WEBVTT 00:00:00.000 --> 00:00:03.000 No. 00:00:03.000 --> 00:00:14.000 Hello, hello, hello. Welcome back. Okay, it is the penultimate session. It is on Geodesy. 00:00:14.000 --> 00:00:23.000 And all maestros of geodesy today are going to be Katherine Guns and Chelsea Scott. Woohoo! 00:00:23.000 --> 00:00:28.000 Thank you, Sarah. 00:00:28.000 --> 00:00:29.000 I think you're going to start us off, Katherine. 00:00:29.000 --> 00:00:35.000 You going for it, Chelsea or should I. You're going to start us off Katherine. Okay, I'm on it. Yeah. All right. So hello, everybody. Welcome to the Geodesy session. So exciting. 00:00:35.000 --> 00:00:48.000 We have a great slate of talks today. That are going to feature four speakers that each have extensive experience in piecing apart different crustal deformation signals from all sorts of different genetic data sets from GNSS to INSAR to optical imagery 00:00:48.000 --> 00:00:59.000 And first up, I'm super happy to introduce Mike Floyd from MIT, who will be speaking about GNSS observations for active faulting, geothermal, and hydrological studies in Northern California. 00:00:59.000 --> 00:01:29.000 Take it away. 00:01:47.000 --> 00:02:14.000 Hey everyone, I'm Katie. 00:02:14.000 --> 00:02:18.000 To talk number two. We'll come back to Mike. Moderate us. Take us to talk number two. 00:02:18.000 --> 00:02:35.000 All right, great. So our first and next talk is then going to be Katherine or Cassie Hannigan who works at the US Geological Survey, and she will be telling us about creep over a step over. 00:02:35.000 --> 00:02:48.000 Hey everyone, I'm Cassie Hannigan and I'm a Mendenhall postdoc at the Earthquake Science Center stationed in Golden. I'm happy to be presenting a specific case study along the central creeping section of the San Andreas Fault that with new geodetic results. 00:02:48.000 --> 00:02:53.000 Highlights capabilities of a rather less typical method for understanding fault processes and deformation. 00:02:53.000 --> 00:02:57.000 And with that, I'll jump right in to creeping through a step over. 00:02:57.000 --> 00:03:05.000 Top-down view from geomorphology, seismology, and geodesy. As the session title states. 00:03:05.000 --> 00:03:15.000 This is a short talk, but I want to provide a roadmap and an overarching question to guide your attention because I'll be highlighting a lot of prior work with our very recent advances that are still a work in progress. 00:03:15.000 --> 00:03:21.000 So it'll take us through a large scale background focused on the central creeping section of the San Andreas Fault. 00:03:21.000 --> 00:03:26.000 Its existing geodetic record and examples of how that has been used. 00:03:26.000 --> 00:03:31.000 Which will funnel into a zoom in of Mustang Ridge, a key sub area along the creeping San Andreas. 00:03:31.000 --> 00:03:42.000 And we'll introduce the geomorphic context of the subarea, updated velocity maps, and a slip model at depth informed by surface deformation and seismicity. 00:03:42.000 --> 00:03:56.000 So here's our overarching question that I want to focus on. How can we combine the detailed geologic and historic records of surface deformation to understand the fall evolution towards current fault displacement from the surface to the subsurface. 00:03:56.000 --> 00:04:00.000 So why here? Why do we care about the central San Andreas? 00:04:00.000 --> 00:04:04.000 Well, creep is a rather interesting phenomenon that serves to decrease seismic hazard. 00:04:04.000 --> 00:04:11.000 It quietly releases the stress and strain that would otherwise build up in the system for future release in an unstable earthquake. 00:04:11.000 --> 00:04:20.000 And there are two commonly cited reasons for creep. It's overall simple fault geometry, generally thought to result from the large over 600 kilometer accrual of slip. 00:04:20.000 --> 00:04:29.000 That tends to simplify the fault geometry. This is shown in this map on the right, highlighting how straight the central San Andreas is at this large scale. 00:04:29.000 --> 00:04:39.000 Multi-hundred kilometer scale. And the presence of weak velocity strengthening minerals, like what I've pictured here, serpentanite, in the field from Mustang Ridge. 00:04:39.000 --> 00:04:50.000 Which the San Andreas Fault cuts through. The same paper that contains the map on the right by Burford and Harsh highlights how remarkably consistent this creep is on decadal timescales. 00:04:50.000 --> 00:04:57.000 From repeat measurements of alignment arrays installed in the 1960s with small 30 to 200 meter across fault aperture. 00:04:57.000 --> 00:05:04.000 Here I've pictured one alignment array that's still there in the ground at me ranch and has accrued over 1.2 meters of offset. 00:05:04.000 --> 00:05:21.000 Across the creeping fault since its installation. Now, creep decreases in the transition zones towards the end of the creeping section of the San Andreas towards the 1857 and 1906 earthquake ruptures on either end. 00:05:21.000 --> 00:05:33.000 Now fast forward to today and many studies have provided estimates of updated creep rates from a variety of newer geodetic data that also document continued consistent creep rates for much of the central San Andreas. 00:05:33.000 --> 00:05:40.000 The most spatially detailed understanding of creep near the fault comes from topographic differencing of repeat lidar presented in sky et al. 2020. 00:05:40.000 --> 00:05:48.000 Which I highlight more later. This provides a 40 meter resolution displacement estimate over the whole creeping section. 00:05:48.000 --> 00:05:54.000 Over an 11 to 13 year time span ending in 2018. 00:05:54.000 --> 00:06:01.000 And these studies provide decades of fault creep rates with an opportunity to observe and understand fault motion on our human timescales. 00:06:01.000 --> 00:06:09.000 Enabling methodologic and scientific advancements without waiting for a major earthquake. 00:06:09.000 --> 00:06:24.000 These deformation measurements we can use. And because they provide a spectacular opportunity for a host of studies, including investigating creep differences through decades and different apertures across the fault. 00:06:24.000 --> 00:06:34.000 Understanding fault coupling and slip depth. And correlating with seismicity to understand seismic behavior and fault properties. 00:06:34.000 --> 00:06:38.000 Upon closer look, though, the simple San Andreas is not as simple as it seems. 00:06:38.000 --> 00:06:54.000 I've zoomed into our sub area of interest for the remainder of the talk where the San Andreas takes an uncharacteristic one kilometer right step through the prominent Mustang Ridge over a distance of a few kilometers. 00:06:54.000 --> 00:07:01.000 And going back to our main question, we can use the already studied geomorphic record of the step over to understand the false evolution here. 00:07:01.000 --> 00:07:11.000 Complementing our geodetic record of displacements. More succinctly, Mossing Ridge might be described as an extensional strike slip. 00:07:11.000 --> 00:07:19.000 And on top of the ridge, the fault rates do noticeably drop, and Monarch Peak is the alignment array from the Burford and Harsh study. 00:07:19.000 --> 00:07:23.000 That is on top of that ridge and is still on top of that ridge today. 00:07:23.000 --> 00:07:31.000 On top of the ridge, the geomorphic expression of faulting portrays a series of en echelon oblique normal faults. 00:07:31.000 --> 00:07:37.000 And two primary structures show evidence for the most slip given the SCARP expression relative to neighboring an echelon. 00:07:37.000 --> 00:07:46.000 At this inferred primary active zone of slip, there is evidence of rotation of the primary San Andreas Fault strand. 00:07:46.000 --> 00:07:52.000 It's pretty neat. And further to the northwest, Garve Expression, Channel Profiles, and SagPond suggest recent activity. 00:07:52.000 --> 00:08:02.000 Whereas to the southeast the initial one structures appear more worn down by time, suggesting they might be relics of the ridge's formational story. 00:08:02.000 --> 00:08:08.000 However, geomorphic timescales in this context are still much greater than our historic and ancestral lifetimes. 00:08:08.000 --> 00:08:11.000 So which parts of this step over are more active today? 00:08:11.000 --> 00:08:18.000 And how do they accommodate such consistent creep in the presence of apparently complex geometry? 00:08:18.000 --> 00:08:30.000 Existing topographic differencing results from Scott et al. That I mentioned earlier hint at the deformation across the unashal unfolds may in fact grade from inactive to active towards the northwest. 00:08:30.000 --> 00:08:37.000 But the 40 meter resolution picture of fault parallel slip is noisy, leaving an ambiguous understanding of detailed displacements. 00:08:37.000 --> 00:08:44.000 Across this region. So can we gain a better understanding with higher resolution displacement mounts? 00:08:44.000 --> 00:08:52.000 Here, I'm showing new fault parallel velocity maps from aerial UAV SAR-based NSAR at 7 meter resolution, courtesy of Jen Liu. 00:08:52.000 --> 00:09:05.000 Which contains data from 2009 to about 2022, as well as results from LIDAR subpixel correlation using the 11 meter time span, 11 year, sorry, time span of 2007 to 2018. 00:09:05.000 --> 00:09:16.000 Half meter resolution. These two maps complement each other to provide a more detailed picture of fault motion, albeit still with some noise. 00:09:16.000 --> 00:09:23.000 And we can compare these maps to records of surface deformation from the existing alignment arrays as a general validation. 00:09:23.000 --> 00:09:30.000 The maps on the bottom right plot existing map fault traces with alignment array velocities in maroon arrows. 00:09:30.000 --> 00:09:42.000 At Mian Monarch, which are in our sub areas. In our sub area. And if we take fault perpendicular profiles of velocity across both arrays from the LIDAR and UAVSAR maps. 00:09:42.000 --> 00:09:51.000 Then compare them to the velocities from recent alignment array occupations. We corroborate that we are at least able to produce similar patterns and rates of fault slip. 00:09:51.000 --> 00:09:56.000 The comparisons aren't perfect, potentially from measurement uncertainties, ground change, and related default creep. 00:09:56.000 --> 00:10:07.000 Or real folk creep differences across different temporal baselines. Nonetheless, we can still infer that the new velocity fields are providing a reasonable fog creep signal. 00:10:07.000 --> 00:10:14.000 And placing this in the context of Burtford and Harsh, we have comparable rates. 00:10:14.000 --> 00:10:20.000 There we go. So I've gone rapidly through snippets of the geomorphic and geodetic records of deformation here. 00:10:20.000 --> 00:10:29.000 Now, let's dive into the subsurface. Here, I'm showing two parallel fault traces derived from the LIDAR and UAV star velocity maps. 00:10:29.000 --> 00:10:35.000 As the distinctly active fault strands which compose the bounds of the step-over region at Mustang Ridge. 00:10:35.000 --> 00:10:50.000 And diving into the subsurface. We can look at seismicity from the Northern California Earthquake catalog to understand where the fault might lie at depth underneath the region. Travis Salonji identified a single surface at depth to help constrain our understanding of the false structure. 00:10:50.000 --> 00:10:55.000 This suggests one connecting plane that bends in a pattern mimicking the surface jog. 00:10:55.000 --> 00:11:17.000 Generally dipping to the northeast of the surface trace. However, between five kilometers in the surface, seismicity is largely absent. 00:11:17.000 --> 00:11:31.000 So, by Occam's razor, we took something really simple and we take the fault traces at the surface and project two triangulated surfaces towards a common trace defined by seismicity at that five kilometer depth. 00:11:31.000 --> 00:11:42.000 With increasing fall patch area with greater depths. Now we have the pieces to put together a very preliminary model of false slip beneath the step over region. 00:11:42.000 --> 00:11:50.000 So what we're going to do is subsample the fault parallel surface displacement maps and invert for strike slip motion within the fault model. 00:11:50.000 --> 00:11:53.000 And for the model, we'll prescribe only right lateral strike slip. 00:11:53.000 --> 00:12:04.000 With loose smoothing constraints on patches within about 100 meters of each other. 00:12:04.000 --> 00:12:10.000 And jumping into the results, the picture of slip looks something like this. On the left, we're looking towards the northeast. 00:12:10.000 --> 00:12:19.000 And on the right, we're looking towards the northwest. The model shows relatively low slip in the immediate subsurface, partly owing to the smooth nature of the displacement maps at the surface. 