WEBVTT Kind: captions Language: en 00:00:01.720 --> 00:00:04.280 Good morning, everyone. 00:00:05.620 --> 00:00:09.340 Welcome to the weekly Earthquake Science Center seminar. 00:00:09.340 --> 00:00:12.340 My name is Grace Parker, and I, along with my colleague, 00:00:12.349 --> 00:00:18.530 Jessie Saunders, are taking the reins from Jeanne and Sara McBride 00:00:18.530 --> 00:00:22.700 organizing the seminar. So I’ll take this opportunity to say 00:00:22.700 --> 00:00:26.970 we’re filling – we’re in the process of filling the seminar schedule for 00:00:26.970 --> 00:00:29.569 August until about the end of the year. So if you have anyone that’s coming 00:00:29.569 --> 00:00:32.579 to visit that you think would like to give a talk, or if there’s anyone you 00:00:32.580 --> 00:00:38.620 want to invite to give a seminar, please get in contact with me or Jessie. 00:00:39.580 --> 00:00:45.139 Next week, there is no seminar because of the UCERF workshop. 00:00:45.139 --> 00:00:47.870 So the next seminar will be on Monday – 00:00:47.870 --> 00:00:51.940 not the usual time – on Monday, June 24th. 00:00:53.920 --> 00:00:57.549 And with that, I’ll hand it over to Andy Barbour, 00:00:57.549 --> 00:01:00.860 who will introduce our speaker today, Noha Farghal. 00:01:01.620 --> 00:01:05.360 - Thank you. All right. It’s my pleasure to introduce Noha. 00:01:05.360 --> 00:01:08.760 She only had to reschedule, like, three times to get this slot. 00:01:08.760 --> 00:01:13.760 So anyway, Noha was born in Cairo, Egypt. 00:01:14.480 --> 00:01:19.220 Got a B.S. is physics from the American University in Cairo. 00:01:19.220 --> 00:01:23.740 She went on to get a master’s in physics from the same university in conjunction 00:01:23.740 --> 00:01:28.820 with the Catholic University of Leuven – did I pronounce that – okay. 00:01:30.980 --> 00:01:35.460 Came over to the states. Got a master’s in geophysics from Stanford. 00:01:35.460 --> 00:01:39.799 And then, in 2018, got a Ph.D. in geophysics from Stanford working 00:01:39.799 --> 00:01:46.079 with Mark Zoback. And part of the talk today will be on her work at Stanford. 00:01:46.079 --> 00:01:49.919 Since September, she’s been with us as a Mendenhall research fellow 00:01:49.919 --> 00:01:54.669 in the ShakeAlert research group, and she’s looking at trying to figure out 00:01:54.669 --> 00:01:57.419 ways that strain data might be incorporated into the system 00:01:57.420 --> 00:02:02.600 and looking at new technologies. So, with that, I’ll pass it to you. 00:02:05.640 --> 00:02:10.960 [Silence] 00:02:10.960 --> 00:02:14.460 - Well, thank you, Andy, for this kind introduction. 00:02:14.460 --> 00:02:17.809 So I’m basically introducing myself today in terms of 00:02:17.809 --> 00:02:23.919 what I’ve worked on before and what I will be doing in the future. 00:02:23.919 --> 00:02:27.789 So the first half of my talk, I will talk a little bit about 00:02:27.789 --> 00:02:31.530 one of the projects I did in grad school. 00:02:31.530 --> 00:02:36.229 I collaborated with several research groups and professors. 00:02:36.229 --> 00:02:40.389 But this was my absolute favorite, and I’d like to acknowledge 00:02:40.389 --> 00:02:46.740 my collaborator, Professor Zoback, who is also my Ph.D. adviser. 00:02:49.200 --> 00:02:53.880 So, all of the projects I worked on shared one common goal, 00:02:53.880 --> 00:02:58.979 which is to understand how faults and fractures affect hydraulic 00:02:58.980 --> 00:03:02.800 stimulation of an unconventional reservoir. 00:03:02.800 --> 00:03:06.780 So let me start by giving you a little background about this. 00:03:06.780 --> 00:03:10.169 So there are currently about 200,000 unconventional oil and 00:03:10.169 --> 00:03:13.460 gas wells in North America, and production from these wells 00:03:13.460 --> 00:03:19.150 is projected – is expected to keep on increasing for at least the next 15 years. 00:03:19.150 --> 00:03:23.350 But despite of the tremendous potential and the enormous resources, 00:03:23.350 --> 00:03:26.489 most of these hydrocarbons will remain unrecoverable, 00:03:26.489 --> 00:03:33.819 which is reflected here in the low recovery factors. 00:03:33.819 --> 00:03:37.749 So remember that these reservoirs have permeabilities in the range of, 00:03:37.749 --> 00:03:41.160 like, hundreds of nanodarcys. That’s like cement. 00:03:41.160 --> 00:03:46.210 That’s very, very impermeable. It’s like, I think, a million times 00:03:46.210 --> 00:03:49.889 smaller than conventional reservoirs. So you need to do some work. 00:03:49.889 --> 00:03:53.109 You need to stimulate them in order for them to be productive. 00:03:53.109 --> 00:03:56.509 So a big part of making these recovery rates better is to make – 00:03:56.509 --> 00:04:01.570 is to enhance our understanding of how different aspects of a 00:04:01.570 --> 00:04:08.960 reservoir respond to stimulation, and this is where my work becomes – 00:04:08.960 --> 00:04:11.850 where my work actually comes in. 00:04:11.850 --> 00:04:14.739 So with this crowd [laughs], I’m sure I don’t need to explain 00:04:14.739 --> 00:04:17.999 what hydraulic fracturing is. But I would like to highlight 00:04:17.999 --> 00:04:21.039 just a few things about this type of reservoir stimulation 00:04:21.040 --> 00:04:24.120 that are relevant and important for the rest of my talk. 00:04:24.120 --> 00:04:29.540 So, for example here, I’m showing – I’m showing a horizontal well, 00:04:29.540 --> 00:04:33.789 which is similar to the wells that I will talk about today. 00:04:33.789 --> 00:04:36.700 This is a case well, so it’s like a plug-and-perf job. 00:04:36.700 --> 00:04:39.939 So what they do is, they perforate the wells, and then they inject fluid 00:04:39.939 --> 00:04:48.200 and propagate hydraulic fractures in the subsurface, hoping that, you know, 00:04:48.200 --> 00:04:52.720 they want the – by doing this, they want the fluid to effectively 00:04:52.720 --> 00:05:00.240 leak out of the hydraulic fractures into a network of pre-existing fractures and 00:05:00.240 --> 00:05:04.999 fault damage zones, and re-activate those. And with this, permeability increases 00:05:04.999 --> 00:05:10.400 and creates conduits from where the gas is trapped – because, you know, these 00:05:10.400 --> 00:05:17.640 are very low-permeability source rocks, and can migrate to the producing wells. 00:05:19.160 --> 00:05:27.080 On your right here is a figure showing a generalized architecture of a fault. 00:05:27.080 --> 00:05:34.140 It consists mainly of a rather impermeable core. 00:05:34.140 --> 00:05:39.059 And a damage zone around that core. And the damage zone is very interesting 00:05:39.059 --> 00:05:46.290 because it’s packed with microscopic fractures that, in some cases, 00:05:46.290 --> 00:05:51.809 can be targeted for production. And, in the literature, there’s 00:05:51.809 --> 00:05:59.279 a lot of cases where people actually intentionally drill – sorry – intentionally 00:05:59.279 --> 00:06:04.309 drill through the faults hoping that the fractures in the damage zone can be 00:06:04.309 --> 00:06:10.780 re-activated with fluid injection and can increase permeability significantly. 00:06:10.780 --> 00:06:14.280 Now, as the fractures are re-activated, microseismic events are emitted. 00:06:14.280 --> 00:06:16.400 And they’re very small-magnitude events. 00:06:16.400 --> 00:06:19.920 They’re equivalent of dropping, like, a gallon of milk off your 00:06:19.940 --> 00:06:22.800 kitchen counter – very, very small-magnitude events. 00:06:22.800 --> 00:06:26.619 But they’re very important because, if you record them, you can tell 00:06:26.619 --> 00:06:30.210 where the pressure is changing. And that’s important for you 00:06:30.210 --> 00:06:34.450 because you – first of all, this is a very effective way 00:06:34.450 --> 00:06:37.790 of monitoring where the fluid is going. 