Suggested Searches

Bron Nelson and Dimitris Menemenlis Talk About Simulating Oceans

Season 1Oct 26, 2017

A conversation with Bron Nelson, a computer programmer with the Data Analysis and Visualization Group at NASA’s Ames Research Center in Silicon Valley, and Dimitris Menemenlis, a research scientist working with the Estimation of the Circulation and Climate of the Ocean (ECCO) project at NASA’s Jet Propulsion Laboratory.

NASA in Silicon Valley Podcast Logo

Transcript

Host (Matthew Buffington): Welcome to the NASA In Silicon Valley Podcast, episode 65. Joining me again today for the intro, we have Kimberly!

Kimberly Minafra: Hey!

Host: Kimberly, tell us about our guest, actually more like guests plural, for the podcast today.

Kimberly Minafra: No problem. Basically, we have with us Bron Nelson and Dimitris Menemenlis, who join us from the NASA Advance Supercomputing Facility here at Ames.

Host: Yeah, and this is a slightly different episode from what we normally do, through the magic of fiber optic connections. We had Bron here in the studio, but Dimitris was actually sitting over at JPL over in Pasadena.

Kimberly Minafra:Right. So Bron, a computer programmer here at the NASA Data Analysis and Visualization Group, he specializes in most of the coding that happens with our supercomputers, whereas Dimitris, he’s a research scientist at JPL, where he actually studies and uses the supercomputing capabilities to analyze global ocean circulation and its interaction with sea ice and all the cool oceanography that happens to be displayed on the hyper wall here at Ames.

Host: This is the really cool thing over at NASA, you always think of space, but you know, when it comes to supercomputing, everyone uses the supercomputers, no matter what they’re studying.

Kimberly Minafra: And it’s great because the visualizations are very helpful in investigating the data they come with, that comes apart from actually using the supercomputers.

Host: And I got a kick out of this one because typically, being NASA in Silicon Valley, we talk about ourselves, but this was a situation with a different NASA center whose using the information, and this is typically how this works. You have other centers, other groups, all working together.

But before too much into the podcast, or into the episode, a little bit of housekeeping. We would love for your comments and suggestions. You can leave us a review on iTunes, Google Play Music, wherever you find the podcast. If you want to participate, or just send us your thoughts, reviews, ideas, we’re using the hashtag #NASASiliconValley. We have a phone number, that’s (650) 604-1400. Give us a call, we’d love to hear your thoughts, and we’ll see how we can integrate that into an episode. But for today…

Kimberly Minafra: Here’s Bron and Dimitris.

[Music]

Matthew Buffington: Welcome Dimitris, welcome Bron. So for folks listening, this is a little bit different because I’m sitting here talking with Bron in our studio, and we have Dimitris on the line, or through the magic of the interwebs, from JPL coming and chatting with us. We haven’t done this way before, so this should be a fun time.

So, Dimitris and Bron, we always like to start the podcast with the same question, and it’s, how did you join NASA? For Bron, I would say, how did you end up in Silicon Valley, but in this case, since this is more NASA California, I’d say what brought you to the Golden State? So Bron, go ahead, man.

Bron: I actually grew up in Livermore.

Host: Okay, local.

Bron: Just 30 or 40 miles east of here. I was actually born in Kansas but my family moved out here when I was like two, so I’m almost a native. I was working for a variety of companies. I’m a computer person and I’ve worked for a number of different companies. I was working for a firm named Silicon Graphics and was assigned here onsite at NASA Ames because they had bought a number of our computers.

Host: And they pulled you in.

Bron: Then after Silicon Graphics went bankrupt again, and cut my salary again, even I could see the handwriting on the wall at that point. So I jumped ship, as it were, and went native as they say in the biz, and started working for the customer that I was previously supporting. So that’s how I ended up here at Ames.

Host: Nice. So you were always into computers, not necessarily — I mean, NASA people are always thinking rockets and, you know, space probes and stuff. But you were always into the computers, so that’s how you came into this.

Bron: That’s right. Like I tell my kids, I am not a rocket scientist, but I work with rocket scientists. I know almost nothing about the physics involved. Dimitris here works on the ocean modeling. I don’t know anything about that, but I know a lot about computers so I’m often a member of a team of people, and I help deal with the computer problems that come up.

Host: It is teamwork that makes the dream work.

Bron: What a horrible saying.

