WEBVTT 1 00:00:06.000 --> 00:00:27.000 Welcome to the GeoMob podcast, where we discuss geo innovation in any and all forms, whether for fun or profit. Hello, everyone, and welcome Back to the GeoMob podcast. Today we're going to be talking about bat distribution modeling with our guest, Tom Foxley from Bath University. Hi there, Tom. How are you doing? 2 00:00:27.000 --> 00:00:29.000 Good, how are you? 3 00:00:29.000 --> 00:00:41.000 Yep, all good, thank you. Looking forward to this. So maybe we can start with you introducing yourself to the listeners a little bit about your background and sort of how you got into bats. 4 00:00:41.000 --> 00:01:50.000 Right, yeah, sure. So, yeah, so I'm an ecologist, conservation biologist. I work at the University of Bath. I'm currently doing a postdoc at the University of Bath. Well, I've worked with bats since 2011 was my first bat job was a radio tracking job for Bristol Uni, and then after that I worked as a consultant ecologist for about 10 years, specializing, I say, I was a consultant, I didn't really do any consulting, I was just doing the field work, the fun bits. I was a freelancer, so I did a lot of radio tracking and tree climbing, looking for bats in trees, stuff like that. And then in 2020, I started my PhD, which is using spatial approaches for bat conservation to develop a better approach for protecting bats in the planning system. Was kind of the focus of my PhD. 5 00:01:50.000 --> 00:01:51.000 Okay, cool. 6 00:01:51.000 --> 00:01:56.000 Now, I'm sticking more with the research side of things now, 7 00:01:56.000 --> 00:02:44.000 and so I'm guessing it was mainly the results of your PhD that I've seen that made me think that you'd be really good to come on the podcast and talk about some of the spatial stuff. But maybe you're carrying on that type of work as well in what you're doing now. So maybe there'll be some opportunity to talk about sort of the latest research that you're doing. But before we get into all of that, maybe we can sort of have a little bit of a chat about why bats are interesting and why they're important. So maybe keeping it to the uk, but sort of where are they found? What are some of the standard tools for monitoring bats and sort of like, why should we be monitoring them? 8 00:02:44.000 --> 00:04:01.000 Well, so I don't think you can restrict it to the UK bats. That's one of the most amazing things about bats. So obviously. But they're just cool because they're mammals that can fly. They use their hands to fly, which is just amazing in itself. And that's enabled them to spread around the world and fill all these amazing niches that they do in other parts of the world where they eat, drink blood you know, eat frogs and just. Yeah, that's part of what's so amazing about that. But so if we're restricting it to the uk, bats are all over the uk. They're all insectivorous in the UK and so they, they're an important part of the ecosystems that we have in the uk, where so most of the UK is incredibly degraded in terms of wildlife. But bats have managed to hang on and they're still fulfilling this really important ecological role. From our perspective, they say they eat a lot of crop pests, is something like an ecosystem service that they provide. From a human centric perspective, that's a reason to protect them, if not just for their intrinsic value. 9 00:04:01.000 --> 00:04:14.000 So are there loads of different species of bat in the UK or is it a relatively small number of species and are they evenly distributed across the country? 10 00:04:14.000 --> 00:04:42.000 So there's 18 species, well, 18 species that are present in the UK, 17 that breed in the UK and no, obviously there's great diversity in the south, particularly in the southwest. You get more species and number of species drops the further north you go generally. Okay, I'm not sure how many you have in Scotland, but it's more like nine or nine to ten. 11 00:04:42.000 --> 00:05:14.000 Yeah. Okay, so. And you've already, you've already mentioned in your introduction as well that part of your previous life was sort of looking in trees for bats and, and radio tracking of bats. Can you say a little bit about those? Sort of. I'm going to show my ignorance here, but those traditional inverted commas methods of sort of looking for bats, like are those the only two or are there more? And what does radio tracking invol? 12 00:05:14.000 --> 00:06:19.000 So radio tracking, obviously. So radio tracking, you trap, you put out a net to trap bats and you glue a tiny little radio transmitter on its back and then set it free and then follow it around for the next. They tend to last about a week or so. So yeah, it's an amazing project. You do get amazing data from it. There's obviously, if you've got a tiny, fast moving animal moving around at night and you're trying to triangulate to get its position, which is what you end up doing, there's obviously some inaccuracy in that. So it's good, but it's limited when you compare it with, with like modern, modern. The latest technology is to use GPS, which is actually what I was doing in my PhD, I was using GPS rather than radio tracking, which is, which is obviously much more, much more accurate. 