00:12:19.000 --> 00:12:29.000 That then coalesces on higher slip rates at depth with reasonable velocities approaching the rates we expect from previous estimates from about 30 to 40 millimeters per year. 00:12:29.000 --> 00:12:39.000 And note that where the surface traces overlap in the relay region, slip indeed tapers on the south tapers on the southwest and transitions towards the northeast Strand. 00:12:39.000 --> 00:12:57.000 And this provides a more detailed picture of slip depth that has been produced in the literature or any study presented so it's a pretty neat picture, I think, and it's concentrated on this detailed region of the creeping section. 00:12:57.000 --> 00:13:05.000 Now, the preliminary slip rate model also reproduces the observed surface velocities reasonably well. 00:13:05.000 --> 00:13:12.000 Leaving behind several features that we know are not fault related, such as these circled earth flows in the residuals. 00:13:12.000 --> 00:13:18.000 And the surface deformation patterns minus the vertical component, which we still don't have reliable constraints on. 00:13:18.000 --> 00:13:26.000 Are largely reproduced. Notably, through going normal faults are not necessarily needed in the step over region. Instead. 00:13:26.000 --> 00:13:36.000 Diffuse deformation between the primary active subparallel fault strands might be able to explain the surface velocities taking place today. 00:13:36.000 --> 00:13:40.000 I've highlighted one region in this transition near the geomorphically rotated portion of the San Andreas Fault. 00:13:40.000 --> 00:13:53.000 Where slip rates transition without any explicit normal faulting. In sum, we're corroborating the zones of active and inactive deformation informed by the geomorphic record. 00:13:53.000 --> 00:14:04.000 But don't require discrete and echelon faults to accommodate deformation. Instead, we introduced the possibility of diffuse near-surface deformation that may be better modeled through methods capable of inelastic processes. 00:14:04.000 --> 00:14:12.000 These combine interpretations bridge a story of formation from geologic to historic timescales. 00:14:12.000 --> 00:14:19.000 And there are similar corollaries in the literature from other field examples, modeling efforts, and experiments that document similar strikes-led evolution. 00:14:19.000 --> 00:14:24.000 And by definition, we are in the late stage strike slip faulting regime. 00:14:24.000 --> 00:14:30.000 Where anastomosing vault strands maintain and trade off to support the present-day deformation field. 00:14:30.000 --> 00:14:37.000 Mustang Ridge records a snapshot of this evolutionary story, and the present fault displacement map. 00:14:37.000 --> 00:14:48.000 Show what's happening today. I'm excited to dive into this evolutionary story more with future time and work. 00:14:48.000 --> 00:14:54.000 With that, I'll leave off. We've gone through this small case example of a geometrically complex region. 00:14:54.000 --> 00:15:08.000 On the conventionally thought of simple creeping San Andreas. We've provided what I hope you might agree is a prime example of how you can use this high resolution geodetic data to complement the geologic and historic records of deformation. 00:15:08.000 --> 00:15:24.000 And methodologically, this presents an advancement in mapping change in detail. Scientifically, we are able to tease apart the details of how fault slip is currently accommodated through the step over with low deformation rates that has a complex geometric signature in the landscape. 00:15:24.000 --> 00:15:31.000 I'm happy to take any questions, comments, and suggestions at the end of our discussion. 00:15:31.000 --> 00:15:38.000 Period. Thanks. 00:15:38.000 --> 00:15:58.000 Thank you very much, Cassie. This looks like there's some good discussion going, so that's great. And we can hopefully come back to some of those questions in discussion period later. So next we'll be hearing from Danny Lenzi and she is at UC Berkeley. 00:15:58.000 --> 00:16:06.000 And she will be discussing annual to decadal creep rates along the Makama Fault. 00:16:06.000 --> 00:16:15.000 Hi, I'm Dani Lindsey and today I'm going to be talking to you about temporal and spatial creep variability. So we're detecting decadal changes in fault behaviour in Northern California. 00:16:15.000 --> 00:16:17.000 And this is work I've been doing as part of my PhD. 00:16:17.000 --> 00:16:27.000 With Takaaki and Roland Bergman. So in the session, we ask what Geodyssey can do for you in Northern California. And I want to tell you everything. 00:16:27.000 --> 00:16:32.000 So we can do plate boundary scale deformation. We can look at volcanic deformation. 00:16:32.000 --> 00:16:37.000 We can look at slow-moving landslides. We can look at subsidence from aquifers. 00:16:37.000 --> 00:16:42.000 And in this talk, I'm going to be focused on looking at fault creep signals. 00:16:42.000 --> 00:16:44.000 So why do we care if something is creeping or not? 00:16:44.000 --> 00:16:50.000 So as something is creeping, it's releasing strain and then there's less energy in the system for big earthquakes. 00:16:50.000 --> 00:16:58.000 Can be a barrier to slip propagation so that creeping patch can inhibit the earthquake's propagation And then limit the magnitude of an event. 00:16:58.000 --> 00:17:17.000 And finally, it can inform the temporal changes in earthquake hazard. So this can be from external forcings or from dynamic processes and these happen across time scales such as annual, so like moderated by precipitation Decadal as in from changes in stressing rates from neighbouring earthquakes or some kind of evolution of the fault itself over the seismic cycle. 00:17:17.000 --> 00:17:36.000 Reflected in coupling changes. So where are we? We are in the northern extent of the San Andreas Fault system. So from south to north we have South of the bay we have the creeping section of the San Andreas Fault, which is well instrumented. And then moving north, we go across where the plate boundary deformation is distributed across these three systems. 00:17:36.000 --> 00:17:49.000 So we have the San Andreas fault system the rogers creek a mechma fault system, the Green Valley and Bartlett Springs fault system, and each of these is generally lesser studied. They've got slower loading rates and is trickier terrain for geodetic imaging. 00:17:49.000 --> 00:18:00.000 I'm asking today, which regions are creeping? And we're looking at surface creep from Elos 2. And so this is an L-band satellite from the Japanese Space Agency, and it's better suited for vegetated regions. 00:18:00.000 --> 00:18:04.000 And we're looking at also Sentinel-1 in SAR and alignment array data. 00:18:04.000 --> 00:18:08.000 And for deep creep, we're looking at that with characteristic repeating earthquakes. 00:18:08.000 --> 00:18:16.000 And I'm also asking, is there evidence for temporal variability? And so we're looking at the alignment array data and then also And Sarah as well. 00:18:16.000 --> 00:18:29.000 So a little intro to the methods. We're looking at characters characteristic repeating earthquakes. So these happen where you have one patch of the fault that's slipping And one side of the fault slides past the other one asseismically. But on that patch there might be these little stuck 00:18:29.000 --> 00:18:44.000 Regions here, and we call these asperities and these pop off for like little earthquakes And each time it pops off you have a seismic wave that goes from the little earthquake to the seismometers and these have a characteristic waveform so if these are 00:18:44.000 --> 00:18:51.000 Have really similar waveforms in a similar location with a similar magnitude. We characterize these as repeating earthquakes. 00:18:51.000 --> 00:19:03.000 And for surface displacement, we use INSAS. We have a satellite that flies over. We generate some kind of map of how the ground is moving. And in this case the ground is moving away from the satellite in one direction. 00:19:03.000 --> 00:19:19.000 One side of the fault and towards the satellite on the other And in this case, we're going to use cross fault displacement time series. So we take a patch on the west side and the east side. We look at how the displacement changes through time. 00:19:19.000 --> 00:19:24.000 Kind of creep signal. And so the slope of this displacement gives us a creep rate. 00:19:24.000 --> 00:19:37.000 And then we can use that across the whole length of the fault and we can look at how the fault accumulates displacement with time And then we can look at how the creep rate changes along the length of the phone. So we are 00:19:37.000 --> 00:19:42.000 For each section of the fault, we're looking at the surface crepe from the insert, we're inferring the location of deep creep. 00:19:42.000 --> 00:19:46.000 With the characteristic earthquakes and then we are also looking for temporal changes. 00:19:46.000 --> 00:19:57.000 So from south to north, here is the rogers creek vault So we start off here and we can see in the insar that there's no clear step across this area when we get to the Santa Rosa area where we have these three alignment arrays. 00:19:57.000 --> 00:20:05.000 There is some kind of step across here and then moving northwards the step continues and we can see that there's more repeating earthquakes here in the northern section. 00:20:05.000 --> 00:20:11.000 If we take a cross section of those repeating earthquakes, we can see again the south has very little seismicity. This is locked. 00:20:11.000 --> 00:20:16.000 And then as you move northwards there's repeating earthquakes along with where we see creep from the alignment arrays. 00:20:16.000 --> 00:20:35.000 If we look at the displacement with time with the ins are we can start to see this coming through. So anything that's moving from pink to green is creeping and right letter of direction and so we can see the southern section here where we have this vertical signal we have this really clear seasonality to the cross fault displacement 00:20:35.000 --> 00:20:43.000 So that's not felt great. And then when we see it go pink to green we can start to see these creep signals coming through. And so here we can see really low creep rates. 00:20:43.000 --> 00:20:50.000 Kind of similar to some of the alignment arrays coming through on the Santa Rosa section. So yeah, I think we see surface green. 00:20:50.000 --> 00:20:57.000 So we can zoom right into Santa Rosa and complement our ALOS 2 with the Sentinel 1. 00:20:57.000 --> 00:21:12.000 The Sentinel one has both ascending and descending and really good temporal coverage through this region. You can see it decorrelates as soon as we get into the vegetated areas but the actual urban area we can extract a quality fault parallel signal from. 00:21:12.000 --> 00:21:20.000 And so because we have both observation angles, I can decompose this into a faux parallel component and a vertical component. 00:21:20.000 --> 00:21:36.000 So on the left we can see that both of the time series, the cross fault time series, have this really clear seasonality. And when we decompose it through time, we can see that that seasonality falls out into the vertical component and we get a clear creep rate of about 3.3 millimeters per year which is 00:21:36.000 --> 00:21:45.000 Slightly less than the nearest alignment array, but a pretty clear signal. 00:21:45.000 --> 00:21:58.000 When we move northwards along the Mekamah Fault. The southern section here we see no clear offset across it and then between Jimtown and Cloverdale here We can see this really clear offset parallel to the geysers. 00:21:58.000 --> 00:22:04.000 As we move northwards, it gets a little complicated. The fault crosses through a few sediment basins and the topography gets in the way a little bit. 00:22:04.000 --> 00:22:13.000 And so we have a less clear signal. The repeating earthquakes though they start to pop off north of Cloverdale and we can see that they follow multiple different structures. 00:22:13.000 --> 00:22:18.000 And this has already been seen an existing repeating catalogues up north. 00:22:18.000 --> 00:22:24.000 When we take this cross section, we can see both the repeating earthquakes and the seismicity shallow towards the geysers here in the middle. 00:22:24.000 --> 00:22:31.000 And we have shallowing of the repeating earthquakes, also where we see some fault creep from the alignment array. 00:22:31.000 --> 00:22:36.000 When we look at the displacement time series, we can start to see there's a few patches that are really clearly creeping. 