Host: I know. I got that from your neck of the woods, Dimitris. I think I heard it somewhere I was walking around in L.A. I don’t know if I was visiting Disney or DreamWorks or something.

Dimitris: What did you hear?

Host: Somebody said, teamwork makes the dream work.

Dimitris: Yes. Well, okay, my story — What I find really amazing, and I don’t know if it happens to everyone, but as you grow up, the dreams that you dreamed as a kid that get realized are the ones that you really remember. So, I’m sure I had tons of dreams when I was a kid, but there were three of them that I remember and that have been realized, and that’s pretty amazing. When I was six, it was 1969, and we gathered around the neighborhood TV. I grew up in Greece, so TVs back then did not exist. Not every house had one.

My grandparents happened to have one. So a lot of people gathered and we watched the first astronaut land on the moon, and that was like a super big impression on me. And of course, a lot of kids who watched that wanted to be astronauts. I’m glad I didn’t become one, because what I’m doing now I think is even cooler.

Host: Nice.

Dimitris: Two more things that, it was a dream, was MIT and Caltech. Those two institutions were just — So, NASA, MIT, Caltech. Somehow, I don’t know, randomly, or accidentally, or because these are the dreams that got realized that I remember, I ended up going from MIT to Caltech NASA. I was doing oceanography as a post-doc at MIT, and there was this opportunity to JPL and work with this satellite that had been launched a few years earlier called TOPEX/Poseidon that observed sea surface height from space. Sea surface height is like a dynamical boundary condition for the ocean. It’s like knowing low pressures and high pressures in the atmosphere, and then you can tell the winds that they’re going to go around the low pressure and the high pressure.

Same thing for sea surface height. If you know sea surface height, you can tell what the surface currents are. The really cool thing that we do is, from space you can only see part of the ocean circulation. You can’t observe everything. You can see surface variables, depth integrated variables, and of course there’s the sampling issue. So in order to make a complete story you need to have numerical circulation models. Those are the really fun models that Bron and others at NASA Ames help us to run on the NASA supercomputer.

Host: I was going to say, that’s probably the perfect transition almost, because I’m sure for folks listening, they think, Bron and Dimitris? You have one person working on a computer, one person looking at the Earth from space, how does that match together?

Dimitris: Bron, do you want to have a go at it first?

Bron: I’ll have a go at it, sure.

Host: What brought you guys together?

Bron: Well, NASA brought us together. Dimitris was working with the people at MIT on this thing that’s called MIT GC, the MIT global circulation model. It does modeling of the weather, if you will, the weather of the ocean as opposed to of the air. But it calculates a great number of things about —

Host: Is it like temperatures and currents and stuff?

Bron: Temperatures and speeds and more things than you could possibly imagine quite frankly. Dimitris would be a much better source on exactly what it does. But as he mentioned, you have this sampling problem. You don’t have sensors everywhere on the earth gathering data every minute, so you have to essentially interpolate between observations. You know it was this temperature on June 21st, you know it was this temperature on June 22nd, what was it like in between? You don’t want to just draw a straight line. That’s hardly very accurate.

Host: That’s hardly what we see in real life.

Bron: Certainly not. And certainly not over the course of a full year, right? You can’t just draw a straight line between July and July and say that was the temperature that it was.

Host: Yeah, exactly.

Bron: That is, of course, a drastic oversimplification. But the MIT GCM essentially applies all the known laws of physics, of climatology, of oceanography, of whatever you want to call it, whatever -ology you happen to like, to try to decide how you got from this point that you know about because you measured it to this other point that you know about because you measured it, and what was it like in between? So you can get a good model of the way the ocean works and what’s going on at, potentially, a very fine resolution. But in order to do that, you have this very complicated computer program, that I did not write, let me make that clear. Other people wrote that. But then you need to run it on a very large group of computers.

Host: I was going to say, I just can’t put it on my PC at home. It’s not going to work.

Bron: No. So we ran this model on typically 30 —

Dimitris: 70 thousand.

Bron: Yeah, we run it typically on 30 thousand, but the particular thing that we work with Dimitris with was 70 thousand processors simultaneously. We were trying to figure out both just how to get it to do that, and how to get it to actually run faster as a result of doing that. The part that I was particularly involved in was writing out the results. So you calculate all these numbers, but then you want to save them so that later on you can analyze them or, in our case in particular, we make movies out of them so you can see this —

Host: Oh, like animations and stuff?