00:06:37.790 --> 00:06:41.040 And because you don’t want to go out of the zone – out of your target zone. 00:06:41.040 --> 00:06:46.380 So it’s important to record and locate these events. 00:06:46.390 --> 00:06:54.640 And that’s – and that’s why microseismicity is kind of important, 00:06:54.640 --> 00:06:59.429 and I will show you, in a minute, how this – how this helps us 00:06:59.429 --> 00:07:03.100 to characterize how we stimulate a reservoir. 00:07:03.100 --> 00:07:09.140 So my – the question that this project was trying to answer was, how do 00:07:09.140 --> 00:07:13.160 fractures evolve with hydraulic fracturing and production? 00:07:13.160 --> 00:07:17.100 And I’m talking about how they evolve away from wells. 00:07:17.100 --> 00:07:22.070 So it’s standard procedure to have – you know, to make image logs of the wells. 00:07:22.070 --> 00:07:26.029 And you can see the natural fractures. But how about after hydraulic 00:07:26.029 --> 00:07:29.249 fracturing? And how about away from the well? So this is where 00:07:29.249 --> 00:07:35.680 the seismic data comes into play. And this is – this is what this project – 00:07:35.680 --> 00:07:39.740 what this research was trying – was trying to find out. 00:07:39.749 --> 00:07:44.019 So the closest that I have [chuckles] – that I know of that we’ve gotten to this – 00:07:44.019 --> 00:07:47.340 to this – to answering this question of how the fractures evolve 00:07:47.340 --> 00:07:53.710 was this paper by Guo et al. So what they did was analyze 00:07:53.710 --> 00:07:57.710 two seismic surveys that are not – like, slightly overlapping. 00:07:57.710 --> 00:08:01.900 They’re not completely overlapping. So one was before hydraulic fracturing, 00:08:01.900 --> 00:08:07.040 and one was after hydraulic fracturing. But this is not really a true time-lapse 00:08:07.040 --> 00:08:12.700 experiment. What I mean by true time lapse is the same location 00:08:12.700 --> 00:08:17.390 before and after hydraulic fracturing, and in my case, after also three months 00:08:17.390 --> 00:08:21.309 of production. This is a true time lapse. Same sources. Same receivers. 00:08:21.309 --> 00:08:28.179 Same everything. Same location. And with this – with this setup, 00:08:28.180 --> 00:08:32.280 we can actually answer the question of what exactly happens to the 00:08:32.280 --> 00:08:37.520 fractures with fluid injection and with production. 00:08:40.200 --> 00:08:45.360 So the data that I’m going to talk today – about today is from 00:08:45.360 --> 00:08:50.580 two stimulated wells – two gas wells in the Barnett Shale around the location of 00:08:50.580 --> 00:08:55.520 the star from the Newark East Gas Field, where most of the gas production 00:08:55.520 --> 00:09:01.220 is taking place. And so these two wells here were stimulated. 00:09:01.220 --> 00:09:02.960 They’re horizontal wells, as I mentioned. 00:09:02.960 --> 00:09:09.030 They were stimulated in eight and 10 stages by zipper frack. 00:09:09.030 --> 00:09:14.830 So what I mean by zipper frack is, they complete – they complete one – 00:09:14.830 --> 00:09:18.590 for example, Stage 1 in Well 1, and then they go on to complete 00:09:18.590 --> 00:09:24.620 Stage 1 in Well 2. Then Stage 2 in Well 1, Stage 2 in Well 2. 00:09:24.620 --> 00:09:30.700 And they were monitored in two monitoring wells. 00:09:30.710 --> 00:09:33.370 One is here at the toe of Well 2. 00:09:33.370 --> 00:09:38.860 And one is right there with 40 three-component geophones. 00:09:38.860 --> 00:09:42.450 Now, as you can – as you can see, this is the microseismic – these are 00:09:42.450 --> 00:09:46.620 the microseismic events that were recorded during all of 00:09:46.620 --> 00:09:48.830 the hydraulic fracturing jobs. 00:09:48.830 --> 00:09:54.740 And you can see that there is this extensive lineations visible here. 00:09:54.740 --> 00:10:01.180 And it’s very remarkable, and it’s quite unusual to have 00:10:01.180 --> 00:10:04.560 microseismic events going to this extent. 00:10:04.570 --> 00:10:10.660 And I will start now talking about how we – just looking at a lot of the 00:10:10.660 --> 00:10:14.700 data that we collected – that was collected during 00:10:14.700 --> 00:10:18.330 the hydraulic fracturing and production, that there is actually evidence 00:10:18.330 --> 00:10:22.560 of some pre-existing faults in this data. 00:10:22.560 --> 00:10:28.860 So I was given three seismic surveys. One was before, after hydraulic 00:10:28.870 --> 00:10:31.570 fracturing and after a month of production. 00:10:31.570 --> 00:10:35.490 And I just did, you know, regular processing, migration, and stacking. 00:10:35.490 --> 00:10:39.720 And then I calculated – I used this seismic volume to calculate 00:10:39.720 --> 00:10:43.710 the variance attribute of the data. So what it does is just scans the traces 00:10:43.710 --> 00:10:48.860 laterally, and it tries to highlight areas of discontinuity, where you – you know, 00:10:48.860 --> 00:10:55.540 we have the faults displacing the layers of the shale. 00:10:55.540 --> 00:11:00.460 And then I used this as input to – and tracking. 00:11:00.460 --> 00:11:07.460 Just an algorithm that makes sure – that only highlights variance 00:11:07.460 --> 00:11:13.580 values that have continuous dip in strike. 00:11:13.580 --> 00:11:21.060 So you make sure that – it kind of makes sure that you don’t highlight any noise. 00:11:21.060 --> 00:11:25.540 You only highlight real faults. 00:11:25.540 --> 00:11:31.340 So this is a time slice of the data. 00:11:31.340 --> 00:11:37.620 Showing here, three faults – Faults A, B, and C. 00:11:37.620 --> 00:11:41.160 This is at the temporal depth of the wells. 00:11:41.160 --> 00:11:47.480 And if you compare the strikes of the faults to all of the natural fractures – 00:11:47.490 --> 00:11:51.480 the strike orientations of the natural fractures along Well 1, 00:11:51.480 --> 00:11:55.340 you see that they are a really good match. 00:11:55.340 --> 00:11:56.900 Okay. 00:11:57.720 --> 00:12:02.280 So the first indication that there were pre-existing faults in this data was 00:12:02.299 --> 00:12:06.920 the fact that there was a lot of fluid communication happening between 00:12:06.920 --> 00:12:12.000 the two wells, and even beyond Well 2. For example, while they were 00:12:12.000 --> 00:12:18.200 hydraulically fracturing Stage 5 of Well 1, a sudden pressure drop 00:12:18.200 --> 00:12:23.169 of about 300 psi occurred in Well 1, accompanied instantaneously 00:12:23.169 --> 00:12:31.740 by a set of – a number of microseismic events occurring beyond Well 2. 00:12:31.740 --> 00:12:39.460 It’s like the fluids somehow traveled – was – Professor Zoback’s term was – 00:12:39.460 --> 00:12:43.540 is hijacked – was hijacked by the damage zone of the fault 00:12:43.540 --> 00:12:49.320 and traveled far fast. And you can see the result – the result of that 00:12:49.320 --> 00:12:53.939 with the microseismic events emanating where it should not actually be. 00:12:53.940 --> 00:12:58.300 They were not fracturing anywhere near Well 2. 00:13:00.680 --> 00:13:05.600 If you look at Fault C, Fault C produced the highest number 00:13:05.600 --> 00:13:09.399 of microseismic events in the whole experiment for both wells. 00:13:09.399 --> 00:13:15.240 This is a favorite lineation that microseismic events seem to be 00:13:15.240 --> 00:13:22.800 showing up as. And if you compare production for Well 1 throughout 00:13:22.800 --> 00:13:26.389 every – all the stages, you will see that, where the fault – 00:13:26.389 --> 00:13:30.720 where Fault C intersects the well, you see the highest – 00:13:30.720 --> 00:13:34.960 the highest gas flow recorded for the whole well. 00:13:35.840 --> 00:13:39.310 Which also suggests that there is a significant damage zone 00:13:39.310 --> 00:13:45.680 there that was targeted by fluid injection and increased production. 