 So Roland is going to wrap this up for us and pull it together and I think his comment to me was that he's more than happy to sort of have this become a discussion as he goes through the slides that he has ready to go so people should feel free to to engage at any point and then we'll obviously continue the discussion once Roland's got to the end of these slides. So, right, Roland. So, thank you for having me and thanks for sticking around. So, the task was synthesis and concluding remarks and I did add discussion to that here so as I go along, please do chime in. I was supposed to talk for about half the time allotted to this to this last session. So, I picked two quotes that actually come from the agenda from the introductory comments. A is societal implication of understanding the range of precursory signals is large and many questions remain that was quite apparent today and then there's that big question are there certain characteristics of precursors that make them more or less likely to assault in a large earthquake or eruption we covered both topics and that obviously I think at the end of this meeting here is still an unresolved question but that's what we should talk about. Now only Emily discussed briefly volcanic systems and clearly our discussions have also focused mostly in earthquake. I did include one slide on that topic eruption precursors and as Emily mentioned due to the kind of physical processes that we understand belong to the process leading up to eruptions magma transport conduit formation and associated seismicity deformation degassing makes that a much more tractable system and in many places where seismic monitoring exists seismic unrest has been observed before eruptions so Iceland is an example where the monitoring is high quality it's the kind of volcanic system that lends itself to that there they quite routinely are able to pick up precursory signals and also issue warnings and alerts that sometimes are as clear as we think there's going to be eruption in this many hours or days. The other systems volcanic systems that seem much less easily understood in that context open conduit volcanoes where there's less deformation of seismicity being produced by magmatic transport making its way to the surface. Most illicit volcanoes are difficult because they have very viscous magmas and tend to be erupting in a more sudden sequence of events and restless caldera's keep moving up and down for decades without end and it's unclear when we should be more more concerned and Emily talked about that some more. So that's it for volcanoes and eruptions unless you have thoughts questions comments Richard. My understanding of the Iceland case is that the places that they do a pretty good job is where there's a they have a volcano that they've seen erupt before they basically have a sense of where the magma chamber is and so there's some sort of combination of knowing where the magma chamber is and having seen a previous eruption and literally timing from when there's a swarm of activity at depth and then the time it takes for that magma to get up. So is that right or is it the most time in these persoatic systems? If they're low so for many of the volcanoes there is a track record and they have eruptions every accident. Simi you might know more about those. Three times is this like we captured the eruption in the Galapagos we knew what had happened the previous three times and when out it happened almost within about a month when we expected so we're a lot better when we have multiple events in the basaltic systems the the others are harder. That's a really good point right I mean here I'm only talking about just erupting but you're right the volcano hazard is manifold you have volcanoes you have earthquakes that relate to the disruption of airspace and all of those require their own type of monitoring system and response so so that's a very good point that it's much more complex when it comes to building and slowly the message is very similar to the earthquake one the key is to have a well monitored and more in-sensor monitored system so that you capture the deformation because sometimes the formation signals have been helped much this incident and vice versa and sometimes all you see is de-gassing with very little positive physical signals. I think you do need a comprehensive monitoring of as many volcanoes as possible in some cases because we have many many volcanoes. Remote sensing techniques, thermal deformation and so imagery, seeing plumes can also play a really important role because we can build networks on every volcano that we believe to be active. Alright moving on so there we go so these are kind of my talking points when it comes to earthquakes and I'll sort of more slowly go through this so we talked a lot about the kind of new observational tools, techniques, measurement systems on the seafloor on land that come into play and with the help of those we have seen many precursors. There's no doubt that earthquakes are preceded by foreshocks sometimes by measurable deformation but those recursive processes seem to involve a wide range of processes over a huge spatial range of spatial and temporal scales and involving maybe sometimes preparation processes that every earthquake thinks it needs to go through to become a seismic event but also simple triggering relationships that are independent that means there's no predictive power in detecting those precursors and there are many