00:22:36.000 --> 00:22:48.000 And so the southern section definitely we don't see the seasonal signal and we don't see any of the For creep, it's pretty much at zero in the southern section we come up north and we get rates as high as kind of seven millimeters a year. 00:22:48.000 --> 00:22:54.000 And these are kind of in agreement with previous estimates as well. 00:22:54.000 --> 00:22:58.000 So I'm just zooming in here between Cloverdale and the MAMR alignment array. 00:22:58.000 --> 00:23:04.000 And we're just taking a look at what's causing some of this noise that we see in this creep rate estimate. 00:23:04.000 --> 00:23:21.000 So the middle panel here is the eastward velocity from the Sentinel-1 decomposition and I've just got the ALOS2 zoom in there for reference. So we can really see how the spatial coverage of the central one rarely drops off as soon as we're in the mountainous regions with lots of vegetation. 00:23:21.000 --> 00:23:40.000 But it still does give us some information. And so we can see that this deep blue patch on the Alos 2 between the two points is also in the sentinel one. And when we look in, we can see that these are actually an existing slow-moving landslide. So the pink polygon there as a previously identified as a 00:23:40.000 --> 00:23:49.000 Landslide and we pull out the time series we can really clearly see that this landslide is moving and it had some kind of acceleration during the 2019 year. 00:23:49.000 --> 00:24:09.000 So although these ones are offset from the fault, it's something that we really need to keep in mind when we're looking at these creep rates. So I'm only taking data from within about a kilometre of the fault trace itself. However, when these landslides or other features get really close they can start to permeate the the fault creep estimates. 00:24:09.000 --> 00:24:16.000 And the northern section finally we can see between the two alignment arrays here between Ukiah and Willits. 00:24:16.000 --> 00:24:19.000 Then we can see this nice clear offset in the NAR. 00:24:19.000 --> 00:24:33.000 However, the seismicity and the repeating earthquakes are offset by the fault. When we take our cross profile, we can see this cluster of repeating earthquakes and this is actually in an area which had a swarm activity in 2020. 00:24:33.000 --> 00:24:39.000 When we look at our cross fault displacement time series, we can see a really clear signal in some of these areas. 00:24:39.000 --> 00:24:48.000 And when we look at the Fall Creek Brights, they're kind of a little lower than the alignment arrays, but the peak rate here is pretty similar, probably around five millimeters a year. 00:24:48.000 --> 00:24:54.000 So then we go to look at decadal changes and the best data we have for this is actually the alignment arrays. 00:24:54.000 --> 00:24:57.000 They go back to the early 90s and they've been taken all the way through to 2022. 00:24:57.000 --> 00:25:15.000 And just when we look at this by eye, we can see that there's a step in the displacement time series here. So I just detrended it to really show that step and we can see that there's a change in the velocity before and after the step and that difference is about 2.5 millimeters per year and it continues through time. So we have this kind of decadal change in 00:25:15.000 --> 00:25:28.000 That creep at Willits. And if we look at the alignment array in ukaio, there is still some stiff, but the change is not as prominent And the data is a lot noisier. 00:25:28.000 --> 00:25:33.000 So something I'm still grappling with is what could drive this change in the creep rate. 00:25:33.000 --> 00:25:35.000 Could it be from a stress shedder from a big earthquake? 00:25:35.000 --> 00:25:48.000 So this was the case after the Loham Prieta earthquake. The Hayward Fault shut down for a number of years and had a suppressed creep, right? However, when we look at the earthquakes along the West Coast for the six months prior. 00:25:48.000 --> 00:25:58.000 There's nothing bigger than a 5.9. Could it be something more moderate and closer by? The largest event was a 4.3 in Hopland about six months prior, and that was 50 kilometres away. 00:25:58.000 --> 00:26:04.000 So I don't think that's the case. Could it be from surface loading, from drought or heavy precipitation? 00:26:04.000 --> 00:26:19.000 The year, the water year of 2020. Of 2002 doesn't stand out particularly for drought or precipitation so I don't think that's the case Is it something we can't see? So is it some kind of transient creep event that's changed the stress state of the fault? 00:26:19.000 --> 00:26:28.000 It could be. I can go back and have a look at the early insert data so we have INSAR for this period from ERS 2 and Radasat. 00:26:28.000 --> 00:26:39.000 One, however, the the magnitude of the step is on the order of five millimeters or something so this could be smaller than the detection of these early satellites. 00:26:39.000 --> 00:26:44.000 So if anybody has any suggestions, feel free to come and chat to me afterwards. 00:26:44.000 --> 00:26:57.000 When we then use the data that has really frequent acquisitions, this was four or five acquisitions a year And we go from displacement to velocity, we can start to see that the changes of displacement are annually modulated. So I've just 00:26:57.000 --> 00:27:06.000 An annual sign here. And when we normalize these velocities through the year, we can really see that the blue here is the mean monthly precipitation. 00:27:06.000 --> 00:27:19.000 The yellow is the model that I fitted and all of the ones in the back are each of the velocity time series. And every year agrees pretty well except for this one that's red and that represents this one with the big step here 00:27:19.000 --> 00:27:29.000 So generally at the alignment array in Willits. The creep appears to be seasonally modulated. So I asked, can we see that insect? 00:27:29.000 --> 00:27:36.000 So we use the Sentinel-1 velocity field to really get at this question of can we see the seasonality. 00:27:36.000 --> 00:27:49.000 We decompose it into a fault parallel and vertical component and the time series I'm showing is just the nearest fault crossing displacement time series to the alignment ray. 00:27:49.000 --> 00:28:04.000 On the top left you can see that both the ascending and the descending has a component of seasonal deformation. And when we decompose that into a vertical and a fault parallel component, we can see that that blue is showing the seasonal component and the green is the fault parallel. 00:28:04.000 --> 00:28:19.000 Time series um And so that seasonality we see really does come out in the vertical component. And the creep rate we get here from the center one is about 3.1 millimeters per year. So it's similar to the ALOS 2 and a little lower than the 00:28:19.000 --> 00:28:33.000 Long-term alignment array. And finally, this is kind of ongoing work, we're looking at whether we can see this decadal change in the creep rates so we can see if this surface rate is reflected also at depth on the fault 00:28:33.000 --> 00:28:40.000 And so here I'm plotting the cumulative false slip calculated from the repeating earthquakes for two previously published studies as well as this one. 00:28:40.000 --> 00:28:46.000 And at this stage, I think… We'll carry on. I'll let you know in the future. 00:28:46.000 --> 00:29:07.000 And so to summarize, we've investigated sources of noise and creep rates at annual indicator timescales and we find We present you spatially continuous creep rate estimates from INSA which is better suited for vegetated and mountainous regions We see these decadal changes in creep rate at the Willits alignment array and we find this annual correlation with the creep rate also. 00:29:07.000 --> 00:29:15.000 And in the future, we want to compare early NSAR estimates with the present day to try and constrain these changes in the long-term creak rate. 00:29:15.000 --> 00:29:23.000 Educational and seasonal skies as well. Thank you. 00:29:23.000 --> 00:29:32.000 Awesome. Thanks, Danny. That was great. Next, we will have Katya Timofieva from NASA JPL, who will talk to us about the upcoming NISAR mission. Yay. 00:29:32.000 --> 00:29:38.000 And what could be coming next for Geodyssey in Northern California. So take it away. 00:29:38.000 --> 00:29:43.000 Hello, my name is Katya Timaveva and I'm from the Jet Propulsion Lab. 00:29:43.000 --> 00:29:48.000 And I work on the nicer mission as well as the surface deformation and change study. 00:29:48.000 --> 00:29:53.000 First, I would like to acknowledge the NICER Project Science and SDC teams. 00:29:53.000 --> 00:29:59.000 Much of the material in this presentation was a collaborative effort together with some of the people listed on this page. 00:29:59.000 --> 00:30:04.000 Today's presentation has two parts. First, I'm going to tell you about the upcoming NYSAR mission. 00:30:04.000 --> 00:30:09.000 And the impact that the new measurements will have on the study of deformation in Northern California. 00:30:09.000 --> 00:30:22.000 Then I will tell you about some of the concepts for potential future missions that we have been considering and how they could contribute to science goals of the Northern California research community. 00:30:22.000 --> 00:30:30.000 Nicer is a synthetic aperture radar satellite mission. That will provide a global data set enabling the study of surface deformation and change. 00:30:30.000 --> 00:30:35.000 Some of the research and applications that will be addressed by NISR are shown on this page. 00:30:35.000 --> 00:30:46.000 They include the dynamics of ice sheets in sea ice. Carbon storage and update dynamics in wooded agricultural, wetland And permafrost systems, earthquake, volcanic, and landslide cycles. 00:30:46.000 --> 00:30:57.000 Including exploring potentials for urgent response and hazard mitigation. The study of aquifer health agricultural activity and the impact of productivity measures. 00:30:57.000 --> 00:31:08.000 Field scale soil moisture with weekly variability Infrastructure monitoring and land level change and other science and application goals. 00:31:08.000 --> 00:31:11.000 The first thing you should know about NYSAR is that it is coming soon. 00:31:11.000 --> 00:31:20.000 Launch is scheduled for 2025 that is this year. We can't share an exact lunch date with you right now, but we can say that it's coming up in the next few months. 00:31:20.000 --> 00:31:29.000 About two to three months after launch, you should be able to start to see first light images and then science ready data will become available about three months after launch. 00:31:29.000 --> 00:31:35.000 Niser will be providing a global data set with a 12-day repeat orbit ascending and descending passes. 00:31:35.000 --> 00:31:42.000 The data will include geocoded terrain-corrected polarometric products and geocoded interferometric products. 00:31:42.000 --> 00:31:55.000 We will also be providing a number of cloud-based tools, services, and tutorials that will hopefully make the data as accessible as possible Even to those members of the community who are not SAR experts. 00:31:55.000 --> 00:32:05.000 This is the nicer observation plan. I apologize for the rather old figure It has been revised somewhat since January 2023, but the changes have been relatively minor. 00:32:05.000 --> 00:32:21.000 What I would like to emphasize here is that all land is covered and also that this observation plan is quite sophisticated. It has been designed and tailored by the NICER Science team over the years to address a variety of observational goals. 00:32:21.000 --> 00:32:25.000 Nicer is designed to operate at two frequencies, L and S band. 00:32:25.000 --> 00:32:29.000 Which correspond to radar wavelengths of 24 and 12 centimeters respectively. 00:32:29.000 --> 00:32:44.000 Most of the time, L-band and S-band observations will be collected separately, but this map shows several areas Most of them corresponding to the NICSR science calibration and validation sites, where S and L band data will be collected at the same time. 00:32:44.000 --> 00:32:54.000 As you can see, this includes all of California. Northern California is home to a variety of geophysical mechanisms that produce surface deformation. 00:32:54.000 --> 00:32:59.000 Some of these have been studied using genetic measurements From current and legacy NSAR missions. 00:32:59.000 --> 00:33:13.000 As well as GMSS. On this slide, I'm showing just a few examples of the many great studies that have been formed here. Mainly, I want to point out the fact that the signals present here span a significant range of magnitudes as well as spatial and temporal scales. 00:33:13.000 --> 00:33:16.000 On the top left is a figure from Shu et al. 00:33:16.000 --> 00:33:22.000 Showing the deformation per year along the San Andreas Fault in the descending line of sight of the Sentinel-1 mission. 