 I decided what to sort of include in this. I had to sort of deal with the fact that our group now works on two very similar, but obviously dissimilar navigation systems. We work on moth orientation to pheromones, but we also work on mosquitoes. And mosquitoes have very different problems to solve and a lot of very interesting cues to use besides odors and odor plumes. So for today, I'm just going to concentrate on moths and the lessons that we might learn in keeping with protocol. I changed the title of the slide slightly because there are two features to think about, one of which is finding the odor plume and the other is navigating along it to its source. And it's really the finding the odor plume is just as important that you do that well and you do it quickly as opposed to just being able to navigate to its source. So let me also reverse orders here in a way and thank many folks who have helped me through the years to understand these processes and a lot of other things that we work on in our lab, particularly for today's talk, Pep Bow, who's in Catalonia, worked with me on modeling Joe Elkinton, who was my former colleague at Massachusetts, who worked on plume dispersion and flight behavior. Also, John Merlis from England, who's an atmospheric physicist in a sense and has certainly enabled me to understand a little bit about what goes on in the atmosphere. And then finally, Anjanar Mofranetto was a graduate student of mine several years ago. And we got into how the fine scale structures of plumes affects behavior. Now, it's pretty clear that in moths with about 140,000, maybe 150,000 species, that they all use pheromones and that almost all of them will use a female-admitted pheromone for the male to find the female. So it's thought that this is a process under a lot of selective pressure. Michael Greenfield called this a race to find the female, because the first male there is very likely to be the one to mate. So you have to find the plume and you have to navigate efficiently. Much of what I've done has been with the gypsy moth. And as Dave has mentioned, some of these insects have very plumose antennae. And this is almost like a basket that receives the odor plume. And you can see the female is rather different in terms of its visage here. Strangely enough, the male doesn't seem to make any use of the female's visual presence. And everything is odor guided in his location of the female. Now, I got interested in this while I was a chemical ecologist working actually as a postdoc in Wendell Roloff's laboratory in the early 70s. When John Kennedy, one of my estimations, one of the leading insect behaviors, started to question the way a lot of people conceptualized how this process worked. And part of it is John was very fond of precise terms. And he really disliked the term attractive because it implied that the chemical itself was inducing the actual response and navigation. And so he got into some very interesting debates with Harry Shory, who was a colleague of not a mind at the time at Riverside, about whether or not insects were using the structure of the odor plume or wind and the presence of the odor to locate a source. And along the way, there's no quiz at the end of this. This is sort of like one of those DNA charts that you can't read. There are just a vast number of terms that have been concocted to try and understand this process. And the problem with a lot of these terms is that they're not based on this. Sometimes they're based on the body orientation. Sometimes they're based on the stimulus. It's possible to avoid all of these terms and just describe the behavior of the stimulus if you think you know what the stimulus is. But I'm just going to concentrate on a few terms here. Taxes is, of course, when the body is oriented in some fashion aligned with the stimulus. Chemotaxis is something that has been invoked a lot, that is using concentration gradients, perhaps, to find the source. We think this occurs very close to the source. But as I'll describe the plume later on, it doesn't seem to contain maybe the right kind of information to permit chemotaxis at a distance. Anemotaxis means that upwind orientation has been initiated. Optomotor anemotaxis refers to the way in which that information about the wind direction enables navigation. I'll come back to that one. Tropotaxis, we're not going to talk about, but that's where you compare right and left inputs. That can occur very close to the source. There is really no good evidence that that's occurring at some distance from the source. For the same reason, I'm going to talk about in a minute with the structure of the plume. And then finally, thanks to one of the organizers here, Massimo, we have a concept called infotaxis. And this avoids a lot of the difficulties in identifying in the term what the actual stimulus or information is. It just says that it uses what's available. It is not necessarily explicitly trying to replicate what the insect is doing. It might or it might not. So John Kennedy had a very little ingenious wind tunnel. He did this in the late 1930s. He used the mosquito 80s, a gypti. He would breathe into this wind tunnel, producing carbon dioxide, of course. And the female mosquito would fly upwind. But what he showed is that you could project a floor pattern on the bottom of the tunnel and that that was how the insect was responding to movement. It was gauging its movement relative to what it could see below. And so upwind heading, in this case, is simply gauged by the fact that the flow of the visual field is front to rear as you look down. And transverse image flow, in other words, if it's you're not headed directly upwind, you also have some side slippage. We call that transverse image flow. So when that kicks in, an