13 00:06:19.000 --> 00:06:28.000 Hang on. Sorry, sorry to interrupt. Am I right in thinking you put GPS trackers onto the back of a bat? 14 00:06:28.000 --> 00:06:31.000 Yeah, yeah, yeah, yeah, yeah. Tiny. 15 00:06:31.000 --> 00:06:33.000 Okay. 16 00:06:33.000 --> 00:07:35.000 About one gram. I think there are some that are a bit smaller than a gram now, but at the time, one gram was the smallest and, and that obviously limits the size of the bat you can attach it to. So it's got to be the weight of the tag has got to be less than 5% of the, of the weight of the bat. So you can only do it with larger UK species because the smaller ones wouldn't be able to carry it. It would affect them too much. So there's still a place for radio tracking because they haven't. Until they develop tags that are smaller. So the other thing about the GPS tag is that you have to retrieve the. You have to retrieve it. The ones that I had, you have to retrieve to be able to get the data from it. So it has to be a species that you know where it's going to go, like. So there are some species that switch roost every night, so you're not guaranteed to get your tags back. And when they cost about, I don't know, four or five hundred quid each. So it's, it's. If you're tagging loads of it can be very expensive. Yeah. 17 00:07:35.000 --> 00:07:39.000 Yes. Okay. 18 00:07:39.000 --> 00:09:19.000 But, yeah, so radio tracking was. So when I was, when I had these, these GPS tags on the bats, I actually also had a radio tag glued to it. And the first year I was doing it, I was out tracking the bats just in case because I didn't know how well the GPS was going to work. And it also helps you if the tag falls off somewhere. The idea was then you could retrieve it using, using the radio tag. But when we, at the end of the season, when I compared the radio tracking data that I got from my tracking and the GPS data, you can really see the difference. The radio tracking data is a lot more. There's a lot of fixes near the roost and then as you go further out from the roost, which is where the bats were roosting, you know, the data drops off because you just have, have trouble finding them. So you get this very kind of biased data and you get good data for a few bats and then some bats, you just have no idea where they went because they had one bat that was. They're trying to track it for about five nights we were trying to work out where it was going and it was, it would fly. It start off going to the north and then, then it would just vanish. And luckily we got the GPS tag back for that and found that what it was doing was turning around and going for 14 km in the other direction, down to. Down to Tu Valley Lake, which is miles from where we started. Yeah. Right. So, yeah, I think with radio tracking, you miss a lot of the data. Those bats that disappear off miles away, you're just much less likely to encounter them. So you get this bias in the data. 19 00:09:19.000 --> 00:09:20.000 Yeah. 20 00:09:19.000 --> 00:09:20.000 So, yes. 21 00:09:20.000 --> 00:09:40.000 So some of them are going quite large distances in the night, then. So if you're looking to do spatial modeling of the sort of distribution, I guess, of where they go when they're foraging, you're going to need. You're going to need to cover a large area. So. 22 00:09:40.000 --> 00:10:25.000 Yeah, yeah, yeah. So we were finding these are greater horseshoe bats, which are. So they're one of the largest species, so they're able to travel quite far. They were going kind of up to 14 km away from the roost every night. Yeah. So it's quite big area. Yeah. This is. I think it's dependent upon the number of bats that are at the roost. So this is a maternity colony in North Somerset and at the time there was. There was six, six or eight, between six and 800 bats at the roost. So I think if there's that many bats, they've got to. Obviously, you know, everything close to the roost, all of the food, the foraging resources get depleted. So I think that means they have to travel further, further to get to find somewhere, you know, to forage. So it's dependent. 23 00:10:25.000 --> 00:10:54.000 Yeah, yeah, yeah. That's amazing. So the next thing I think we should probably discuss is, I've seen some of the results of the spatial modeling that you've done for North Somerset and the sort of area around the Mendip Hills. Could you explain a little bit about your models and how you built them and sort of like what the driver was for doing that work? 24 00:10:54.000 --> 00:14:10.000 Yeah. So I'll start with the rationale. I think it makes for the modeling. So all bats are protected in the uk, but when I say the protective, for most species, it means that their roots are protected. So you're not allowed to. And the individual bats themselves, you're not allowed to disturb them when they're in their roosts or destroy their roost. But then for some of the rarest species, there's four species have this extra level of protection. Greater horseshoes and lesser horseshoes. Being two of those there, you have to have. You have these special areas of conservation set up to protect. To protect them, to protect their species. And so what that Means is so you'll have. The actual sac itself is quite small, so it'll be the roost that they're using. So in North Somerset it's a maternity roost. A single maternity roost is the actual ssc. But then the habitat