Subscribe
Share
Share
Embed
Documenting the monumental discoveries of researchers, and the role Pawsey plays in supporting their breakthroughs.
Welcome to HPC Hearts & Minds, where we talk to the people behind some of today’s most fascinating discoveries.
High Performance Computing, or HPC, allows researchers to tackle incredibly complex questions. This series is all about finding out the real-world impact that comes from using this large-scale computing.
We want to showcase the hearts and minds behind the technology.
HPCHM - Gretchen Benedix 1.mp3
Transcript
Pawsey sits on the Whadjuk country of the Noongar nation, and we'd like to pay our respects to their elders past, present, and emerging.
That's just like the very beginning of the solar system. And maybe what we want to do is we want to be able to see, well, when did all those rivers start flowing, right? If they're rivers, they should have started flowing at some point. Why is the northern hemisphere of Mars like essentially gone?
What's the quickest way of counting every crater on Mars? Hello, and welcome to HPC Hearts and Minds, where we talk to the people behind some of today's most fascinating discoveries. High-performance computing, or HPC, allows researchers to tackle incredibly complex questions. This series is about finding out what can be done using this kind of large-scale tech. More importantly, it's the hearts and minds behind that tech that makes it special. In this episode, we sat down with Professor Gretchen Benedix. Associate Deputy Vice Chancellor of Research at Curtin University. Gretchen is an astrogeologist who studies craters. Her studies at university started in psychology, but one elective unit changed all of that. She's gone on expeditions to Antarctica to pick up meteorites, can analyse rocks that made the journey from Mars all the way to Earth, and even has an asteroid named after her. Using supercomputing, Gretchen was able to train an algorithm to recognise craters. mapping the whole surface of Mars and its many, many more craters than they originally thought. Gretchen, thank you so much for joining us today. I'm very, very excited to talk about your work.
It's very nice to be here. Thanks for inviting me.
You've had an incredible start into what you currently do, but I really want to know what got you excited about the stars.
I've always been interested in the sky. The places I grew up had very big open sky. So I lived in a state in the United States called Idaho, which is a very rural state. It's very, very much like regional WA in a lot of ways. And when I lived there as a young, really tiny, I was there for, I lived there twice, but you have access to the sky and you can see brilliant stuff at night. Now, I was a child, so I was like, pretty, that's cool. Yay. But I do remember my uncle bringing a telescope around and showing us, oh, there's that star and there's Jupiter. And I thought, oh, yeah, that's cool. Those are interesting. And then I never thought about it again when I was in school. I wasn't too bad in maths and science. I wasn't too bad in drama. I wasn't too bad in speaking. I wasn't too bad I really didn't enjoy history. That was about the only one I was like, whatever. But I was, mostly I was like, okay, I can, whatever. Maybe I'll try. So I would like reel things off. I'd be like, maybe I'll try engineering. No, today I'm going to do medicine. I'm going to be a doctor. And then I actually applied to go to university and I applied, I thought, I'll do psychology. That seems good. All right. So off I went to go do psychology, kind of the first two years really is breadth. So you can define a direction you want to go, so you can declare yourself a psychology major. But in your first couple of quarters, you have to take classes that will satisfy A quantitative assessment, that will satisfy a writing assessment, that will satisfy, you know, science. It makes you kind of stretch outside of the thing that you think you want to do, and you can just have a look around. So I didn't feel confident in math, so I thought, well, I'm going to do a queue, but I'm going to do a really easy queue. I'm going to do astronomy 101. I'm going to go and I'm going to do the basics of astronomy and stars and things like that. And so I did that, and it was just amazing, right? So I was like, oh, this is really cool. And they took us to an I went to university in Santa Cruz, California, which is right on the water, but there is an observatory inland about two, 2 1/2 hours. So one of our field trips was to this observatory. So they took us up there, it's nighttime, blah, blah, blah. It's this old, old telescope, like historical. And I was like, yeah, history, whatever. But they showed us Saturn. So they like, pointed it around and said, okay, here's Saturn. And I looked in the telescope and I was like, whoa, that looks just like the pictures of Saturn. And my brain was like, it's really out there? What? And.
Kind of solidifies it a little bit, doesn't it?
Yeah, because you're kind of, you're kind of in two places. It's like, it looks exactly like the pictures that you see of it. So did they just take one of those pictures and stick it at the end of the telescope? Yeah, see that? That's what it looks like. No. So it's just, it's really that kind of, yeah, it's like, oh, it's real. It's like that perspective shift of, yeah, I've seen pictures of it before. It's like when you watch a movie and it's in a place you've never been and you're like, oh. But when you watch a movie and it's in a place you've been, you're like, oh, I know that and I know that and I know that.
That's it.
And you get that kind of connection to it. So I think that was partly it. And then They didn't have an astronomy degree. They had a physics degree.
Right, okay.
So I just shifted to a physics degree, which then led to me becoming interested in particle physics instead of astronomy. But...
But it's funny because you started off with psychology. There's this idea that you think, okay, I think I'll go into this, but it was just this one elective that like sort of piqued your interest. And that obviously gave you such an interest that you pivoted into this new field. I feel that I guess, like you said, you weren't so sure about maths, but you know, physics is quite heavy on maths. And for some reason, this thing inspired you enough to go, I'm going to give it a go, because that's what college is all about.
