 This is Think Tech Hawaii. Community Matters here. It's one o'clock on a Monday afternoon, so you must be watching Think Tech Hawaii. I'm your host, Pete McGinnis-Mark. Now, every week we bring you scientific results from the UH Manoa campus, and it's either in space science or it's in earth science, so it gives me great pleasure today to introduce our guest, who is a returning guest. Estelle Bonney is going to be telling us a little bit about both earth science and space science today. So Estelle, welcome back to Think Tech Hawaii research in Manoa. I'm really excited to hear what you have to say today, because I understand that primarily we're going to be talking about a new scientific paper, which you've just had published, and this relates to... I'm trying to predict the end of the love flow from the interaction. And our backdrop, for example, shows a moving RR flow, which is these thick hot lava flows, which we have in Hawaii, and your satellite observations are trying to predict how not that lava flow, but other ones in general. So tell me a little bit about what the background is for your paper. All right, so it's basically the idea that a lot of people actually care about when an eruption starts, right? And that's really important, and we do care about it, but for long-lasting lava flows, it also is important to know when it's going to end, especially for some places where you have a city close by, and it can reach the city or not. People might have to evacuate or not. So knowing when an eruption would end, it's actually also a pretty big importance for how they're managed. So a hazard manager on the Big Island, the Hawaii Volcano Observatory, might be particularly interested in certain eruptions actually on the Big Island, but in your satellite study, presumably, you can look anywhere around the world, is that correct? Yeah, exactly. We use a certain satellite called MODIS, which is a moderate-resolution imaging spectrogeometer that's on board two different spacecraft, and it can go over the years and can image the same place four times a day. Four times a day. At what level of detail? The satellite I use has like a one-square-meter pixel, so it's really, really large, but it's still usable and useful for the repeaters. So you cannot see a person in this image? Probably not. But volcanic eruptions, presumably, are white enough, or how do you detect the eruption? We use thermal remote sensing, so the heat of the lava flow is detected from speed. Just to say these night vision goggles, which people can wear to see a person because of the temperature that they are emitting. Yeah, exactly. It's the same kind of principle, and do you make the observations during just the daytime or? We also have night time. You also have night time, so which is better? Night time, because then you don't have any solar reflection. So the sun reflecting off the surface of the line might conspire? Yeah, you need to think about into account. All right, let's take a look at the first slide. Even though we've got a nice backdrop here, let's take a look at our first slide, and I believe this is an eruption killer wear or this could be monolow. I'm not too sure which one, but this is the kind of thing which presumably would be detectable from your spacecraft. Yeah, so we look at lava flows. This is a nice fissure, and you can see those lava fountains that then get on the ground, and it creates those black line that then flow downhill, and that's what we are interested in. And it's not just the orange fire fountain, as it's called, part of the eruption which you're observing. It's the black bit as well, which is also hot. This black bit is still really hot, so compared to the surrounding, the temperature difference is high enough to be detected from this. Any idea of what sort of temperatures these flows are? So the red part is between like up to 1200 degrees Celsius, and the black part, the crust, might be, it definitely changes a lot, but between 100 degree and 500 degree Celsius. And of course, we look to the left, what we're seeing are fully grown here trees, so this must be quite a big eruption, so it's identifiable from space. Yeah, for sure. All right, and so far over the last few decades, we've only had eruptions of killer wear, but of course we also worry, and HVO worries a little bit about monoloa activity, and so your models might actually be applicable for that scale of eruptions. Yes, so full eruption is actually not really working for the model because it's been going on for the past 34 years, but monoloa is an episodic style of eruption. So having eruptions the last one was in 1984, and it lasted about 20 days, and that's exactly the case of what we could use the model for. All right, and our second slide I think actually shows the flowpants from the March 84 monolo eruption. Monoloa is down on the bottom left, A, correct? Yes. And then all those colors, the orange and the red, their individual lava flows, which seemed almost to get to the outskirts of downtown here. Yeah, it stopped almost, I think, about four miles away. So how would your model actually work then? You're making observations of these short-lived eruptions. Do you need to get there quickly, or what's the challenge in using satellites? So the challenge is to get near real-time