around the roost is in effect protected because you're not allowed to do anything that can negatively impact the bats roosting in the SoC. So then suddenly you've got to protect the habitat that the bats are using, which if you've got 600 Greater Horseshoe bats roosting in, in the center of your county, which is what they have in North Somerset, and these bats can go up to 14km every night, that means basically everywhere in North Somerset these bats is being used by these bats. And so anytime anyone wants to build anything, it has to go through a habitat regulations assessment. And it's a big headache for the developers and for the council. It's just a loss of admin and it makes, makes things difficult. And then it's also quite hard to understand or to know how a development might be impacting them if you don't understand how the bats are using the landscape. So this is where this is. So my supervisor, Emma Stone, she developed the idea for my PhD in collaboration with North Somerset Council and Natural England to try to develop, you know, a better approach to dealing with that. And so yeah, so that was, that was kind of the rationale behind it. So to give the council or local decision makers this amazing understanding of how bats are using the landscape so that they can make better informed decision development decisions. So the main threat, so development, the main threat that development poses to bats is through habitat loss and through loss of landscape connectivity, which essentially can result in the same thing. If the bats aren't able to access the habitat through a lack of connectivity, then it's the same kind of impact. So those were the two modeling approaches I took to, to understand how the bats are using the landscape. 25 00:14:10.000 --> 00:14:23.000 Sorry, I was just going to say is connectivity really important for bats given that they're relatively small? Are they using sort of linear features in the landscape in order to get around the landscape? Is that key? 26 00:14:23.000 --> 00:15:52.000 Yeah, yeah. Some species will than others. So horseshoe bats particularly, they've got, they use very high frequency echolocation which so high frequencies attenuate quite quickly in air. So they don't really, if they're record locating, they don't really have a very kind of short sighted, I suppose is one way of saying it. So they're known to use linear features as navigation routes and the other thing is that they're kind of a clutter specialist. They're adapted to flying clutter. So it's a predation, it's security. It stops them being predated. If they can dive into a bush quickly, it makes them a bit safer when they're flying around as well. So, yeah, they definitely do depend. There are other species that fly really high and will just fly over everything. So it's species specific, the amount that, like, linear features and stuff are used. Yeah, so. Right. So the 2 model, the 2 types of models that I built, so they were habitat suitability models and landscape connectivity models. So your listeners have probably heard of the set of these. Like, these people are aware of it. And the package that I used were Maxent for habitat suitability modeling and circuitscape for landscape connectivity models. 27 00:15:52.000 --> 00:15:56.000 Both of those. NR. Yeah, 28 00:15:56.000 --> 00:16:04.000 the MaxEnt is in R. You can do Maxent in our. But circuitscape is in Julia. 29 00:16:04.000 --> 00:16:05.000 Oh, right. 30 00:16:04.000 --> 00:19:49.000 Okay. The latest one's in Julia, which makes it a lot quicker because that's the one I use. So, yeah, so I had this GPS tracking data. I did two summers of GPS tracking of two different maternity booths of greater horseshoe bats. And then I use that data as the input for the habitat suitability model. So the habitat suitability model, it needs species locational record. And then you take a load of environmental layers, like all the GIS layers describing the landscape, and it looks for associations between the habitat that the species are using and what the environmental conditions are at those location points. And then you use that to predict, you know, where suitable habitat will be over the whole of your study area. So I built two different habitat suitability models, one based on one of foraging habitat and one of commuting habitat. I did that from the GPS fixes. I applied a clustering algorithm to look for clusters of GPS fixes. So I defined a cluster as being if a bat was staying in one spot within kind of a kilometer for more than half an hour, then I was defining that as foraging activity. Because one of these bats could fly 14 km quite easily in half an hour, probably, if it had to. So if it's staying in one spot, then it's clearly doing something there. That was kind of the rationale for that. And then all the other points, the other points were then commuting habitat. So you're differentiating those two behaviors in the GPS data. And then so. So the foraging habitat, that was obviously, that was the result. That's what was one out. One of the outputs that we wanted was to know where the best foraging habitat for the species was. And then the commuting habitat went in as the. As the base layer for the connectivity model. So do you know circuitscape? Have you heard of Circuitscape? It's