And it wasn't easy and I didn't like I wasn't perfect. So I experienced failures, but I learned from them and I retook classes multiple times. And what's really interesting is that I had a really hard time with electricity and magnetism as a concept when I was in the university. And in my professional, like as a researcher, I use instruments all the time that are relying on electricity and magnetism, and I have a much better understanding now of it. But at the time, I was like, these equations are hard.
But that's it, right? Learning a new concept. Like, I guess it is kind of scary to jump into something like that. And like you said, it's about failure and practice, but having the the ability to grab the foundations, I guess, if you will, physics being a very, very key element to other studies.
Absolutely, and that's exactly what happened is that I got to the end of the physics degree. I had done, I actually was gonna go do particle physics and try to figure out, you know, the origins of the universe because that's like the big bang, how amazing is that? And got to do this internship. at a linear accelerator center, and that was absolutely amazing. So I was like, yeah, I'm going all in. I ended up extending my undergraduate for another year. So I ended up doing two different research projects, and I did one on particle physics, but then I did a second one that was based off an internship where I started looking at asteroids. And I was trying to understand, there are a bunch of asteroids that have orbits that are pretty close to Earth's orbit, and we call them, we now call them potentially hazardous asteroids. We used to call them near-Earth asteroids or near-Earth objects. And so essentially, the internship I was doing was trying to predict how many of these objects should be out there because you can't see them, right? You can see a few. So at that point, We had about 132 total.
Okay.
And I was given, like, I was given actual log books from the people who were painstakingly trying to identify these asteroids night after night. And they would have, okay, so we're observing all night, and the next night we're observing all night, and then we're gonna look at the pictures the next day of the two nights, and we're gonna look for the thing that moved.
Right.
And this is 1989.
Okay, yeah.
So the optics of the telescope is fine. Not great, but fine. But the fact is that basically to do what I did for that, which would have gone a lot faster, they didn't do that. They literally would have to look, they'd have to compare by eye photos, and then they'd have to go back and they'd have to look for that. same object in other photos to be able to get the information, the mathematical information that described those orbits, so that then they could work out where these orbits are. So there were 132 at that time. Now there's like a million. So over the course of this time, the technology has gotten to the point where we can send a telescope out into space And it can just look around for a while, because atmosphere is a problem. So if we get out into space, then the atmosphere isn't so much a problem. So we can see darker things a little bit easier. Still not great, but a little bit easier. And so then we have a better way to get faster information that we can then interrogate to figure out, are there potentially hazardous things we should keep an eye on? Turns out, yeah.
I guess, because there is obviously an interest to have anything that's floating the earth. It's obviously many things can happen from it. And so it's just a good way of keeping an eye on it. But there's only so far, I guess you can look. And then doing that manually, what happens if you miss something? Exactly.
And this was also like, so I think it was the late 70s or early 70s was when the Alvarez father-son team were the ones that first suggested that the dinosaurs were made extinct. possibly by an impact, a giant impact. So then there was this whole kind of, oh, wait, if that happened then, well, what does that mean now? And so this whole, and we started, because we looked at the moon, and we're like, there's impact craters everywhere. Especially over the years, science has kind of found ways to explain things that previously were unknown and kind of scary, and we didn't understand it. And so by being able to explain it, kind of gives it less, it's less scary. It helps us kind of find our place, I guess.
What was going through your head at the time in terms of like your, I guess your interests lining up with what you were doing? Because you were discovering these particle physics is really interesting, but then as you start working on that, you might notice maybe another project or maybe something that was really dragging you along this trajectory, right?
Yeah, so basically that internship project around, you know, trying to figure out probabilities of asteroids. and how many there should be, that kind of morphed into a master's degree. But the master's degree was completely different. It was still around asteroids, but it started to incorporate the rocks more. So it was, we knew from spectroscopy, light reflected off the surface of an asteroid tells us something about what it's made out of. And we can do experiments by saying, okay, well, we'll shine light that mimics sunlight on a rock on Earth, and we'll read its spectrum, and we'll look at it at different wavelengths, and the pattern it gives us, we can compare that to the rock, and then we know what that looks like, and then we can do it for the same out there and do these comparisons. On Earth, we have a bunch of rocks from space, meteorites. And those two had been characterized by their kind of compositions and the things they looked like. And so, one of the things was that, over the years, we knew we had come to understand that meteorites come from space, meteorites most likely come from asteroids, so the obvious thing is, well, which meteorite came from which asteroid? And so when you look at the spectroscopy and the groupings based on the spectroscopy, and then you look at the meteorites and you can look at their spectra as well, you start to see some patterns that emerge. And one of those patterns led us to think there's a group of rocks we have here on Earth that we think definitely came from the asteroid Vesta.
Okay, interesting. Yeah.
We have a connection that we're like, yeah, that's like 99.999. We're good with that one.
Yeah.