observations of seismicity or deformation events that fail to precede anything but they're not they don't know to be precursors meaning we can't just see them and say our that's going to be in those days and so therefore I do think as Emily with a big red note implied the the prospects for prediction are pretty poor so far and so instead we need to focus on probabilities and that obviously was the main theme in this last session and I think there's a lot of motivation for more research based on public scientific curiosity better understanding how faults work and the whole process but also slightly understanding them in the context of hazard so we had a lot of talks that did talk about various techniques some of them have been allowed now for some a number of decades and she honestly we have satellite remote sensing more recently we've learned that even satellite gravity measurements might be able to capture what could be precursory dynamic signal deep in the earth we've talked about seismological technologies and methods including the optical fibers but also Emily talked a fair bit about new measurement what we do with our seismic data screens using finding low frequency earthquakes repeating earthquakes tremors earthquake swarms using template matching to get an order of magnitude larger numbers of earthquakes with the same data that we we have before doing time-dependent ambient noise imaging of systems using machine learning to try to see if there's information in there that we don't recognize by just thinking about them the way we do so far and then clearly a big theme of this meeting was the need for and the steps we're taking towards c-claw monitoring both seismic and geodetic and and I like the quote from Spar no sensor is adequate for all signals so this clearly will require a combination of different methods pressure centers to be as the crispy whole tilt meters ideally interrogated in a real-time sense using cables or other methods so thoughts recommendations discussion of just observational techniques that really have been able to a lot of what we might call it all right so yep we're thinking ahead where should we be putting our money in terms of the future of in types of instrumentation I mean is it seafloor you know is it iroptics online like what what's the recommendation if you think about it that way both scientifically and societally right subjections on the biggest breaks to do stupid as fault-related has a tsunami shaking and so the lack of knowledge from closer to the earthquake source and the source of possible precursors or what we know to be because this makes that I think the biggest frontier it's very clear that you know if we can take the data we have already in hand right so much more valuable by using some of these template matching etc machine learning techniques that's kind of easy and less expensive so in that sense that's the bang for the buck is high but I think that the frontier is under the oceans the oceans covers 70% of the earth and most of the plate boundaries both transform spreading and just to amplify that I don't think you said it you would talk but also that's where all of these interesting new observations are being made as well so it's not just that we have the technology capability to go and do this now we're actually starting to see I mean that's what the whole beginning of this morning right subjections are also of the ocean transform thought tend to be culturally coupled and make the earthquakes but they also have more slow-slipped tremor etc phenomena there are more fluids involved so that makes them in that sense more interesting along those lines of funding the need for more continuous observations so just you know not just running to the time of the scene but before during and after as well as the portfolio the range of observations that have to be done especially as you noted in volcanic systems simultaneously I think we'll make important more collaboration between agencies in the future these are not really campaigns like we're used to funding I suppose from individual agents so it's just interesting thought in terms of the future of how all this equipment will be purchased deployed and maintained and and remain in that place for very long time I think you know the most synergies there are right away so it's from optical fibers to utilization of these offshore networks it's combined USGS NSF NASA like every every agency that contributes to this committee is involved in all of this I think in this room that we should also tackle is the fact that it's not necessarily easy to get like long-term funding for long-term instrumentation in remote areas so so probably in Cascadia in the US or in Japan or in New Zealand are very good observation instruments but in lots of places if there are data they are not necessarily available and they are not open which is something that probably collectively we should work on and also trying to find fielding schemes that allow installing networks that's why I think in the US it has to be more than NSF right we're learning that now with PPO for example yeah I bet I would my comment would be this is what Alan and I have been speaking about earlier is you just said in the US that we're talking about these international collaborations that are required the constant and also the reward to all of us is facilitating international collaborations for multi-disciplinary research just seems so really important the more you can