insect realizes it's not heading directly upwind. And it can correct its course toward due upwind. And therefore, to follow a plume, there is no need to monitor changes in odor concentration or gradients. Although insects do generally assume a not always perfect straight upwind course, but there is some zigzagging as it moves along. And the reasons for that will become clearer later on. So this is a concept that is not intuitive to us because we're anchored on the ground. And when we feel the air coming toward us, we have very instinctively know which way it's going by tactile sense. But when you're immersed in the medium, it's a very, very different situation. And it's hard to conceive of this as an armchair scientist. Now here's sort of the overview of some of the things I'm going to talk about today. So this is a hypothetical moth. And one of the important things that it needs to do is to have a ranging flight or a searching flight that enables it to contact the plume. And this needs to be timed to match when the female is producing pheromone. And it also needs to be effective in the sense that, remember, that male is competing presumably with other males to find the female. When it encounters a plume of odor, it begins to turn upwind. There may be some zigzag movements. And if you notice here, this moth has managed to find its way out of the plume. How did that happen? Well, besides having this turbulent diffusion process, the wind is changing direction. We're looking down on this plume. And it's changing direction. The best way I can offer an analogy of that is if I could have a nice hose of water here, turn it on, and spray it back and forth across the room, you would have this path that's zigzag, or that circulate its path. But the actual trajectory of the water would be pretty much in a straight line. So as this insect is coming up to this point here, where it sort of reaches a spot where this is due upwind and it now has lost the plume, so what does it do? Moths cast. That is, they repetitively go back and forth without much upwind progress. And they have a sort of a giving up time, presumably. And if they don't contact the pheromone, they go back to ranging flight. As they approach the source, there may be much more straight upwind flight. And this is occurring because of the way in which the internal structure of the plume is changing. So here is sort of a fake plume. This is a visual plume in a forest. And it gives you a sense of just how chaotic this plume actually is. So this is not very far. This is over just a couple of meters, fairly still air. You can see all these wispy orders. What happens is that turbulent diffusion, of course, tears the plume apart. But there are also little pockets that persist of fairly high concentration. And these get carried many meters downwind. And this partly explains why distance of communication can be fairly substantial. So besides that, what you see there, I did mention this notion of wind shift. And this has been actually one of the more interesting things that we've looked at. It was occasioned by a paper published in Nature by Charles David and John Kennedy a few years earlier. And it's supposed that the wind, in fact, is constantly shifting its direction. And that influences how the plume is dispersed, the overall envelope. Now, here we are in a forest. And the way in which we would follow these plumes, we'd sometimes use puffs of smoke. In this case, we're using a neutrally buoyant balloon. And you can follow this for fairly long distances in the forest. And we use very high-tech methods. As this plume or bubble was going through the forest, we'd have little bamboo stakes that we put in the ground that signify where it was in time. And then we would go back and measure it. And this is fairly typical of what we would see in the forest. This is the point of origin, dead center. And there is one trajectory here that's not too bad from the viewpoint of straight lines. But most of them end up sort of being very circuitous. Here's a great one for a straight line. So you can see the problem that ensues when an insect is trying to fly upwind. And the plume is constantly changing its angle where it goes. Remember, this is only 15 meters here. And lest you think that that's sort of idiosyncratic to that forest, John Brady looked at the same problem. He was, I guess, inspired or maybe thought we were wrong. And he looked at this in the savannah of Zimbabwe. And he found exactly the same thing. Patterns are just precisely the same. And that's a fairly open kind of plant area. So it seems to be something that's fairly routine. There are cases where there are relatively long fetches. And that tends to occur more in open fields, places where you don't have obstructions and you don't have a canopy up high. So this, I guess, reflects the fact that when someone like an entomologist or a behaviorist gets involved in meteorology, that we really don't understand what we're doing. Because we had assumed when we were actually trying to understand this whole process that if you look down on the way the plume is moving in nature, you would assume that this is what it was. That as it was snaking along, the upwind direction would follow the long axis of the plume. But in fact, this is what happens when the wind holds at a steady velocity. This is the analogy I gave you using a water hose. And you can see that in those cases, you end up going right out of the plume if you head upwind. And if the wind changes velocity, then these can get bent around and make it even harder to find the source. So the way in which we looked at this experimentally in the forest was to use gypsy moths. And what we would do is we would take a rack of gypsy moths, 30 gypsy moths, and place them in stands