00:13:46.470 --> 00:13:51.880 Also, microseismic events were not so – when you have – when you’re – 00:13:51.899 --> 00:13:55.580 when you’re hydraulically fracturing a stage, you expect microseismic 00:13:55.580 --> 00:14:01.840 events to – you know, to start emanating perpendicular to that – to that stage. 00:14:01.840 --> 00:14:05.180 That was not the case in this – in this experiment. 00:14:05.180 --> 00:14:09.210 For example, while they were hydraulically fracturing Stage 2 of 00:14:09.210 --> 00:14:14.190 Well 1, events were showing up kind of near Stage 1 of Well – I’m sorry – 00:14:14.190 --> 00:14:22.000 2 of Well 2, microseismic events were showing up near Stage 1 of Well 1. 00:14:22.000 --> 00:14:27.370 Maybe that’s because this stage was already fractured. 00:14:27.370 --> 00:14:33.290 But then we saw more evidence that this was not because a previous stage 00:14:33.290 --> 00:14:36.519 was fractured, but actually because of the damage zones of the fault. 00:14:36.520 --> 00:14:43.960 So if you look here, for example, they were hydraulically fracturing Stage 1 – 00:14:43.960 --> 00:14:48.700 Stage 2 of Well 1, and microseismic events start showing – started showing 00:14:48.710 --> 00:14:58.460 up kind of connecting Stage 2 of Well 1 and between 5 and 6. 00:14:58.460 --> 00:15:03.160 There was no – absolutely no hydraulic fracturing occurring there. 00:15:04.960 --> 00:15:13.440 Another interesting case – Stage 4 – Stage 4 of Well 2. 00:15:13.450 --> 00:15:18.170 Again, microseismic events starts emanating from close to – 00:15:18.170 --> 00:15:21.600 kind of between 5 and 6. Now, if you put all of these 00:15:21.600 --> 00:15:27.509 microseismic events together, and you look at – as shown in this 00:15:27.509 --> 00:15:33.880 figure, you see that the microseismic events are tracing Faults A and B. 00:15:34.730 --> 00:15:42.000 So it seems that, although they were injecting fluid away from the faults, 00:15:42.019 --> 00:15:46.760 the damage zones of the faults caused – hijacked the fluid, and the fluid 00:15:46.760 --> 00:15:56.920 traveled far fast and did not – it was not what we expected to see. 00:15:56.920 --> 00:16:01.139 Because the events were not emanating from close to where 00:16:01.139 --> 00:16:03.909 the hydraulic fracturing was occurring, but actually 00:16:03.909 --> 00:16:08.000 along the damage zones of the faults that we identified. 00:16:10.350 --> 00:16:15.960 So this data was actually perfect for looking at fractures. 00:16:15.970 --> 00:16:18.070 They provided – so ConocoPhillips Company 00:16:18.070 --> 00:16:23.440 provided us with three 3D time-lapse surveys. 00:16:23.440 --> 00:16:26.360 It was full azimuth surveys. I’ll talk about that. 00:16:26.379 --> 00:16:30.850 And the incident angles – the incident angles were up to 40 degrees, 00:16:30.850 --> 00:16:35.500 which is perfect for doing azimuthal AVO work. 00:16:35.500 --> 00:16:42.620 So, by full azimuth, I mean that, if you – if you look – if you consider all the – 00:16:42.629 --> 00:16:47.390 all the source receiver pairs, and you plot the angles, and you 00:16:47.390 --> 00:16:51.290 figure out the angles between the source-receiver pair and the north, 00:16:51.290 --> 00:16:56.920 you will get a full 360 azimuth. So this is the full azimuth data. 00:16:58.390 --> 00:17:02.200 What they did is that they first started with a baseline survey, this – with this 00:17:02.200 --> 00:17:06.849 acquisition geometry, that does not change with all the surveys acquired. 00:17:06.849 --> 00:17:10.369 So they started with a baseline survey, then they did all of the hydraulic 00:17:10.369 --> 00:17:16.570 fracturing treatments in the wells. And then they did a post-frack survey. 00:17:16.570 --> 00:17:19.690 And then, after three months of production – so water flowback first, 00:17:19.690 --> 00:17:24.760 and then three months of production, and then I did a post-production survey. 00:17:25.340 --> 00:17:32.340 So the question was, how do I – what kind of fractures do I have there? 00:17:32.340 --> 00:17:35.920 How do I figure out the fracture parameters? 00:17:35.920 --> 00:17:42.050 So I started by – I just obtained the image log of Well 1, and I plotted 00:17:42.050 --> 00:17:49.009 all of the – all of the – these are poles to planes, so they correspond to fractures – 00:17:49.009 --> 00:17:53.660 to natural fracture planes. And I realized that most of the 00:17:53.660 --> 00:17:58.640 fractures were actually vertical, near-vertical, like in this situation. 00:17:58.649 --> 00:18:03.040 And this – the fact that we have a set of aligned vertical fractures will 00:18:03.040 --> 00:18:08.759 introduce anisotropy in the subsurface, and the wave will encounter – 00:18:08.759 --> 00:18:13.420 the seismic wave will encounter different median properties based on 00:18:13.420 --> 00:18:21.779 the source-receiver azimuth. So, if you look here at this synthetic, 00:18:21.780 --> 00:18:24.620 this is a COCA gather, so it’s a common-offset-common-azimuth 00:18:24.620 --> 00:18:32.620 gather. You first bin the data by offset, and then, within this offset bin, 00:18:32.620 --> 00:18:37.320 you sort by azimuth. And if you look at the far offsets, 00:18:37.330 --> 00:18:41.269 where you actually have the effect of the anisotropy more pronounced, 00:18:41.269 --> 00:18:46.389 you see that you can actually – with different – with different azimuths, 00:18:46.389 --> 00:18:48.680 you can see that the amplitude is changing. 00:18:48.680 --> 00:18:53.519 And the same thing for the velocity. The travel times, they’re changing, 00:18:53.519 --> 00:18:57.960 depending on the azimuth of the source-receiver azimuth. 00:18:57.960 --> 00:19:03.320 So, depending on whether the wave is encountering the aligned fractures 00:19:03.320 --> 00:19:06.409 perpendicular to their plane or parallel to their plane. 00:19:06.409 --> 00:19:10.839 So this is the more stiff direction. And this is the more compliant 00:19:10.839 --> 00:19:13.210 direction, which effects the wave. Because, as the wave – 00:19:13.210 --> 00:19:18.810 say this is the aperture. As the wave crosses the fracture, the 00:19:18.810 --> 00:19:23.149 fracture kind of responds like a spring. Because it’s more compliant in 00:19:23.149 --> 00:19:27.599 this direction, it takes some of the – some of the energy of the wave, 00:19:27.600 --> 00:19:30.300 and it delays the wave a little bit. 00:19:31.720 --> 00:19:37.960 So, in a rock with more randomly oriented fractures, you kind of expect 00:19:37.960 --> 00:19:44.340 the azimuth of the maximum horizontal stress to dominate the anisotropy, 00:19:44.340 --> 00:19:48.740 because it’s going to close a lot of the fractures that are perpendicular to it. 00:19:48.740 --> 00:19:53.360 However, when you have a structural anisotropy – have a set of aligned 00:19:53.369 --> 00:19:58.979 fractures, yes, S-Hmax will try to close some of these fractures, 00:19:58.979 --> 00:20:03.869 but you might still end up with an anisotropy in subsurface 00:20:03.869 --> 00:20:07.980 that’s controlled by the presence of fractures. 00:20:08.800 --> 00:20:16.200 So I used – I used both properties of the seismic wave that changed – 00:20:16.200 --> 00:20:22.249 that changed with azimuthal anisotropy. The first was amplitude. 00:20:22.249 --> 00:20:31.099 So I used a linearized form of this Ruger equation to find out two parameters. 00:20:31.099 --> 00:20:36.239 The first was the anisotropic gradient, which is proportional to fracture density. 