demonstrate multi-disciplinary and multi-institutional interests more likely there might be a chance to get that I just want to make one comment that I actually have to be rude and you know Emily this morning laid out these two sort of in member hypotheses for four shocks that we've been talking about for decades and you know that the general mood observations on land for precursors has not been terribly optimistic pessimistic and we have these sort of intriguing observations from Japan and Chile the question is there's just something fundamentally different about subtraction zones and I think I don't know the answer I don't think anybody knows the answer to that but we also have a situation now where I think we've gotten to a state where people see earthquakes that seem to evolve in space-time in a way that people point to them and say slow slip event without a deformation signal and I'm not convinced that that's the only way you can get the space-time evolution of earthquakes that there could be cascades of just triggering of events that happen to evolve in a space-time pattern just the same way that I'm not convinced with every time the space-time pattern seems to fit square root of time that it means it's fluid diffusion you know maybe it does but not always which is pretty confident that so if we could answer one question you know in the next decade or so or two questions that would be is there something fundamentally different about subduction zones and are these probably earthquake sequences really indicating a seismic slip we could do that I think that would push us a lot farther yeah I totally agree I mean it's really clear I'm afraid from the examples for either the cascade model where people look very carefully at patterns of foreshadowing in space and time continental earthquake, lander sector mine, El Mayo, Cucabal, Cormomodo, Ismael there there's relatively little indication of pre-slip or slow slip progression that leads up to the earthquake whereas the examples where we see that to hopefully kick it and with the others with two dead exit masks included even the friction is there so certainly in terms of it being more likely that seems to be the case 1966 prop field that creeped the day before so that's that's one of the reasons why so many periodic instruments went as well and that turned out wrong in 2004 so yeah that's a good point I won't see you fall. Okay so did you understand anything? Yeah so in any case right with all these observational capabilities we have an increasing number of interesting precursors big-scale gravity, moderate sized flip episodes from geodesy, repeating earthquakes, sea flow, geodesy pressure sensors etc indicating pre-slip before to Hoku, the long-term acceleration that Paul described to us, foreshocks it seems like more and more it seems like most earthquakes have foreshocks and they may just be exactly as H explained just what we should expect based on how we understand interaction of seismic events so one has to look at these in the context of are these foreshock sequences more special to make them have any information about the size of the impending earthquakes but with this new catalogue that Zach Rust produced they now see that almost all 75 percent of magnitude-form larger earthquakes have foreshocks so that's really pretty exciting and we see a lot of indications of slow slip above and beyond foreshocks from geodesy where we have it but also seismic indicators of slow slip that we think of pretty diagnostic, tremolo-frequency earthquakes, repeating earthquakes, some of swarm-like seismic incidents. So there are precursors that's exciting. The challenge is that the processes involved in these precursors seismic events versus slow slip may be also fluid flow, may be also even geodynamic accelerated strain rates in the mantle as would be implied by this observation. Span a huge range in time foreshocks can happen seconds before an earthquake, definition, accelerations can take decades and in space meter-scale preslip versus hundreds of kilometers scale slip episodes like we saw in New Zealand recently and so that suggests there's there's not a very simple way of thinking about the precursor as being due to a single process that all earthquakes have to go through. Before I move on thoughts on that would you agree with that assessment that there's no coherent theme something we would say well we've seen it but for enough earthquakes I think every earthquake has this particular process. We saw for each of us parameters they often come out very close to critical and be correct for work in aftershocks so foreshocks the ones you just don't know back because they're too far away or before your catalog or too small. I guess you explained right foreshocks tell us something about earthquake probability but they don't tell us something about we're gonna have a big earthquake. Yeah so even though it could be a hundred percent in that number that doesn't make the prediction. I just want to so people have a bunch of people online I just want to make sure they're able to let us know if they want to make a comment or question. They can raise their hand and I can unmute them. Okay so for people who are online if you raise your hand electronically that's how you can get back to them. Sorry. So with related processes I quickly showed that slide right so the end-member models what we're really hoping to find is a process that could be pre-slip, it could be precursory dilatancy