And then there are other ones where it's like, well, this is the biggest group of meteorites on Earth. It's like 90% of meteorites on Earth basically have slightly different compositions, but essentially look like this one thing. And then there's a group of asteroids that back then there was just a single kind of characterization that had a very similar compositional spectrum. But in space, they were reddened. So they were kind of, like they reflected more light in infrared wavelengths. But they had all the same bumps. And so it was like, well, what would be causing this difference? Because everything else lines up. It's only the fact that this has this weird slope. My M-fill was to try and determine experimentally if The solar wind interacting with the surface of an asteroid might causeway this, what we call reddening. So, does that increase the reflectance in the infrared? Does that change some aspect? of the mineral composition or some other thing of how we're reading the spectrum. I had a little linear accelerator, so I was firing a beam of helium atoms at little pieces of rock that are similar to the rocks that were in these meteorites, but they were from Earth. And so I was doing before and after measurements, and I was like tracking all this stuff. And It's really fun because I was like, it's like particle physics. I got a linear accelerator and I'm smashing this thing. And asteroids are really cool and spectroscopy. And then I was like, these rocks, these rocks are really cool. I'm kind of really enjoying these rocks. And I took a geology class as an elective for my master's degree. And they took us out in the field. I got to tell you, field work is a really, really good way to get people to come into a degree.
Oh, for sure, yes.
You go out and you see things in action. And geology is just so satisfying. In physics, it's like you have to imagine particles interacting. You have to imagine what a magnetic field might look like. And geology is you pick up a rock and you look at it and you go, This rock cooled really, really slowly, and it is very primitive.
It says the tangibility, I guess, the holding of things. So you learn just through this one elective, you start just deciding. rocks have this really amazing property to her that you think I could probably continue doing this.
Exactly. And then I got to the next stage, which is, okay, you're going to graduate. You've done your master's thesis. Now, what are you going to do? And I'm like, okay, I think I'm ready. I think I'm going to do a PhD. My advisor had gotten her PhD in Hawaii, left the university I was at right just around when I was finishing. And she went to the University of Maryland, which has an amazing astronomy department. So she went into the astronomy department there. And I was like in two minds. So I applied to both places. And then I went and visited both places. And once you visit, you kind of go, oh.
Didn't make this choice sound like it.
Beaches, sunny.
Maryland.
In the middle of the DC Beltway, freezing cold in the winter. I'm going this way. I'm going. And when I did the visit, they, were showing me meteorites that are, absolutely fantastic. It's so great to see meteorites in person. The research environment was great and the environment environment was great.
Well, yeah. There's a lot of, obviously, like Hawaii's got kind of volcanoes as well.
There's different rocks and observatories. And it's basically, if you want to see rocks being born, you can do that. And then later that night, you can go up and try and figure out the, how the universe started. It's all within 14,000 feet of each other. Right. It's pretty close, which is great.
It's really close. It allows you, I guess, to dabble into these different places again and sort of see what jumps at you.
I had colleagues at the Smithsonian Museum of Natural History, and so second biggest meteorite collection in the world, essentially, or maybe the biggest, hard to say, and a year and a half of basically just tackling all kinds of fun and exciting things, and also just learning how to do things. with multiple different meteorite types. So one of the part of the job is that the US has a program called ANSMED, which is the Antarctic Search for Meteorites. And it has traditionally been a program that's supported by NASA, by the NSF, and by the Smithsonian Institution. And NASA funds and NSF funds groups to go to Antarctica and pick up meteorites. And sometimes it is, 1000 meteorites in a season. The infrastructure is there, right? Antarctica is a research place. It's dedicated to research only. It has no political agenda and nobody owns it. Everybody kind of has a piece and there's a treaty that says you will not, there's no commercialization. This is research, research, research. And so these teams will go down and they will collect meteorites for six weeks. And then those meteorites get housed at Johnson Space Center initially. And then finally, the final repository is the Smithsonian. So the collection of the Smithsonian houses a lot of the US Antarctic meteorites. Amazing.
Okay.
And so part of the job that I had was in addition to doing all kinds of other things, like experimental and trying to figure out different types of meteorite evolutions, was to classify these rocks that had come back from Antarctica. And so when you do that, 500 pieces of a little rock, and you start to notice things, right? And that really builds your experience and your capacity to make those connections across different things. I went to St. Louis, Washington University in St. Louis, and I had a part-time gig as the laboratory director of the electron microprobe and XRD labs. I was running the lab and teaching people how to use the lab, but it was half, and then I had half for research.
Oh, excellent.
And Washington University was a planetary science hub. That was really beneficial to be able to just interact with a bunch of other people. So that's when I started looking at moon rocks and started talking to people about moon rocks and starting to better understand the chemistry. So it was very chemistry oriented, but there were Mars people there as well who were associated with Martian missions. So then I was able to start thinking, okay, I'll go back to what I did originally, but now that I know all this other stuff, I can build on that. and start building on the research side of things. Yeah.
At that stage, you're so comfortable with the work that you've been doing with all the meteors that you're able to now start being sort of like a leader in terms of being able to show these students.