based on. It's kind of developed from connectivity modeling. Sorry, what was that? Electrical modeling. So it treats a map of the landscape, kind of like a circuit board. So you'll have areas of high resistance and areas of low resistance. The areas of high resistance representing areas that are hard for wildlife to pass through. And low resistance being areas that are easy for wildlife to pass through. And then you apply a current across the board from a start point to an end point and it will show you all the various ways in which. Like the kind of path of least resistance, almost at the current being the flow of animals or wildlife, the path it would take. Does that make sense? Okay, so I use the commuting habitat suitability to define ease of movement. With that as that kind of base map and then knowing where the roosts were, you can then have the roost as the start point and then the bats can move out into the landscape to all the areas of best foraging habitats you have. You end up building this model of the routes that bats are likely to take between the roost and their foraging ground. Does that make sense? It's kind of hard. 31 00:19:49.000 --> 00:19:56.000 So is that your start point and your end point is the roost is the start point and the foraging ground is the end point? 32 00:19:56.000 --> 00:20:37.000 Yeah. So the bats are moving. Actually it is in reverse, but essentially it's the exact same thing. You actually have the roost as the ground point. And then to have the bats starting in the areas of foraging habitat. So higher suitability foraging habitat will get more current injected to it than lower suitability foraging habitat. Which makes sense because the bats can be going to. You presume the bats be traveling more towards the better foraging areas. So created by foraging habitat. And then. Yeah, then you, you run it and it all the current travels back to the roost, which is the reverse. But it's exactly the same. Yeah. Yeah. Okay. 33 00:20:37.000 --> 00:20:45.000 And do you just iterate that for as many start points as you have? Is that the idea that for each roost. 34 00:20:45.000 --> 00:21:10.000 Yeah, that's. That's then every roost in the study area. So it depends. That was fine for greater Hawaii, where their roots are reasonably well known. But then there are, there are other species where you wouldn't be able to do that because they, you know, they move roosts every night and people, especially tree roosting species People don't really know where they. You don't really know where they are all the time. They have to go. 35 00:21:10.000 --> 00:21:11.000 Right. 36 00:21:11.000 --> 00:21:58.000 Which is why I spent years as an ecologist going around looking for roosting trees. They tend to be a bit more ephemeral. I mean, the big ones might be easier to spot, but I don't know, like. Yeah, it depends on the species. So, so the project I'm working on at the moment is, is, is using. I want to be able to build these, these same models for, for every species of bat using, using an acoustic approach rather than a telemetry approach is the, is the input data input. So it's going to be. I'm going to have to rethink, rethink the approach for, for a lot of species because it's, it's different, you know, if you don't have good roost data, you're going to need to take a different approach. 37 00:21:58.000 --> 00:22:33.000 So given the area that we're talking about, I mean we're talking about the whole of North Somerset, which, I mean it's not a large county, but it's still a county. So. And you have sort of running your model over, that is. So you've mentioned already that you're using Julia to speed things up, but did you find that you needed to run it on like a high powered computer or was this like, is it sufficient to run it on something like a decent laptop or whatever in order to do it or. It sounds like there's a lot of processing involved. 38 00:22:33.000 --> 00:23:11.000 Yeah, there's a lot, yeah. So it was bigger than North Somerset. It was North Somerset, Bath, Northeast Somerset and then the former county, former districts of former Mendip and Searchmore districts, which are now just part of Sunset. So it's quite, it was quite a big area. And yeah, the first time I ran it, it was on my laptop and it took two weeks to complete and I had to do that twice. And then I started using AWS where, you know, you can Amazon Web Services where you can just say I want loads of processes and ram. And then it did it in about two hours. Yeah. 39 00:23:11.000 --> 00:23:42.000 Wow. Okay, that's really interesting just to understand that sort of processing because, yeah, I guess if you had really good input data in terms of where bats were roosting and good quality sort of environmental variables as well. So I don't know, let's say for the whole of England. Could, could this be run as a, as a method for the whole of England or it. Does it need to be done at sort of like local Scale, I'd like 40 00:23:42.000 --> 00:23:56.000 to do it one day. If you could get, you'd need very good data. So that's, that's the problem. So the environmental data is there. So you've got like. So right where this is. Just to clarify, this is, this is. I was doing everything at a 10 meter resolution which is. 41 00:23:56.000 --> 00:23:57.000 Right. 