or combination thereof where we have a physical process that grows, evolves into an earthquake. This is a slide from Brett McClaskey from a presentation he just gave at SSA and he was making the distinction because he makes earthquakes now on his three- million machine in Cornell and he's able to create these as kind of a critical scale of slip patch below which falls like to slip slowly and then as they grow beyond that they grow. So if that L sub C has a natural meaning that the scale of that would determine the size of the pre-slip and how it relates to the subsequent earthquake. And so when we see slow slip before big earthquakes that's exciting because it could be that nucleation phase that relates to this recent process. On the other hand for many events where people have looked very carefully at the foreshort sequences and how they've often faced in time they conclude that those evolve as a cascade of events trickling each other without any coherent slip process driving them, making them essentially just interacting presumably forming interest like statistical characteristics. Now this I added sort of this trigger model. We know we have slow slip of phase sizes, durations and shapes and sometimes they're followed by a slip and then we call them slip or precursors. And so in this case the size location timing of that precursor is only related to the eventual earthquake in the sense of possibly having triggered it. A bigger precursor doesn't make a bigger earthquake, the duration of it wouldn't necessarily come into play. You would have to model it, the physics, the statistical characteristics to understand. And I think many of the precursors we've seen that really don't seem to be something right around the hypocenters growing into the next earthquakes such as that that's at least as common as a nucleation phase that we're trying to put our fingers on. So the on-shear is a slow slip so it's kind of similar to the in that last slide but its size and location is not necessarily related to it and there are also other foreshocks doing so it's a cascading sequence of seismic and as seismic events not a nucleation phase. Just a comment. So you were mentioning that probably subduction zones are quite different from intracranational platforms and I definitely agree with this. And when you look at precursory deformation or seismicity before subduction earthquake what is really clear is that the area that is activated before the earthquake is large. Right, so if you think for example right the foreshock sequence really kind of wraps up. I was actually thinking about it a little bit. Yeah and so when you were showing the premise slide from the prism tree they would be like extremely large, involved in a very large area and then concentrate. And this is true also with the hooku depending on where you think that this long term slow slip is associated with the preparation but also when you when you look at the foreshock sequence it is not a very large area including so in the case of the hooku right that that pre-slip did occur near the hypersandricks. I worked on the Karatsky replica of the Gamehunter Chetka. It was a big slow loop, you know, a hundred magnitude 5-5 phase. But it was clearly at least 50 kilometers away from Chetka. So I think that was a starting triggering relationship probably but it was not a pre-slip. So pre-cursor candidates those are the events seismic events, slow slip events that could become a pre-cursor. But I think the message we've learned they just as often don't. Right, so the pre-tohoku pre-cursor that was captured both theoretically by pressure sensors and from seismicity. That was a very similar than the 6.8 equivalent marble equivalent made that wasn't followed by the tohoku. And there may have been more of those we don't know. We know from Nankai, from Cascadia, pretty much whatever we look now, we see slow slip events, short-term slow slip events, little tremor sequences, long-term multi-year, multi-month, multi-day slow slip events that happen all the time. And pretty much all the great majority, 90% of those are not followed by earthquakes. So they're pre-cursor candidates that failed to proceed anything. Which means the ability to say, oh, we need to, you know, there's going to be a big earthquake in the sense of prediction that seems difficult. Because then we need to be able to distinguish the kind of pre-cursor candidates from the real pre-cursors, even though we say so many that are indistinguishable really within our current capabilities. So here I use good news because prospects for prediction are poor. And prediction really in sort of the rigorous sense of you calling the place, time, magnitude, and a very narrow window of ranges. And given the diversity of processes, scales, etc., we really haven't seen anything yet that would make us feel like that's just around the corner. We just need to invest another few billion dollars in civology. Obviously, we can promise people that. So that's probably not the direction we want to go. But I'd be interested in hearing more of the mystery of the use of that going once. So I have a question. So we've been discussing all these pre-cursors, say SMIC and Geodetic. Are all the other pre-cursors of other times completely now swept away? And when you say others, do you mean fluid, slow? Any other physical observations? So I deliberately did not get into attempts to talk about pre-cursors from ionospheric disturbances, some other