WEBVTT Kind: captions Language: en-US 00:00:00.800 --> 00:00:05.440 Hello. I’m Dan Gittins, and today I’ll be talking about characterizing the 00:00:05.440 --> 00:00:09.200 along-strike rupture lengths of creep events along the San Andreas Fault. 00:00:09.200 --> 00:00:12.480 I would like to thank the organizers for inviting me to speak with you today. 00:00:12.480 --> 00:00:18.056 I’ll also thank my collaborator Jessica Hawthorne for her help with this work. 00:00:18.080 --> 00:00:22.400 So the San Andreas Fault has been observed to be creeping at the surface 00:00:22.400 --> 00:00:27.176 between San Juan Bautista and Cholame since the 1960s. 00:00:27.200 --> 00:00:30.800 This creeping section is bound by two 100-kilometer-long locked sections. 00:00:30.800 --> 00:00:34.080 The one north of San Juan Bautista last ruptured in the 00:00:34.080 --> 00:00:37.440 1906 magnitude 7.9 San Francisco earthquake 00:00:37.440 --> 00:00:42.160 and also the 1989 magnitude 6.9 Loma Prieta earthquake. 00:00:42.160 --> 00:00:44.480 And the locked section south of Cholame last ruptured 00:00:44.480 --> 00:00:49.016 in the 1857 magnitude 7.9 Fort Tejon earthquake. 00:00:49.040 --> 00:00:53.680 So this creeping section itself has some smaller magnitude approximately 6 00:00:53.680 --> 00:00:56.160 earthquakes that occur near the edge of the creeping section, 00:00:56.160 --> 00:01:00.080 typically around Parkfield. But there’s been no major magnitude 7 00:01:00.080 --> 00:01:05.096 or above earthquakes reported along this creeping section. 00:01:05.120 --> 00:01:08.720 So, since creep was observed at the beginning of the 1960s, this section of 00:01:08.720 --> 00:01:11.120 the San Andreas Fault has been heavily instrumented with creepmeters, 00:01:11.120 --> 00:01:15.736 strainmeters, GPS, alignment arrays, etc. Lots of instruments were installed. 00:01:15.760 --> 00:01:18.560 And the installation of these instruments revealed that creep wasn’t occurring 00:01:18.560 --> 00:01:22.456 at a steady rate, rather than, it was a steady rate that was 00:01:22.480 --> 00:01:27.040 repeatedly punctuated by bursts of slip, known as creep events. 00:01:27.040 --> 00:01:30.080 So these creep events typically have a few millimeters of slip. 00:01:30.080 --> 00:01:33.440 They typically last a few hours to days and recur every few weeks to months. 00:01:33.440 --> 00:01:37.760 So a fairly regular kind of burst that occurs. 00:01:37.760 --> 00:01:41.120 So, despite these creep events being pretty abundant, we’ve known about 00:01:41.120 --> 00:01:45.520 them since about the mid- to late 1960s, we don’t know how spatially large these 00:01:45.520 --> 00:01:49.360 creep events are – like, how far they extend along strike, how far they extend 00:01:49.360 --> 00:01:54.960 along depth, or the kind of fault zone process behind these creep events. 00:01:54.960 --> 00:01:57.840 So these events are kind of important to understand because they may be 00:01:57.840 --> 00:02:01.200 playing a key role in the physics of this creeping section as well as 00:02:01.200 --> 00:02:04.400 other creeping faults and may also have a wider implication for 00:02:04.400 --> 00:02:08.000 understanding fault physics in general. If your fault is creeping as well as 00:02:08.000 --> 00:02:10.800 slipping seismically, then you need to understand both parts of that 00:02:10.800 --> 00:02:16.456 system to be able to assess the hazard for that fault fully. 00:02:16.480 --> 00:02:20.320 So today I’m going to be looking at how long creep events are. 00:02:20.320 --> 00:02:24.160 So there’s kind of three categories of previous scenarios for how 00:02:24.160 --> 00:02:26.640 big creep events are. There’s these short, shallow ruptures, 00:02:26.640 --> 00:02:29.840 we are only located at one creepmeter. That’s the only place you’d record it. 00:02:29.840 --> 00:02:32.080 Anywhere else is just kind of a coincidence. 