00:20:36.240 --> 00:20:41.120 And the second was – is the fracture perpendicular. 00:20:45.320 --> 00:20:49.660 This is my workflow. I’m just showing it because people ask about it. 00:20:49.670 --> 00:20:53.200 And I’m happy to talk about it offline if you would like to. 00:20:53.200 --> 00:21:00.020 It’s a lot of – it’s a lot of – a lot of small processes – a lot of small steps. 00:21:00.020 --> 00:21:06.520 And it took a while to do it. But the only thing I can say about this 00:21:06.520 --> 00:21:14.140 is that there are – there are processes that do affect – that have to be done 00:21:14.140 --> 00:21:20.399 before azimuthal binning, before and after azimuthal binning. 00:21:20.399 --> 00:21:25.969 If – unfortunately, if these processes are not taken into account, they can 00:21:25.969 --> 00:21:28.580 completely – you can completely mask the anisotropy or 00:21:28.580 --> 00:21:32.580 introduce anisotropy that is not there, that is not in the data. 00:21:32.580 --> 00:21:36.760 So you need to be – you need to be careful about that. 00:21:38.590 --> 00:21:42.060 So these are the results for all three surveys. 00:21:42.060 --> 00:21:44.119 So the first is the baseline survey. 00:21:44.120 --> 00:21:49.360 The colors indicate B-ani, the fracture density parameter. 00:21:49.360 --> 00:21:54.959 And the orientations of these glyphs here, they indicate the 00:21:54.960 --> 00:21:59.160 fast direction or the fracture – the fracture parallel. 00:21:59.540 --> 00:22:04.340 And so I made some assumptions when I was – when I was working on this data. 00:22:04.349 --> 00:22:08.089 I was kind of worried. Because, first of all, I don’t know 00:22:08.089 --> 00:22:11.840 how good these assumptions hold. And secondly, I was also worried 00:22:11.840 --> 00:22:15.820 about the processing. It’s possible – if all of the surveys 00:22:15.820 --> 00:22:19.649 look nothing like each other, then that doesn’t really make sense. 00:22:19.649 --> 00:22:26.629 So for a QC, I looked first at the trends of the fractures. 00:22:26.629 --> 00:22:30.599 You see that they are – in all three surveys – here you need to kind of 00:22:30.599 --> 00:22:35.290 change the scale a little bit to see it, but it’s there. 00:22:35.290 --> 00:22:39.759 You see that these kind of S-Hmax parallel trends are 00:22:39.759 --> 00:22:43.860 actually there in both surveys. So this is – this was a good QC for 00:22:43.860 --> 00:22:53.080 me because the fracture strikes are kind of consistent in the – in the surveys. 00:22:53.080 --> 00:23:00.930 And this different – this north-northwest orientation is also visible in all three – 00:23:00.930 --> 00:23:04.959 in all three surveys. So this is an indicator that, 00:23:04.959 --> 00:23:09.230 one, the fracture strikes are consistent in all three surveys. 00:23:09.230 --> 00:23:13.100 And, two, that actually, they don’t change much 00:23:13.100 --> 00:23:16.140 with stimulation, which kind of makes sense. 00:23:17.760 --> 00:23:22.570 The fracture strikes remained the same. The density, though, changed. 00:23:22.570 --> 00:23:30.170 Our ability to detect these fractures has increased after fluid injection. 00:23:30.170 --> 00:23:37.430 And I’ll talk about that more. But the fact that the strike orientations 00:23:37.430 --> 00:23:39.550 were consistent in all three surveys was 00:23:39.550 --> 00:23:44.540 very important for me to trust – to trust these results. 00:23:45.720 --> 00:23:50.020 So, if you look at what happens to the anisotropic gradient, B-ani, which is 00:23:50.029 --> 00:23:56.300 an indicator of fracture density, you see that, first of all, in the baseline survey, 00:23:56.300 --> 00:23:59.399 you have anisotropy to begin with, which is what we expect. 00:23:59.399 --> 00:24:02.050 I showed you – remember the stereonet with all the fractures? 00:24:02.050 --> 00:24:06.640 So we know that this is a highly fractured part of the Barnett. 00:24:06.640 --> 00:24:14.339 And this is – this is an expected result, the fact that we have – we do have 00:24:14.339 --> 00:24:20.259 some fracture anisotropy to begin with. Then, after fluid injection, you see that 00:24:20.260 --> 00:24:27.980 this has become – the increase in B-ani has become very extensive and – 00:24:27.980 --> 00:24:31.999 has become very extensive, and it reminds us of all the 00:24:32.000 --> 00:24:37.180 microseismicity that was emanating far away from the wells and 00:24:37.180 --> 00:24:40.519 confirming the fact that we had a lot of communication – 00:24:40.519 --> 00:24:43.399 a lot of fluid communication between different locations, 00:24:43.399 --> 00:24:46.249 even away from the wells in this experiment. 00:24:46.249 --> 00:24:50.589 But what was most striking to me was the fact that, after only 00:24:50.589 --> 00:24:53.570 three months of production – waterflow back, they retrieve 00:24:53.570 --> 00:24:56.119 as much as the injected water as possible, and then three months 00:24:56.120 --> 00:25:03.260 of production, you have significant fracture closure around the wells. 00:25:03.540 --> 00:25:07.680 - Can everybody hear me now? No? Hello? 00:25:07.680 --> 00:25:10.520 [muffled sounds] 00:25:11.720 --> 00:25:16.740 Can everybody hear me? It’s kind of difficult to talk in this mic. 00:25:16.740 --> 00:25:18.220 Okay. 00:25:22.800 --> 00:25:26.540 [muffled sounds] 00:25:27.780 --> 00:25:29.260 Okay. 00:25:29.260 --> 00:25:31.540 Better? All right. [laughs] 00:25:31.540 --> 00:25:35.680 So, yeah, so I just wanted to confirm that, not just by 00:25:35.699 --> 00:25:40.049 visual inspection, but I wanted to confirm that the anisotropy gradient 00:25:40.049 --> 00:25:44.480 was actually a good indicator of fracture density. 00:25:44.480 --> 00:25:49.120 So how about fracture orientations? How good are my estimates of fracture 00:25:49.120 --> 00:25:54.999 orientations? So I plotted all of the natural fractures along Well 1. 00:25:54.999 --> 00:25:59.640 And, if you compare these fractures – the orientations of these fractures 00:25:59.640 --> 00:26:05.650 to the – estimated from the seismic by – from seismic data, compare them to 00:26:05.650 --> 00:26:09.979 the strike orientations from the natural fractures along Well 1, 00:26:09.979 --> 00:26:15.979 you’ll see that they’re kind of maybe parallel to S-Hmax, and this trend 00:26:15.979 --> 00:26:22.109 is really sound along the well. But I was worried about this trend. 00:26:22.109 --> 00:26:26.229 This is not parallel to S-Hmax. It showed up on all of my surveys. 00:26:26.229 --> 00:26:29.890 The question was, was it real? Is it real? 00:26:29.890 --> 00:26:36.210 So what I did was isolate the fractures along Well 1 in the first, like, 00:26:36.210 --> 00:26:42.210 few hundred feet of Well 1, and actually, this trend was really there. 00:26:42.210 --> 00:26:50.440 The natural fractures along the heel of Well 1 did have – do have this 00:26:50.440 --> 00:26:56.780 non-S-Hmax parallel trend. And it was – I was fascinated – 00:26:56.780 --> 00:27:01.140 this was captured – was actually captured by the seismic data, and it 00:27:01.140 --> 00:27:07.360 was very – it was a very good indicator of efficiency of this method. 00:27:08.440 --> 00:27:14.500 Although I know – I knew that using travel times is not as robust 00:27:14.500 --> 00:27:20.000 as using amplitudes to estimate fractures, I still was interested in 00:27:20.000 --> 00:27:29.170 doing it because a lot of researchers in the literature mentioned that with – 00:27:29.170 --> 00:27:34.500 if you use velocity anisotropy to estimate anisotropy – if you use 00:27:34.500 --> 00:27:38.840 velocity and estimate anisotropy, you end up with more of the – 00:27:38.840 --> 00:27:43.320 more of the stress-induced anisotropy than fracture-induced anisotropy. 