electromagnetic fields. Rain on emissions, that could be, I wait and see. A lot of them don't have physical foundations. So I tried to steer away from that a little bit, but we couldn't talk about that. I mean, people have launched satellites, which are new satellites, just recently, the Demeter satellite from your other friends, that were explicitly meant to use ionospheric variations as a pre-cursory process you can study and I don't think there's anything to that yet. Paulin? On those lines, the HECI pre-cursor is one of these, since I TEC pre-cursor, it's very HECI, H-E-K-I. Yes, yes. It's very interesting. They claim to see it before all magnitude 8 earthquakes. It happens fairly often, so it would be a lot of cause and effect. So he has, by now, written about six, seven papers and I let them off. So we're actually working on a paper trying to, you know, there are already several papers reporting what he's done. I think, even though he says he found ways to get around that, he's still looking at the data in a way that includes the cosides, the fitting. So he finds roughly one hour long pre-cursor from the analysis of ionospheric delay changes or electron content changes in the atmosphere that he offers of the cursor. You're not meant to depend on, I still don't believe it, but that's just my opinion. Is there a question for somebody online? Yes, this is Emily Brodsky, back here in Santa Cruz. So, and I apologize if my question is off base because I missed a chunk of the day. But going back slightly, Roland, you said, is anybody a little bit more optimistic? Yeah, yeah, I know you left me an opening. And I think, you know, when I first started doing volcanoes, what the old volcano observatory people would say is when the deformation of the seismicity go together, that's when you get worried. And I think we're kind of in the same place on the earthquake problem that, you know, now it's a tiny bit of evidence. I don't want to get overexcited about it. But, you know, the type of observation that's before Kike or Tohoku, where you have a very large scale slip with a very large scale foreshock sequence, as Anne was pointing out, it's enormous. I don't think we've seen that with the slip of company get not followed by a large earthquake. Now it may simply be we haven't seen it because we haven't had the instruments to see the slip, you know, so, you know, caution, certainly. But I don't, is there a negative result there yet? So I mentioned, I don't know if you heard that before, Tohoku, there was a similar sized slow slip episode in 2008. Oh, well, yes. It could play the precious senses. That's true. That is certainly true. That's, of course, how they knew about the one in 2011. It's really maligned on that 2008. But did it have the seismicity with it in the same way? Not the same way, right? OK, so you get the slow slip. It's when they kind of come together that we seem to be seeing something more interesting. Yeah. So we'll have to watch more examples to see. It's a fair point. I totally agree with you that the more, you know, observations we make of something in modelism. And even when we talk about ETSs in Cascadia, people have developed models and have suggested that things will change when either the recurrence interval or the size of the ETS suddenly explodes in a way that we've never seen before. And that's a fair thing to then take as an indication now things are going in a different direction than they have in the last several hundred times since 1700. No, I ask a question, too. Yeah, Christmas. That's not. Go ahead, Matt. I could probably be asleep, but I guess my question or comment is really just following on what Emily was saying. And also something that Laura said, and I mean, I understand the skepticism about using any kind of prediction term, but I guess my question is really, we know enough now from these examples in Japan and Chile to really to get a little nervous and to have governmental agent responding and asking questions about probabilities. And I guess my question is, you know, there's still a lot of research to be done. There's a lot of false positives and so forth. But what do we need to know to be able to give better answers to society about what the probabilities really might be based on what we've learned from these recent events? With, of course, I don't know if anyone's mentioned the word alakala with that, you know, sort of, Paul of responsibility put upon scientists that is sometimes not correct, but or is sort of unrealistic expectations, but it's the world that we live in. Yeah, that's exactly right. And that came up in the discussion with Laura and also Morgan that when it comes to how we, you know, put information out that takes into consideration the occurrence of precursor candidates in terms of probabilities. And I'll get to that in a couple of slides, right? Laura said we need some kind of OEF that includes the consideration of these precursor candidates in a way that we can use. And, you know, alakala is really what got a lot of this started, right? So I'm only saying that short-term prediction in the sense of saying there is going to be an earthquake of this size in this place that I don't think we will get to easily and wouldn't want to when it comes to the societal aspect. You know, I definitely agree. I mean, I guess I think we all agree on that. And I think there is question of is what is the new space that's been opened up after Tehoku that have we fully prepared