WEBVTT Kind: captions Language: en-US 00:00:02.000 --> 00:00:06.000 Hello. I’m Kathryn Materna, and I’m excited to talk to you today about rates 00:00:06.000 --> 00:00:09.976 of interseismic deformation near Humboldt Bay in southern Cascadia. 00:00:10.000 --> 00:00:14.080 This is a collaborative project with Jessica Murray at the USGS and 00:00:14.080 --> 00:00:18.376 Lucy Sandoe and Roland Bürgmann at UC-Berkeley. 00:00:18.400 --> 00:00:22.960 To begin, I’d like to bring us all to the study area of this project, 00:00:22.960 --> 00:00:27.680 the Mendocino Triple Junction. The triple junction is at the intersection 00:00:27.680 --> 00:00:30.080 of San Andreas Fault, the Mendocino Fault zone, 00:00:30.080 --> 00:00:34.640 and the Cascadia subduction zone. It’s a very seismically active area 00:00:34.640 --> 00:00:38.536 that hosted a magnitude 7 earthquake in 1992. 00:00:38.560 --> 00:00:41.280 North of the triple junction, there are number of Quaternary 00:00:41.280 --> 00:00:45.976 active structures in the USGS Faults and Folds Database that serve, 00:00:46.000 --> 00:00:49.030 to some degree, as motivation for this project. 00:00:50.000 --> 00:00:54.160 In the map on the right, I’m showing the overall tectonic overview of this 00:00:54.160 --> 00:01:00.800 area plus a number of annotations. I’m adding the horizontal GPS velocity 00:01:00.800 --> 00:01:04.880 vectors with respect to a stable North American Plate, and I also annotated 00:01:04.880 --> 00:01:11.096 a number of the major faults in the Quaternary Faults and Folds Database. 00:01:11.120 --> 00:01:14.880 We can see, based on the colors and the annotations, that some of these 00:01:14.880 --> 00:01:19.440 faults we don’t know very much about. But the fault that is thought to be 00:01:19.440 --> 00:01:22.800 most active out of this set is the Little Salmon Fault, which is 00:01:22.800 --> 00:01:26.936 thought to have a slip rate of between 4 and 12 millimeters per year. 00:01:26.960 --> 00:01:30.240 Trenching studies on this fault have suggested that it has 00:01:30.240 --> 00:01:34.626 experienced at least three earthquakes in the last 1,700 years. 00:01:35.440 --> 00:01:38.080 So the Little Salmon Fault and some of the others in this area 00:01:38.080 --> 00:01:40.704 are thought to be quite active. 00:01:41.280 --> 00:01:46.560 One of the exciting features about the Mendocino Triple Junction is 00:01:46.560 --> 00:01:51.710 its vertical deformation, evident on long and short time scales. 00:01:52.960 --> 00:01:55.040 On long time scales, vertical deformation 00:01:55.040 --> 00:01:59.176 at the triple junction is evident in the geologic record. 00:01:59.200 --> 00:02:03.360 Vertical deformation rates from marine terrace dating suggest that, 00:02:03.360 --> 00:02:07.760 in certain places, the uplift rates are up to 4 millimeters per year. 00:02:07.760 --> 00:02:13.016 In particular, these high uplift rates are located near the King Range terrane, 00:02:13.040 --> 00:02:15.988 where I’ve annotated in red on this map. 00:02:17.120 --> 00:02:20.960 But, throughout the area, there is evidence for uplift along the coast. 00:02:20.960 --> 00:02:24.493 And we can see in some of the photographs of the area that there 00:02:24.493 --> 00:02:28.160 is geomorphic evidence for uplifted terraces 00:02:28.160 --> 00:02:31.416 and young and steep topography. 00:02:31.440 --> 00:02:35.920 For context, 4 millimeters per year of long-term uplift is quite fast 00:02:35.920 --> 00:02:40.136 and is among the fastest of any region in the lower 48. 