.000 showing there – were just showing there. - Yeah. 00:40:02.060 --> 00:40:03.630 I can pull up my paper. [laughter] 00:40:03.630 --> 00:40:05.610 - Oh, okay. - But I think you actually 00:40:05.610 --> 00:40:09.900 reviewed this paper. [laughter] 00:40:11.140 --> 00:40:14.480 Let’s see if I can find – oops. - [inaudible] 00:40:14.480 --> 00:40:16.640 - Okay. [laughter] 00:40:18.760 --> 00:40:23.100 - I can pull up my revised version. All right, this is fresh off the press. 00:40:24.120 --> 00:40:28.140 Look at all my mistakes and corrections. Look at all that. 00:40:28.140 --> 00:40:31.600 - I can’t believe you … - Yeah. [laughs] 00:40:32.840 --> 00:40:36.500 - All right, so actually – so although it’s been published that there were 00:40:36.501 --> 00:40:39.530 very few Pawnee foreshocks – all right, so this bottom plot. 00:40:39.530 --> 00:40:42.050 All right, so the red dots are the cataloged earthquakes. 00:40:42.050 --> 00:40:45.400 The black dots are our catalog. All right, so we actually see that 00:40:45.400 --> 00:40:48.780 there’s plenty of foreshock activity and plenty of aftershock activity. 00:40:48.780 --> 00:40:51.610 All right, so I think that the statements that have been made that there were 00:40:51.610 --> 00:40:55.450 very few foreshocks is not accurate. Right, so there were very few magnitude 00:40:55.450 --> 00:40:59.502 3 and larger foreshocks, right, but this – I mean, this is kind of the benefit 00:40:59.502 --> 00:41:02.580 of using these kind of correlation algorithms and having this better catalog. 00:41:02.580 --> 00:41:06.120 You know, we can see this case’s history much better, right? 00:41:07.650 --> 00:41:12.580 So let’s see. To actually answer your question, I want to – hopefully, I put it 00:41:12.580 --> 00:41:18.880 into this paper, but if not, I can – I can show you the improved figure later. 00:41:19.640 --> 00:41:22.080 Okay, so here were those sequences. 00:41:22.080 --> 00:41:25.800 All right, so this is the Cushing, Fairview, Prague, and Pawnee sequence. 00:41:25.800 --> 00:41:30.300 So they’re pretty close to that kind of – that boundary line, right? 00:41:30.300 --> 00:41:34.560 But … [laughs] - [laughs] 00:41:34.560 --> 00:41:37.880 - So, yeah, it’s not – it doesn’t demonstrate a huge characteristic 00:41:37.890 --> 00:41:40.430 of being a swarm or an aftershock. - Okay. 00:41:40.430 --> 00:41:42.890 - Middle of the road. - That’s not a very clear-cut 00:41:42.890 --> 00:41:46.220 answer to my questions, but okay. [laughter] 00:41:46.220 --> 00:41:50.860 This is as good as we can do. - This is a fairly arbitrary classification. 00:41:50.860 --> 00:41:54.820 Right, you’re welcome to try to draw the line somewhere else, right, but … 00:41:54.820 --> 00:41:57.160 - Okay, thank you. [laughter] 00:41:59.180 --> 00:42:01.000 - Hi. Good talk. - Thank you. 00:42:01.000 --> 00:42:04.480 - This 1 to 2% in Oklahoma. - Mm-hmm. 00:42:04.480 --> 00:42:07.080 - You’re not saying the other 98% are natural. 00:42:07.080 --> 00:42:11.100 You’re saying that this is – the only ones you can demonstrate are induced. 00:42:11.100 --> 00:42:12.410 - Yeah. - Is that right? 00:42:12.410 --> 00:42:15.260 - So, yeah, 99.9% of Oklahoma earthquakes are induced. 00:42:15.260 --> 00:42:18.060 All right, I mean, that’s the scientific consensus. 00:42:18.060 --> 00:42:21.010 There’s really no other reasonable explanation. All right. 