00:33:22.000 --> 00:33:33.000 If you look closely enough, the figure captures quite a few different deformation mechanisms that are present in California, including intersseminic motions that vary slowly in time and space. 00:33:33.000 --> 00:33:39.000 As well as the rapid and localized deformation Due to the hydrologic subsidence signals in Central Valley. 00:33:39.000 --> 00:33:44.000 This figure also highlights the particular challenge of studying deformation in Northern California with INSAR. 00:33:44.000 --> 00:33:53.000 Namely, the data loss due to the correlation in vegetated areas, specifically when using data collected at C-band. 00:33:53.000 --> 00:34:04.000 On the bottom left is an older study from Zhenlu and others using ERS 1 and 2 data to build the time series of the Time-varying uplift and deflation of the Long Valley Caldera. 00:34:04.000 --> 00:34:12.000 Spanning the years 1992 to 2001 and showing total displacements on the order of about plus or minus 10 centimeters over that time span. 00:34:12.000 --> 00:34:22.000 You may notice that the measurements are fairly sparse. Sometimes only a few per year because of a limited number of observations provided by the ERS mission. 00:34:22.000 --> 00:34:33.000 And finally, on the right is a figure from Chelsea Scott and others showing one of the many studies of creep along the central San Andreas Fault using Sentinel-1 data in the line of sight of the satellite. 00:34:33.000 --> 00:34:40.000 The creep is localized at the fall trace and characterized by mostly horizontal fall parallel motions on the order of millimeters per year. 00:34:40.000 --> 00:34:49.000 In addition, in this image, you can also see deformation from other sources such as oil fields, which would likely be characterized by mostly vertical motions. 00:34:49.000 --> 00:34:57.000 Again, these are just a few examples of deformation that can be observed in Northern California and have been studied using data from past insert missions. 00:34:57.000 --> 00:35:07.000 Next, I will try to make the case that the NICSAR mission will enable significant improvements in the availability and quality of the geodetic imagery in the region. 00:35:07.000 --> 00:35:14.000 Here I would like to show how NICSER can address some of the current challenges of studying surface deformation in Northern California. 00:35:14.000 --> 00:35:18.000 As well as what sets it apart from existing INSAR missions in general. 00:35:18.000 --> 00:35:24.000 As I mentioned in the previous slide, the vegetation cover in Northern California is one of the challenges for insert deformation measurements. 00:35:24.000 --> 00:35:33.000 As it can cause de-correlation and loss of data. Especially at smaller radar wavelengths On the left is a figure from a study by Wei and Sanuel. 00:35:33.000 --> 00:35:38.000 Comparing L-band and C-band imagery. Over several active faults in Northern California. 00:35:38.000 --> 00:35:46.000 It is clear that the LBAN data shown in the top panels is much more effective at maintaining correlation than the C-band data in the middle panels. 00:35:46.000 --> 00:35:52.000 Because the L-band radar signal is capable of penetrating the agitation cover. 00:35:52.000 --> 00:36:01.000 Niser will provide geodetic measurements at L band, which will help mitigate data loss Due to decorrelation in vegetated areas. 00:36:01.000 --> 00:36:06.000 Nasa will also provide regular observations with a 12-day interferometric revisit time globally. 00:36:06.000 --> 00:36:15.000 For both ascending and descending geometries, which will enable the robust estimation of time series of observations for transient deformation processes. 00:36:15.000 --> 00:36:20.000 Ionospheric noise is one of the most significant sources of error in INSAR. 00:36:20.000 --> 00:36:38.000 A couple of slides ago I showed that California was one of the sites where Joint observations at two radar wavelengths, L and S band, will be available, which will significantly improve the deformation measurement accuracy by enabling the estimation and removal of ionospheric noise. 00:36:38.000 --> 00:36:47.000 Finally, one of the greatest challenges of studying deformation with INSAR measurements in general is the fact that INSAR is inherently a scalar measurement of deformation. 00:36:47.000 --> 00:36:51.000 Always made in the line of sight of the observing satellite. 00:36:51.000 --> 00:37:03.000 One way to mitigate this is to collect data from multiple lines of sight, for example, using ascending and descending observations of a single satellite, or possibly by combining data from different satellite missions. 00:37:03.000 --> 00:37:07.000 Most data that is currently available has been collected with a right-looking satellite geometry. 00:37:07.000 --> 00:37:15.000 And even combining data sets from multiple missions does not provide sufficient geometric diversity for a good constraint on 3D deformation. 00:37:15.000 --> 00:37:35.000 Nicer will be left looking and so combining its data with that of right-looking missions like ALOS or Sentinel could provide significantly better constraints on 3D deformation, making it possible to better separate, model, and understand The different deformation mechanisms that may be superimposed in a given study area. 00:37:35.000 --> 00:37:40.000 Here I'm going to switch gears and talk a bit about what we hope might happen after NISAR. 00:37:40.000 --> 00:37:51.000 So first, a bit of background. The surface deformation and change study was commissioned to come up with a synthetic aperture radar mission that would address the observation goals from the 2017 Decadal survey. 00:37:51.000 --> 00:37:59.000 Some of the geophysical observables and research areas are shown on the left and some of the desired capabilities of the mission are shown on the right. 00:37:59.000 --> 00:38:09.000 As you can see, there is a wide range of science and applications identified by the Ducata survey requiring not only SAR interferometry but also SAR radiometry. 00:38:09.000 --> 00:38:27.000 Some of the other desired capabilities include continuous global coverage, deformation time series measurements with accuracies of one millimeter per week to one millimeter per year, and interferometric repeat pass measurements ranging from subweekly to sub daily rates. 00:38:27.000 --> 00:38:35.000 We began our study with over 40 different potential architectures and we have narrowed them down to these satellite constellation concepts. 00:38:35.000 --> 00:38:46.000 I will go through each of these concepts briefly. On the top left is our multi-squint architecture The science needs driving this architecture include any geophysical research that requires the knowledge of 3G deformation. 00:38:46.000 --> 00:38:51.000 Such as the earthquake cycle volcanic systems, groundwater flow, and many more. 00:38:51.000 --> 00:39:00.000 The basic premise of this concept is that we have a group or groups of three satellites that observe a point on the ground simultaneously at different angles. 00:39:00.000 --> 00:39:10.000 This enables the measurement of 3D deformation in real time and also the partial removal of the tropospheric propagation effects leading to increased measurement accuracy. 00:39:10.000 --> 00:39:20.000 On the top right we have the split swath concept Basically, there are a number of satellites spaced evenly in orbit. In this example there are five. 00:39:20.000 --> 00:39:24.000 In order to get global coverage every 12 days, the equivalent of NYSAR. 00:39:24.000 --> 00:39:27.000 Each one of them covers one fifth of a larger swath. 00:39:27.000 --> 00:39:31.000 However, if a disaster happens, such as an earthquake, volcanic eruption, or other event. 00:39:31.000 --> 00:39:36.000 The satellites can be temporarily repointed to observe the same footprint on the ground. 00:39:36.000 --> 00:39:42.000 And dramatically reduce the local revisit time. On the bottom left is a sub-daily concept. 00:39:42.000 --> 00:39:46.000 For certain specified areas, a number of interferograms with six-hour revisit. 00:39:46.000 --> 00:39:53.000 Can be made every 10 to 12 days. Depending on the number of satellites in this constellation. 00:39:53.000 --> 00:40:05.000 This enables the study of certain specific dynamic phenomena such as some cases of rapid transient deformation I shall flexure tidally driven coastal processes and soil moisture change. 00:40:05.000 --> 00:40:16.000 Finally, on the bottom right is the low duty cycle concept where we have a number of small satellites with a somewhat reduced data quality but with a flexible observation strategy. 00:40:16.000 --> 00:40:21.000 These can be configured in a variety of ways and used to target specific areas for observational campaigns. 00:40:21.000 --> 00:40:27.000 For example, reducing your visit time In certain chosen areas to a daily revisit. 00:40:27.000 --> 00:40:31.000 Each of these concepts can be applied to studying a variety of deformation mechanisms. 00:40:31.000 --> 00:40:49.000 In the next few slides, I will show some examples of deformation signals that have been observed elsewhere but are just as likely to be observed In Northern California, and I will explain how the study of these phenomena would benefit from observations made by some of these architectures. 00:40:49.000 --> 00:40:59.000 This first example explores the capability of responding to a disaster such as an earthquake using one of the mission concepts where multiple satellites provide background global coverage. 00:40:59.000 --> 00:41:04.000 But can be repointed when needed to observe a specific area at a more frequent rate. 00:41:04.000 --> 00:41:15.000 The example shown here specifically highlights The low duty cycle architecture where there are many In this case 12 small satellites, and revisit times can be reduced down to one day. 00:41:15.000 --> 00:41:25.000 This can allow for mapping of damage after an earthquake. And providing vital information to first responders, significantly supporting their efforts in rescue operations. 00:41:25.000 --> 00:41:46.000 The imagery here is from Turkey where two major earthquakes happened nine hours apart on February 6th 2023, with the low duty cycle architecture and its capability to provide Revisit times of as little as one day Depending on the timing of observations, it may have even been possible to study these two events separately. 00:41:46.000 --> 00:41:57.000 To better understand the co-seismic rupture corresponding to each event. The second example shows the benefit of a split swath architecture for measuring rapid transient information such as monitoring. 00:41:57.000 --> 00:42:09.000 Magma movement. This can likewise be said for the low duty cycle architecture from the previous slide since they are both concepts that allow for campaign measurements with more rapid revisit time. 00:42:09.000 --> 00:42:17.000 The split swath architecture can achieve two-day interferometric revisit over an erupting volcano in case of a need for urgent response. 00:42:17.000 --> 00:42:22.000 The example shows insert measurements of surface deformation caused by dyke propagation at depth. 00:42:22.000 --> 00:42:28.000 The two-day revisit capability could allow for better estimation of the shape and extruded volume of the dike. 00:42:28.000 --> 00:42:34.000 As well as a possibility for a prediction of how long an eruption might last. 00:42:34.000 --> 00:42:48.000 This final example focuses on the multi-squent architecture concept. Were groups of three satellites observed the ground at the same time Resolving deformation into three dimensions in real time Rather than studying deformation in the line of sight of one satellite. 00:42:48.000 --> 00:42:51.000 The examples shown here are from the Kilauea volcano in Hawaii. 00:42:51.000 --> 00:43:00.000 And tectonic deformation in Southern California. These examples highlight the potential for separating and understanding different mechanisms that are driving the deformation. 00:43:00.000 --> 00:43:12.000 To create more accurate models. The multi-screen concept also provides a way to partially estimate and remove the tropospheric noise enabling significantly better accuracy of observations. 00:43:12.000 --> 00:43:20.000 To wrap up this discussion, I would like to pass along to this community The current status and outlook for a future surface deformation and change mission. 00:43:20.000 --> 00:43:26.000 As I mentioned previously, over the course of five years we studied over 40 different possible satellite architectures. 00:43:26.000 --> 00:43:35.000 In a variety of configurations and assess their capability for achieving the science goals outlined by the 2017 decadal survey. 