Absolutely. And that's a huge part of my whole journey from the time I was a grad student has been outreach, interacting with all levels of general public students, teaching them about meteorites. I think the first thing, the first one I ever did was when I was a grad student and I had to teach about meteorites and I used Rice Krispie treats because you can map them, you can put stuff in and you can say, well, this is what this is like. Now you describe it like this. And so that was just super fun. I really enjoyed that. Don't eat the thing. And that's the other aspect of working at a museum is you, Yes, there's a collection there, and the research is a huge part of what a museum is there to do. But the biggest remit of a museum is to get the public to understand what all is involved with natural history. And there's an art to it, because you can't just kind of launch into, I studied the spectroscopy of us, and I was looking at 0.3 microns earlier today, and none of that works. But what you can do, and this was the cool trick at the Smithsonian, was you can get them to have their gloves on and you hand them two rocks. And you go, okay, so what do you think of those rocks? Oh, those are interesting rocks. And then you get to say, that one's the moon and that one's Mars.
Wow, okay.
And everybody goes, oh, my brain is blown up. Oh my gosh. At a museum, the thing that brings people into a natural history museum is the dinosaurs and the space rocks. So we always had a really interested audience, essentially. And so.
It started from the get go. It was, you know, you already got a keen room of people, which is amazing.
Yeah, exactly. And you just have to... You have to talk about it at the level that's relevant. They would say, Well, what was it like to go find meteorites in Antarctica? So then you talk about that and everybody's, You went to Antarctica, and I'm like, Yeah, and the meteorites are blah, blah, blah, blah, blah. You know, I was like, And the meteorites are cool too. And this is why we study meteorites, because They tell us lots about this and that and those and the other. And so that was the next six years was that, and I was getting research grants and I was building a team and starting to build up the leadership side of things there. And that would have been kind of the lifestyle, the life plan, but my partner had a research project here. So he was based in the UK, but he had a big research project here in WA. Yeah, so he had brought a reasonably sized team from the UK. I came along. It was kind of jumping from, okay, I run this team and I have this money and I do this to Okay, I'm a part-time senior lecturer now and I have to teach. Oh, okay. So the teaching was my biggest hurdle. I found though, that as I got better at it, what I gained from it in terms of expanding that tool set was absolutely amazing. It was helping me think, because I could say, all right, well, I'm teaching this about the earth. So this is how this works for the earth, but hey, wait, that thing is in this rock. on Mars. That doesn't make sense because it can only occur like this on Earth. So then what does that mean about the environment on Mars? And then it's like research area.
Right.
No one's ever addressed this. No one's ever said, oh, that's weird. So I got research funding. I got a couple of fellowships. And then that was kind of right at the beginning. So I got, we got here in 2012 and I think 2014. was when I got my first fellowship, and that's when I started my interaction with Pawsey.
How did Pawsey come into your space?
When we got here, because of the Desert Fireball Network, my partner was introduced to all these different opportunities. And so he started learning about Pawsey, because essentially, Desert Fireball Network is a network of 50-odd cameras out in the desert. operating autonomously, collecting huge amounts of optical data. Every single night, these cameras are taking pictures all night long. And initially, they just left the camera open all night. They now have, they've updated the engineering, so it does it a little differently. But Pawsey was a huge part of that partnership of having a place to store data, but also a place to do the high performance, like how do we interrogate these images better, easier, things like that. So he had been in contact with Pawsey. And so, you know, we'd sit around the dinner table going, what do you want to do? And essentially, I was going back to that connection between a place and a rock, but instead of it being a certain type of meteorite and an asteroid, it was now a Martian meteorite, and where does it come from on Mars?
Right, okay.
And so people had already been trying to map these kind of things together, and they built up databases of rock spectra, but the rock spectra has always been terrestrial rock. And then they were using that to compare to what they were finding on Mars. And they could never really get fully satisfying. And they weren't fully things that you would go, we can pinpoint exactly that area. And that is where this rock was launched off the surface.
Gotcha.
And it has to do with resolution. It has to do with spatial resolution, spectral resolution, all these things that you have to kind of figure out how you're going to map all that together. Part of how they were getting the spectroscopy of the terrestrial rocks where they were taking whole rocks and they heat them up in an oven and then they measure the infrared signal coming off of it. So this was a big, big group at Arizona State. Huge group. They had multiple missions had gone with their instrumentation. And so, and the spectroscopy you get from that particular type of, it's called emission spectroscopy, is really good for distinguishing different minerals.
Okay, yep.
Really, really good. Whereas the stuff that we use to look at the surface of asteroids doesn't distinguish compositions of minerals very well.
Okay.
You can tell the difference between this mineral and this mineral.
And that's it.
But you might not be able to get into like, okay, where exactly in terms of composition. It was kind of around the time that Google was getting really good with facial recognition and Apple was getting really good with facial recognition. And so it was kind of like, well, if you can recognize a face, surely you can recognize stuff on the surface of other planets. Seems pretty straightforward. And there was this other way to figure out, to kind of map the rock to the surface. And it was through age, determining ages. And so we can determine ages in both. So if we have the rock, we can do it with radioactive isotopes. We can work out time that way. And on the surface of a planet, we can look at how the craters are distributed and how they have built up over time. Basically, a body out in space, Impact is by far the number one geologic process in space. On Earth, we have lots of stuff going on. We've got volcanism. We've got like water flowing everywhere. We've got tectonics. We've got all kinds of stuff going on.