00:02:40.160 --> 00:02:43.576 It’s thought to be second only to Ventura, California, 00:02:43.600 --> 00:02:47.200 in southern California. And the uplift rates on this order 00:02:47.200 --> 00:02:54.242 of magnitude in Ventura were associated with significant hazard. 00:02:55.440 --> 00:02:58.480 On shorter time scales – geodetic time scales, 00:02:58.480 --> 00:03:04.530 there’s also interesting evidence in vertical deformation. 00:03:06.160 --> 00:03:11.040 In this map here, I’m showing geodetic data that comes from leveling in the 00:03:11.040 --> 00:03:15.656 circles, tide gauges in the squares, and GPS in the triangles. 00:03:15.680 --> 00:03:19.840 These come from various – from different data centers. 00:03:19.840 --> 00:03:24.080 The GPS comes from the USGS, and the leveling in the tide gauges 00:03:24.080 --> 00:03:27.736 are courtesy of the Humboldt Bay Vertical Working Group. 00:03:27.760 --> 00:03:33.440 When taken together, these vertical velocities show an interesting pattern 00:03:33.440 --> 00:03:37.440 with systematic subsidence around the area of Humboldt Bay 00:03:37.440 --> 00:03:40.651 in a block that seems to be fault-bounded. 00:03:41.760 --> 00:03:50.400 So this vertical deformation field in the geodetic domain is the 00:03:50.400 --> 00:03:54.880 motivation of our project. And we’re interested in understanding, 00:03:54.880 --> 00:03:57.920 what are the implications of the three-component geodetic data 00:03:57.920 --> 00:04:00.320 for fault slip and for seismic hazard in the region, 00:04:00.320 --> 00:04:04.696 and especially for these upper crustal faults. 00:04:04.720 --> 00:04:08.400 In order to answer this question, we are going to start by looking at 00:04:08.400 --> 00:04:12.640 the two-dimensional data alone. And the analysis technique we’re 00:04:12.640 --> 00:04:17.816 going to use with two-dimensional data is to investigate strain. 00:04:17.840 --> 00:04:23.840 So why are we doing this? The strain rate is a measurement that 00:04:23.840 --> 00:04:27.920 we can obtain from GPS velocities, and it’s convenient because it’s 00:04:27.920 --> 00:04:31.520 independent of the reference frame, and it removes – it removes for us 00:04:31.520 --> 00:04:35.736 the solid-body rotation component of the velocity field. 00:04:35.760 --> 00:04:42.080 The strain rate is giving as a tensor – epsilon-ij – and it is computed from 00:04:42.080 --> 00:04:45.440 displacement gradients. In the most general sense, 00:04:45.440 --> 00:04:51.040 the tensor is three-dimensional and has 3-by-3 components, 00:04:51.040 --> 00:04:54.720 but, because we’re looking at 2D horizontal velocities, 00:04:54.720 --> 00:04:59.600 and we’re only looking at 2D strain, we only use the 2-by-2 in the 00:04:59.600 --> 00:05:05.576 top corner – the X and Y components. The diagonal components are 00:05:05.600 --> 00:05:10.160 normal strains, and the off-diagonal components are shear strains. 00:05:10.160 --> 00:05:16.776 Because this is a tensor, it can be fairly difficult to plot on a map. 00:05:16.800 --> 00:05:21.760 And so we often plot invariants of the tensor because they are 00:05:21.760 --> 00:05:25.120 more convenient. There are some standard invariants – 00:05:25.120 --> 00:05:29.016 the first invariant and the second invariant that I’m showing here. 00:05:29.040 --> 00:05:33.016 The first invariant is simply the trace of the tensor. 00:05:33.040 --> 00:05:37.656 And it is also known as the dilatation. It’s a very physical quantity. 00:05:37.680 --> 00:05:41.520 When it is positive, it means there is extension and the volume 00:05:41.520 --> 00:05:46.696 of that domain is growing. And when it is negative, it means 00:05:46.720 --> 00:05:51.176 there is shortening and the volume of the domain is shrinking. 