data. Explain what that is. So basically, as the eruption is going on, we will get the data right away. There's no lack of time between when the acquisition is made and the time where we actually can use it. So right now, the data that I use is about 12 to 18 hours behind. So there's quite a large gap, especially if it's, let's say, a couple of days long eruption. So that's a challenge. But as we get those data, we can get it almost every six hours, which is way better than going on the field every six hours. Particularly in hazardous conditions, presumably. Yeah, and the measurement that we make is actually pretty hard to get on the field. There's a lot of assumption, and being in a lot of places is hard. But from space, you have this really wide view, so you can see the whole thing. So typically, you can get hold of the data of a newer eruption within 12 to 18 hours. Is that after the satellite has made the observation? You say it comes by every six hours? Yeah. Well, so about every six hours, because we have two satellites, and they come at different times. And they do one image per day and one image per night. So you have to look forward. So it's not every six hours exactly. And you get the data when they are above the ground station, they can send the data back to the ground. And I understand in your paper, you were doing what are called retrospective studies, correct? You're not working with a disaster manager, at least not yet. Not yet, no. So what we've done so far is actually just looking at the database from MODIS that we had since 2000, trying to see all the eruptions that have been detected since then, all the overflow eruptions, and then try to retrocast and use our certain method to try to see if we could predict the end of the eruption. And retrocasting would mean that you're looking back in time. Yeah, we already know when it ended. We already know when it started, but we pretend that we don't, and just use the data as if it was coming in real time. I understand. And how many eruptions have you been able to study in this way? About 100 eruptions. 100? Yeah. On many different volcanoes then? About 104 at 34 different volcanoes. 34 different volcanoes. So that many eruptions since the satellites were operating in 2000? Yes. And those are the resulting? About two a year. Yeah. Well, let's see. I think we've got another slide of another eruption. And this is the one which you have actually worked on, all right? Explain to the viewers what it is we're seeing in this black and white image. Yeah. So right now you're actually seeing Iceland. You can see in the larger image the shape of the island. And there's a zoom in on this those bright pixel right there in the middle of Iceland. And this is from the 2014 to 2015 holo-horon eruption in Iceland. And you see those really bright pixels, so the square pixels that are really bright, those are really hot, and that's where the lava flow is. I see. All right. And then I think in the next slide we'll actually see how you start interpreting some of these data. So here we've got a curve. Yeah. And again, tell us what the curve is showing. So this is the black line. This is a theoretical line called the watch curve. So this is based on the 1981 paper that he published. Jeff Watt. Jeff Watt, a UK volcanologist, where he defined this asymmetric behavior of the effusion rates over time. And the effusion rates is the instantaneous measurement of volume flux of lava that comes out of the ground. So we're basically showing a diagram. Time goes from the beginning on the left-hand side to the right-hand side. And effusion rate is in units of cubic meters. So the spike on the left-hand side. Yeah. So you can see, basically he defined that most episodic, basaltic effusion eruptions would behave like this with a really sharp increase at the beginning of the eruption. And then a slow decrease over time due to the elastic release of energy. And this is the part that takes the longest to do. And that's the part that we care about in the predictive way. And of course, in real life, time might be measured in days, in weeks, and a month. Okay. Okay. And the effusion rate, any idea at what scale of eruption, you say killer wear isn't a very good volcano to model. But monolow back in 84 was hundreds of cubic meters a second. Yeah, that's the sort of thing that you're studying here. All right. And I think the next diagram will actually show us in a bit more quantitative manner. All right, this looks really complicated. So help the viewers and myself understand what it is we're looking at here. So again, it's effusion rate on the y-axis against time or the eruption duration in days in this case. You have all the black crosses are basically synthetic data points. So I made this graph to try to understand how we use our model to predict the end of eruption. Is this for my specific eruption? No, this is just the general idea. So the idea is that time zero, the eruption starts, you get a first data point that's pretty low in effusion rates. And then you get a couple more data points, but you cannot do anything until your rates reach the maximum effusion rate. That's q max on the left hand side, the red cross. The red cross, yeah. So there are three red cross that are important to start, the maximum and the