We're busy.
We're busy. Very busy. Other planets, not so much. And so kind of to get things to really shift and change, it requires smacking into each other. And so what was realized even before we went to the moon was that you should be able to use that as a clock. If you imagine that you have a brand new baby planet, it's got a smooth surface. And if the only thing that can affect it over time is impact, you can imagine that you would be building up the numbers of impacts. Now, the other really cool trick is that they're not all the same size. So the things that are hitting these surfaces aren't the same size. So, and they behave in a really predictable way to a degree in that whenever you break anything, like you can take a piece of ice, throw it on the ground, and you'll see this phenomenon. If you break something big, it will break into a couple of big pieces, and then it will break into a few more kind of medium-sized pieces, and then a whole lot of tiny pieces. So it's a power law. So it's essentially you've got this kind of size range off of a single object. When you look at the surface of another planet, you see different sizes of craters, and you see that there are very few big ones, and that there are There are a lot more medium-sized ones than there are. tons of tiny ones everywhere. And people had been looking at the surface of Mars and someone had actually calculated, had not calculated, had actually manually, similar to the whole asteroid, determining if an asteroid exists or not, had manually gone through and looked at the highest resolution available data, which at the time was 100 meters per pixel, to identify and catalog the size, shape, depth, everything, location of every crater on Mars.
Wow, it's a big job.
It was, it was, it was a PhD.
Right, okay, of course.
So, PhD. And what he ended up with was this database of 385,000 craters bigger than one kilometer. Because essentially, if you have The resolution dictates the smallest size that you can actually see, and it's the resolution in order to be confident that you have a correct identification. it has to be at least 10 pixels. So if your resolution is 100 meters per pixel, to have really good confidence in the size of something you've picked has to be 100, or 10 times 100, so 1000 meters. So that's a kilometer. So they have theirs going like that. Now, the issue with that is that when you look at, when you convert to time, and the conversion to time is, very complicated because it involves knowing how old, you have to have a comparison point. So you have to have a rock that came from an area that you have done this crater analysis of. So we had done that for the moon. And then there's like really cool mathematical tricks you can do to kind of say, well, here's what it should be for Mars. So we were using that. And that still is the case, right? We don't, we don't, we haven't been to Mars to actually collect, well, we haven't been, we have been and we have collected samples, not as humans, but as robots, but we have not measured the dates, the ages of those rocks on the surface. So we still, in order to look at that, we still have to look at the rocks we have on Earth. And so, The issue was that if you're only counting the big ones, if you're only looking at the big ones, the time region or the time era you can get at is really old, like 3 billion years and older.
Okay.
Right? And that's fine. But that's just like the very beginning of the solar system. And maybe what we want to do is we want to be able to see, well, when did all those rivers start flowing? Right? If they're rivers, they should have started flowing at some point. Why is the northern hemisphere of Mars like... essentially gone. it's a much smoother area. So when did that happen? And we can kind of pinpoint that, but what are the kind of smaller points in time, discrete, the more discrete points in time? So you could actually say, oh, well, this is how something has changed over time, because you can actually look at a million years instead of a billion years as your time stretch. And to do that, you need to count the small things. And by the time I came around, there was a global map at 5 meters per pixel, which means you could get down to craters 50 meters across. And if you're looking at the power law, it means you're going to go from 385,000 to a lot more. And so this was a perfect test bed for HPC, for using high performance computing and to really build up the automated detection of an image. So when we first started off, we were trying to do how other people had done it. We were using these old algorithms that were really clunky. And It turns out teaching a computer to recognize a circle is ridiculous.
Is it?
Yes, because if you're doing it, if you're not doing it as like an analysis of an image. it's just the number of pixels, right? And so it's going, well, this is the value of this pixel, this is the value. And it starts to, and you have to say, okay, I want you to take a line that's this many pixels long, and I want you to find all the places where that pixel line works. And you have to map it around a whole bunch of different ways. So you're taking a line to create a circle. And it was really clunky, but we tried it. And we got sort of somewhere, but then in combination, so that was with one of the Pawsey interns. So he was one of the Pawsey interns, the first one, and he was really good. He was like very into it, did lots of stuff, limited by the technical kind of feasibility of it because you can't just take the image as output. You have to take the image and then you have to do all of this kind of stuff to the image. You have to stretch it and you have to kind of make it. into highs and lows, really increase the contrast so you can really see where those circles are. So then when the computer is looking, it can see that, okay, that's a much higher value than that. And then I can see that this line has this high value here and the high value here and low values in the middle. So I've never fully understood that algorithm, to be perfectly honest. So my description of it is very basic. But essentially that was it. But this was right about the time that the Curtin Innovation, Curtin Institute for Data Science, which wasn't called that then, came online. And so there were data scientists that were there to just help anybody at the university. And their job was just to kind of offer options. And so we were working here and then they were working with Pawsey. So we had this kind of conduit. So we had the Pawsey intern and then that was kind of okay. And we had put together some engineering stuff and we had done a paper, a very small paper that kind of described it. And I had gone to a workshop that had described it, found that other people had been doing that particular algorithm in the past. And so then the focus became, well, okay, this isn't working that well, what do we need to do? And so the people at CIDS were, oh, it was CIC then, Curtin Institute of Computation. And the folks there were, we presented them with the problem and said, okay, we're trying it this way, but... we're open to suggestions.