here. And we would have sort of a tree here, tree-like stove pipes, actually, with pheromone source here. And what we would want to understand is how far away could the plume be detected and how good a navigation system as a gypsy moth have for finding that odor source. So this is an example here of actually two panels. Here is the rack of gypsy moths, all set in here. And there's a fast-talking undergraduate student here with a little microphone and a recorder indicating how quickly the actual wing fanning ensues. Because this particular moth wing fans for a little bit before it takes off, raises its body temperature so that it can fly. So they're quiescent, they begin wing fanning. And when they wing fan, they are let loose. And they're marked at each of these distances so that we know exactly which rack they came from. And then down here, where the pheromone is, we have our trusty moth catchers with a butterfly net. Remember, these are day-flying moths, which the only way we could have ever done this is with a day-flying moth. And they catch them here and record where they came from in the time. The only problem that you have with this is one that you might have imagined. If you see all these strings radiating out, so we have to guess a hope that we can figure out which way the wind is going to be blowing when we set this experiment up. Sometimes we set it up and we never figure out which way the wind is going to be blowing the time. You can look up in the canopy, through the canopy, see the clouds going by. But in the forest, the winds may be coming from all directions. So it's an interesting methodological problem. So how successful are we? Well, let's say if you look at distance here and you look at percent departing, which means that they started wing fanning, clearly they sense pheromone and would take off, that it's very high, even up to 80 meters. It only drops off somewhat at 120 meters. But quickly, you begin to see that the percent arriving drops off. We waited quite a time. And the transit time gets longer, as does the minimum. Now, they could fly that 20 meters in less than, well, under a minute, very, very, very fast flyers. So what this tells us is that the distance of detection of the plume is much, much farther than the math's ability to navigate. Why do we think that that navigation ability is impaired? It's because we think of the turbulent diffusion and the shifts in wind direction, which means that you're catching little packs of odor, filaments of odor. But you're not necessarily able to find. Sure. What about the letters in order? Oh, those are statistical significance. So if you have statistical significance here, that would mean that these fall into the same bin at the 5% level, and these are different. So obviously, in order to interpret whether or not we have any, whether these are just random pieces of bad information, we always subject things to statistical analysis. Sorry. Thanks for asking. So in our view, with the gypsy moth, what limits the distance of communication is not how much pheromone she admits, and possibly not how low the threshold of response is, although clearly those are very important parameters. But the changes in wind direction and turbulence seem to be important. OK, let's get back to the fine scale features of the plume. Back in the forest, we used a system which is actually, oh, I think I might have misplaced that slide. It'll probably pop up in the wrong spot. We used an ion system to actually measure fine scale structure. Open field looks very different. There are two timescales here. This is one second. This is five seconds. And in the forest, what you see is that if the wind is blowing and it doesn't change direction, you have a very spiky, very intermittent signal. And then you have long periods when, perhaps, you'd have no signal at all. Open field, again, slight changes in direction. You get this. These are moderate distances, like 10 meters away from the source. So in the wind tunnel, we can begin to understand and manipulate these features. And we use a stimulus generator, which we can set to have little puffs of odor, which, again, you see represented here by hydrochloric acid from titanium tetrachloride that we use as a visual marker. And you can see that you can create very, very definitive puffs. And we work with this particular insect here, another simple little moth, very simple pheromone communication system. And one of the things you can do is you can give it a stream of pheromone. It flies up wind. And then you can turn it off. What does it do? It begins to cast. And when that happens, you see it right here. You're looking down at the path of the moth. Wind is coming this way. Now the moth is casting. And at some point here, probably about 200 milliseconds before this actual turn, we give it a puff of odor. And then it shoots, acceleration shoots straight up wind. But it does not encounter another puff of odor, so it goes back to casting. The length of this is a little bit dependent or can be varied by the concentration of odor in the puff and how big that puff is. But you have that same process. Now the idea is that in this turbulent world where you have these puffs of odors, you're continually contacting these individual filaments. And if you contact them frequently enough, then you head up wind. So we also, and I won't go into this, we've measured a lot of these features of plumes using a photo ionization detector system. We also know that the antenna of at least five months that we've looked at are capable of resolving these puffs up to 25 hertz. So at least at the peripheral level, we know that those can be easily perceived. And again, looking at tracks, this I think is somewhat helpful. What you see here is that if you have