00:02:32.080 --> 00:02:35.680 Or you have these long and shallow ruptures that rupture multiple 00:02:35.680 --> 00:02:39.840 creepmeters but are quite shallow – only a kilometer or so deep. 00:02:39.840 --> 00:02:45.440 Or you have these kilometer-long, kilometer-deep big creep events 00:02:45.440 --> 00:02:48.320 like in Scenario C. And each one of these scenarios 00:02:48.320 --> 00:02:53.816 will have a different implication for the slip of the creeping region. 00:02:53.840 --> 00:02:56.640 So, in order to understand how big creep events are, you kind of 00:02:56.640 --> 00:03:02.776 need to know when they occur. So you can look at different records 00:03:02.800 --> 00:03:05.600 manually – visually inspect them to try and pick out all these 00:03:05.600 --> 00:03:09.576 creep events, but this is exceedingly time-consuming. 00:03:09.600 --> 00:03:14.160 So what we did was we did a bit of visual inspection first to get a feel of 00:03:14.160 --> 00:03:18.240 what these creep events were like, identified template events for each 00:03:18.240 --> 00:03:20.880 creepmeter, and then, using a cross-correlation approach, 00:03:20.880 --> 00:03:24.696 we identified these creep events within the creepmeter record. 00:03:24.720 --> 00:03:27.360 So we tried to identify times that had a similarity to the template, 00:03:27.360 --> 00:03:30.720 therefore we kind of get it was a creep event, and also had a significant 00:03:30.720 --> 00:03:35.976 slip, so it wasn’t just something to do with the instrument themselves. 00:03:36.000 --> 00:03:40.480 Doing this, we were able to identify 2,120 creep events across the 00:03:40.480 --> 00:03:43.520 18 creepmeters that have a variety of different statistics, 00:03:43.520 --> 00:03:48.880 which I’m kind of displaying here. So we have – for example, XSJ, 00:03:48.880 --> 00:03:53.016 we have few events, under 100 events, 00:03:53.040 --> 00:03:58.720 however, these are events which are larger and much longer, typically. 00:03:58.720 --> 00:04:03.440 And then we also have, so like XMM, Middle Mountain, which has lots of 00:04:03.440 --> 00:04:09.694 small short events, but it has over 500 events between 1985 and 2020. 00:04:10.560 --> 00:04:13.600 But the thing that they kind of all have in common, despite whether there are 00:04:13.600 --> 00:04:17.440 lots of small events or large fewer events is that they typically 00:04:17.440 --> 00:04:22.240 accommodate over 50% of the slip at the surface for most creepmeters. 00:04:22.240 --> 00:04:26.000 There are some exceptions of this, where the creepmeter might have 00:04:26.000 --> 00:04:30.640 left-lateral slip, so it slips the other way, so it’s kind of hard to determine what 00:04:30.640 --> 00:04:35.816 the maximum actual overall percentage of slip is accommodated in the events. 00:04:35.840 --> 00:04:38.960 So now we have identified when these events occur. 00:04:38.960 --> 00:04:41.920 We need to know how long they are. 00:04:41.920 --> 00:04:47.016 So here is an example of a creep event at Cienega Winery in May. 00:04:47.040 --> 00:04:51.840 And then we look for the next creepmeter, which is the Harris Ranch. 00:04:51.840 --> 00:04:54.480 It’s 4 kilometers away. And if you search within 24 hours 00:04:54.480 --> 00:04:57.280 either side of this start time for the creep event at Cienega Winery, 00:04:57.280 --> 00:04:59.736 you see this creep event at Harris Ranch. 00:04:59.760 --> 00:05:03.760 If you go further – 11 kilometers even further north of the creepmeter 00:05:03.760 --> 00:05:07.656 at San Juan Bautista, you do not see a creep event here. 00:05:07.680 --> 00:05:10.640 So this is kind of implying that, at this moment in time, we have 00:05:10.640 --> 00:05:14.216 a creep event that might be 4 kilometers long, but it’s probably not 15. 00:05:14.240 --> 00:05:16.960 This isn’t a very systematic way of doing it, if you just want to go through 00:05:16.960 --> 00:05:20.480 and manually look at every single creep event, so we kind of took this 00:05:20.480 --> 00:05:24.080 and then made a more statistical framework to base this on. 00:05:24.080 --> 00:05:28.960 So we identified times that were coincident between one creepmeter 00:05:28.960 --> 00:05:32.880 and every other creepmeter that had a creep event within 24 hours 00:05:32.880 --> 00:05:37.280 either side of that event. This led to a possible detection 00:05:37.280 --> 00:05:42.456 of 306 possible larger multi-creepmeter events. 