00:43:35.000 --> 00:43:39.000 We eventually narrowed down the possibilities to the four concepts that I presented here. 00:43:39.000 --> 00:43:44.000 We had originally expected that the next step would involve moving forward with one of the concepts. 00:43:44.000 --> 00:43:56.000 Toward building and launching the next INSAR mission. But unfortunately, we were directed to wrap up the study in 2023 And at this time, no program start is planned as we wait for a better budget outlook. 00:43:56.000 --> 00:44:07.000 The nicer Prime Mission is expected to end between 2028 and 2030, Possibly extending to 2033 or 35 If the satellite remains healthy and NASA continues funding. 00:44:07.000 --> 00:44:14.000 To get the next instar mission off the ground. The program must start at least six years prior to launch at the latest. 00:44:14.000 --> 00:44:29.000 This implies that if we want to have continuous insert measurements after NISARS, The program for the next INSAR mission should start any time between 2022 which has already passed to 2029, depending on how long Nyson remains healthy. 00:44:29.000 --> 00:44:33.000 The bottom line is this, as a member of the earth science community. 00:44:33.000 --> 00:44:38.000 You can play a role in defining and advocating for the next INSAR mission. 00:44:38.000 --> 00:44:47.000 You really do have a voice. Not only that, but we actually really need you and your help in order to make sure the next mission happens. 00:44:47.000 --> 00:44:54.000 The way to go about it is to write and submit a white paper to the next decadal survey outlining your specific research needs. 00:44:54.000 --> 00:44:59.000 And the measurement capabilities that you require in order to reach your scientific goals. 00:44:59.000 --> 00:45:07.000 These white papers are expected to be solicited sometime approximately a year from now, so it may be a good idea to start planning for them soon. 00:45:07.000 --> 00:45:23.000 If you have questions and if you would like to get more information about this or any other SDC related topics, please contact us by visiting our website which is listed here on the bottom of this page or reaching out to me or anyone else on the SDC team. 00:45:23.000 --> 00:45:29.000 Thank you. And of course, many thanks to the organizers of this workshop for the fantastic opportunity to give this talk. 00:45:29.000 --> 00:45:34.000 I'm looking forward to your questions and discussion. 00:45:34.000 --> 00:45:46.000 Awesome. Thanks, Kathy. That was great. So now we're going to try Mike's talk again And see how this goes. So let's try to take it away. 00:45:46.000 --> 00:45:50.000 Good afternoon, everybody. Thanks very much for the opportunity to talk. 00:45:50.000 --> 00:46:01.000 Before starting this, I'd like to just acknowledge support from USGS over the years and in collaboration with Gareth Funning at the University of Riverside. 00:46:01.000 --> 00:46:09.000 I'm going to be talking about three different aspects of the GNSS contribution, mostly about active faulting, particularly our survey work. 00:46:09.000 --> 00:46:19.000 That has increased coverage over the North San Francisco Bay Area, specifically 2022 and 2023 surveys and the updates that we've made to our velocity solution there. 00:46:19.000 --> 00:46:25.000 The influence of the geothermal field in the geysers, looking at decadal velocity changes potentially there. 00:46:25.000 --> 00:46:44.000 And then a little bit about hydrology and the way that geodetic observations may actually help Because it's obvious that the geothermal and the hydrological deformation must be influencing the active faulting, the stress changes as we gain more and measurements of these things. 00:46:44.000 --> 00:47:01.000 So to start, we'll look at the usurf3 a rupture forecast model where we have geologic constraints of the North Coast San Andreas of anywhere between 15, 16 millimeters per year and about 25 5, 24, 25 millimeters per year. The Rogers Creek fault 00:47:01.000 --> 00:47:08.000 6 to 10, the Green Valley a few millimeters per year. Those geologic measurements are made at these green dots here. 00:47:08.000 --> 00:47:13.000 On the right you see the geodetic GNSS solution from that report. 00:47:13.000 --> 00:47:18.000 The G Odyssey wasn't really used to constrain the deformation model at that time. 00:47:18.000 --> 00:47:24.000 But you can see here what I'm going to be talking about in the North Bay area, the relative sparsity. 00:47:24.000 --> 00:47:34.000 Of GNSS velocity vectors relative to the Bay Area itself and Southern California at large as well. Most of the studies have been focused on the Bay Area. 00:47:34.000 --> 00:47:40.000 Or there and we're going to be focusing in a little bit on these population centers where the faults go through. 00:47:40.000 --> 00:47:45.000 Marin, Sonoma, Napa counties. And where they also creep. 00:47:45.000 --> 00:47:57.000 A little bit more of an updated model is the National Seismic Hazard Model of 2023, Although in this case, again, with the green dots here, the geologic estimates that are used to constrain the model haven't changed very much. 00:47:57.000 --> 00:48:03.000 The Rogers Creek is still the same. The Green Valley fault has a slightly wider range of uncertainties than the San Andreas's. 00:48:03.000 --> 00:48:18.000 Is basically the same as it was. Similarly, on the right hand side we see the GNSS compilation of zeng 2022 and this is on a slightly different velocity vector scale than the previous plot, so it's a little bit more difficult to discern. 00:48:18.000 --> 00:48:29.000 But basically the idea still is that we have a concentration in Southern California in the Bay Area, but not so much in the North Bay area in Northern California in general. 00:48:29.000 --> 00:48:42.000 So that's what we're going to try and address. If we look at the continuous networks, they are the red dots and they are the notar sites or PBO as it was previously known. 00:48:42.000 --> 00:48:56.000 Barred sites are in blue. And USGS sites are in yellow. Here is a zoom in of the density in the Bay Area itself and then you can see the relative density at the same scale throughout the rest of Northern California. 00:48:56.000 --> 00:49:02.000 With the continuous network only, we have an inter-site median distance of about 16 kilometers. 00:49:02.000 --> 00:49:20.000 And the white dots show the network of survey benchmarks that we've actually incorporated by remeasuring or accessing historical data and reprocessing it to build more dense network of GNSS velocities. And at that point we get to about seven kilometers 00:49:20.000 --> 00:49:30.000 Of median inter-site distance, and with faults here being about 20 to 30 kilometres apart or two to three locking depths, we feel as if we need that additional density to really add a geodetic constraint. 00:49:30.000 --> 00:49:34.000 To the geologic constraints on slip rates that we already have. 00:49:34.000 --> 00:49:49.000 And start to incorporate geodesy into these seismic hazard models. Our results to date are on the right here, which show that increased density The red vectors are the continuous sites. You'll see the same color scheme throughout some other plots in the talk. 00:49:49.000 --> 00:49:57.000 The dark blue vectors are the survey sites that have velocities that are below one millimeter per year. 00:49:57.000 --> 00:50:18.000 In uncertainty and the light blue have velocities where the uncertainty is greater than one millimeter per year. We tend to use this arbitrary number of one millimeter per year precision for an indication of whether the velocity itself is tectonically useful when it comes to constraining things like slip rate. 00:50:18.000 --> 00:50:36.000 So here on the right on the left is the GNSS solution in a pacific plate reference frame and if we invert along a couple of profiles with elastic dislocations for the faults we see in this Bodega Bay Santa Rosa profile a 00:50:36.000 --> 00:50:44.000 The San Andreas Fault is estimated to have a slip rate of about 17 millimeters per year and the Rogers Creek Vault about 14 millimetres per year. 00:50:44.000 --> 00:50:54.000 In italics, in the square brackets on the right you see the comparable You serve three national seismic hazard Model 23 geologic slip rates. 00:50:54.000 --> 00:51:00.000 And we can see that there is also this big trade-off between the San Andreas Vault and the Rogers Creek Fault. 00:51:00.000 --> 00:51:09.000 Slip rate negative 0.9 correlation. So as one goes down the other goes up obviously in this inversion scheme. 00:51:09.000 --> 00:51:17.000 Looking now at the southern profile, profile B through Point Res and Nappa We see similar numbers. 00:51:17.000 --> 00:51:21.000 About 17, 18 millimeters per year for the San Andreas Fault. 00:51:21.000 --> 00:51:37.000 The Rogers Creek Fault is about 10, a little bit less than it was further north. And then we have a West Napa fault and a Green Valley fault both around five millimetres per year to the south of of Lake Berryessa again in the italics and square brackets you see those geological rates that are used 00:51:37.000 --> 00:51:51.000 In the National Seismic Hazard Model 23. In this case, we see that there is a couple of millimeters per year left relative to the Pacific at the very end of the profile, the tip of Point Rez. 00:51:51.000 --> 00:52:03.000 And that indicates that any offshore faults relative to the Pacific plate are very likely to have relatively low slip rates, about two to three millimeters per year. 00:52:03.000 --> 00:52:10.000 Now if we do the inversion by using a Monte Carlo simulation, we can see the comparable results. 00:52:10.000 --> 00:52:16.000 And here we are looking at the point raise profile again, profile A. 00:52:16.000 --> 00:52:22.000 And in this case, we get about 19 millimeters per year for the San Andreas Fault. 00:52:22.000 --> 00:52:25.000 About 10 to 10 and a half for the Rogers Creek Fault. 00:52:25.000 --> 00:52:41.000 And about nine, nine and a half millimeters per year for the Green Valley fault although in this case We do not model the intervening West Nappa fault and so It may well be that the Green Valley Fault has a distributed slip or 00:52:41.000 --> 00:52:59.000 Incorporated onto itself here in this inversion. This raised an interesting question for us about what the border faults to the east of our area are incorporating in terms of relative slip between the Central Valley and those inner coastal ranges. 00:52:59.000 --> 00:53:16.000 So we decided in 2022, September, to do a survey which incorporated measurements on top of survey sites that we know had been previously measured about 10 years ago by the Department of Water Resources. 00:53:16.000 --> 00:53:23.000 And we came up with a set of velocities just by these two observations. 00:53:23.000 --> 00:53:29.000 Which show a little bit of slip rate in the eastern area to the east of the Green Valley Fault. 00:53:29.000 --> 00:53:41.000 Along this border area. Of one or two millimeters per year and certainly with a sense of contraction between the Central Valley and the coast ranges as well rather than just strike slip motion. 00:53:41.000 --> 00:53:51.000 So this is very interesting and it may well be that we have a little bit of velocity left over not only on the Pacific plate side but also on the Central Valley side that we need to take into account. 00:53:51.000 --> 00:54:03.000 When doing our seismic hazard analyses using geolitic constraints. In March 2023, we also tried to improve our solution in a couple of different areas. 00:54:03.000 --> 00:54:07.000 One up towards Point Arena, between Point Arena and Clear Lake. 00:54:07.000 --> 00:54:14.000 Along the very northernmost tip of the coastal san andreas And the mark of fault in land. 00:54:14.000 --> 00:54:20.000 And also along our sort of Russian River, Bodega Bay, Santa Rosa profile area. 00:54:20.000 --> 00:54:35.000 And you can see here the difference that that March 2023 survey made. We now have much many more dark blue velocities, which means that they have those uncertainties less than a millimeter per year and are now much more useful for doing 00:54:35.000 --> 00:55:00.000 Inversions of fault slip rates using geodetic constraints. Let's focus in now on the geysers to see what influence that area of rapid deformation might have on there's a fault slip rate inversions are going to be like with the geodetic constraints. We have measurements from levelling in the 1970s, which along this profile in yellow 00:55:00.000 --> 00:55:04.000 On the right here, sorry, this is a scan from an old report. 00:55:04.000 --> 00:55:13.000 But you can see showed very rapid subsidence between a few years in the 1970s, that's 1975 and then 1977. 00:55:13.000 --> 00:55:20.000 And slightly less subsidence in the central southeastern area, but still a number of centimetres per year. 00:55:20.000 --> 00:55:24.000 Is the kind of rates that we're talking about here in the geysers. 