the fastest track at 25 hertz, 17 hertz, it's pretty much straight up wind. Even at five hertz, it's not bad. However, if you look at the average track as a statistical way of computing, which is the average amongst the tracks, what you find is that you get a lot of zigzagging at five hertz, but as you start going to 10 hertz, starts to straighten out. And so this is, in fact, how we conceive this works. And you can analyze these tracks on video as we've done many, many different ways. You can decompose them. This is in two dimensions. You have the vectors of wind speed and direction. And the important thing to realize here is that to achieve a particular track here with this wind speed, the moth must actually head this way because it's being blown back. So we can calculate all of these. We now have good computer programs. This used to be a very laborious hand on process frame by frame, but we now have very good computer programs that do this pretty much automatically and give you lovely accelerates with vast amounts of information. But in any case, again, here's what you see. You see that the ground speed accelerates when you get up to about 17 hertz of puffs per second and the track angle diminishes, meaning that it's going faster and it's heading straighter upwind. And is this a universal principle? This is what the Quadra, Tom Baker, and his group looked at this with another moth in another family, a noctuid. And they found virtually the same response to this. So this appears to be somewhat of a universal principle with moths, but in fact, there are 140, 150,000 species. We've only looked at a couple. But I think given that these two lineages are more than 100 million years apart, that's probably a pretty respectable way to think of it. So now I'm going to launch into something about finding odor plumes very quickly. Remember, it doesn't do you any good to have a race along the plume to find the female if you haven't found the plume. So there are strategies that you might imagine. You might sort of have a random strategy. You might hit upwind, downwind, and crosswind. We always thought random would be best. But some colleagues in the Netherlands proposed a theoretical model that suggested upwind, or maybe downwind would be better. And the reason they did that was based on this fact that the plume changes its angle continually because of the shifting wind direction. And if it does it by more than 60 degrees, what you find is that the crosswind area is, in fact, longer than this area here. Now, there are a lot of other assumptions that are based in this. But they propose that that should be what moths do, or any insect for that matter. But they didn't maybe account for the fact that moths are pretty stupid in the sense that they don't keep track of where they've been in space over a very long period of time and where they may or may not have encountered pheromone. So to make use of a model like this, they would have to keep track of an awful lot of information. So we looked at this. Actually, we looked at it with gypsy moth. And we made another experiment where we looked at it in an even more sophisticated way with this moth. And this moth is another one of these that's ideal for this because the male does not have functional moth parts. So the male, just like the male gypsy moth, has nothing to do but look for a calling female, pheromone-emitting female. And it, yeah. Is there a sense of this memory that you just mentioned that you're short or can you put some numbers on that here? A little bit with some of the casting. For casting, we can. But for the rest of it, I'm not sure I know how to do it. We're going to have a chat. But motorhawk is what? Pardon? The memory in time, for example, you said they don't remember that much. Well, they may remember recent encounters, but they would also need to remember often where they were if they weren't using ophthalmotor and even when you say recent is what? It's second, 10 seconds. 10 seconds, 5 seconds. That would be my seat of the pants. We can chat more about that. So all flights are meant to be in search of the female. We did this in this lovely place in Acadia National Park, where I will be in a week looking at this moth again. And there's a sonic anemometer here and a video camera. And these are the kinds of tracks that we could see. Most of them are fairly straight. There are a few weird ones that you get where they change course, but we were able to get a fair number of these. And I should maybe point out, let me go back and point out one other thing I think in here. This is a way in which we analyze directionality. It's called circular statistics. It's not circular reasoning. It's circular statistics. And this is actually the wind direction. This is a coastal bog. And so in the afternoon, the air comes in from the ocean into the bog. And that's why there's that preferred direction of airflow. And this arrow denotes the strength of statistical association. It's a very weak arrow, or there's no arrow. That means there's no statistical value in the distribution. So quickly, what do we find? If we analyze it vector by vector, meaning each 30th of a second, or we analyze the entire track and we correct for wind always being at 0, you see a totally random distribution of trajectories, either moment by moment or over the entire trajectory, meaning that this moth has no particular sense of paying attention to direction of the wind. And you actually see some other things that you can pick out of this, because the mean ground speed will vary depending on whether it's going upwind a little bit or downwind. But not so much, because the moth again uses this optomotor reaction. It uses that to set its course. So it's looking down and picking up the same rate of visual