with society the answers that they're demanding? Or what more do we need to do to be able to make them? I think it would be very timely, and maybe people have done that, to revisit Tehoku. Everything that was observed and done and think about what could we have, should we have done or what should have been done, obviously there, but in that context, right? As to what kind of communications could have been issued that would incorporate everything we know now. I think that would be very useful. No, no, no. I just wanted to build off Emily's comment. She was talking about how the 2011 precursor to the Tehoku event had more seismicity than the 2008 social event. And I just wanted to put this point out that, and this came up in large talk too, that parts of New Zealand and like Hoso in Japan and probably other locations, there are social events that are known to be associated with increased seismicity rates that do not end trigger precursor rates. So that's not by itself. Right, so maybe the way to put it, the precursor candidates, little guys, and precursor candidates, big guys with seismicity, we've never seen them before. Yeah, so like if we know that an area produces seismic increases in social events, regularly, then it's less scary to see them. All right, so precursor candidates, including foreshocks, even the ones that just behave the way it does for an action model, predicts do increase probabilities. And so an earthquake forecast, right, tries to take that into consideration. Earthquake probabilities change, and Morgan showed that very nicely how as you keep track of seismicity rate in a given area, such as Bombay Beach, your probability of large events also will be extremely time dependent and very often by orders of magnitude. And so supervents and these other precursor candidates also come into play. Currently, OEF, OEF, operational aftershock forecasting, only consider seismicity, right? So OEF is only initiated when, and that's the currently active system when there's an earthquake, above a certain magnitude, then you do the calculations using the different kind of models and approaches that Morgan described. And you calculate what is the time dependent seismicity rate and probability as a function of time. But we're not doing that yet with slow, split events. So as Morgan pointed out, and this is from email she sent me before, changing seismicity, have predictive information about earthquake rate and predictive power about the chance of a large earthquake. However, foreshocks are not predictive of earthquake science. So they change the probability, but they don't allow us to predict, they don't allow us to tell that it's going to be an earthquake of a certain size. And then Laura, this is that Crowdy put in from her presentation, we need to develop OEF models that incorporate slow supervents and not just make ad hoc decisions. And they learned that lesson in real time in New Zealand when they had this immense slow, split events, slow, split events near a critically stressed section of the subduction zone. And they were told by their government, we need something more from you. We need numbers. And you remember the whole way they tried to do that. And I think it's actually in a very admirable fashion. So an operational earthquake forecasting approach that would incorporate these kind of precursor candidates is clearly challenging. It's challenging enough to do this with seismicity, even the most sophisticated use of ETAs kind of OEF models. They're not that easy to do them if you throw in static stress changes, physical ideas about how slow, split relates to the occurrence of earthquakes makes that even more difficult. But I think the experience of New Zealand suggests it's something that it should be considered, especially in subduction zones, where we know these slow, split events cover a big part of the split behavior of the system. It's not that we have mostly just earthquakes and slow, split events are an exotic animal. We shouldn't worry about too much. But they're really a big part of how these systems seem to work in many of the world's subduction zones. OK, so Richard also asked me to mention NIPEC. So I'm currently chairing NIPEC. So NIPEC is a USGS committee. It's a council just to advise the director when it comes to A, people's attempts to predict earthquakes and providing some input for the director with regards to what those mean, but also talking about design. So this committee here discussing precursors, suggests that's an important topic, and that is one of the things we've been tasked to do. And so the kind of discussions and reports we've issued have been about operational earthquake forecasting, providing advice on how that might be best implemented in a way that builds that up in a useful way, addressing some of the societal concerns. The whole question of, well, if people do predictions, how should they be done in a way that allows for quantitative analysis that were common some years ago about CSAP, the Collaboratory Study of Earthquake Predictability, which essentially is a place where people can send their predictions in a form and format that is evaluatable that can be quantitatively assessed. And that's obviously an important aspect. So I just included that to follow Richard's advice due to mention that. But there's probably, you know, there would be interesting discussion between whatever comes out of this and what NUPEC does. And so this is my talking point list