00:55:24.000 --> 00:55:32.000 We reprocessed some data from the mid-1990s taken by Tony Mossup and Paul Siegel from Stanford. 00:55:32.000 --> 00:55:42.000 And produce this geodetic velocity solution in the horizontal relative to the surrounding area of the geysers you can see here. 00:55:42.000 --> 00:55:50.000 Our P203 site which is a notice site, has almost zero velocity relative to this surrounding reference frame. 00:55:50.000 --> 00:56:01.000 And everything else shows essentially an inward horizontal contraction towards the center of the geysers. 00:56:01.000 --> 00:56:22.000 In a vertical sense, we decided to try and do two solutions one about 2000 when there were wastewater injection projects initiated to maintain the steam reservoir in the geysers and one after the 2000 sort of time period. So on the left here we see as well 00:56:22.000 --> 00:56:33.000 Insolar from Vasco and others 2013 projected into a vertical and the red vectors show the vertical of the GNSS velocities. 00:56:33.000 --> 00:56:41.000 Again, a number of centimetres per year here, that's 50 millimeters per year that you're seeing. That's several centimeters per year. 00:56:41.000 --> 00:57:05.000 A little bit less after 2000, which we can see in this comparative plot here most of the dots here fall above the one-to-one ratio line Which means that their velocity, their vertical subsidence velocities were higher In the 1990s than they are in the early 2000s. Everything has shifted 00:57:05.000 --> 00:57:11.000 A little bit towards the lower end of the velocity space. 00:57:11.000 --> 00:57:28.000 We can also see other changes. So a little bit later in the 2000s we have this site P203 which is a continuous site that allows us to see changes in the velocity after the end of those surveys, so from sort of 2010 onwards 00:57:28.000 --> 00:57:41.000 Here you can see the northwestward motion relative to North America and some minor subsidence. But if we now detrend before this green line, which is the start of an enhanced geothermal demonstration project. 00:57:41.000 --> 00:57:57.000 We'll just detrend using that first period and see what happens afterwards. We can see this big shift in not only horizontal motion but also vertical motion as well. We've gone from a little bit of minor substance to a change where there's a little bit more uplift 00:57:57.000 --> 00:58:06.000 And an increased westward motion. So we've gone from motion towards and down to motion away and up. 00:58:06.000 --> 00:58:12.000 Relative to the geysers. And that happened in about 2011. 00:58:12.000 --> 00:58:23.000 So we also decided to install three continuous sites of our own within the geysers itself to try and track this motion a little bit better tg01 in the west, TG02 in the northeast. 00:58:23.000 --> 00:58:32.000 And TG03 in the southeast. And here are some of the results from that up to the present day, the end of 2024. 00:58:32.000 --> 00:58:47.000 You can see that there is this initial uplift that we saw in the P203 time series as well, up to around about 2014 when we then start to see the site subside again. 00:58:47.000 --> 00:58:52.000 So we've got a second change. In the vertical rate. 00:58:52.000 --> 00:59:06.000 Of the geysers area. We also see this in TG02, although it was set up a little bit later so that uplift segment is a little bit shorter but you can see very clearly the subsidence thereafter. And TG03 was set up 00:59:06.000 --> 00:59:15.000 In 2014 itself. So actually you don't see any signal that is uplift you only see the subsequent subsidence. 00:59:15.000 --> 00:59:23.000 We also record earthquakes. You can see very clearly the magnitude 5, 2016 December earthquake west of Cobb. 00:59:23.000 --> 00:59:33.000 And so we think that this pilot project and potentially in the geysers can see three potential velocity changes that have been observed geodetically. 00:59:33.000 --> 00:59:39.000 Rate of subsidence slowed around the turn of the millennium, likely due to the start of wastewater injection projects. 00:59:39.000 --> 00:59:45.000 And there is some possible reversal in the horizontal contraction and the subsidence, especially in the northern half. 00:59:45.000 --> 00:59:51.000 Since mid-2011. And then a second reversal going back down in early 2014 onwards. 00:59:51.000 --> 01:00:02.000 This sort of stress pattern might be affecting the way that we look at decadal changes in seismic hazard on the faults that surround the area. 01:00:02.000 --> 01:00:10.000 It's also possible that we may need to consider the effect of hydrological signals, which can also have quite a large geophysical signature. 01:00:10.000 --> 01:00:17.000 In this case, this slide is taken from Stacey LaRochelle's work 2024 SCC Annual Meeting Talk. 01:00:17.000 --> 01:00:30.000 Where she has used insol projected into the vertical along with vertical time series from continuous GNSS sites, which are fairly sparse in this area. 01:00:30.000 --> 01:00:35.000 To look at poroelastic effects in the northern half of the Central Valley. 01:00:35.000 --> 01:00:56.000 And we've considered that potentially densifying the network in these kind of circumstances as well may actually help decompose into three dimensions rigorously the insar that may help provide on a very dense spatial scale those kind of constraints that that can help with seismic hazard analysis when it comes to hydrological effects. 01:00:56.000 --> 01:01:02.000 One example would be the Department of Water Resources network that was set up over the last decade or 15 years. 01:01:02.000 --> 01:01:19.000 In that northern part. This is from a report. All of the survey work and we could for example re reassess and reprocess this data to provide a much denser constraint on the NSAR in an area like this. 01:01:19.000 --> 01:01:25.000 So just to summarize, densification of GNSS networks with historic and ongoing survey measurements we think is critical. 01:01:25.000 --> 01:01:33.000 To constraining the deformation and the earthquake hazard models, particularly in region like Northern California, where there's a relative sparsity of geodesicy. 01:01:33.000 --> 01:01:47.000 Particularly as we try to build geodesy into our seismic hazard models. There is evidence for decadal scale changes in deformation around the geysers and in other areas like the Central Valley to do with poor elasticity and water resources. 01:01:47.000 --> 01:01:58.000 So we think the usefulness of these observations across decadal time scales and geographic extents is important, not just for those processes in terms of studying the geo. 01:01:58.000 --> 01:02:14.000 Physical, geothermal and hydrological processes, but also that influence on the seismic hazard analyses themselves. 01:02:14.000 --> 01:02:18.000 All right. Thanks, Mike. That was great. I'm glad we got the talk to work. 01:02:18.000 --> 01:02:32.000 So thanks, John, for making that happen. But now we get to have a great discussion And there's been a ton of activity in the chat, which is great to see. Thank you guys so much for asking awesome questions and to the speakers for answering them very promptly. 01:02:32.000 --> 01:02:37.000 I think Chelsea has a question to sort of start us off for the discussion with our speakers. 01:02:37.000 --> 01:02:38.000 Chelsea, do you want to ask that? 01:02:38.000 --> 01:02:43.000 Yes, so this is a question to all four of our speakers. 01:02:43.000 --> 01:02:54.000 So I'd like you to imagine for a minute that you are able to collect one data set to answer a question that you posed in your presentation. 01:02:54.000 --> 01:02:59.000 We don't need to have the technology to collect this data set. So you could need a time machine. 01:02:59.000 --> 01:03:05.000 Or possibly small error. Or sensitivity to a part of the crust that's hard to measure. 01:03:05.000 --> 01:03:13.000 So the question I'd like you to answer is what question that you discussed previously would you want to answer? 01:03:13.000 --> 01:03:23.000 And what data set would you collect? 01:03:23.000 --> 01:03:28.000 Yes. 01:03:28.000 --> 01:03:29.000 When… 01:03:29.000 --> 01:03:40.000 Can we go back in time? Because then I would just collect the same data But back in time because I feel like data continuity is super important. And if we can just like have the same type of acquisition like type of data through time for a long time, then we can start to see changes. And it gets tricky when things are 01:03:40.000 --> 01:03:54.000 Like measured slightly differently and they can't be like exactly compared to see these temporal changes. So I would just deploy the allos to satellite back from the year 2000 all the way through exactly the same the whole time. 01:03:54.000 --> 01:03:55.000 Okay. 01:03:55.000 --> 01:03:57.000 You mean nicer? 01:03:57.000 --> 01:03:59.000 Oh, yeah, of course. I actually mean nicer. Sorry about that. My bad. 01:03:59.000 --> 01:04:00.000 Yes. 01:04:00.000 --> 01:04:06.000 I'll be. 01:04:06.000 --> 01:04:13.000 I'm trying to pay attention to the chat at the same time because there's still a few questions coming in on my talk. 01:04:13.000 --> 01:04:21.000 But I think one of the questions that I was answering there actually was regarding the constraint that we have on the west side of the San Andreas Fault. 01:04:21.000 --> 01:04:42.000 Where it runs along the north coast there and obviously one thing that I think any geodicist would love is seafloor geodesy, underwater geodesy uh you know we gnss in sa You know, even strain meters, things like that, they're all essentially glam based 01:04:42.000 --> 01:04:47.000 Techniques or require Skyview. 01:04:47.000 --> 01:04:59.000 For satellites so that that would definitely help in a lot of areas just to provide an additional constraint where you know it's not just the San Andreas fault this happens a lot of the time we have these very asymmetric 01:04:59.000 --> 01:05:05.000 Geodetic data sets that we're trying to make sure inference with with 01:05:05.000 --> 01:05:27.000 I would follow up on that. Laura Wallace had a great talk at AGU and she was presenting. They have like one or two I don't exactly know about it, but the GPS sites that are like on the a granky subduction zone offshore and just having one or two data points has changed her interpretation of what is locked and what is not locked just from having these two points but the 01:05:27.000 --> 01:05:39.000 The work in terms of like being able to get three or four GPS data points over time offshore is huge. 01:05:39.000 --> 01:05:50.000 I'll chime in with something different. I wish we had slightly better quality and preservation of old aerial film scans. 01:05:50.000 --> 01:05:52.000 Hmm. 01:05:52.000 --> 01:06:00.000 Those are a ton of fun to use. And I think we could constrain some events up in Northern California a little bit better. 01:06:00.000 --> 01:06:09.000 We also might having older data, maybe I get a little bit into Danny's request of longer time series of NSAR. 01:06:09.000 --> 01:06:18.000 The data wouldn't be obviously with quite that time sensitivity But you could go back in time before SAR came around. 01:06:18.000 --> 01:06:34.000 Let's have LiDAR three dip like 1967. That'd be pretty cool. 01:06:34.000 --> 01:06:37.000 Adia, do you want to add anything? 01:06:37.000 --> 01:06:47.000 I think I agree with all these things. I think a lot of things come to mind and you know, you know. 01:06:47.000 --> 01:06:54.000 Something that we've always talked about was, you know, everywhere all the time. Wouldn't that be cool? 01:06:54.000 --> 01:07:12.000 How you could just have if you could just have you know insert close to really close to the fault let's see if each of the faults How it would be cool to have you know faster to enter things faster when you have like in Ridgecrest when you had two earthquakes that you couldn't separate how you could 01:07:12.000 --> 01:07:30.000 You know, but I, to be honest i think I think the longer time series is really the thing that we um that we need because that we need because The earthquake cycle is long and our emissions are short. 01:07:30.000 --> 01:07:46.000 So having longer time series. Would really would really be great. And this is why I didn't intend this answer to be another plug for a nicer follow-on, but it is, you know, we have to, that's why we have to make sure 01:07:46.000 --> 01:07:50.000 That we keep them going. 