00:05:42.480 --> 00:05:46.160 So the reason we look at 24 hours either side of the main creepmeter is because 00:05:46.160 --> 00:05:50.480 it provides a balance between robust statistics and slowly propagating 00:05:50.480 --> 00:05:55.760 creep events. If we go for a time less than 24 hours, we will exclude 00:05:55.760 --> 00:05:58.000 any creep event that might be slowly propagating. 00:05:58.000 --> 00:06:01.360 But if we go longer than 24 hours, especially in the few years after the 00:06:01.360 --> 00:06:04.880 2004 Parkfield earthquake, where the recurrence time of creep events 00:06:04.880 --> 00:06:07.600 is reduced, we end up with double correlations, and therefore we don’t 00:06:07.600 --> 00:06:10.560 know whether the creep event is going from one creepmeter to the other and 00:06:10.560 --> 00:06:13.520 back again, or whether it’s only one of them is actually correlating, or 00:06:13.520 --> 00:06:18.696 whether neither of them are correlating. So this is why we went for 24 hours. 00:06:18.720 --> 00:06:24.320 So, to get a better understanding of the statistics behind this, we first off 00:06:24.320 --> 00:06:28.560 take one creepmeter record and bootstrap it and then compare that time 00:06:28.560 --> 00:06:32.480 to every other – those bootstrapped times to other creepmeters to work out 00:06:32.480 --> 00:06:38.536 a distribution of creep events that were observed at both within 24 hours. 00:06:38.560 --> 00:06:42.160 So you get this distribution, but then it could be the case that these might 00:06:42.160 --> 00:06:45.280 just be occurring by chance. It’s a coincidence that they’re similarly timed. 00:06:45.280 --> 00:06:49.976 So then we took the creepmeter record and time-shift it by multiples of years 00:06:50.000 --> 00:06:53.840 with looping back around to the start to cover the beginning to look at whether 00:06:53.840 --> 00:06:57.040 these just kind of happened by chance. This time-shifting over multiples 00:06:57.040 --> 00:07:04.696 of years allows for the consideration of annual weather 00:07:04.720 --> 00:07:08.469 and long-term effects of rainfall. So this is kind of 00:07:08.493 --> 00:07:13.120 a by-chance or a coincidence distribution that we end up with here. 00:07:13.120 --> 00:07:15.760 And then, to understand how much is actually occurring, we need to 00:07:15.760 --> 00:07:19.496 take these by-chance events out of the observed distributions, 00:07:19.520 --> 00:07:23.200 which we do by randomly sampling both the by-chance distribution and the 00:07:23.200 --> 00:07:27.360 observed distribution and taking the by-chance away from the observed. 00:07:27.360 --> 00:07:32.000 So this leaves us with a real distribution here, which can be seen in Panel C, 00:07:32.000 --> 00:07:35.840 which gives you an idea of whether these group events are actually 00:07:35.840 --> 00:07:39.336 a coincidence or whether there’s something relating them. 00:07:39.360 --> 00:07:42.720 So, in this bottom Panel C, for example, we have creep events between 00:07:42.720 --> 00:07:47.440 CWN and XSJ, and this distribution centers around zero. 00:07:47.440 --> 00:07:50.136 So these events are likely to be unrelated. 00:07:50.160 --> 00:07:54.960 However, if you look at the CWN-XHR ones, they hover around 25, 00:07:54.960 --> 00:07:59.576 meaning these events are probably related to each other. 00:07:59.600 --> 00:08:01.920 So then, doing this for all the creepmeters, 00:08:01.920 --> 00:08:07.040 we end up with these five different behaviors for creep events. 00:08:07.040 --> 00:08:10.160 So we end up with these isolated events. These creepmeters that only have creep 00:08:10.160 --> 00:08:15.416 events at that one creepmeter, and you don’t see them anywhere else. 00:08:15.440 --> 00:08:21.600 So, for example, we have XMR, or Melendy Ranch, which we only 00:08:21.600 --> 00:08:24.400 see creep events there within 24 hours with itself. 