field. So you don't see very much variation in this, even though there's some variation, quite a bit of wind. So the flights are not aimed with respect to contemporaneous wind flow. We actually tried that and also found a true gypsy moth. However, there's kind of a funny de facto arrangement here, because when you go crosswind, you actually there are two crosswind sectors, one upwind and downwind. So if it's randomly distributed in all four quadrants, you actually are spending more time going crosswind. So now I'm going to move into the last area, modeling. It's pretty difficult to do this in the field. There have been a few cases where they have been able to put harmonic radar transmitters on moths and follow them for a bit. The precision of knowing exactly where that insect is in time and space is decent. It's sort of somewhere in this area here, maybe. But the problem, in part, is you don't know what it's experiencing in terms of its current wind flow. So there's a lot of things we can't yet do with minimization of tracking systems. But we can sort of explore these concepts in a virtual world. And we then can sort of apply sensitivity testing to see what parameters affect the outcome. And there, we've been able to apply critical actual flight parameters for the gypsy moth, which we know an awful lot about now because we've been working on it for so long. So that's where we sort of focus some of our work. And again, gypsy moth doesn't feed as an adult. We know from wind tunnel and video records flight speeds. We know how far away it can detect pheromone. I just showed you those data. And very importantly, we have a plume model now, a simulation plume model that my colleague Jay Ferrell and John Merlis and some others at UCR developed that enable us to simulate the plume. It's a model that gives you a sort of an instantaneous slice of odor density. It also has a wind direction change built into it so you can set the parameters. And this model is available online. It's been cited several hundred times already. And a lot of people have used it a bit in simulations. So our virtual world is a grid of traps within 100 meter inner-trap distance. And we release virtual moths in the center. And then we give them various strategies for either finding a plume and the actual way in which they navigate along the plume is pretty straightforward. But we do have to sort of say they have to find the plume. Navigation up is pretty straightforward unless you lose it, then there are various casting strategies. So we run thousands and thousands and thousands of moths through these experiments. And here's a way in which this thing is set up in a sense. You can see here that when we can't keep expanding this area computationally forever. But what we do is we end up looping an animal back into this model if it gets to the edge. And we think this is a reasonable approximation. It's a fairly large area. And the model itself, as I said, is pretty straightforward. The details here are not terribly important other than when you're in the plume and you detect a reasonable concentration of odor, you head upwind. If you lose the plume, you cast. We'll explore what you do to cast. And then you go back to the searching strategy, ranging strategy. And again, we can vary those parameters as well. So here's a little bit of experimental data from the wind tunnel showing casting behavior. Now you're looking down on the flight track of a moth heading upwind. Here's the wind going this way. Pheromone source is over here. And here's the gypsy moth navigating along the plume. In all the other cases, we shut the pheromone source off at the point where these dots become a little bit thinner. And then you can see the casting behavior ensuing. And the casting behavior in the wind tunnel seems to be fairly defined into maybe a couple of meters most, partly because if they went any farther, they'd hit the wall of my wind tunnel. So we don't really know how that affects it. But field data suggests that the casting may be a little bit wider in the field, but not a lot. So let's look at casting strategies. We actually had a whole bunch that we looked at. We have sort of three turns per second producing a very small cast, which you would expect to be ineffective, perhaps. Larger, getting up to two meters. Final sweep followed by two four meter wide ones. And then a large final sweep by six meters. So the idea is, of course, for the moth to, the virtual moth to relocate the plume. Casting behaviors last for about 20 seconds, which pretty much matches what we saw on the wind tunnel. And it's fairly straightforward. These are just some selected tracks. This is a very narrow cast. So here, this animal is flying along the plume. It loses it here. It regains it here. It comes up, it loses it. Then it goes into this casting behavior. In all of these, the strategy then is to go downwind. So that's the second part of that strategy. Here's a wide cast and then downwind. So all of these things can be varied. And casting is effective. The wide sweep seems most effective, but I'll show you a little data on that. But we don't really know that it's occurring that much in the field. For finding the plume, there are many strategies to use. They have these little lovely terms, random walk, downwind based correlated random walk, crosswind bias, pardon me, correlated random walk, intermittent. The basic idea is that you either are going totally at random direction or the direction you take is dependent on the previous direction. You sort of vary somewhat. You don't sort of pick a random direction. And another strategy that people are very fond of considering is a lay they walk. And this is a power law distribution, meaning that there are a lot of short distances interspersed