end. Then you have, you need to have a couple more points as the eruption goes on. And you have different colors. So you have orange, green, blue, purple and red. That tells you how much data point we used. So at first you only have a few data points. Then with that you can start fitting an exponential curve. And then as the eruption goes on you have more data points. So you fit a new exponential curve, et cetera, et cetera, until you reach the end of the eruption. And then we try to estimate this delta t of the difference between the predicted time and the observer time and see how it changes. Okay. And nobody had thought about doing this before or the data once available one? It's a mix of both. Mix of both? Yeah, because the data set that I used is only since 2000. And in the past, they've measured effusion rates, but at really resolution. So go there a day and then you go there 10 days later. So you don't have a lot of resolution tempo. You don't have a lot of data points to do it. So this presumably is why you published the paper, right? Yes. Well, we're getting close to the first half of the show but I want to pick up on that point when we come back because if nobody else had done this before, that explains why it's been such a popular paper. So let me just remind the viewers that you are watching Think Tech Hawaii research in Manoa. I'm your host, Pete McGinnis-Mark, and our guest today is Estelle Bunny, who is a graduate student within the Hawaii Institute Geophysics and Planetology at George Manoa. And we'll be back in about a minute. See you then. This is Think Tech Hawaii, raising public awareness. Greetings. I'm Martin Despang, the long-time host of human-humane architecture here on Think Tech Hawaii. Think Tech is important to our community because think about how awesome our natural environment is here in Hawaii and we need to make our built environment equally awesome exotically and inclusively. So because of that, for the first time, Think Tech Hawaii is participating in an online web-based fundraising campaign to raise $40,000. Your thanks to Think Tech will run only during the month of November and you can help. Please donate what you can so that Think Tech Hawaii can continue to raise public awareness and promote civic engagement through free programming like mine. I've already made my donation and look forward to yours. Please send in your tax-deductible contribution by going to this website thanksforthinktech.causevox.com. On behalf of the Think Tech community and enriched by Think Tech Hawaii 30-plus weekly shows, thank you so much for your generosity. And welcome back to Think Tech Hawaii research in Manoa. I'm your host Pete McGinnis-Mark and my guest today is Estelle Bunny, who is a graduate student within the Hawaii Institute Geophysics and Planetology at UH Manoa. Now Estelle, in the first half of the show, you describe this model which you've written up and it's been published. And as our next slide is going to show, you got quite a bit of publicity from this. So if we can briefly go across and maybe it's a little small print for the viewers to read, but on the left hand side I put together this is the cover of your your paper and it was published in a journal called Bulletin of Volcanology. And then on the right hand side just a glimpse of what the media attention is. Now for our viewers, I should explain that Estelle is a graduate student. She's working on a PhD and this paper which you published earlier this summer got quite a lot of attention, right? Yeah, it was actually my first paper. Your first ever paper. First ever paper and we have this person at school that make press release on what's going on at SOEST. And SOEST is the School of Volcanology and Technology. Yes, sorry. And so she put together a small text about my paper and this got a lot of attention and I got calls from HPR, the radio, from AOS and Science Magazine. So it was kind of surprising and exciting. Really? Yeah, it must be fabulous. I mean sort of your first paper and lots of people are paying attention. What kinds of things did they want to know? They really just asked basic question and everything I told them was what I wrote in the paper. There wasn't much things new but I don't know. It was like telling them how we got the idea, why is this so new and how this relevant to Hawaii. Any interest from disaster managers? Because presumably your methodology could be quite useful to people on the ground. Right, so far not really. I do have a contact IATO that is interested to use that model in case Manaloa wakes up. But that's about it. So are you expecting with all your future papers that you're publishing when you get this kind of media attention? It must be fascinating as a new graduate student to actually see that your research has got such broad interest and also some of the relevance. Yeah, I had friends on social media posting things from me and I was like, hey I found your name and I was like, oh my gosh. Even across the ocean in Europe people told me that they seen my paper. Really nice. All right, so you may not have included it in your actual paper but what are the limitations of your technique? You say that you've tried to apply this methodology to over a hundred different eruptions. Yeah. How often did it work? Do you know why