00:05:51.200 --> 00:05:55.680 The second invariant, however – we also plot the second invariant 00:05:55.680 --> 00:05:58.960 on occasion, but it doesn’t have a catchy name. 00:05:58.960 --> 00:06:00.936 We just call it the second invariant. 00:06:00.960 --> 00:06:05.120 And it displays the overall magnitude of strain. 00:06:05.120 --> 00:06:09.280 It does not tell you whether the strain is shortening or shear, 00:06:09.280 --> 00:06:12.296 but it simply displays the overall magnitude. 00:06:12.320 --> 00:06:18.960 We also occasionally view the principal strains, which are the 00:06:18.960 --> 00:06:22.560 eigenvectors of the system. And they’re plotted as these 00:06:22.560 --> 00:06:26.056 sort of crosshairs on certain plots. 00:06:26.080 --> 00:06:29.280 So how do we compute strain from GPS velocities? 00:06:29.280 --> 00:06:33.760 The first technique – the simplest technique 00:06:33.760 --> 00:06:38.000 is known as Delaunay triangulation. It’s quite simple, but it is not smooth 00:06:38.000 --> 00:06:41.280 or robust to outliers, and it’s not robust to oddly shaped networks, 00:06:41.280 --> 00:06:44.056 as you can see in these elongated triangles. 00:06:44.080 --> 00:06:47.840 So are there more robust ways to compute a strain rate? 00:06:47.840 --> 00:06:55.120 It turns out there are quite a few ways that are smoother, or they incorporate 00:06:55.120 --> 00:07:02.240 an interpolation scheme or a weighting scheme to produce a robust strain rate. 00:07:02.240 --> 00:07:06.560 I have seen probably at least 10 in the literature, and so there’s 00:07:06.560 --> 00:07:09.440 a number of ways that one can go about doing this. 00:07:09.440 --> 00:07:13.040 And so what we decided to do for this project was to construct an ensemble 00:07:13.040 --> 00:07:18.000 strain rate map by performing this calculation in a number of – 00:07:18.000 --> 00:07:21.096 using a number of methods and then averaging their results. 00:07:21.120 --> 00:07:25.576 The idea here is to give us an ensemble strain rate map 00:07:25.600 --> 00:07:31.200 that displays the characteristics common to all of these methods 00:07:31.200 --> 00:07:35.336 and does not display any of the artifacts of a single method. 00:07:35.360 --> 00:07:39.520 This is similar to a strain exercise that was performed by SCEC 00:07:39.520 --> 00:07:45.200 a number of years ago. So when we use four methods here, and we average 00:07:45.200 --> 00:07:49.576 them together on the same grid, here are some of our results. 00:07:49.600 --> 00:07:53.736 Here is our ensemble second invariant map. 00:07:53.760 --> 00:07:56.880 And we can even calculate an uncertainty – 00:07:56.880 --> 00:08:01.440 an empirical uncertainty on quantities like the second invariant by taking 00:08:01.440 --> 00:08:05.679 the standard deviation of those four methods that we’ve tried. 00:08:06.800 --> 00:08:10.800 So, to interpret this, near the triple junction, it appears that we have high 00:08:10.800 --> 00:08:16.536 interseismic strain through the second invariant throughout the forearc faults. 00:08:16.560 --> 00:08:19.360 We can tell that the empirical uncertainties are higher in 00:08:19.360 --> 00:08:23.336 the forearc than they are around the San Andreas Fault. 00:08:23.360 --> 00:08:27.655 Higher in this area than in these values. 00:08:29.280 --> 00:08:32.400 The uncertainties in some places are on the order of magnitude 00:08:32.400 --> 00:08:36.160 of the strain rates themselves. And this may be surprising, 00:08:36.160 --> 00:08:40.856 but similar results were actually found in the SCEC exercise too. 00:08:40.880 --> 00:08:45.921 We find similar rotation values in each of the methods that we tried. 