42 00:23:57.000 --> 00:26:37.000 Which is really high res, but it means you can pick out individual hedgerows and stuff which are potentially important on a really fine scale for foraging. So if you could, if you kept it at that resolution, I mean it would, I don't know how you'd do that. You'd need a lot of processing power to do that on a large scale and, and a lot of input data. So it's that yeah, you couldn't get GPS data, you know, on that kind of scale is you wouldn't, you would have to use an acoustic approach. Basically. You couldn't be too disruptive to be tracking bats all over the country. And also you can't do it for every species like I mentioned before, just because the, the tags are too big for the smaller species and radio tracking is not accurate enough. If you've got 100 meter kind of margin of error in a radio tracking fix, then it's either in on the hedgerow or in the middle of the field and you can't distinguish it. And if it's really important for the modeling then that's not going to work. Okay, but so yeah, so this is what I'm working on at the moment is using acoustic data as the input instead. So in North Somerset, so my supervisor, the bat lab and North Somerset Council run together. It's the North Somerset Bat Survey. So it's a big citizen science project. So they've got this acoustic data that they'd be gathering for the last five years. And it's, it's what I'm working on is being able to feed that into the models to, to use that because there's this ton of ton of data that is, there's definitely data in there about how bats are using the habitat. But it's, it's, the challenge is going to be with the biases in the data. Obviously like people tend to put them in their gardens a lot because that's what, that's what's, that's what they want to find out, you know, what's flying in my garden. So she's got a lot of gardens and obviously because when you're attaching a bat detector you have, you have to put it on something. So it's always attached to a hedge or, you know, a tree line or a fence or something. So you don't have any data. You don't have data from the middle of field. So I'm going to need to work out a way around that. Might be some additional surveys. Putting detectors out in the middle of fields and somehow cow proofing them like this. Yeah, just. Just. Yeah. And. Yeah, so this kind of big data, if you have these huge, huge acoustic data sets, you can get data from the whole year. Like, if. If you could have detectors out all year round, there's. There's a lot of information you can. You can be getting there. Which. Yeah, on a temporal scale that you could. You could. You could do this all year round. So you could. Yeah. 43 00:26:37.000 --> 00:27:40.000 I find it fascinating that you've started off in the field as an ecologist and now you've got the classic data scientist problem, which is loads of data, but it's really messy and you need to try and work out how to make it tidy and get what you want. It's brilliant. So the paper that explains what you've just been talking about will be in the show notes as well. And I really want people to go and have a look at it. Particularly the connectivity outputs, because they're amazing. The way I guess I would describe them is a bit like those Van der Graaff machines that you get where you sort of put your hand on them and then all the electricity flows to your hand. You can sort of see that, the whole analogy, in the way that those outputs are generated. They certainly are very visually stunning. What sort of feedback have you had from people that you've shown about these two different types? So the habitat indexes and the connectivity. 44 00:27:40.000 --> 00:29:05.000 Well, it's all being used by the council at the moment, so they used it in developing, in their new local plan, in developing a special planning document for how to deal with the greater horseshit pattern or some. And they're developing a scheme called North Somerset Nature Parks, where they. It's like a. Basically, it's an offsetting scheme where you pay into. You buy habitat credits from them. And so the modeling is always used to find places, the best places to put these parks. And, yeah, it's been used. What else has it been used for? A few different things, directing local planning decisions. So it's been useful. I think at least that was kind of the best feedback, really. And, yeah, I'm working with the council to keep them up to date on this new project now, to keep it going indefinitely into the future, because, yeah, Obviously the radio tracking data, it was collected in at one point in time and then you can't keep using that same model forever. So keeping the models up to date is going to be important for it to be useful for the modeling to be useful in the long term. 45 00:29:05.000 --> 00:29:23.000 Yeah. Are there any of the sort of geospatial and environmental input datas sets that need to be updated as well, or can you continue to use that? I'm guessing you're using things like LIDAR and maybe the land cover map from ceh, stuff like that. 46 00:29:23.000 --> 00:30:03.000 Exactly those, yeah. So the land cover map comes out every year, so that's, that's good. And then NDBI was one of the things. So I need to write a script to automate downloading and processing Landsat imagery to keep. Keep that. Yeah, NDVI upstate. That's fine. So the light LIDAR data is. The national. Is from the National LIDAR project, which I don't know is. Do you, do you know, is it. Is that still. Is it going to keep getting updated? Because from what I've read they said that they didn't, they didn't have plan to resurvey everything, but I don't know. 47 00:30:03.000 --> 00:30:21.000 That is a good question. Yeah, I also don't know if any of our listeners know that would be very useful for, for them to get in touch and let us know. But yeah, I've sort of seen the same thing as you that it got collected and there's not funding in place for continual updates. 