it did work or not work? All right, so the first thing I need to specify is that out of those 100 eruptions, first we looked at the shape of this effusion rate. Does it actually look like it's doing this behavior all the time? And it's not. So nature is complicated and so it's actually, we only found that shape or that behavior 30% of the time. Okay. So that's definitely a limitation and we would not be able to tell the different shape before like quite a lot of the eruption happened. So you need to collect data perhaps for a week before you can tell the general characteristics of the eruption? So, but the part of it is that even if the shape is different, I could eventually update the model or the prediction as the eruption goes on. So we have different shape that we identified. So it's either completely random, but sometimes you do have a couple peaks. So instead of having just one, you have a second and we could update that as time goes on. Or sometimes it just has different shape that we could also try to model differently. So that's one limitation. And presumably it's critical to get measurements right at the peak. Yeah, exactly. How easy or difficult is that? Difficult because you can miss the peak. Satellite measurements are great, but they are affected by cloud. So if there is a cloud, if there is a cloud on top of it, it might make the measurement, but have a lower effusion rate that it actually is and miss the peak. And if you miss the peak, presumably this curve that you were talking about would start off at a lower level. And so you'd think it would end earlier than in fact it would do. That's quite useful. I mean, we are here a little bit about monolow when that might erupt again. But presumably that would be a very useful kind of eruption for us to be. Very good. So where do you take this model next? What do you do to help define it? Are you going to have more discussions with disaster managers so that it's actually of societal importance? What are your plans? Well, it's still need to work always. And the main thing that we want to do is actually, instead of retrocasting it, it's actually using it for a real eruption. But so far, I haven't had the chance to look at one as it goes. So I need to be on top of it and make sure that I start as the eruption starts to try to predict and make my prediction and see if that works. That's the one thing that I want to do. But the thing and thing is, we haven't talked about disaster management people. And I think it's still something that we need to take in. It's a good thing, but it doesn't have to be like, okay, the stylist is going to stop then. So now you're good. But it's still like maybe that's the sort of thing your advisor might get to help with. Yeah, hopefully. Interesting. Yes. So if monolo was to erupt, you have to get ready to work with the data from day one. Yeah. And we have no understanding monolo in past. The 19th century erupted for 200 days, whereas the 84 eruption only lasted 23 days. How quickly can you actually get some kind of information to even the scientific community if not the disaster managers? Well, I think really quickly, because this increase part is actually really fast. So it depends, but it can be from a couple of hours to a few days. Then after that, after this maximum is reached, that's when you can start feeding curves through the data. And so it might not be very good accuracy, but still can get us an idea. Would it help if you had either a greater level of detail, the spatial resolution was improved, or the frequency at which? Definitely better temporal resolution. Temporal resolution. More opportunities to image per day. Yeah. And like maybe using more than one satellite or using other satellites, and then importantly, having the data in real time. Can you augment your satellite observations with aircraft or with drones? Would that be possible? Yeah, sure. Why not? It could be something different data processing, but you can definitely calculate the same thing. So as long as the entire eruption is within the same field of view, you can't go over one place on one location and then elsewhere. It would be all the data said that we can include. And are you working on these ideas for your PhD thesis, or is this it? You've had enough publicity from this particular study. Right now, I feel like I want to do other stuff and then maybe I'll try to include more. But it's great to see that this first investigation that you've published got such wide publicity and that people see the significance of your work. And hopefully your advisor can actually start talking to not only Hawaii Volcano Observatory, but maybe elsewhere around the world. So congratulations on the work. Thank you so much. Thank you so much. Well, unfortunately, Estelle, we've got to the end of the show. But thank you again for being on Think Tech Hawaii. Maybe we can get you back at a later date. But let me just remind the viewers, you have been watching Think Tech Hawaii Research in Manoa. I've been your host, Pete McGinnis-Marc, and my guest today has been Estelle Bunny, who is a graduate student within the Hawaii Institute of Geophysics and Planetology at UH Manoa. And I hope that you will join us again next week for another show starting one o'clock on Monday. So see you then. Goodbye.