00:27:43.320 --> 00:27:48.820 So I wanted to put this to the test and see whether using velocity will be able – 00:27:48.830 --> 00:27:53.080 because I have two different trends – one that’s parallel to the stress – 00:27:53.080 --> 00:27:55.320 the maximum horizontal stress, and one that is not. 00:27:56.120 --> 00:27:59.120 So I was interested in knowing whether using velocity anisotropy 00:27:59.130 --> 00:28:03.000 will be able to capture the difference in the two trends. 00:28:03.540 --> 00:28:11.340 So I used the NMO stacking velocities to estimate beta, 00:28:08.100 --> 00:28:12.080 - Can everybody hear me now? No? Hello? 00:28:11.340 --> 00:28:14.660 which is the orientation of the slope velocity, which is the 00:28:12.080 --> 00:28:14.920 [muffled sounds] 00:28:14.660 --> 00:28:19.760 fracture perpendicular. And delta here, if you’re wondering, 00:28:19.760 --> 00:28:26.370 is this the Thomsen parameter, it is. But it’s for HTI medium to estimate – 00:28:26.370 --> 00:28:31.660 so this would be the indicator of fracture density in this case. 00:28:33.380 --> 00:28:40.110 I plotted here – I did the same thing that I did for amplitudes, and I plotted delta, 00:28:40.110 --> 00:28:44.820 the indicator of fracture density, with the natural fracture density 00:28:44.820 --> 00:28:49.620 along the wells. And you see that you get a rather smaller – not a – 00:28:49.620 --> 00:28:52.700 a correlation that’s not as good as what you got with amplitude, 00:28:52.710 --> 00:28:57.110 which is expected. But still, you can – you can say – you can tell that – 00:28:57.110 --> 00:29:01.840 you can say that delta here is a good indicator of fracture density. 00:29:03.700 --> 00:29:10.660 But how about fracture orientations? So, as expected, you know, I expected 00:29:10.660 --> 00:29:17.360 that the velocity anisotropy will be able to capture this S-Hmax parallel trend. 00:29:17.360 --> 00:29:25.360 And close to the heel of the well, it did show that there is a non-S-Hmax 00:29:25.360 --> 00:29:30.760 parallel trend here, although it did not really show an accurate angle 00:29:30.760 --> 00:29:38.200 as the natural fractures or as the amplitudes indicated. 00:29:38.200 --> 00:29:42.280 But it was able to capture a difference between the two trends along the well, 00:29:42.280 --> 00:29:46.880 which is also – it was also very remarkable for me. 00:29:46.880 --> 00:29:55.560 So, to conclude this project, this part, we saw, by examining the baseline, 00:29:55.560 --> 00:30:01.020 post-frack, and post-production surveys that the orientations of the fractures, 00:30:01.020 --> 00:30:05.080 which kind of makes sense, do not change a lot. But what changes 00:30:05.090 --> 00:30:10.590 is our ability to detect them. And this has to do with the increase in 00:30:10.590 --> 00:30:18.660 fracture compliance with fluid injection. And then with pressure. 00:30:18.660 --> 00:30:26.370 And the decrease when the pressure goes – becomes lower, the decrease in 00:30:26.370 --> 00:30:32.940 pressure reduces the compliance of the fractures, and our ability to detect them 00:30:32.940 --> 00:30:36.920 decreases, and the anisotropy decreases. So you can say here, for example, 00:30:36.920 --> 00:30:44.020 that anisotropy can be viewed as a proxy for pressure – for pressure change. 00:30:44.030 --> 00:30:48.640 And for people who are concerned about where you’re stimulating, 00:30:48.640 --> 00:30:51.920 about stimulated volume, about hydraulic fracturing, 00:30:51.920 --> 00:30:56.400 this is – this is something to be really investigated, the fact that 00:30:56.400 --> 00:31:01.480 you can use anisotropy to indicate where pressure is changing. 00:31:03.780 --> 00:31:10.760 So now I will move on to the work I’m doing here as a postdoc at the Survey. 00:31:10.760 --> 00:31:15.559 I’m working with Andy Barbour. And this is about our upcoming 00:31:15.559 --> 00:31:24.210 paper with John Langbein. And today I will – I will – 00:31:24.210 --> 00:31:29.500 I hope to present a proof of concept for the use of strain data in earthquake 00:31:29.500 --> 00:31:34.360 magnitude estimation for early warning, which is part of an ongoing research 00:31:34.360 --> 00:31:37.920 that aims at enhancing the accuracy and timeliness of magnitude and 00:31:37.920 --> 00:31:41.320 ground motion estimates and utilize as many of the already-deployed 00:31:41.320 --> 00:31:46.820 strainmeters as – as many of them as possible for data acquisition. 00:31:47.180 --> 00:31:51.120 So this map on your right shows locations of 74 plate boundary 00:31:51.120 --> 00:31:55.680 observatory, or PBO, strainmeter stations, some of which 00:31:55.680 --> 00:32:00.340 have co-located geophones. And all of those stations are Gladwin [inaudible] 00:32:00.340 --> 00:32:04.060 strainmeters. I’ll talk about that more in my next slide. 00:32:04.060 --> 00:32:14.900 And this cluster here is an array of five Sacks-Evertson dilatometers – 00:32:14.900 --> 00:32:21.819 the ones in green. And the ones in black are 00:32:21.819 --> 00:32:25.789 five mini-PBO – what we call mini-PBO, or three-component 00:32:25.789 --> 00:32:30.560 strainmeters, which is a variance of the Sacks-Evertson dilatometer. 00:32:30.560 --> 00:32:35.240 So I’ll start by briefly walking you through the – how we calculate peak 00:32:35.260 --> 00:32:40.200 dynamic strain to use in estimating our earthquake magnitudes. 00:32:40.200 --> 00:32:43.900 I’m using here the Gladwin strainmeter as an example 00:32:43.910 --> 00:32:48.299 because that’s the dominant type of strainmeter used in the network. 00:32:48.299 --> 00:32:53.950 So each consists of four channels. Mainly, there’s two capacitors – 00:32:53.950 --> 00:32:56.850 one with a fixed gap and one with a movable gap. 00:32:56.850 --> 00:33:02.340 And, as the housing, or the case – the housing of the instrument deforms, 00:33:02.340 --> 00:33:06.000 the capacitance bridge ratio will represent – will represent 00:33:06.000 --> 00:33:14.620 the average strain across the sensor. The gauge output here is basically 00:33:14.620 --> 00:33:21.259 a linear combination of areal and shear strains. 00:33:21.259 --> 00:33:27.410 And you can obtain instrumental strain from this relationship using the gauge 00:33:27.410 --> 00:33:37.870 output and instrumental – the gauge – using the gauge output and knowing 00:33:37.870 --> 00:33:44.332 the gap – the gap – the gap diameter and instrument diameter. 00:33:44.340 --> 00:33:48.080 Then we high-pass the data to remove static offsets. 00:33:48.080 --> 00:33:54.380 And we combine the channel outputs very simply, so it’s RMS – 00:33:54.380 --> 00:33:59.140 as an RMS strain. And this gives us a time series, 00:33:59.140 --> 00:34:03.670 the maximum of which is the peak dynamic strain. 00:34:05.700 --> 00:34:11.019 So, in their paper of 2017, Barbour and Crowell suggested 00:34:11.020 --> 00:34:15.220 a relationship between peak dynamic strain and earthquake hypocentral 00:34:15.220 --> 00:34:18.800 distance and magnitude. And they formulated four different 00:34:18.809 --> 00:34:22.759 linearizations of this relationship with varying complexity. 00:34:22.759 --> 00:34:27.370 So, for example, they can account for station bias. 00:34:27.370 --> 00:34:30.731 They can account for earthquake bias, which is a combination of source 00:34:30.740 --> 00:34:35.320 and path effects. And they can account for both. 00:34:35.320 --> 00:34:39.260 So depending on what kind of accuracy and complexity you need, 00:34:39.260 --> 00:34:43.340 you can use any of these different variants. 00:34:44.660 --> 00:34:48.380 So they used about 1,800 records for earthquakes with magnitudes 00:34:48.380 --> 00:34:52.730 greater than 4.5 and hypocentral distances smaller than 500 with 00:34:52.730 --> 00:34:56.860 an average hypocentral distance of around 351 kilometers. 