Okay.
And essentially what they came up with was, it was right about the time that GPUs were really kicking off in terms of how you analyze imagery. And so they found... they knew enough to be able to say, okay, we need a GPU because it's an image. And then they also knew enough that there was an algorithm that had just come out called YOLO. You only look once. So it was based around facial recognition. And so facial recognition software and algorithms had morphed And so this particular algorithm was an open source algorithm that was used to identify multiple objects in a single image. And the example they had on their page was an example of three animals. So it was a dog, a cat, and a horse or something. And it identified all of those things individually on the same image. And you just had to train it to identify the single thing with enough data so that it would know. So the next steps were as part of this research, I then got another grant to kind of focus on developing that whole machine learning side of it. And so that is what resulted in this, map, which is essentially after building the training data set, and that's the hardest part. And we started by looking at the existing manually determined data set, but found pretty quickly that it wasn't going to be good enough to use as a training data set by itself. We actually needed to identify individual craters on images, and you had to do it at the right size, And so building all of that up is what took the time, because you needed a training data set that would go across multiple resolutions to be able to get the different sizes. So we had this whole kind of layered setup, and you had to kind of flip images around with craters manually identified. so that the computer could start to see, okay, that where the square is, the thing I'm looking for. And this is what it looks like here. This is what it looks like here.
Right, okay, yep.
So building up the training data set was the hard part and the time-consuming part. And once we had that built, we were able to run the whole 5 meters per pixel global data set. here through Pawsey, and it returned this data of 94 million craters in 24 hours.
That's, oh, in 24 hours as well.
So that meant it interrogated every single, so it had to take the whole mosaic image, it had to tile it into subtiles, then it had to look at different resolutions because you could have a bunch of small craters, but you might have a big crater over it that you wouldn't see. So if you want to get as close as possible to 1 kilometer at the top end, because we didn't care about above 1 kilometer, that's done, manually done. Don't care. We don't need that. And so we were focused on the 50 meters to 1 kilometer. And so that was the kind of That was the bread and butter of the algorithm was getting those smaller things, and then we extended it because we also started using a really high resolution data set to identify or to create a training data set, so... That data set is called high rise and it's 30 centimeters per pixel.
Oh, wow. That's very detailed.
But it is not global. So it is very specific areas, but you can use it to help define, you know, up to the range you want. So it's very helpful for getting the super small ones. And the algorithm also got updated over time. And then it got updated to a point where it was really good with small things rather than just being good with different things, really good with small things. So all of that came into play. And so this is the result of that connection between the folks at CIC, the folks here at Pawsey, and the ability for us to interpret that. And then we were able to take the data that we get from this And we confirmed that we could get the same numbers as manual information.
Excellent.
That's the beauty of it is we didn't just produce a database and say, okay, we're done. It works. Whatever. We've now used it to identify areas where on Mars we think craters are associated with the rocks we have. So we've used it to say, okay, detective story. I have a rock and it has this age and it has this age and this age. So the beauty of radioisotopes is that they date different events in a rock.
Okay, yep.
So you can do things like, okay, I can figure out when it actually solidified from a magma to a rock. I can figure out when it actually got launched off the surface of Mars and then I can figure out when it landed on the Earth. So these three things are really useful. And then when you look at the ages that you can find from the craters, you can figure out, okay, this is the age of the crater. This is the age of the area surrounding the crater. So the area surrounding the crater is probably related to the age of the rock from when it formed. The age of the crater is probably related to the age the thing got launched off. Should be within, you know.
Wow.
And then you can work out like where's the best place to look. And so we knew that any meteorite getting off the surface of Mars is done by an impact. Any impact that's 3 kilometers across would have been formed by an impact with enough energy to get material to escape velocity so that they could get off Mars. So when we first started looking at that, there were 70,000 craters, 3 kilometers and bigger.
Okay.
And with our data set, we were able to knock that down to 19 craters that were the most likely craters that possibly were the launch sources of the meteorites.
That's amazing to think. Now that we know this, I guess all your questions start getting bigger or more detailed. You can start learning things that you didn't Everything we were able to get to that point, you have this number and you can wind it all the way down to 19.
Yeah.
It's incredible.
Exactly. And it's, you know, there's still skeptics that are like, oh, those can't be the ones. But we have identified 2 areas that have enough of those similarities across that seem to be related to very specific groups of rocks that look like those are the areas they'd come from, they fit, that we're pretty happy that it's telling us something that's reasonable. One of the big open questions is, there's this thing called early bombardment. So there's this idea that when we went to the moon, we picked up a bunch of rocks. We came back and we dated them. And we picked them up from all over, right? So it wasn't just one spot. And we brought them back and we dated them and they were all the same age. So we were like, every single crater on the moon was made at this time. That's weird. So the only way you can explain that is that you had a whole bunch of stuff that headed out that way and smacked into the moon. But there has never been a way to get any dates younger than that, because that was a very kind of bespoke sample group. And we haven't really gone back to the moon enough to get new samples. And so people over time were looking at, okay, but is that real? And then there were starting to be cracks in, because as technology gets better, the aging, the dating mechanisms get more refined. And so you start to see that, wait, maybe these are all the same rock, right? Maybe these are all from the same crater, because they're all bits of that. When that impact occurred, you see those, like, it's got streaks coming off of it. Those are called ejecta lines. So when you smack into the surface, a whole bunch of material goes up and then it flies away and it goes into this concentric kind of rays. One possible theory is that on the moon, all the stuff we brought back in the Apollo samples was basically the ejecta from a single crater forming and launching material everywhere. And there are enough craters on the moon that you can see how this ejecta has like gone around the moon twice.