01:07:50.000 --> 01:08:09.000 I guess in the same way there's also like old inside data sets that aren't very easily accessible. So I've sent out a few emails in the last few weeks being like, hey, do you guys have this data set or that data set? And in the same way that maybe aerial photos aren't cataloged in the same kind of place, having older data sets in a central location um 01:08:09.000 --> 01:08:27.000 Which either already processed or because the codes to process ordinarily also that knowledge is being lost as like i don't know how to process ARS and I don't know what those codes are so like making sure that these historic archives are available. 01:08:27.000 --> 01:08:28.000 I just got some SIRC data. I have no idea what to do with it. 01:08:28.000 --> 01:08:33.000 Great. Thank you. 01:08:33.000 --> 01:08:54.000 But yes, I agree that we shouldn't lose our messes and things it's so easy to like i processed ERS and then we saw data in grad school But I would have no idea how to do it now. It's completely different. 01:08:54.000 --> 01:08:56.000 I see Scott, you have your hand up? You have a question? 01:08:56.000 --> 01:09:03.000 Yeah, great. Cassie, I had a question. I obviously have interest since I'm one of the developers of the community fault model. 01:09:03.000 --> 01:09:15.000 The modeled fault geometry with the step over on the San Andreas that you modeled, how does that differ from the The SCEC model and the NSHM representation. 01:09:15.000 --> 01:09:27.000 So it's just one single fault plane going through that region. Because on the scale of the whole San Andreas, central San Andreas, it looks like pretty much straight as an arrow. You just can't see it. So I think at that larger scale. 01:09:27.000 --> 01:09:28.000 Yeah. 01:09:28.000 --> 01:09:39.000 That step over is kind of ignored. But since we were just looking at like that one kilometer scale, it becomes pretty critical when you're looking at the high resolution displacement data sets. 01:09:39.000 --> 01:09:46.000 Okay, that's sort of what I thought. How big was that step over again? I can't remember from the figure. 01:09:46.000 --> 01:09:47.000 Okay. 01:09:47.000 --> 01:09:55.000 It was only about one kilometer across and it was like five kilometers sort of where that gradual zone of deformation occurs. 01:09:55.000 --> 01:10:03.000 Well, definitely share the mesh with the CFM group because we can um see if we can always make improvements to the CFM. 01:10:03.000 --> 01:10:07.000 Yeah, I need improvements to my mesh too. I think that became apparent. 01:10:07.000 --> 01:10:16.000 It's fixing meshes is so boring. So if I can save you any headaches on that, let me know. 01:10:16.000 --> 01:10:22.000 All right. Thanks. Great talk. 01:10:22.000 --> 01:10:28.000 If you have a question, please raise your hand. 01:10:28.000 --> 01:10:44.000 Yeah, in the meantime, we had Sarah actually beat me to this question in the chat We wanted to ask you all sort of, we live in this this uh community of Insarge, Odyssey, lots of different data sets coming in, lots of different ways to process those data sets. 01:10:44.000 --> 01:10:50.000 And we wanted to ask you all sort of what extra resource would be the most useful to you all. And this again came up in the chat a little bit. 01:10:50.000 --> 01:10:56.000 Would new measurements be helpful? Would it be best to have easier to access data like we just chatted about? 01:10:56.000 --> 01:11:07.000 Or more software or more training for software, what would be sort of like the one thing you would want to have like right now to be able to process your data better and better find your signals better. 01:11:07.000 --> 01:11:23.000 So I want to answer this first, and I think this came up a lot in the answers to that question, but I want to emphasize that we you know nicer nicer is coming, but there are other insert satellites that are coming. 01:11:23.000 --> 01:11:31.000 Nysar is going to be a global data set that's going to be bringing in something like 100 terabytes of data per day. 01:11:31.000 --> 01:11:47.000 This is unprecedented. We've never had this before. And so to deal with that data we really need to start thinking about how to not do this on our laptops anymore. 01:11:47.000 --> 01:11:58.000 Honestly, that's what we, the most, the number one thing that we all need is computing resources that are somewhere on the cloud because hard drives are not going to be big enough. 01:11:58.000 --> 01:12:21.000 I think I'm going to add to that and be like, to be able to manage all this big data, we also need to learn how to utilize machine learning because like we can't be manually mapping landslides for global data sets or all of California data sets and like i think there's such an opportunity to learn about how things work when we actually have these big data 01:12:21.000 --> 01:12:38.000 To map identify and help us pick out features in these big data sets means means we'll find results that are globally applicable because at the moment, like a lot of insight studies are focused where the scientists are or where the money is. So there's like big spots of the world 01:12:38.000 --> 01:12:51.000 That there was inside data for, but we haven't done big landside studies about, or we don't know as much about whether there's creeping faults there. And they probably do exist but like If we have to manually go and find them all, then we're limiting ourselves. 01:12:51.000 --> 01:12:57.000 Yeah, just to follow up on that a little bit from my perspective from the sort of GNSS side of things. 01:12:57.000 --> 01:13:21.000 I think one of the big things that I would like is time or additional people who can go and search old data sets essentially from a time before we were so organized with data sharing and big data centers and things like that, because a lot of these surveys, the GNSS surveys 01:13:21.000 --> 01:13:30.000 A lot of them are archived. Pis were pretty good back in the day at making that data available but The more I search. 01:13:30.000 --> 01:13:44.000 For things like the Department of Water Resources data set and things like that. The more you realize that people have been making these measurements in a number of different contexts for a long, long time. And they are just either not available 01:13:44.000 --> 01:14:04.000 Or not easy to find or they've been lost to time or whatever And when we're starting to talk about these decadal changes in geodesy, we need that long time span. And those older observations, the survey GNSS as well, are critically important. And we could really fill in 01:14:04.000 --> 01:14:17.000 A lot, I feel. If we could just find that data, have it archived properly, and then have some people who are dedicated to actually processing it all in a rigorous, self-consistent fashion. 01:14:17.000 --> 01:14:31.000 That's just not sexy science though it's kind of, you know, and that's a really big problem because It's such a shame to me to think about data being lost. 01:14:31.000 --> 01:14:46.000 Despite the fact that data gathering and data production is not necessarily what scientific programs are going to fund But it's really the foundation, I think, for everything that we do on top of it. 01:14:46.000 --> 01:14:52.000 Makes me sad. 01:14:52.000 --> 01:15:02.000 My only chime in would be like kind of reiterate, I'm just going to piggyback off of both of those main points, the temporal aspect, the long span. 01:15:02.000 --> 01:15:19.000 As well as the ability to deal with the big data sets because I have an over constrained fault model Right. With all of those surface data points. And I just can't let my laptop do it. So it's just a matter of bigger computing resources. 01:15:19.000 --> 01:15:26.000 That's why the fault model still looks so bad because I had to figure out how to use 50,000 by 50,000 Greens functions. 01:15:26.000 --> 01:15:29.000 Matrix. 01:15:29.000 --> 01:15:49.000 I'm on the HPC system, but I'm like one of the biggest users on the scratch drive because I can't figure out where to move stuff afterwards. So I've got like 100 terabytes of data that's just like sitting there and I don't, I'm like, how much of this needs to be archived with the paper? Like, can I put this on Zenodo like what happens now 01:15:49.000 --> 01:15:54.000 Where do I put it? And I can't delete it because I can't make it again because it's months of work. 01:15:54.000 --> 01:15:55.000 I don't know what to do. 01:15:55.000 --> 01:16:07.000 Especially with Insar, where do we put these data sets how do we archive them for people who might want to access them? We've processed, we spend years of our lives processing this data And it's here and we'd love people to use it, but how the heck do we get them to access it? You know, yeah, no, 100%. 01:16:07.000 --> 01:16:20.000 And then also like how much data do they want? Do they want all of my different corrections? Do they want all of my different things or do you just want the velocity field at the end or the time series in the middle? 01:16:20.000 --> 01:16:35.000 And like, I can't publish the raw data with it because that's like like each date for the LLOS 2 is like 64 gigabytes per date and there's like hundreds of them. So it's like just too big, which is why NYSAR is going to be better because they're just going to give us the pre-made products 01:16:35.000 --> 01:16:40.000 And the stuff that I've been battling with is going to become obsolete. 01:16:40.000 --> 01:16:43.000 That's what I was going to bring this up, actually, that, you know. 01:16:43.000 --> 01:16:52.000 Insar in particular is a little bit newer than GNSS and a little bit more difficult in some ways. Radar is difficult. 01:16:52.000 --> 01:17:09.000 And one of the things that we really need to start thinking about for the future of this is is agreeing, finding ways to agree on best practices so that we can make higher level products so that instead of You know, instead of 01:17:09.000 --> 01:17:15.000 Slcs, which already sound No. 01:17:15.000 --> 01:17:24.000 Difficult we can we can start providing things like like time series and everybody will be that that everybody would be happy with that we that we know are reliable We're nowhere near nowhere near that, not even close because we there's still 01:17:24.000 --> 01:17:29.000 I. 01:17:29.000 --> 01:17:43.000 A lot of things to work out in that sense but we have to this something we have to start moving towards so that You know, so that it's easier for us, of course, ourselves, that we don't have to you know. 01:17:43.000 --> 01:17:46.000 We don't each have to write up our own SBAS code, sorry. 01:17:46.000 --> 01:17:57.000 Sorry for not answering our people, but also easier for people who are non-experts That was one of nicer program applications leads. 01:17:57.000 --> 01:18:06.000 Basically, one of the things I had to do is find agencies and people who could benefit from SAR data but had no idea what it is. 01:18:06.000 --> 01:18:24.000 And uh it did you know thinking about that made me realize how many great uses there is for geodetic data sets, not just INSAR, but geodetic data sets and people have no idea no idea what to do with them. So higher level products is 01:18:24.000 --> 01:18:30.000 It's a good suggestion too. 01:18:30.000 --> 01:18:44.000 I'm going to also be like, now that Nysol is coming and our band is going to be way more available, it really opens up what we can do in Northern California because like geodesic specifically, like NZAR has been really, really limited. You could see in my sentinel stuff how de-correlated it was 01:18:44.000 --> 01:18:52.000 The L band is going to do so much better. So if you have got a question about a landslide or about a volcano or about something that you think is subsiding. 01:18:52.000 --> 01:19:00.000 Then like, I think we can ask way more of those questions in Northern California in the coming year So. 01:19:00.000 --> 01:19:06.000 Definitely. 01:19:06.000 --> 01:19:14.000 Let's see. If we have other questions, if there's any other questions that anyone has or is thinking about, please don't hesitate to raise your hand. 01:19:14.000 --> 01:19:18.000 There's a question we have about sort of getting into uncertainties a little bit. 01:19:18.000 --> 01:19:40.000 So we were thinking about what are the largest impacts to the uncertainty in genetically constrained fault slip rates and fault slip models and what is the best way to decrease those uncertainties 01:19:40.000 --> 01:19:42.000 Sorry, can you repeat that? There were questions in the chat 01:19:42.000 --> 01:19:51.000 Of course. Oh, I know. Yeah. What are the largest impacts to the uncertainty in geodetically constrained fault slip rates? 01:19:51.000 --> 01:19:59.000 Really what causes the most uncertainty in trying to get at fault slip rates and fault slip models And what's the best way to try to mitigate those? 01:19:59.000 --> 01:20:00.000 How do we do better? All right. 