Bron: Yes, and very detailed ones. We have a piece of equipment called the hyper wall, which is essentially a big array of TV screens, and a single frame, a single moment in time, is about a quarter billion pixels of imaging. We have salt concentrations, and temperatures, and velocities, an enormous amount of data that the model, MIT GCM, is producing. Just storing it all and saving it all is a much bigger task than you might think off hand. So we needed to not only produce these numbers at some relatively fast rate, but then also to store all those numbers at that same rate and not slow down the calculation.

This was a whole team of people. I’m sitting here in this chair but there is of course a whole bunch of people that were involved both in writing the code and getting it to work. Then all the support people who made the computers themselves, and so on and so on and so on.

Host: Of course. So, I was going to say, Dimitris, is this just a matter of, you give Bron, or the team, some raw data, some stuff that you do know, and then he works on that model and sample?

Dimitris: Yeah, I’ll answer that question, but first I want to go back to something that Bron said earlier and I think is a fantastic segue into explaining a little better what we do. So, a line. Bron said that our model is more complicated than the line that it is. But a line is a model, it’s a model with two parameters. And let’s say you have observations of that line and they’re all over the place, they have some noise. And then you try to adjust these two parameters, the place where it crosses the zero axis and its slope and you try to adjust these two parameters in order to fit these points as well as you can because the observations have errors, right?

In a way, and a very efficient way of doing it, a very good way of doing it, is called least squares. You try to find the line that minimizes the distance between the observations and the line in a least squares sense.

Host: Okay.

Dimitris: That is exactly what we do with satellite observations and with our model. Now, it’s a hugely more complicate problem because, as Bron said, the equations of the model are non-linear as opposed to a line is linear. There’s a lot more observations, but the degrees of freedom of the model are hugely greater than the number of observations. So it’s a so called under determined problem. We’re trying to fit a description of the large-scale ocean circulation that passes to within some distance of the observations from space, and also there are instruments in the water, floats that profile the temperature and salinity.

So, I like the fact that Bron mentioned a line and I was waiting to pick up on that. Your second question was in terms of how we operate. We have this numerical model which is called the Massachusetts Institute of Technology, it’s actually a general circulation model, so MIT GCM.

Bron: Sorry.

Dimitris: No, that’s alright. Global is good because we do a lot of global things.

Bron: I was pretty close.

Dimitris: You were pretty close. You know, we can actually run that thing on almost any platform. We can run it on our laptop, we can run it on workstations. However, to do really interesting problems where you — The way that you run this model is you break up the ocean into little boxes of water. The more of these little boxes of water you have, the more realistic your model is. You’re capturing more and more of the physics of the ocean. At some point you can’t just do it on your laptop. That’s when you go to people like Bron and many, many, many others at NASA Ames — the magicians, we call them — who show us how to scale up that problem.

That’s the first thing that they help us with, which is just on its own is unbelievable. But the second thing that happens is once you’ve run that thing, you have no idea what’s in it because there’s so many numbers. There we also need help in figuring out how to look at those numbers. So the second thing that those magicians at NASA Ames do is help us to animate, cut, look at the physics, look at processes.

You know, one of the things I have to admit that they do is find all the problems, all the bugs, all the things that are wrong with the model. When they look at it, hey what’s this, hey what’s this. Things, we had no clue. So it’s really fun to work with them.

Bron: It’s a very good point. When you visualize something, when you make a movie out of the data and then your eyes look at it. Your eyes are really good at picking out things that are bad, whereas if you were looking at pages and pages of numbers, it would be almost impossible to tell that something was amiss. Or that something was good, for that matter. I work with the people that do the visualizations although I personally don’t do the visualizations, but I work very closely with those people.

Host: And I like to grab those visualizations, turn them into a GIF, and put them online.

Bron: Yes. I will be happy to supply you with unending images I’m sure.

Host: So, I’d imagine, sometimes does it go both ways? I think of, over at Ames, the aeronautics model where they have these theories of how, and the models in the supercomputer, of how airflow works. But then sometimes you put a plane in a wind tunnel to test it, kind of check the answers in the back of the book. Is there a similar thing going on with you guys where, yes, you’re using the model to find things that, for you Dimitris, that you didn’t know before, but also I’m guessing that there’s some real data from the sensors in the ocean that then can help modify and tweak that model as well?