00:08:47.120 --> 00:08:53.120 To expand on our results, we can plot the first invariant in the dilatation. 00:08:53.120 --> 00:08:58.240 And, again, the negative values are shortening, and the positive 00:08:58.240 --> 00:09:02.880 values in the blue are extension. The dilatation map – the mean 00:09:02.880 --> 00:09:06.080 dilatation map here shows a strong shortening feature 00:09:06.080 --> 00:09:11.200 around the forearc faults, where the azimuth of shortening 00:09:11.200 --> 00:09:14.645 is approximately given by these vectors. 00:09:15.680 --> 00:09:19.736 So we see shortening perpendicular to the forearc faults. 00:09:19.760 --> 00:09:23.120 When we look at the direction of shortening, we can actually plot 00:09:23.120 --> 00:09:26.936 the direction of shortening in this very colorful plot here, 00:09:26.960 --> 00:09:32.776 where it may be easiest to look at some of these colors by their little keys. 00:09:32.800 --> 00:09:38.936 The blue, which is at about 50 degrees clockwise from north, 00:09:38.960 --> 00:09:46.240 represents shortening along this axis. Purple – closer to 20 degrees 00:09:46.240 --> 00:09:51.040 from north – represents this kind of tensor and is indicative 00:09:51.040 --> 00:09:53.440 of the San Andreas shear system. 00:09:53.440 --> 00:09:57.336 And red would indicate this kind of tensor. 00:09:57.360 --> 00:10:06.375 So we can see, in the forearc, there’s agreement that the tensors are in this – 00:10:06.400 --> 00:10:10.640 in this shortening direction, but there’s actually not as much 00:10:10.640 --> 00:10:14.400 agreement in the methods compared to what we find 00:10:14.400 --> 00:10:17.840 further south in the San Andreas. This is where the uncertainties are low, 00:10:17.840 --> 00:10:21.280 and in the north in the triple junction is where the uncertainties 00:10:21.280 --> 00:10:25.920 are a little bit higher. We can also study the maximum shear, 00:10:25.920 --> 00:10:29.040 which is another type of – which is basically another invariant we can 00:10:29.040 --> 00:10:35.920 pull from the tensor, whom we find, not surprisingly, a lot of shear strain 00:10:35.920 --> 00:10:38.560 in the San Andreas Fault system and then a fair bit 00:10:38.560 --> 00:10:41.416 of shear strain to the north. 00:10:41.440 --> 00:10:46.488 And we find low uncertainties on each of these quantities. 00:10:47.920 --> 00:10:53.896 To put some of this together, we can plot the dilatation 00:10:53.920 --> 00:10:58.536 together with vertical velocities from geodetic measurements. 00:10:58.560 --> 00:11:04.640 So, in this map here, the background color is the dilatation derived from 00:11:04.640 --> 00:11:11.896 horizontal GPS velocities, and the symbol color – the triangles and the 00:11:11.920 --> 00:11:19.656 circles – are the GPS and leveling vertical velocities from the earlier slide. 00:11:19.680 --> 00:11:23.920 This map shows a reasonable degree of correlation between areas 00:11:23.920 --> 00:11:28.320 where there is shortening and areas where there is vertical uplift. 00:11:28.320 --> 00:11:31.840 And similarly, a correlation between areas where there’s extension and areas 00:11:31.840 --> 00:11:37.040 where there’s vertical subsidence. And that’s the kind of pattern 00:11:37.040 --> 00:11:41.256 we would expect from an elastic strain accumulation process. 00:11:41.280 --> 00:11:45.680 However, there seems to be an interesting outlier in the Humboldt Bay 00:11:45.680 --> 00:11:50.640 area, where there is subsidence in the same area that there is shortening. 