00:34:56.860 --> 00:35:01.920 They got a good – a good correlation between observed and predicted strains. 00:35:01.930 --> 00:35:07.979 However, the data was lacking near-source events, which is kind of 00:35:07.979 --> 00:35:11.020 very important for earthquake early warning. 00:35:11.020 --> 00:35:16.620 So we decided to look at a more local data set and reformulate the – 00:35:16.620 --> 00:35:22.360 and recalculate the coefficients for this – for this data set. 00:35:22.360 --> 00:35:28.520 So the events that we used are here, indicated by stars. 00:35:30.630 --> 00:35:34.320 The maximum hypocentral distance in this case was about 62 kilometers 00:35:34.320 --> 00:35:37.920 compared to 500 kilometers in the Barbour and Crowell work. 00:35:37.920 --> 00:35:40.759 The mean hypocentral distance was about 30 kilometers, 00:35:40.759 --> 00:35:43.170 compared to 351 kilometers before. 00:35:43.170 --> 00:35:47.850 And we lowered the minimum magnitude to 3.5 to be more suitable 00:35:47.850 --> 00:35:53.600 and to match the ShakeAlert current alerting threshold. 00:35:55.580 --> 00:35:58.759 This figure shows results of testing the near-source magnitude 00:35:58.759 --> 00:36:01.770 distance relationship on strain records from the 00:36:01.770 --> 00:36:06.060 Berkeley 4.4 earthquake of January 2018. 00:36:06.060 --> 00:36:10.000 Here we show magnitude estimates of the – from the two closest stations. 00:36:12.020 --> 00:36:15.720 From the two closest stations to the earthquake. 00:36:16.820 --> 00:36:21.120 And we can see that we started getting magnitude estimates 00:36:21.120 --> 00:36:26.920 right after the earthquake reaches the station, and it kind of plateaus. 00:36:26.920 --> 00:36:31.740 Reach the maximum peak dynamic strain once the S wave arrives. 00:36:31.740 --> 00:36:35.380 This shows – this line – the black – the thick black line shows 00:36:35.380 --> 00:36:37.960 the response of ShakeAlert. 00:36:37.960 --> 00:36:44.360 And remember that this is based on at least data from four – from four stations. 00:36:44.360 --> 00:36:49.180 So, even – looking at this, even if you account for data latency, 00:36:49.180 --> 00:36:54.460 and you kind of shift the station – the estimates from these stations by, say, 00:36:54.460 --> 00:36:58.740 1 second to the right, you can see the potential of incorporating magnitude 00:36:58.740 --> 00:37:02.360 estimates from the strainmeters, which may also help reach the 00:37:02.360 --> 00:37:08.280 four alerting station thresholds faster, especially in locations where 00:37:08.280 --> 00:37:12.200 the density – the station coverage is not very high. 00:37:12.200 --> 00:37:16.190 Here is another example from another pair of close stations from 00:37:16.190 --> 00:37:22.180 the PBO network, also showing good magnitude estimates. 00:37:24.580 --> 00:37:29.719 So now that we have the original Barbour and Crowell catalog and 00:37:29.719 --> 00:37:38.130 also the near-source catalog, in addition to more earthquake 00:37:38.130 --> 00:37:43.780 strain records that John provided, we have enough data to develop a 00:37:43.780 --> 00:37:50.700 relationship between magnitude – earthquake magnitude and strain. 00:37:50.710 --> 00:37:56.160 And Andy is leading this effort, and it will be – it will be part of 00:37:56.160 --> 00:37:59.519 our upcoming paper as well. 00:37:59.520 --> 00:38:03.860 So we’ve been using raw, uncalibrated strains to simplify 00:38:03.860 --> 00:38:06.020 the math as much as possible. 00:38:06.029 --> 00:38:11.630 However, we did calibrate 43 out of the 68 sites used by Barbour and Crowell, 00:38:11.630 --> 00:38:16.760 and we concluded that – as you can see here, the model misfit has 00:38:16.760 --> 00:38:24.720 not changed by calibration, which means that it makes sense to use raw, 00:38:24.729 --> 00:38:31.440 uncalibrated strains, at least in this case, which is – which is really significant 00:38:31.440 --> 00:38:36.540 in simplifying the processes required to obtain earthquake 00:38:36.540 --> 00:38:38.900 magnitudes from strain. 00:38:40.080 --> 00:38:45.940 One issue with the data, though, which we made – we made UNAVCO 00:38:45.940 --> 00:38:50.370 aware of, and they are collaborating with us on solving this issue, 00:38:50.370 --> 00:38:55.440 is the fact that – realize that, when you’re very close to an earthquake, 00:38:55.440 --> 00:39:00.520 or if the earthquake is a large magnitude, the strain rates are high, 00:39:00.520 --> 00:39:04.440 you tend to get gaps in the data. And these gaps are usually 00:39:04.440 --> 00:39:09.900 replaced by spikes like that. So here I’m showing – this is just 00:39:09.910 --> 00:39:17.299 the raw data obtained from IRIS. And this is the same data – 00:39:17.300 --> 00:39:19.900 it’s for the Napa earthquake. It’s one of the channels. 00:39:19.900 --> 00:39:23.220 And this is the same data after removing the spikes. 00:39:23.220 --> 00:39:27.360 And, of course, the location of these gaps, whatever that 00:39:27.360 --> 00:39:30.720 was there, the data that was there, is irretrievable. 00:39:30.720 --> 00:39:35.240 So we pointed that out to UNAVCO because this problem needs to be – 00:39:35.240 --> 00:39:38.580 needs to be solved before the data is available in real time, 00:39:38.580 --> 00:39:40.880 which they’re also working on. 00:39:42.880 --> 00:39:48.580 So, looking, you know, at scalability, another really great thing about strain 00:39:48.580 --> 00:39:56.440 is the fact that it can be scalable if you – if you utilize fiber optic – 00:39:56.440 --> 00:39:59.559 fiber optic sensing to do it. So, looking to the future, 00:39:59.559 --> 00:40:03.569 we investigate – part of my work is to investigate the use of telecom 00:40:03.569 --> 00:40:09.360 optical fibers to provide thousands and thousands of earthquake strain records. 00:40:09.360 --> 00:40:14.660 This gentleman here, he is applying stress along an optical fiber. 00:40:14.660 --> 00:40:18.999 The screen behind him is showing how strain can be determined 00:40:19.000 --> 00:40:22.820 very densely at different points along it. And in fact – in fact, I think, 00:40:22.820 --> 00:40:27.760 for 10 meters of fiber, you can get about 1,000 sensing points. 00:40:27.760 --> 00:40:30.840 So we are collaborating with Stanford. 00:40:30.840 --> 00:40:40.220 These images are from earthquakes that they recorded with their DAS array. 00:40:40.220 --> 00:40:43.870 And they are providing – they are going to provide us with a lot of data 00:40:43.870 --> 00:40:52.580 and to process and to compare with that, with the data obtained from – for the 00:40:52.580 --> 00:40:58.620 same earthquake but from strainmeters. And I’m very hopeful about this, 00:40:58.620 --> 00:41:04.880 and I hope we can get a lot of high-quality work done with that. 00:41:04.880 --> 00:41:09.780 So, to conclude, I hope I convinced you that strain data has good potential 00:41:09.780 --> 00:41:12.900 in providing reliable estimates of earthquake magnitudes from raw, 00:41:12.900 --> 00:41:17.320 uncalibrated strain, which simplifies the processing and the math a lot. 00:41:17.320 --> 00:41:21.160 Strainmeters already deployed in multiple locations can supplement 00:41:21.160 --> 00:41:23.920 seismometer networks and enhance station coverage. 00:41:23.920 --> 00:41:28.280 Resolution of issues such as data gaps and data availability in real time, 00:41:28.280 --> 00:41:33.040 which UNAVCO is working on both, also need to be done before the data 00:41:33.040 --> 00:41:36.880 is being used – utilized for this purpose. 