again. So I'll open it up for other comments, questions, discussions that you guys might have. Thanks. We are running a little over time, but I think we can take a few more minutes if people have other questions, comments, or things that they want to put on the table as takeaways from today's discussions. Yeah, Mike. We're only just a general question to you. You're not optimistic about prediction in that strict sense, but are you optimistic about the prospects for some of this knowledge, making it into sort of public decision making and risk mitigation and so forth? Given our current state of knowledge, can we do you think we can move in that direction without it? Yeah, I have sort of strict minds. Part of me says we need to do it. We need to offer it based on the demand, like what Laura described, based on experiences like in Italy, how that truly gets translated in a meaningful way into decisions by governments. And mitigating of risk exposure, that's a much more difficult problem. But I think to do our best of expressing what we know about the change in probabilities to the best of expert knowledge, that's part of our responsibility, it seems. But I'm much less certain about how useful that is in terms of reducing losses, et cetera. That's beyond my insight. Roland, can I just ask, how do you imagine that happening? In terms of how it would be part of an OEF system? Yeah, I mean, how do you imagine that just sort of continuation of what things like Morgan or Laura were already talking about? I mean, I guess the question is, does there need to be some research program or even a study or something to help? And with that question, how do you move this forward? Right, so they currently write different flavors of OEF, more or less physical. Some are purely statistical, very simple. You know, Riesenberg-Jones kind of aftershock statistics, some take into consideration for geometries, stress changes. And I think to implement slope events, you need to have a physical component. I think there has to be a stress calculation included in the process, stress calculation adding to stressing on hazardous faults and taking them from there. But it's globally not easy, right? Because if you want to do that in real time, that gets pretty challenging. It might involve utilizing earthquake simulators, right, that try to incorporate the basic first order physics and interaction, the triggering aspects of the system. But I think that's a reasonable way to approach it. I have to take a comment. I think in some cases, I think the city hall example was a good one. It wasn't the probability that spurred the decision. It was the fact that the scientists said so. I think I'll make that assertion. I don't know if it's a sensible one. But I think we do have the danger of wrapping ourselves around the axle of feeling that we can't calculate an accurate enough probability when 80% or some large percent of the committee is just us agreeing as a community that something is important enough to say something and agreeing how to express that in the absence of having a number. And I think that we need to be maybe a little more brave or organized about how we provide information and that's a very broad set of categories. I think some sections are right with those examples of cases in which we can't assign a hard number that we'll agree on. But we think that something's up there. And if you take, again, the New Zealand example, if you have half a megapascal of stress increase on a critically stressed large rupture of disparity and you say nothing, that seems cause for more lawsuits. Yeah, fine. I thought it was this thing with Italy, right? They basically said, oh, don't worry, right? Have a class of one. Some did, right? Yeah, so they did not express that when you have racist, messy rates, your probabilities will always go up. They said the opposite, which is unscientific. So I guess I'd like to make a comment. I mean, we were beginning to be pulled a little bit into a doldrums. We're not going to be able to predict earthquakes. So what are we doing? I should make a comment. Mike just started to kind of pull us out of it. Thank you, Mike. But I mean, I think that to me, as I think of myself as not being somebody who really studies earthquake processes and seeing these observations that we've heard a lot about, particularly in the first half of this morning, about processes that are obviously clearly part of the earthquake cycle, these slow slip events, these sequences of seismicity that are occurring on critically stressed patches of faults, it seems to me that there are some very exciting new observations that are telling us a lot more about that earthquake cycle. And it seems to me that there was a real opportunity for us as a community right now to point to that and say, look, we really can learn a lot about earthquakes. We are not telling you that we're going to be predicting earthquakes, but we can understand a lot more about this process, about this cycle, and this will probably help us to reduce the overall hazards associated by understanding long-term forecasting, all of the things that we do already, but just doing it that much better. I 100% agree with you. Right, I'm not saying this is not important, exciting and meaningful. I'm only saying that these precursor processes have enough non-uniqueness that the prospect of short-term prediction of time-place magnitude, that's the least likely outcome of all the great things that we can