00:42:21.010 --> 00:42:23.460 But of those earthquakes, 1 to 2% are induced by 00:42:23.460 --> 00:42:26.201 hydraulic fracturing, all right? - Hydraulic fracturing. 00:42:26.201 --> 00:42:28.109 - Almost all the others are induced by wastewater disposal. 00:42:28.109 --> 00:42:31.080 - Oh, okay. - Sorry I didn’t make that clear. 00:42:31.080 --> 00:42:33.600 - You probably did. - [chuckles] 00:42:36.460 --> 00:42:38.780 - Any more questions for Rob? Oh. 00:42:40.880 --> 00:42:42.240 - Nice talk, Rob. - Thank you. 00:42:42.240 --> 00:42:46.510 - One quick technical question. On your plots kind of toward 00:42:46.510 --> 00:42:51.410 the end where you’re going through and doing the tradeoff – that one – 00:42:51.410 --> 00:42:53.950 optimizing detections. - Yes. 00:42:53.950 --> 00:42:57.770 - Could you go into a little bit more, just shortly, detail on how you 00:42:57.770 --> 00:43:03.060 went through this? Is this, like, a summation over all of your templates? 00:43:03.060 --> 00:43:06.910 Or is this just a – yeah, how did you go about doing this? 00:43:06.910 --> 00:43:11.030 - Yep. So I took that magnitude 1 event. Right. I correlated against that 00:43:11.030 --> 00:43:13.760 day of data – all right, so February 14th. 00:43:13.760 --> 00:43:17.620 And I think I took, like, about, like, 10 events, right? 00:43:17.620 --> 00:43:20.980 And then, of those 10 events, then I ran it through this optimization. 00:43:20.980 --> 00:43:23.320 All right, so I was cross-correlating them all against each other, 00:43:23.320 --> 00:43:26.340 varying those parameters. - Okay, so just a very small number 00:43:26.340 --> 00:43:28.940 of events that you feel confident should be somewhere. 00:43:28.940 --> 00:43:32.100 - Yeah, yeah. It’s just some sort of sub-catalog, right, that should 00:43:32.100 --> 00:43:34.300 hopefully represent the signals that you hope to detect. 00:43:34.300 --> 00:43:36.180 - Okay, thank you. - Yep. 00:43:42.200 --> 00:43:44.800 - Hey, that was awesome. So you’re seeing a lot of foreshocks 00:43:44.800 --> 00:43:48.280 that haven’t been seen before. So could you say something new 00:43:48.280 --> 00:43:51.200 with further investigation about the size of the nucleation zone 00:43:51.200 --> 00:43:53.780 before some of these bigger ones? 00:43:54.360 --> 00:43:56.260 - I definitely have not looked at that. 00:43:56.260 --> 00:43:58.840 All right, I’m still at the catalog stage. [laughs] 00:43:59.580 --> 00:44:01.740 I mean, so that paper that is – that showed – you know, 00:44:01.750 --> 00:44:03.870 that hopefully will be in review soon and hopefully 00:44:03.870 --> 00:44:08.500 other people will use that catalog to do those type of analyses. 00:44:08.500 --> 00:44:12.280 I guess the answer is, potentially. But I don’t know. 00:44:17.000 --> 00:44:19.860 - All right, any last questions for Rob? 00:44:20.400 --> 00:44:25.400 Well, before we let Rob go, we’re probably going to go to 00:44:25.400 --> 00:44:29.820 lunch at around 11:30, 11:45, or so. Meet at the flagpole. 00:44:29.830 --> 00:44:32.990 We’ll just go locally. And if you do want to find Rob 00:44:32.990 --> 00:44:36.470 for more in-depth discussion, he’s in Building 3A in Ross’ old office 00:44:36.470 --> 00:44:40.460 for those that have been here for a while. But anyway, thanks, Rob. 00:44:40.460 --> 00:44:41.820 - All right. Thank you, everyone. 00:44:41.820 --> 00:44:46.480 [ Applause ] 00:44:47.940 --> 00:44:51.940 [background conversations]