That's neat.
Yeah. So maybe that story isn't quite right. So what happened between 3.2 and now? Like what's the story? Because then the next point we have is we've got some data that's much more recent. And so it's like, what's going on in between these two? So it's one of the recent moon missions by the Chinese. They sent a robotic mission up there, gathered some samples, and then brought them back to Earth. Those have been dated. Those have been shown to be much younger than the rocks that were all brought back from Apollo. So we have now this new point. We then formulated the same kind of map of craters that we did for Mars for the moon. as well. So now we are able to look at a whole range of different ages that we weren't necessarily able to access before. And then we can make comparisons, like what does it look like at this age on Mars? And what does it look like at this age on the moon? And are there differences? Is there, do you see kind of an effect that's happening? Can you see the evolution? So we've done a few more research projects around what that looks like. But then there's all this stuff we can do. We don't have to just date craters. We can date any feature that we see. So for instance, we just started looking at one of the things that we found was for the moon. I had always, in all of my sci-coms, said, impacts hit everywhere. So meteorites fall everywhere around the earth. There's no special, like they don't follow magnetic lines into the poles. Antarctica is just a great way to find them because they're preserved really well. And they're not hidden in the ocean. But it may be that the moon, there is some kind of concentration effect and that numbers of craters may vary depending on what latitude the impacts are occurring. Because it may be that the orbits of certain things will potentially get affected because the Earth is this giant thing in the way.
Right, okay, yeah.
So there is some evidence that there could be a concentration mechanism for the moon. I don't, we don't know if there's one for Mars, but that's another thing we can kind of interrogate as well.
But it's another, it's another question that, you know, 10, 20, 30 years ago, you would never even think to ask. And I think as things evolve, you get to ask these new questions and going, okay, if there is something on the moon that's affecting, and then what does that mean for Mars? What does that mean for Earth? And how our galaxies are formed and A whole lot more opens up just because you found this new piece of this gigantic puzzle.
Exactly. And that's the thing is that the puzzle, the puzzle is huge. Each piece is super tiny and on its own, you may not, you may not see where it fits. But over time, those pieces start to fit together and you can add a new piece that says, oh, we're going to, that picture is going to move over there now. So I think we've tried to apply the same algorithm to a bunch of other bodies as well. And it turns out that at that point when we were doing that, and it was YOLO and you had to have a training data set, you couldn't have a generic training data set. You had to have a bespoke training data set for that body. because the impact dynamics are different enough that the circles look different enough. So it's not just circle. It's circle on Mars, circle on the moon. And so we tried it for Mercury as well, because there's a mission that's coming, that's on its way to Mercury and about to go into orbit. This could be really interesting to be able to say, okay, well, we take this data and see what we can find in terms of the cratering rate at Mercury, which is like this far from the sun. So really close to the sun, there's a whole weird gravitational... scheme there that may or may not affect impact onto Mercury.
Of course, yeah. It's a totally different thing to play with, I guess. A lot more questions that come from it.
It will. And there's physics you can get out of this. Like for instance, you look at that crater and it's got rays going around it, but the rays aren't equidistant in terms of their distance from the center of that crater. So you can do a few different things. You can work out potentially an angle of impact into it. And then the distance that the stuff flies might give you something about the energy that was imparted to make it fly that far. Because you can see this other one with the smaller rays still has, to the right, it still has rays that are noticeable, but they're definitely shorter. That's a smaller hole in the middle. But then if we go, if we were to zoom over to the other side of Olympus Mons, which is the other way, there's another crater with hugely long, but it's probably the size of that in terms of crater. Right, And that's one of the ones that we think is a source region.
What's that? What's the source region?
Source region. That is when we think we've narrowed down, like we think we know exactly which meteorites come from that hole.
That's amazing. That's what you were talking about before, about like being able to find this source. That's...
Yeah, because that's the thing that has been, you know, it's the connection back, right? I think one of the things that people are always surprised by is that they're kind of like, why do you, why do you, why do you need to know where it comes from? And it's like, well, okay, so say we have a rock from space and it has all this gold in it. Wouldn't you want to know where that gold is actually from? You don't go out in, you know, you don't go out in the outback or, you know, if you're out exploring and you find something that has a resource that's useful, You don't just say, oh, well, there's just that one. I don't know where the rest of it is. You want to know what the context of that is. You want to know, well, why is that concentrated there? Did that, what happened? So this is essentially meteorite. And, you know, planetary science in a nutshell is we have only ever had the pieces that come to us. They're gifted to us and we love it and it's great. And we can understand so much about that. but where did it come from? How do we know what does that mean about the body overall? Is the rest of the body like this or is this just the one part that looked like this? And so it's that kind of...