learn studying precursors. What we can do better, right, is to get to Laura's factors, right? He has a factor two from statistics, and has a factor 10. But that's my prediction, that's on the structure. Right, yeah. We can develop a more better instrument. Right? Yes. Yeah. So if I came across this sort of thing, I know some of the things I was going to hold. But this is important just for other people. I mean, one of the roles of this committee is to talk about what are the needs, and, you know, there's this question, you know, that there really are new opportunities here based on some of the observations. If we can get out there and observe them, I think that's a critical thing. And understand them in a physical model, right? Absolutely. And observe and open access, and real-time, whatever. But that's not the case for large portions of the world. And I think it's a really important one. That's a really good point, right? On the international scene, that is a school of big issues in many parts of the world. So call for open data should be part of... Well, I think in practicing New Zealand, I can tell you the station's spacing is very large and reliant on temporary arrays in many places as well. Just... It's a really broad problem. Chef. So... Could I say something? Yeah, go ahead, Emily. So the framing of this workshop included volcanoes, and maybe that's too broad an umbrella. Maybe we really should be just focusing our attention on all these new earthquake observations. But I do think there is something to be learned from the cross-fertilization that in a lot of ways these communities are moving in the opposite direction. That earthquake precursors and prediction has been a subject that people have been hesitant to even talk about, let alone say they're doing, and people will not make an earthquake prediction with actual, out-of-actual numbers, as Mike was just pointing out. Whereas volcanic predictions are done routinely based on fairly qualitative data with fairly qualitative statements about what's going to happen next. And they're trying to working hard to tighten it up, but it seems like that has certainly been societally useful. It's hard to argue that those have not been useful predictions. So maybe we've got something to learn there. Yeah, I agree. I mean, as you described, volcanic systems are different than important rays, but they're also very promising and challenging in different ways. And I like the idea of keeping those two topics together and informing each other. So obviously, I like your cartoon, and Paul was pushing us a little bit to think about that as a problem to go after. And I think one of the really interesting places this conversation ends up is sort of the quality versus quantity trade-off of what you do. And so in writing the SC4D report, Emily and I had a little devil's advocate thing where Emily took the widespread low-density approach and I took the dense but only one or two places approach in putting together boxes. And I don't necessarily know the answer to that trade-off if you can make good arguments either way, but I think the thing that really could use a lot of thought right now is as this momentum is building for significant investment in convection zones is what we're trying to pull up in the deep in the sense of, is it one really well-captured example that really lets us get the physics captured? So at least we know what it is that we're talking about. And in places where we can recognize that we're more than effective to bump into probability or is it in this statistics of getting, my feeling is that we have a lot of these examples where we have these data sets that we're pretty much into something, but not good enough to really nail it down off-sharp. And I think that trade-off between getting good number of statistics with those kind of data sets versus something like that active ejection experiment where you know everything. That's the conversation to try and steer among the seduction zone crowd is this momentum builds towards doing something. I mean, I'm a big fan of the natural laboratory approach. And there's also the question like if the problem is scale and variant and looking in great detail in one place where we think we can do so optimally would be really, really important. That doesn't mean we should not also follow Emily's suggestion, right? She showed the whole earth and saying, we can go to all these places and we should and try to capture things because they might be relatively rare, for example. But I do think that a natural laboratory component that's really important and takes advantage of the scale and variant so that we can maybe in one place see a dozen nucleation phases if they exist in a more general way could tell us a lot more than if we failed to scratch the surface, everyone. You're proposing the Parkfields now. Yeah, some people are saying we should go back to Parkfields. The next one is just around the corner. No, that's exactly right. And it was a good lesson, right? That even though people put in strain meters because we wanted to see a nucleation phase that we put them in a place where the previous two are really 5% or 34, 66 were. And then the earthquake started on the other side where there's little slow slip and we didn't see anything. So that's a good lesson. That's a good point. One more good fight. Yeah. All right, I think that's a good place for us to put. Thanks to start. Thank you, Roland, for that great introduction.