Dimitris: Yeah, absolutely. What I like to say when people come to me and they say, oh, you’re a modeler. I say no, and they say, oh, he’s an observationalist. You can’t use a model without observations, and you cannot use observations without a model. Basically, the way science works at a very basic level is, you look at data, you look at observations with your senses and your augmented senses. You feel things around you and then you try to explain them. And the way you explain them is you make models.

The models can be very simple. They can be a line or they can be something conceptual or something back of the envelope, or they can be very complicated. They’re never the last models. So with the models what you want to be able to do is you want to be able to reproduce the observations that you see. That’s the very first thing. You adjust, you change, you tweak your model. You change the equations, you change the boundary conditions until you can reproduce the observations to the degree that you believe the observations.

Bron: As Dimitris said, the observations themselves may have errors too, so you’ve got to be a little bit careful. You don’t want to necessarily reproduce it exactly.

Dimitris: Exactly. And then, once you have that, now you can make predictions. You can say well, given this, I expect such and such events to happen, or such and such processes. Then you can go and make focused observations to see if it’s happening. Or you can go and gather observations that you had thrown away and hadn’t used and use them to see if they support or if they disqualify, invalidate, your hypotheses. So that’s one way that models are used.

The other way, of course, is to try to better understand the physics just from a scientific curiosity perspective.

Host: Giving another shout out to another NASA center on the other side of the country over at NASA Goddard Space Flight Center. I remember when I visited there, they also had a hyper wall and they had some visualizations set up. I’m thinking this is along the same lines where it was like, they had the globe, they had Earth, and then they would dive down into their visualization and it would get into the ocean and it had all these arrows and different things. And it was just showing the different currents and the different flows. It’s the same things.

Bron: Yeah, very similar sort of thing. Exactly so.

Dimitris: Actually, some of the very nice Goddard visualizations are based on simulations that we did at NASA Ames.

Host: I’d image that they’re all shared back and forth and that all these teams —

Dimitris: Yeah, and one of the things we would really like to do, they have a very good atmospheric model at Goddard. And obviously I believe we have a very good oceanic model —

Host: With their powers combined.

Dimitris: It would be absolutely amazing to put the two together. Because some of the most important things, actually the things that make, why are we looking for oceans on other planets, on other moons? Because one of the key things that makes life possible is the presence of liquid, of ocean, to start with. But in our case, since we don’t live in the ocean, the interaction of the ocean with the atmosphere. The ocean allows climate to be moderate, meaning that it doesn’t get super-hot and super-cold. If you go to the desert, you’ll realize at night it can freeze even though in the daytime you can bake an egg, right?

The oceans kind of store heat when it’s very hot, release it when it cold. They have a moderating impact on climate. At the same time, they do the same thing for chemical quantities like carbon dioxide. Most of the carbon dioxide that we might burn through fossil fuels and put in the atmosphere eventually will be absorbed by the ocean. The ocean is helping the atmosphere from really exploding in greenhouse gasses, for example.

There’s many other examples. Therefore, what you really want to understand very well is the exchange of properties between the atmosphere and the ocean. Therefore, if we were able to put those two models together at very high resolution to make them realistic, you would gain a better understanding of how things are transferred from one fluid, the atmosphere, to the other, the ocean.

Host: My mind immediately goes into the practical application. If I was talking to my family in Ohio, explaining, oh, this is so cool. My brain first goes to weather patterns, like hurricanes. Understanding the ocean flow, understanding the atmospheric flow, and computing this craziness and to understand it. Are there realistic applications in that way?

Bron: It’s not quite the same thing as predicting where a hurricane is going to make landfall. This is much more retrospective about, you take already existing data and try to munge it and try to understand. The application really is to gain deeper understanding of how these processes work. Hopefully you’ll be able to use that make predictions, but at the very least, to be able to understand how and why things are occurring the way they are.

So, a lot of the data that we worked on was actually gathered several years ago.

Host: Oh, really?

Bron: It’s not like last month, but we’re trying to use that to gain an increased understanding of the physics of the model. To refine the model. You know, a straight line is not so good, maybe a curve isn’t so good, maybe it’s got to be really squiggly. Whatever that model might be, how things behave, we want to refine the understanding of that. So it’s somewhat more theoretical than, you know, is it going to be raining tomorrow? That’s not really the kind of questions that we’re trying to answer. But it is sort of more fundamental science about how and why do these things work.