01:20:00.000 --> 01:20:23.000 I think I have an answer for you. Which is that each of which is our different techniques like measure different spatial scale right so the gps is like wider and then my incel cells are much bigger than Casey's lidar and so like being able to use the ensemble of measurements to understand how it changes over the dimension of the fault is going to be more useful for helping constrain any one particular 01:20:23.000 --> 01:20:33.000 Data set. And that kind of is seen by the alignment array and the insar as well so like each of our data sets have their own problems. But when we start to like collect and ensemble then we can like 01:20:33.000 --> 01:20:50.000 Met that heart changes with distance and what the actual width of the fault zone is and I think it comes back to Christie's question as well of like how do you define these things like is just a single line on a map which is kind of what I'm inferring with my Enzai which is not correct and like how do we 01:20:50.000 --> 01:20:58.000 So when we collaborate together, we can have a better picture despite the uncertainties in our data, I guess. 01:20:58.000 --> 01:21:11.000 Yeah, Scott Marshall just put a response in the chat as well just saying fault geometry is probably the biggest source of error And second is the fact that there's a number of different ways that you can invert for slip rate from the geodesic. 01:21:11.000 --> 01:21:29.000 As well as the full geometry there's also this inescapable problem which Dave Shelley and Jess Murray also mentioned in the chat earlier that I was responding to is we fundamentally have this issue in Northern California and Southern California as well for that matter 01:21:29.000 --> 01:21:48.000 Where, as I showed in my presentation, you have faults that are two to three locking depths apart if we just take a simplistic view of elastic dislocation model type thing. I know that geologists will argue with this, but from a geodetic perspective, when we're doing this large scale inversion 01:21:48.000 --> 01:22:00.000 So it does become very difficult. To separate out the slip rate on individual faults when they fundamentally have deformation fields that overlap with one another. 01:22:00.000 --> 01:22:08.000 More data doesn't necessarily help with that because you still have an overlapping deformation field. 01:22:08.000 --> 01:22:24.000 So as well as the fault geometry, sort of the exact position of your you know elastic dislocation there is also the fact that you just have elastic dislocations that are close together here. So in that sense, you know, I kind of agree with 01:22:24.000 --> 01:22:42.000 Well, Danny, I think, was saying there that a combination of methods, the geodesy covering different spatial and temporal wavelengths, but also those trenchings any other methods that we can use that maybe pinpoint on the fault surface itself a little bit more what might be going on 01:22:42.000 --> 01:22:58.000 In the long term that we can even if we average over several seismic cycles combination of these is I think it's the only way we can do it in many ways. 01:22:58.000 --> 01:22:59.000 Cassie or Katia, any thoughts? Oh, Danny, go for it now 01:22:59.000 --> 01:23:06.000 I'm going to go back to just data being acquired regularly because I also have uncertainty my data because I've got big data gaps. 01:23:06.000 --> 01:23:13.000 So when NYSI gets here and it's every 14 days and it's always every 14 days then that will reduce the uncertainty. 01:23:13.000 --> 01:23:14.000 12, 12. 01:23:14.000 --> 01:23:15.000 12. I'm so sorry. 01:23:15.000 --> 01:23:21.000 Is your uncertainty in the like the you know or the the seasonal rate or the average? Yes. 01:23:21.000 --> 01:23:38.000 For my answer? When it's like… So in the LS2 data set, there's like a data gap of like a year. So I'm forming interferograms that are spanning like multi-year. So my network has like one to the next four and then 01:23:38.000 --> 01:23:43.000 Year, two year, three year, four year. So I have to like create lots and lots of pairs across this big data gap. 01:23:43.000 --> 01:23:52.000 And the longer time frame you have, the coherence drops off. And so then you add in more errors from unwrapping errors and all kinds of other problems that you have. 01:23:52.000 --> 01:24:07.000 So… Yeah, it's from like decorrelation and from probably like cycle skipping. So you're missing like maybe a whole whole thing but the way I get about it is just by doing lots and lots and lots of pairs. So I just have like 01:24:07.000 --> 01:24:24.000 I think I have like 92 dates, which is a lot less than you would have in the same kind of sentinel time series that like they're connected by 600, 700 kind of interferograms so like and each of those interferograms goes from like monterey to the Oregon border. 01:24:24.000 --> 01:24:29.000 Huge. So… It's a, yeah. 01:24:29.000 --> 01:24:36.000 Frequency of data acquisition is helpful for reducing. Uncertainty and errors. 01:24:36.000 --> 01:24:41.000 Can I ask you guys a question? Is that okay? 01:24:41.000 --> 01:24:53.000 I wasn't planning on this, but I thought of this. So Paul Rosen asked me, I think this was yesterday, so I'm going to pass on his question to you. And I think this is a question that's interesting for us on the mission side of things, right? 01:24:53.000 --> 01:25:01.000 We're talking about looking at regularly acquired data. Yes, that's important. We're talking about long time series that's important. 01:25:01.000 --> 01:25:08.000 We're talking about better data sets, maybe three data permission. Okay, all these things are important. 01:25:08.000 --> 01:25:16.000 To solve your problems, to solve your scientific problems How long? 01:25:16.000 --> 01:25:35.000 Of a time series would you need? Ideally. How long? Is it 100 years? Is it 300 years? Is it 3000 years? Is it… I mean, how many faults, how many creep events do we look at to figure out how creep works? 01:25:35.000 --> 01:25:52.000 How many earthquakes do we look at to figure out something, at least something about how many how do we figure out how much data do we need to figure out all geometry. And I gave him a kind of a lame answer. So I hope you guys come up with better ones. Go ahead. 01:25:52.000 --> 01:26:02.000 Oh, no, I only have a lame answer to that one. I think we want it all because we're all looking at different processes and scales and deformation rates. 01:26:02.000 --> 01:26:03.000 Yes. But how much? 01:26:03.000 --> 01:26:24.000 I would add to that that like I would add like the more data we have, the more we learn about how these processes work. So like these pulse-like creep events we may not have seen or known about before, but now we have more data so we know them. And so if we have a longer data set, then we will learn something else about maybe how the fault evolves for time. And then we'll be like, oh, well, now we need like at least seven seismic cycles. 01:26:24.000 --> 01:26:28.000 So every time we get more data, we understand more about how these processes work. So how long is a piece of string? 01:26:28.000 --> 01:26:34.000 Do you think there's a do you think there's a way to quantify that, at least attempt to quantify How many events do you need to see? 01:26:34.000 --> 01:26:36.000 I think Christy does. 01:26:36.000 --> 01:26:43.000 In order to have a better understanding of something, whether it's false geometry, whether it's creep, whether whatever it is. 01:26:43.000 --> 01:26:50.000 How many? Is there you know? Not to figure it out right now, but is there a way to do that for us to answer that question? 01:26:50.000 --> 01:26:53.000 Is there a comment on this, Christy? 01:26:53.000 --> 01:26:57.000 It's a follow-on, so… You can just go ahead with that and I'll try to… Come late. 01:26:57.000 --> 01:27:01.000 You don't want to answer this question for everybody? 01:27:01.000 --> 01:27:14.000 No, I mean, my question was about like how useful is looking at slip slip deficit if we don't have a kind of a closed system or a total… number of structures to look at, structures and things, structures and volumes or whatever. 01:27:14.000 --> 01:27:21.000 Yeah, so related to that, I guess, because we're talking about closed system and space as well as time. 01:27:21.000 --> 01:27:27.000 Right. Like we have different sort of time cycles of completion there? 01:27:27.000 --> 01:27:30.000 But… Yeah. 01:27:30.000 --> 01:27:31.000 Yeah, it's a… Go on. Yeah, yeah. 01:27:31.000 --> 01:27:36.000 I'm not a, oh, go ahead mike 01:27:36.000 --> 01:27:51.000 Oh, I just had a thought about, I'm not a statistician, but is there a statistically relevant number that if we had X number of earthquakes, we could generalize something about like we had 10,000 earthquakes all the same magnitude. Here's what happened. Here's the you know 01:27:51.000 --> 01:27:58.000 Gaussian distribution of what has happened with this particular earthquake, would that answer our question or would we always continue to need more data after that? 01:27:58.000 --> 01:28:02.000 How many samples do we need? To get at this question. 01:28:02.000 --> 01:28:15.000 Yeah, at least not to figure out everything, but at least to answer something, some question, one question about earthquakes how Yes, that's exactly what we're getting at. That would be an interesting study to do, I think. 01:28:15.000 --> 01:28:16.000 This is… Go ahead. Sorry, go ahead, Cassie. 01:28:16.000 --> 01:28:35.000 And part of that, oh, sorry, I was… I was just going to say that part of that also wraps back to the uncertainty question. What's the uncertainty in our data and also the other processes that are going on in the earth that are going to mask the signal in our data? 01:28:35.000 --> 01:28:42.000 I've been looking at the Calaveras and thinking about like climate change and deforestation over the last few hundred years. 01:28:42.000 --> 01:28:47.000 A few thousand years and how much it affects the geomorphic expression in the landscape. 01:28:47.000 --> 01:28:53.000 Of creep, of earthquakes, this kind of thing about how, you know, how wide is a fault zone, like how many visible fault strands can you count? 01:28:53.000 --> 01:28:57.000 And then can you study all of them? And if you can. 01:28:57.000 --> 01:29:04.000 Do they add up to something that's predictive? Like if you were able to describe everything that had happened in the last 5,000 years. 01:29:04.000 --> 01:29:16.000 Would that be informative about the next earthquake? And looking at Chelsea, she's actually been exploring this like in as much detail as you can with the best possible best case scenarios, right? 01:29:16.000 --> 01:29:35.000 And yeah, with all this handing off between faults, as we were discussing in the chat, there are probably some areas where there's a finite number of faults that we need to look at. And there are other areas where we just need a really different representation of deformation in order to describe what's happening right now. 01:29:35.000 --> 01:29:47.000 Like, yeah, I think we need to start examining that and not just try to scale up like high precision local ideas to like bigger systems and maybe the bay area is a great workshop for doing that. 01:29:47.000 --> 01:29:55.000 Yeah. More of a comment than a question, but I don't know what that new kind of representation might look like, especially not for geodicy. 01:29:55.000 --> 01:30:10.000 But I think that's something that we can kind of grapple with together as we get more focus here and we get more observations hopefully to work with. 01:30:10.000 --> 01:30:11.000 Great. End the meeting? Are we at the end? 01:30:11.000 --> 01:30:18.000 I was going to say there's more questions in the chat, but we are at our time Chelsea, do you want to lead us out here? 01:30:18.000 --> 01:30:26.000 Thank you, everyone. And we really appreciate the presentations And then the discussion. 01:30:26.000 --> 01:30:33.000 Certainly going to go back and look through the chat. So thank you all for those questions in the chat. 01:30:33.000 --> 01:30:40.000 And I think the next and final part, I believe, starts in half an hour. 01:30:40.000 --> 01:30:43.000 So come back. 15. 15. 01:30:43.000 --> 01:30:44.000 15 minutes, 15 minutes. 01:30:44.000 --> 01:30:48.000 You can turn it to Sarah. Yeah. Thank you, speakers. Oh, 15 minutes yeah 01:30:48.000 --> 01:30:52.000 That was an amazing session. That was really wonderful. Thank you. 01:30:52.000 --> 01:30:56.000 Thank you, Sarah. Thank you, everyone. 01:30:56.000 --> 01:31:04.000 And as you were just saying, there is now a 15 minute break until we come back for our last session of the workshop at 3 p.m. Pacific. 01:31:04.000 --> 01:31:17.000 So everybody gets a 15 minute break unless you are speaking in the next session. So if you are Kathryn, Gordy, Luis or Jia or our moderators, please stick around.