00:11:50.640 --> 00:11:56.456 And that is an interesting feature that we intend to look at further. 00:11:56.480 --> 00:12:00.480 We can also plot the maximum shear zoomed in on the Humboldt Bay area, 00:12:00.480 --> 00:12:04.960 and we can see this feature of shear strain extending to the north. 00:12:04.960 --> 00:12:11.280 But we can expand this tensor and look at the plane on which that 00:12:11.280 --> 00:12:16.480 maximum shear is experienced. And that is more a north-south 00:12:16.480 --> 00:12:21.440 oriented plane. It’s not exactly aligned with these particular faults. 00:12:21.440 --> 00:12:24.936 So it’s an interesting feature as well. 00:12:24.960 --> 00:12:26.936 So what have we learned? 00:12:26.960 --> 00:12:29.680 We’ve learned that, near the triple junction, at the latitude 00:12:29.680 --> 00:12:33.680 of Humboldt Bay, we have high interseismic strain with 00:12:33.680 --> 00:12:38.240 fairly large empirical uncertainties. So that makes it difficult for us to 00:12:38.240 --> 00:12:43.436 use this exact calculation for slip rates or moment accumulation. 00:12:44.480 --> 00:12:48.160 But it’s still useful, as we have identified that there is quite high 00:12:48.160 --> 00:12:52.080 margin-perpendicular contraction around the faults north of 00:12:52.080 --> 00:12:54.640 the triple junction. And there’s well-resolved 00:12:54.640 --> 00:12:58.177 shear strain, as well, north of the San Andreas Fault. 00:12:59.280 --> 00:13:04.456 So what are the implications of this result for fault activity? 00:13:04.480 --> 00:13:09.176 We are interested in understanding further these implications through 00:13:09.200 --> 00:13:13.736 our next steps, which would be fault modeling through dislocation models. 00:13:13.760 --> 00:13:17.120 We would like to estimate slip rates on some of these faults by building 00:13:17.120 --> 00:13:20.800 a suite of dislocation models. There has been previous work 00:13:20.800 --> 00:13:26.936 in dislocation modeling in this area from Williams et al. and Pollitz et al. 00:13:26.960 --> 00:13:32.240 And we would like to extend that work by including the leveling data, 00:13:32.240 --> 00:13:36.480 the tide gauge data, and the continuous – as much continuous 00:13:36.480 --> 00:13:40.776 and campaign GPS data as we can reasonably include. 00:13:40.800 --> 00:13:45.120 And our goal is to incorporate both vertical and horizontal 00:13:45.120 --> 00:13:47.840 deformation into these models. 00:13:47.840 --> 00:13:52.080 We will also focus on the tradeoffs inherent in a network of closely spaced 00:13:52.080 --> 00:13:57.600 faults in doing this type of modeling. So, to conclude, thus far, on this 00:13:57.600 --> 00:14:02.456 ongoing project, we have shown that GPS and leveling, tide gauges, 00:14:02.480 --> 00:14:05.280 display systematic vertical deformation near faults 00:14:05.280 --> 00:14:08.136 north of the Mendocino Triple Junction. 00:14:08.160 --> 00:14:14.800 And our analysis of horizontal GPS velocities and strain rates shows high 00:14:14.800 --> 00:14:19.976 strain rates north of the triple junction with residual right-lateral shear, 00:14:20.000 --> 00:14:24.776 active shortening in the forearc perpendicular to these forearc faults, 00:14:24.800 --> 00:14:27.360 and we have derived empirical uncertainties on each of 00:14:27.360 --> 00:14:30.960 these quantities. The dislocation modeling work is ongoing, 00:14:30.960 --> 00:14:35.280 and we’re aiming to perform that working with vertical data. 00:14:35.280 --> 00:14:38.720 So thank you for your attention, and I’m happy to take questions 00:14:38.720 --> 00:14:41.501 and discussion. Thank you.