00:41:36.880 --> 00:41:41.500 And, ending on a positive note, fiber optic strain sensing can provide dense, 00:41:41.500 --> 00:41:46.600 cheap, and highly scalable strain sensing capabilities, which can be 00:41:46.600 --> 00:41:56.980 really significant and really meaningful for cheap – for cheap strain sensing 00:41:56.980 --> 00:42:01.640 in the future and for this particular application of earthquake early warning. 00:42:01.640 --> 00:42:05.360 And with that, I would like to end and take any questions. 00:42:05.360 --> 00:42:07.720 And thank you very much for listening. 00:42:07.720 --> 00:42:14.200 [Applause] 00:42:15.360 --> 00:42:18.940 - Does anyone have any questions for Noha? 00:42:21.640 --> 00:42:36.700 [Silence] 00:42:36.700 --> 00:42:41.110 - Hi, Noha. Thanks. That was a great overview of your – both projects. 00:42:41.110 --> 00:42:47.630 In the first half, you sort of concluded by saying that there were – 00:42:47.630 --> 00:42:50.890 we were observing fewer fracture – lower fracture density, and that 00:42:50.890 --> 00:42:54.410 was reflecting our ability to observe the fracture density. 00:42:54.410 --> 00:42:57.110 But you also sort of hinted at the fact that maybe the fractures were 00:42:57.110 --> 00:43:01.520 opening and closing, and so therefore, there were fewer fractures. 00:43:01.520 --> 00:43:05.740 - Yeah. So what I meant was that they become less detectable. 00:43:05.750 --> 00:43:07.390 - So they’re – you think they’re … - Because they’re closing. 00:43:07.390 --> 00:43:09.960 - They’re closing and becoming smaller or nonexistent, but … 00:43:09.960 --> 00:43:13.820 - Yeah. They’re becoming – so, yeah, so as compliance increases with – 00:43:13.820 --> 00:43:21.120 because they’re held open by the – by the pressure – as compliance 00:43:21.120 --> 00:43:24.720 increases, anisotropy increases, and then you get – this gets reflected 00:43:24.720 --> 00:43:27.100 in the – on the anisotropic gradient, for example. 00:43:27.100 --> 00:43:32.719 But, as they close, this ability to detect their presence decreases. 00:43:32.720 --> 00:43:34.780 - Okay. - And – yeah, that’s what I meant. 00:43:34.780 --> 00:43:37.060 - Okay. I see. - Thank you. 00:43:37.360 --> 00:43:38.360 Yes? 00:43:39.820 --> 00:43:43.760 [Silence] 00:43:43.760 --> 00:43:47.000 - Thank you. Nice talk. Both parts were really interesting. 00:43:47.009 --> 00:43:52.440 My question kind of follows up on the – on the same point about your first talk 00:43:52.440 --> 00:43:56.019 and sort of a pragmatic question. I’m not really too familiar with these 00:43:56.019 --> 00:43:59.660 methods, but I assume, when the fractures close, 00:43:59.660 --> 00:44:02.549 your productivity goes down pretty quickly. 00:44:02.549 --> 00:44:04.930 - Yes. - And I wondered what proportion – 00:44:04.930 --> 00:44:09.420 you talked about the recoverability of these resources in your 00:44:09.420 --> 00:44:13.719 early motivating slide. What proportion of the oil are 00:44:13.719 --> 00:44:16.930 we recovering by these methods? And is this – can this process 00:44:16.930 --> 00:44:21.709 be renewed? Can they – can they go in again and frack again and go through 00:44:21.709 --> 00:44:25.910 the cycle multiple times? - That’s a – that’s a great question. 00:44:25.910 --> 00:44:30.759 Thank you. So, actually, productivity in general goes down for the first – 00:44:30.759 --> 00:44:34.610 like, it declines rapidly the first three years. 00:44:34.610 --> 00:44:41.969 But, all over, like, if you look at the whole lifetime of a well, 00:44:41.969 --> 00:44:46.380 you can say that, for a particular well, recoverability is around 25%. 00:44:46.380 --> 00:44:49.560 That’s what I – what I know. 00:44:49.560 --> 00:44:55.690 So, yes, so whether going back and re-injecting is feasible or not is 00:44:55.690 --> 00:45:02.360 up to the energy company. But the fact that fractures close 00:45:02.360 --> 00:45:06.700 after water flowback – so, when you inject fluid, you also inject proppants. 00:45:06.709 --> 00:45:12.390 You inject – there are different sizes of sand to keep as much of the 00:45:12.390 --> 00:45:18.240 fracture as open as possible. But some of them, as you see, will close. 00:45:18.240 --> 00:45:22.610 And a lot of them will close. And the fact that production declines 00:45:22.610 --> 00:45:28.840 the first – the first two to three years, for most wells, is well-established. 00:45:28.840 --> 00:45:33.160 But what I’m – what I want to say is that, not being able to detect these 00:45:33.160 --> 00:45:41.039 fractures with seismic methods does not mean that they are completely gone 00:45:41.039 --> 00:45:44.650 or they’re completely closed. It’s just – it’s our ability to detect 00:45:44.650 --> 00:45:46.789 their presence that – they’re open, they’re there, 00:45:46.789 --> 00:45:50.540 they’re affecting the anisotropy declines. 00:45:53.760 --> 00:45:56.980 - Do we have any other questions for Noha? 00:45:59.500 --> 00:46:03.080 [Silence] 00:46:03.080 --> 00:46:07.500 - What is the anticipated life expectancy of these wells? 00:46:09.480 --> 00:46:17.420 - I am not sure I have a – I don’t know I have a number in my head. 00:46:17.420 --> 00:46:22.920 I can – I can try to find out. But … 00:46:25.680 --> 00:46:31.380 I’m – yeah, I don’t know. I’m sure – I’m sure it’s – 00:46:31.380 --> 00:46:35.900 it can go up to, like, 20 years or – so it depends – so it depends on 00:46:35.900 --> 00:46:41.299 how the reservoir was stimulated. So if you look at all of the wells 00:46:41.300 --> 00:46:46.380 that are being stimulated, you can categorize them into three categories. 00:46:46.380 --> 00:46:51.259 First, wells that are complete flops. That’s one-third of the wells. 00:46:51.259 --> 00:46:55.910 And then you have one-third – another one-third of the wells that are 00:46:55.910 --> 00:46:59.500 economical only if the gas prices are high enough. 00:46:59.500 --> 00:47:02.989 And then a third of the wells actually carry all the others, 00:47:02.989 --> 00:47:07.059 that are well-performing – they perform very well, 00:47:07.060 --> 00:47:11.600 they are very economical in their production. 00:47:11.600 --> 00:47:17.719 So, but about the lifetime of each well, I think it depends on which category 00:47:17.719 --> 00:47:27.140 the well lies in, but in general, I don’t have – I don’t know which category – 00:47:27.140 --> 00:47:30.280 what each – what the lifetime of wells in each category is. 00:47:30.280 --> 00:47:33.900 - All right. Well, it seems that some of this activity has been going on long 00:47:33.910 --> 00:47:38.569 enough already that there would be some kind of history to it already. 00:47:38.569 --> 00:47:39.990 And also, I was wondering about the proppants. 00:47:39.990 --> 00:47:42.780 Have you looked into the proppants much? 00:47:43.260 --> 00:47:47.020 - I looked into the proppants? - Well, I think it was in the – 00:47:47.020 --> 00:47:53.480 well, it was in the poster, and I was just wondering 00:47:53.480 --> 00:47:59.500 how effective are proppants being in keeping these fractures open? 00:47:59.509 --> 00:48:05.340 - So how effective the proppants – it depends on what – so the proppants 00:48:05.340 --> 00:48:07.370 come in different sizes and different types. 00:48:07.370 --> 00:48:11.220 So, for example, if you know that – if you have – for example, 00:48:11.220 --> 00:48:17.779 you can use ceramic proppants if you know that the S-Hmax, for example, is – 00:48:17.779 --> 00:48:22.110 the stress is very compressive and you’re worried about the sand getting 00:48:22.110 --> 00:48:25.360 crushed. So it depends on what type of proppant you choose. 