Dimitris: Bron is absolutely correct that our specific investigations are more theoretical. They are nevertheless important for weather patterns, eventually, in the sense that if you want to predict hurricanes and where they’ll make landfall and whether they’ll grow or they won’t grow, you need to have a good understanding of air/sea interaction, and of mixed layer depth, for example. The amount of warm water that’s stored near the surface of the ocean.

One way that I think of our work is a model, a numerical model, is a reservoir of knowledge. So, as you learn more and more about processes, you adjust things, change things in the model to make it a better representation of reality. Then these models in turn can be taken by more operational agencies, like NOAA for example, and used for very practical applications. I would say the most practical applications that we work on are not at the edge. The kind of model we’re developing now will be used for practical applications maybe in 10 to 15 years.

Right now really we’re pushing the envelope, we’re exploring what’s possible. We’re learning. Ten years ago, or even 25 years ago, we were also pushing the envelope, but with models that now are really easy to run because of the increased computational power. So the models that we’re actually using in quasi-operational capacity as part of one of the projects that I’m involved with are models that were cutting edge 15 or 20 years ago. So there is this progression where you improve the model and then you start using it for more practical applications.

Bron: There are certainly plenty of analogies one could paint. If you say the wind tunnels, if you’re doing, shall we say, fundamental research in aerodynamics, do you want to know about turbulence, do you want to know about streamlining? That’s not the same thing as designing a car that gets good gas mileage. But eventually you hope that because you did all these experiments to gain increased understanding of the fundamental principles behind it, eventually that knowledge will get incorporated into, as you say, more practical, every day applications.

So, no you’re not going to see the results of the stuff that we work on on your local weather channel next week, but it is still a very important investigation.

Host: Talk a little bit about what you guys see in the future. Looking five years, ten years from now, what are you guys going to be sitting around working on? What numbers are you going to be crunching? Or where would you like to see things go I guess?

Bron: I’d like to be retired, myself.

Dimitris: We’re not going to let you retire, Bron. You’re too good.

Bron: As soon as my kids graduate from college then I’ll think about retiring. Until my kids graduate and my mortgage is paid, I think I’m kind of stuck.

Dimitris: Come on, you like working with us.

Bron: Yes, you’re right. I do. It gets me out of bed in the morning. Or sometimes in the afternoon. Right now I really see the thing that Dimitris mentioned, which is trying to couple this to other pieces. This is an ocean model and it’s very large. We’re doing whole earth simulations, not just —

Host: Yeah, this isn’t small scale. We’re doing the big things.

Bron: The big thing, yeah. And as Dimitris said, you do this by essentially cutting the ocean up into little boxes and studying the boxes. Right now the boxes are about a kilometer on a side, which, when you’re talking about the whole Earth, that’s a lot of boxes.

Host: I was going to say.

Bron: We just recently did a simulation where the boxes were 250 meters on a side.

Dimitris: And 25?

Bron: Oh, that’s right. But that one didn’t work for some reason, right?

Dimitris: No, it did. It did.

Bron: But in any event, it’s not over the whole Earth, just over a small portion. So increasing the resolution, more processors, better resolving of all of these factors, that’s certainly a place. That’s sort of a quantitative difference rather than a qualitative one. It’s the coupling of it with atmospheric models, or with ice and so forth. Dimitris is heavy into ice.

Dimitris: Ice is nice.

Bron: That, I think, is the direction that we want to — the thing that will be new and interesting, if you will. So Dimitris, as long as I have you here on the phone, could explain to me what the difference between coupling with a Goddard model and, say, the MM5 or six, or whatever their up to now at NOAA is? Because that’s coupling with land, water, and everything.

Host: This is a good way to get him to answer that email you sent.

Bron: Yeah, it is.

Dimitris: No, there’s no difference really. I want to beg to differ on one point. As you increase resolution, things change. If you want to think of the ocean not in space time, but in frequency wave number, and those are big words that — Frequency has to do with wavelengths in time, and wavenumber is wavelength in space. You can draw bubbles, if you like. Bubbles for different processes that occupy different lengths and timescales. With one kilometer, what we’re capturing very well is what’s called geostrophic eddies.

They are motions that feel the rotation of the Earth. In the atmosphere these would be the storm systems, if you like, which have thousand-kilometer scale. In the ocean, because the fluid has a different density, and also the stratification. The fluid and also the stratification of the density from the surface to the bottom of the ocean, the scales are much smaller. They range from like ten to a hundred kilometers. The scales that feel the rotation of the Earth, that it.