We've talked about how wonderful field work is, but unfortunately it's quite an expensive trip just to go out. So the amazing thing is the amount you're able to gather from other parts. It's not going onto the surface itself, but you're studying meteorites, asteroids, these things that aren't there, they're here, or there's some ways in which we can gather them, are able to answer questions that you would assume you'd have to be there on the ground pulling that out. So you're able to learn without having to actually be on the planet itself, which is incredible.
It's amazing the things we can infer from the remote data that we have. And just having a way to connect it to a rock really, really expands the value. So I think that's one of the best things. the two, well, there have been 4 asteroid sample missions. three have returned material. So the two Japanese missions and the American mission. And so they brought quite a bit of material back. And then the American mission, they got quite a bit of material, but quite a bit of material is at the level of grams.
Right, okay.
Right, so we are trying to interrogate this body based on these tiny particles that have come back. It's not a useless thing. You have to do it. But there's always going to be kind of questions until we can actually go, okay, we need the instrumentation to be miniaturized that will let us get this specific piece of information that we can actually really compare. And I think one of the great things about space has been the challenge to get things off-world has made interesting things happen for Earth. It's just a really interesting, like when you set your mind to do something hard, the stuff that's really helpful will fall out of it. So there's some interesting things that I think will come out of some of these studies. Trying to figure out how you will go survive on the moon as a human will lead to amazing new technology here on Earth. It just will. But we want to understand these other bodies in our solar system because we need to understand the solar system as a whole. The Earth doesn't exist in a bubble. It exists because of what the solar system looks like. And what's weird is why the Earth is the way it is. It's different. The other planets, there are more similarities amongst the other planets. than to Earth. Right. And across, if you look across the entire solar system, you're seeing this change from the interior where things are small and rocky to the exterior where things are huge and gassy and icy. And so what does that mean? Some of the modeling has suggested that Jupiter may have formed roughly where it is now, but it did this wander And it came in close to almost Earth's orbit.
Oh, wow.
And it brought Saturn with it. And that will cause a huge amount of disruption. But this is modeling. So how do we find the evidence that would help shore that up? We still really don't understand why and how life formed on the Earth. We know that really single-celled things existed very early on in the history, but we don't know why that chemistry was able to convert to biology. Europa has this ocean, this gigantic ocean, even though it's small, has more water on it overall than the surface of the earth. Wow. And we're pretty sure that water is briny.
Okay.
And we also know that it's got, or we have surmised with some confidence that it's got an icy rind that's about 10 kilometers thick. Cool.
That's neat.
And we can see all kinds of cracks and all this stuff that's happening. There is a mission going from NASA to Europa. It's called, I think it's called Europa Clipper. And it It has solar panels that are the size of a tennis court. That's huge. Because it, I guess.
Because it's just, you need that.
You have to have it. Because if you don't have a radioisotope generator, a thermal generator to drive, there's no petrol. You can't just fill up on the way. And so once it got into the like outgoing orbit, in order to keep keep energy, it has to have these giant solar panels. But it also has to do this amazing orbital journey, because when you're at Europa, you're within the radiation bands of Jupiter. which are nasty, nasty things. And so you can't stay there that long. Otherwise your instrumentation will go, nope, I'm out. And so they have this really cool, like they've got this orbital journey planned where they go, okay, we're circling. No, we're not. We're going, we're going out here for a while. We're going to cool down. Okay, now we're going to come back. And that's kind of just a huge. there's a lot of interesting maths that goes into that, which I can only assume will result in all kinds of fun understanding that we didn't know. Titan is the other one we want to go to, because Titan is the only other place in the solar system that has a nitrogen-rich atmosphere, like Earth. Okay. And it has, it may have liquid something on it, not necessarily water, probably liquid methane. And it looks like from the one thing that we've had land on Titan, it's, there's like, it looks like there's rivulets of some kind of thing. So there's a mission going to Titan, hopefully. called Dragonfly, which is basically going to be a drone. And they're going to send it to Titan and have it. Now, these are long-term missions. But these are the things that we need to understand. Why does Europa have this water? Why does Titan have a nitrogen atmosphere when other planets have carbon-rich atmospheres? Yeah, why is Jupiter the way it is? Jupiter is weird. It's got really weird things on the top and the bottom and all the swirlies.
I feel like with everything that you've been able to work on up to this point, it's only a matter of time before what you've discovered with your team has incredible implications to understanding not just our planet, but the planets around us. And so it's very, very exciting to know that something that you've worked on is going to influence the next discovery that comes along. So whether it be Titan or Europa, this is quite an exciting journey to witness, but I appreciate you spending some time to tell us about your journey.
Well, thank you.
Thank you so much.
Wonderful.
Thanks for listening in to our chat with Gretchen. There's so many more stories for us to share, and we simply cannot wait for the next one. If you're interested in what supercomputing can do to research, visit our website for more stories like this. Until then, we'll talk soon.