with longer distances. So it sort of follows this kind of a distribution of the distance between turns. And this is very commonly thought to occur in a whole variety of birds and other organisms foraging for resources. So when you see flight tracks here, you can sort of see that very idea. So over on B, we have a lay by walk where you have lots of short ones interspersed with longer ones. So the idea is that you end up sort of sampling an area fairly well, and then you move on a longer distance away to resample another area a little more carefully. And then these other ones here sort of again carry you over greater distances in sampling areas over that period of time. And the crosswind or downwind or upwind, of course, is not probably very successful. So what you can see, I have a couple of slides here very quickly to sort of show you how these things pan out. So on the bottom is wind variability, meaning how many degrees per minute the direction is changing. So here it's not changing very much. It's 13 degrees. Here it's changing a lot, 65 degrees. And so this is the finding the plume. And what you see is that the downwind correlated behavior doesn't do very well. Most of the others do pretty well. And the lay by walk does very well. So there's certainly a big difference in picking a strategy here if you're going to find a plume. And then this complicated slide sort of puts everything in one big perspective. It tries to explain this from the viewpoint of the walk and whether you actually contact the plume in the virtual world. And then whether or not having contacted this plume, you actually find the source. And then it plays it out at different distances from where you actually contact the source initially. And again, you can see some things don't work very well. The correlated downwind walk is very, very poor. But most of the others do fairly well. And you can see that, say the lay by walk is actually one of the best performers here when the wind is shifting at 13 degrees. And it does fairly well, again, when it's shifting at 65 degrees. So we think we can understand how to program a virtual moth to find a virtual plume. It's going to be a little more difficult to sort of imagine how we're going to use that in the field to understand things. So very quickly, because I'm running into question time soon and I stand between you and coffee. I don't want to do that. So here the crosswind and downwind strategies are not highly effective strategies for finding a plume and location. Lay by walk and other random walk permutations are all effective. And m, because of the way in which crosswind has got twice the area of upwind and downwind in terms of the angles, it ends up being an effective system. So this is what we expect to find, perhaps, in nature. And we have used this now to understand, and I don't show you these simulations, but one of the major uses and one of the reasons I work with pheromones is it's very useful in insect control, either for mating disruption, direct control, but also surveillance of invasive aliens. And the real problem is understanding how many traps to put out in a landscape so that you don't fail to detect an insect that's there. And so we are now using this method to sort of say, what are the effective distances with traps? And we have some very, very good data from Mark's release recaptured so that we can sort of make sure that this is a true story at the end. The last part is to sort of go back to something. We'll leave the air and go down into the water. Jay Farrell and I did receive funding for our work based not so much on aerial plume following, but underwater plume following. And there's been a long history of this in the Navy to look for, for example, unexploded ordnance. You might use divers, marine animals, towing bodies. And now what they want to use is some of these submersibles. They sort of look like a torpedo. And they would launch these into the water and they would attempt to find a source. And what you see here is fluorescein, which because we don't have, at the time these experiments were done, we didn't have any sensors that were capable of picking up, for example, unexploded ordnance and detecting it in the marine environment at the levels it would be found. So what they would use is fluorescein and use an optical detector to say whether or not we found the source of the pollutants, for example. Could be oil. It could be whatever was coming and you had to find. So Jay has published quite a bit on this. There's quite a bit on his website. I'll just show you one example of his work in collaboration with the Navy. And they ran these out in San Diego Harbor. And what you see here are tracks. And the track is presumably starts here and you find that there's a detection of the odor, but then there was never any further detection of the odor fluorescein in this case. This is a boundary that was prescribed by the Navy to search. And these are miles. This is not a small area. So it's now searching in this area and comes back. And here it picks up the odor plume or the fluorescein and it makes its way here. And eventually it gets to this spot and they have a little routine for saying the odor source is found. So this actually works. Obviously this is not an optimal reaction. So how did they do this? Well, they just used GPS. So the little submersible could keep track of where it was by GPS. But it's the same idea. Now you can program a lot more into that if you wish because now you have that information and getting back to infotaxis, you could program into that where you detected it and what the probability of finding it at some other spot would be. So I think I'm more or less on time. Thank you for your attention. And I hope I haven't gone through too much too quickly.