00:48:25.360 --> 00:48:30.670 I depends on the size of the proppant. It depends on how the – how many 00:48:30.670 --> 00:48:38.840 fractures you re-activated. And it also depends on the difference – 00:48:38.840 --> 00:48:43.840 how compressive – how compressive the stress environment is. 00:48:43.840 --> 00:48:49.380 Because, as I said, if you – if you choose, for example, sand in this – 00:48:49.380 --> 00:48:54.200 if the sand is not suitable, if it’s too compressive, you can go to – 00:48:54.200 --> 00:48:58.150 you can choose ceramic. You know. So it depends on what type of proppant 00:48:58.150 --> 00:49:04.220 you use, and it depends on the stress environment as well. 00:49:04.640 --> 00:49:06.860 - Thank you. - Thank you. 00:49:11.140 --> 00:49:15.420 - So there’s also this sort of ongoing discussion about the role of poroelastic 00:49:15.420 --> 00:49:20.451 effects in these stimulations and how you might, you know, re-activate 00:49:20.451 --> 00:49:25.299 a distant fault by inducing stress changes, not just pore pressure. 00:49:25.299 --> 00:49:28.870 - Yes. - I mean, the damage zone explanation 00:49:28.870 --> 00:49:34.319 is nice, and I can appreciate that, but how – just not having a good sense 00:49:34.319 --> 00:49:41.469 of how this all works, how badly do you suppose microseismic locations, 00:49:41.469 --> 00:49:46.069 for example, would be biased by anisotropy in the medium? 00:49:46.069 --> 00:49:48.960 Do you have any sense of that? 00:49:49.500 --> 00:49:51.980 - How the locations of microseismic events … 00:49:51.990 --> 00:49:55.559 - Right. Because it’s really critical to know where the – you know, 00:49:55.559 --> 00:50:00.690 where the locations of the earthquakes are so that we can determine whether 00:50:00.690 --> 00:50:05.110 it’s – you know, what the role of poroelastic effects are. 00:50:05.110 --> 00:50:08.630 - Okay. - So I’m just curious if that anisotropy 00:50:08.630 --> 00:50:14.320 that you’re measuring, or identifying, plays any role into those locations. 00:50:15.740 --> 00:50:22.509 - So the anisotropy – if you – if you think that anisotropy – if I succeeded 00:50:22.509 --> 00:50:29.340 in persuading you that anisotropy can be a proxy for pressure changes, then yes. 00:50:29.340 --> 00:50:33.930 If you – you can – you can estimate where you’re fracturing – you can – 00:50:33.930 --> 00:50:37.890 you can get a better idea about the simulated volume without 00:50:37.890 --> 00:50:41.049 resorting to the seismic – microseismic events if you 00:50:41.049 --> 00:50:44.950 look at the pressure changes reflected in anisotropy. 00:50:44.950 --> 00:50:49.090 So it’s true that – you know, you can actually – you can have pressure 00:50:49.090 --> 00:50:55.299 changes in locations where there were no fluid communication there. 00:50:55.299 --> 00:50:58.789 You can – you can – and that can affect the anisotropy, and it can actually – 00:50:58.789 --> 00:51:03.759 the fact that there were pressure changes there can cause microseismic events 00:51:03.759 --> 00:51:09.140 to be produced from – because the anisotropy means that 00:51:09.140 --> 00:51:12.220 you’re affecting the fractures. If you’re affecting the fractures, 00:51:12.220 --> 00:51:14.759 you might end up having microseismic – detectable 00:51:14.760 --> 00:51:20.200 microseismic events in that location. That’s a great question. 00:51:24.920 --> 00:51:27.520 - A couple comments. A great talk, first of all. 00:51:27.520 --> 00:51:32.680 - Thank you. - The one thing that, you know, 00:51:32.690 --> 00:51:38.780 we sort of hear now in the public is that it’s the injection of the wastewater 00:51:38.780 --> 00:51:42.920 that’s causing the induced seismicity and not the – by and large, and not 00:51:42.920 --> 00:51:48.229 the actual fracking process. But it looks to me like it’s both. 00:51:48.229 --> 00:51:51.009 And maybe you can comment on that. And then how we should – 00:51:51.009 --> 00:51:55.210 when we get – people come up to us and ask us about it, how we should respond. 00:51:55.210 --> 00:52:01.880 But the other thing is, your map view of the fractures and this dense layout 00:52:01.880 --> 00:52:06.489 over just, you know, a few hundred meters, it looks like – maybe 500 – 00:52:06.489 --> 00:52:11.529 half a kilometer or something, anyway, it’s not clear how deep down 00:52:11.529 --> 00:52:18.760 is the things that you’re imaging. And then, is the – this gets to 00:52:18.760 --> 00:52:23.840 Andy’s question about, do you have an idea for the uncertainty in the 00:52:23.840 --> 00:52:26.660 location of the earthquakes, and are those actually determined 00:52:26.660 --> 00:52:27.969 as part of your process? 00:52:27.969 --> 00:52:32.900 Or do you get those from somewhere else and just map them on there? 00:52:32.900 --> 00:52:35.680 So I guess that’s a couple of questions. [chuckles] 00:52:35.689 --> 00:52:39.580 - Thank you very much. Yes. So the earthquake process – 00:52:39.580 --> 00:52:41.979 the earthquake locations – the hypocenters were actually 00:52:41.980 --> 00:52:46.380 located by – were given to us by ConocoPhillips company. 00:52:46.380 --> 00:52:51.560 And they do play a role in my analysis concerning the faults, but not my 00:52:51.560 --> 00:52:56.160 analysis concerning the fractures. This, I did everything from scratch, 00:52:56.160 --> 00:52:59.600 and I took classes in seismic processing to do that. 00:53:01.540 --> 00:53:07.460 So, yeah, to answer this question, the microseismic locations were 00:53:07.479 --> 00:53:11.979 not determined by us and did not play a role in the fracture [inaudible]. 00:53:11.979 --> 00:53:16.229 - How deep is all this? - 8,750 feet. 00:53:16.229 --> 00:53:21.620 So if you think about – that’s pretty deep, so that’s the lower Barnett. 00:53:21.620 --> 00:53:25.120 And so I also get asked this question a lot. 00:53:25.120 --> 00:53:31.540 Like, does hydraulic fracturing, you know, cause induced seismicity. 00:53:31.540 --> 00:53:35.400 It seems that it depends on where you’re fracking. 00:53:35.400 --> 00:53:40.059 So if you’re fracking near a fault that, you know, can – it can become critically 00:53:40.059 --> 00:53:44.380 stressed and can slip and cause a large earthquake, so it depends on – 00:53:44.380 --> 00:53:48.249 depending on how big the fault patch would be in this case, then yes. 00:53:48.249 --> 00:53:53.809 But, in this case – in most – at least, to the best of my knowledge, 00:53:53.809 --> 00:53:59.949 these locations are surveyed first. And, depending on, for example, 00:53:59.949 --> 00:54:03.799 whether they want – there was a particular paper that I read where 00:54:03.799 --> 00:54:08.720 the faults were – by Hennings et al., was in Suban field, I think, in Sumatra. 00:54:08.720 --> 00:54:11.860 So what they did is they actually targeted the faults. 00:54:11.860 --> 00:54:18.900 And they ended up causing a lot of mud loss, so they had to stop drilling. 00:54:18.900 --> 00:54:22.620 However, where they drilled, the other faults that intersected the well, there 00:54:22.620 --> 00:54:26.610 was a lot of increase in production. So sometimes they are targeted. 00:54:26.610 --> 00:54:30.590 But that was in locations where they don’t care about induced seismicity. 00:54:30.590 --> 00:54:33.890 So it depends – so, yeah, it can happen, of course. 00:54:33.890 --> 00:54:36.249 I mean, you’re changing the pore pressure. 00:54:36.249 --> 00:54:40.299 So it depends on the faulting environment, and it depends on 00:54:40.299 --> 00:54:44.590 whether there are large enough faults that become critically stressed 00:54:44.590 --> 00:54:51.120 and can slip with fluid injection. But this is – this is pretty deep. 00:54:52.520 --> 00:54:55.140 This is about 9,000 feet. 00:54:59.140 --> 00:55:01.220 - Any remaining questions? 00:55:03.560 --> 00:55:06.260 Okay. Then let’s thank our speaker one more time. 00:55:06.260 --> 00:55:11.720 [Applause] 00:55:14.420 --> 00:55:20.940 [Silence]