00:08:24.400 --> 00:08:26.400 However, it’s quite isolated from other creepmeters, 00:08:26.400 --> 00:08:28.880 so if there was any creep events getting there, they might be 00:08:28.880 --> 00:08:34.536 propagating slower than the allotted 24 hours that we allowed for. 00:08:34.560 --> 00:08:39.680 The interesting pair here is that, in Carr Ranch and Gold Hill, 00:08:39.680 --> 00:08:42.000 we have two creepmeters that are only 2 kilometers away from each other, 00:08:42.000 --> 00:08:45.840 but they don’t have any interaction between each other. 00:08:45.840 --> 00:08:48.400 One seems to slip and the other one stops, or vice versa – 00:08:48.400 --> 00:08:51.840 one slips and the other one stops. 00:08:51.840 --> 00:08:54.800 So then we also have this next group of events, which are these short events, 00:08:54.800 --> 00:08:58.320 which are less than 2 kilometers long. These happen around Middle Mountain 00:08:58.320 --> 00:09:03.950 and Melendy Ridge, as well as the two creepmeters around Highway 46. 00:09:05.760 --> 00:09:09.920 So then we have these medium-size events, which happen around – 00:09:09.920 --> 00:09:13.440 there’s a section of five creepmeters around Parkfield between 00:09:13.440 --> 00:09:18.320 Middle Mountain and Taylor Ranch, which sit in two sections. 00:09:18.320 --> 00:09:21.680 So we have a Middle Mountain- Middle Ridge-Varian section, 00:09:21.680 --> 00:09:24.320 and then a Varian-Parkfield- Taylor Ranch section. 00:09:24.320 --> 00:09:27.280 So this is two 5-kilometer-long stretches that both slip 00:09:27.280 --> 00:09:30.080 independently from each other. And we also have this pair of 00:09:30.080 --> 00:09:32.800 creepmeters in the north, Cienega Winery and Harris Ranch, 00:09:32.800 --> 00:09:37.120 which slip at – about 4 kilometers away from each other, 00:09:37.120 --> 00:09:40.296 and they often slip together. 00:09:40.320 --> 00:09:44.080 So next we have the larger slip events. We have these events that look to 00:09:44.080 --> 00:09:47.200 be over 10 kilometers long. So these are typically around, 00:09:47.200 --> 00:09:50.880 again, this Parkfield section. So there’s – September 2000, 00:09:50.880 --> 00:09:54.880 there’s a creep event that occurs – that joins those two sections around 00:09:54.880 --> 00:09:59.176 Parkfield, that Middle Mountain to Taylor Ranch section. 00:09:59.200 --> 00:10:02.560 And also there’s some evidence for longer relations between events 00:10:02.560 --> 00:10:05.656 between Slacks Canyon and Work Ranch, 00:10:05.680 --> 00:10:10.216 however I will explain more about these events later. 00:10:10.240 --> 00:10:14.480 And finally, there is, around Parkfield, the San Andreas has two strands. 00:10:14.480 --> 00:10:17.200 We have the main strand and we have the southwest trace. 00:10:17.200 --> 00:10:20.720 And there appears to be evidence of creep events propagating on this 00:10:20.720 --> 00:10:26.320 southwest trace between the two creepmeters on that strand as well as 00:10:26.320 --> 00:10:31.360 evidence of interaction between creepmeters on the main San Andreas 00:10:31.360 --> 00:10:35.920 Fault and creepmeters on that southwest trace, predominantly a relation between 00:10:35.920 --> 00:10:42.136 the Hearst Southwest creepmeter and the Work Ranch creepmeter. 00:10:42.160 --> 00:10:45.120 So the elephant in the room when talking about these creep events is because 00:10:45.120 --> 00:10:47.920 we’re looking at creepmeters, and creepmeters are buried quite 00:10:47.920 --> 00:10:53.336 shallow, is also that these events could be driven by rainfall. 00:10:53.360 --> 00:10:56.560 So here we look at the short-term effects of rainfall. 00:10:56.560 --> 00:11:01.360 So we’re only looking at up to about two weeks’ time scale for rainfall. 00:11:01.360 --> 00:11:03.920 So we repeat the previous analysis that we’ve done already, 00:11:03.920 --> 00:11:07.280 but we remove events from the main creepmeter that have rainfall 00:11:07.280 --> 00:11:14.960 in the 3, 7, or 14 days beforehand. These typical time scales are 00:11:14.960 --> 00:11:17.600 kind of arbitrary. They don’t really mean anything physically, 00:11:17.600 --> 00:11:21.360 however they are longer than the typical time scale for association 00:11:21.360 --> 00:11:24.720 with rainfall because we’re also allowing for varying time scales of 00:11:24.720 --> 00:11:30.456 rainfall infiltration, as this is not a very well-constrained quantity. 