With a one kilometer grid we’re capturing those incredibly well, which is very nice. Because before that, we had to create Band-Aids because we could not really resolve these in our models, we had these Band-Aids, they’re called parametrizations that would try to approximate how these things would work if there’s a lot of them. But these so-called parametrizations, they just don’t do justice to the complexity of the circulation of the ocean. So now with one kilometer we’re capturing these features, but then there’s other bubbles that we’re not capturing.

There’s something called sub-mesoscale processes. There’s something called internal waves. We’re starting to touch on those, we’re starting to see them in the same way that ten years ago, we could start to see eddies in our simulations, but we were not fully resolving them. So we were kind of in this no man’s land where, should we be representing them, or parameterizing them, or should we trust these crude representations in the model are useful?

Now, there are a bunch of processes that we’re not resolving and that we are still representing in a crude way in the model. So as you increase resolution, it’s not just more of the same. There’s different processes that kick in. So that’s kind of really fun and instructive.

Bron: Where are you going to go from here? I agree with you, I’m reminded of a maxim of computer science that a factor ten in quantity, is a change in quality.

Dimitris: Exactly.

Bron: When something is ten times bigger, things are different in some sense.

Host: No judgement, but different. Better, smaller, better worse.

Bron: When your computer is suddenly ten times faster than it used to be, it’s not just that you can do the old things ten times faster. You can now suddenly do new things that you couldn’t have done before. So in the same way, I think Dimitris is saying, it’s not just that you can see the same old things better, but there are these new things that you didn’t even know were there. Or that you knew were there but couldn’t see before, but now suddenly your magnifying glass is ten times more powerful than it used to be, and you can actually see these processes. So, that’s a very good point, and thanks to Dimitris for correcting my off-hand comment.

Dimitris: Like I said earlier, a model makes predictions. So one of the things we’re super interested in is, we’re going to make some predictions, and NASA is actually launching a very nice satellite in 2020, or 2021, that’s called surface wave ocean topography. We’re going to make some predictions and that satellite is going to tell us whether our predictions are correct, and also allow us to change the model in order to better represent what the observations see.

In terms of quasi-practical applications, a couple of things that I’m really interested in and I’m involved with is application of these simulations to study interaction of the ocean with ice. And when I say ice I mean both sea ice, which is ice that forms when the air is very cold, which is formed from ocean water, and floats, and cracks, and it’s actually really beautiful both in the real world and also in the simulations. That sea ice is important because, think of it as a piece of Styrofoam on top of the ocean. Where that sea ice is, it inhibits exchange between the atmosphere and the ocean.

When you remove it, you start exchanging things, and that’s important to know for many processes that have to do, for example, with regulation of the weather patterns, and of how warm or cold the atmosphere is. But also in terms of biology. As soon as you remove the sea ice, some biology that wasn’t there can start to grow. In terms of uptake of carbon, sea ice is important for that.

A second type of ice we’re very interested in is land ice. That is ice that is formed by accumulation from snow. If you have a region where the amount of snow that falls every year is a little bit more than the amount of snow that melts every year, you form what’s called glaciers, or ice sheets, and these ice sheets are covering, for example, Greenland and Antarctica, and they’re on land. If these were to melt and to return to the ocean, or if they were — You know, we assume that they’re in some sort of steady state and the amount of snow that falls on them every year is about the same as the amount of ice that melts at the edges.

That’s good. That means the sea level won’t change. If they start melting a little faster, well we care about that because it means sea level will rise and we need to know about it so we can take action in terms of protective coastal environments from erosion and other things. So, I think that, and also the interaction of the ocean currents with biology, ecology, and carbon cycle. Those are some of the things that really interest me.

Host: So everybody should stay tuned for more to come, especially in 2020, as the work gets further complicated, and Bron here is trying to kick his kids out of the house.

Bron: They’re just going to come back.

Host: So for folks that are listening that want to get more information, we’re on Twitter @NASAAmes. We use the hashtag #NASASiliconValley. Until we change the podcast name to NASA California, that’s what we’re using in the meantime. But thanks a lot Bron for coming on over. And Dimitris, this has been awesome, thanks for calling in from beautiful Pasadena.

Dimitris: Thank you very much.

Host: All right, thanks a lot guys.

Bron: Sure thing.

[END]