48 00:30:21.000 --> 00:31:10.000 Yeah, I think one of the main objectives was to make a dtm, but then the terrain is not going to change necessarily so much. But for me, what's interesting is that the vegetation layer, which is going to change. And so even like when I, when I was using it at first, like some of the, some of the squares were, there's quite a range. Some of them are from 2017, so some of them, when I was using it, was even already 5 years old. So, you know, if you've got a hedgerow and you've got the height data of that from five years ago, that's going to become irrelevant quite quickly, isn't it? So there's other approaches you can use. There's, you know, the Sentinel 2 micro micro microwave kind of data that you can use to calculate height of stuff, I think. 49 00:31:10.000 --> 00:31:41.000 Yeah, there's all sorts of various bits and bobs you can put in, I guess. Do you make the decision about what data goes through into the model or is there some way of being like a process at the Beginning of the model to work out which ones cross correlate and which ones would be most useful, have the most information in there that would be useful for the model. 50 00:31:41.000 --> 00:32:18.000 No, it was just me kind of tweaking and tweaking it. Yeah, you've got to kind of think you've got to. You've got to choose something that there's, you know, there's a good biological reason for you to include it in the model. It's not just trying stuff. And I wasn't doing any kind of automatic automated like process for choosing the best ones, which is like model selection or anything. No, you've got. Just based on the species and you know, you've got to think what's actually relevant and. And yeah, and there's. Yeah, there's all sorts of different ways to describe. Describe the same things, isn't there? So like the land cover and the lidar and it all kind of overlaps a bit if you're looking at vegetation. 51 00:32:18.000 --> 00:32:19.000 Yeah. 52 00:32:19.000 --> 00:32:23.000 So it's. Yeah, getting the combination right and yeah, 53 00:32:23.000 --> 00:32:24.000 it's always tricky with this thing. 54 00:32:24.000 --> 00:32:40.000 Kind of the same thing because there's a lot of those. So. Yeah. So I'd like to simplify it in future. And probably. I'll probably have to remove the lidar from whatever just because it's going to come up, become out of date, isn't it? If it's not. If it's. 55 00:32:40.000 --> 00:33:46.000 That's true, it might be worth. Well, yeah, that is a tricky one. Trying to find out how you can get very high resolution sort of elevation information that's continually updated because you're right a lot of. Well, particularly with all the development that's planned as well, not only are you going to get things being removed and added into the environment, but also things will be cut back in order to allow some of that development, I guess. So you're going to get those height changes. Well, there you go. That's an interesting problem for you for the next couple of years. Keep you on your toes. Yeah, it's been brilliant talking to you. Our time is coming to an end, unfortunately. Is there anything that you want to add around either the modeling or a particular message that send our listeners about bats and sort of bat conservation and that sort of thing? 56 00:33:46.000 --> 00:33:48.000 I can think of. No. 57 00:33:48.000 --> 00:34:01.000 Okay, put you on the spot there. But if listeners did want to know more, how can they get in touch with you and your team? 58 00:34:01.000 --> 00:34:05.000 My email is probably the best. I'm not on any social stuff anymore. 59 00:34:05.000 --> 00:34:06.000 Okay. 60 00:34:06.000 --> 00:34:12.000 There's the lab website that's got all my contact details back. Conservation Research Lab. 61 00:34:12.000 --> 00:34:15.000 And that'll be in the show notes. Yeah. 62 00:34:15.000 --> 00:34:23.000 Yeah, no, I'm not on Twitter or anything. I did have a Twitter, but no, not anymore. 63 00:34:23.000 --> 00:34:36.000 No. Yes. Okay. That's been absolutely brilliant. Tom, thank you so much for coming on the podcast and talking to us about this. And yeah, I really genuinely look forward to see where the models go. They are fascinating. Cool. 64 00:34:36.000 --> 00:34:40.000 Thank you. 65 00:34:40.000 --> 00:35:10.000 Thanks for joining us today on the GeoMob podcast. We hope you enjoyed the discussion. You can get the show notes over on the website, which is@thegeomob.com and while you're there, you can sign up for our monthly newsletter and find links to our accounts on Mastodon and LinkedIn. So you can stay up to date on all our upcoming events. We hope to see you at an event soon.