00:11:30.480 --> 00:11:32.880 But this does allow us to have a broad overview of the effects 00:11:32.880 --> 00:11:37.440 of rainfall without having a proper, thorough investigation into it. 00:11:37.440 --> 00:11:42.800 It just kind of gives us the idea of how this rainfall is affecting these 00:11:42.800 --> 00:11:47.040 creep events. So we come up with these three kind of behaviors. 00:11:47.040 --> 00:11:50.160 The creep events are either affected by rainfall, unaffected by rainfall, 00:11:50.160 --> 00:11:55.016 or the percentages kind of increase when you remove rainfall. 00:11:55.040 --> 00:11:59.280 So the main events that are affected by rainfall is these longer creep events 00:11:59.280 --> 00:12:03.840 between Work Ranch and Slacks Canyon, which typically, when you 00:12:03.840 --> 00:12:07.736 remove events that have rainfall, you remove the relation between the two. 00:12:07.760 --> 00:12:12.080 So this would suggest that the closely timed events occurring at these 00:12:12.080 --> 00:12:16.160 creepmeters is a coincidence driven by rainfall rather than 00:12:16.160 --> 00:12:20.456 a longer creep event that’s connecting both regions. 00:12:20.480 --> 00:12:23.280 The second one, which is the most common, is that, when you remove 00:12:23.280 --> 00:12:27.680 rainfall, the change in percentages between the related two 00:12:27.680 --> 00:12:30.800 creepmeters doesn’t really change. It just stays pretty constant, 00:12:30.800 --> 00:12:32.960 doesn’t move. Which kind of implies that 00:12:32.960 --> 00:12:37.096 the effects of rainfall here are minimal. 00:12:37.120 --> 00:12:44.296 Finally, we have this kind of uncommon sort of behavior whereby, 00:12:44.320 --> 00:12:49.120 the pair of Cienega Winery and Harris Ranch, when you remove the rainfall 00:12:49.120 --> 00:12:54.560 at longer periods – at 7 or 14 days – we have this increase in relation 00:12:54.560 --> 00:12:58.160 between the two creepmeters. So it could be that we are removing 00:12:58.160 --> 00:13:02.000 these rainy intervals, which are already having shallow single creepmeter 00:13:02.000 --> 00:13:06.760 events, and then what’s left is these deeper, longer creepmeters that – 00:13:06.760 --> 00:13:10.856 creep events that span both creepmeters. 00:13:10.880 --> 00:13:15.200 So, to summarize the study that we have done is that we’ve identified 00:13:15.200 --> 00:13:20.080 and correlated 2,120 creep events along USGS creepmeters along the 00:13:20.080 --> 00:13:23.736 creeping section of the San Andreas Fault between 1985 and 2020. 00:13:23.760 --> 00:13:28.080 We’ve identified 306 potential multi-creepmeter events, which spans 00:13:28.080 --> 00:13:31.040 various along-strike lengths. And the vast majority of these are 00:13:31.040 --> 00:13:35.440 not correlated with rainfall. So this lack of correlation with 00:13:35.440 --> 00:13:38.560 rainfall would imply that something else is driving these creep events, 00:13:38.560 --> 00:13:42.056 and it’s not just a perturbation of a rainfall kick. 00:13:42.080 --> 00:13:45.440 So finally, what we’re working on currently is we’re working on 00:13:45.440 --> 00:13:48.240 currently to determine the depth of these creep events in order to 00:13:48.240 --> 00:13:52.856 better understand the driving process of these events. 00:13:52.880 --> 00:13:57.600 So this work is now available for you to read at your leisure if you’d like to, 00:13:57.600 --> 00:14:02.240 but also we have made public the creep event catalog, 00:14:02.240 --> 00:14:06.400 which some of you may also find useful. If you have any questions, 00:14:06.400 --> 00:14:10.616 please feel free to contact me on the above email address or on Twitter. 00:14:10.640 --> 00:14:14.320 And thank you again for this opportunity to speak to you today.