Naturally speaking. Alright, welcome to another episode of Naturally speaking. Today we've got a couple of the new, relatively new members of our faculty here in the department, the University of Glasgow. Before we get into that, I'm here with James Long, Sean Killen.
And I'm James P. James, and I'm here with me, as usual. But who do we have with us here today? We have Jason Matthew-Opless, who comes to us from the University of St.
Andrews. And how long have you been with us now? Four months. Four months now.
How are you looking so far? I'm fantastic. Where do I'm at? I'm from Oxford.
Let's go. Or the worst. And Grant Hotcraft. You've been here for a few years.
I was in November. Well, actually I was in St. Andrews for a short stint as well. And then before that was in Mr.
Gronegan. So I guess what you guys have in common really is that you're both spatial ecologist, sort of a spatial ecology-themed episode here. So I have a challenge for you guys. A lot of people might not exactly know what spatial ecology is.
So define it for us without using the words spatial or ecology. Go. Okay, well. As long as you need, you can edit out the awkward pauses.
Well, I would say that it's looking at distribution of events and in particular we're interested in animals or organisms and how those patterns are falling across landscapes and the interaction between those patterns, it's between different types of organisms and between those patterns themselves. Would you agree with that, Grace? Yeah. I would definitely agree with that.
I would also say that from a more conceptual perspective, the spatial ecology is the area which examines what are the implications of short-term decisions for origin or any other behavior decisions that animals make, but also dispersal attributes that plants might have on the long-term and large scale. And it becomes really, really interesting when it starts to deviate, when real biology starts to deviate from the assumption of perfect mixing and the assumption where everybody is just as likely to interact with everybody else. So from a theoretical perspective, it's really interesting to examine what happens to classic models of population dynamics when you introduce this heterogeneity of interactions between individuals. And so you can look at different scales as well, regional or local scales or the global scale and landscape patterns and other effects.
Yeah, one of the interesting things I find with spatial ecology is that quite often the patterns you observe in small scales don't necessarily scale up in the way that you would anticipate them. So they're not necessarily kind of like the sum of the parts. So you end up with these weird things like Jason's talking about where suddenly you end up with very strange population dynamics simply because animals are choosing to associate with each other in different ways. Well, that's funny, that's almost an interesting parallel with what you get with individual animals as well with body size issues where behaviors or processes that happen in small animals don't happen the same way in large animals.
Yeah. That's actually a nice link for talking about some of your work, Grant, because some of your previous work has looked at how large and small animals and their physiological processes how they differ moving from a small room into a large room and how that affects the way they interact over a landscape. Right? You want to talk a little bit about some of your previous work?
Sure. You know, well, I guess one of the central questions is how can it be that you can have some ecosystems with 28 species of herbivores all eating the same thing? And then you're talking about it for now. I'm talking about, yeah, exactly.
We're talking about the searingate ecosystem. This is an ecosystem that's very diverse in terms of like ungulate community and herbivore community, but also in the predator community. So one of the questions we were really intrigued us is how is it possible that you can have so many species all in the same, or using the same resource, right? When we talk about the seagull system, how big is the searingate system?
We talk about things getting from small scales to large scales. The searingate top to bottom, for example. Okay. So top to bottom, it's about 260 kilometers.
Okay. And east to west, it's about 180. So the whole thing comes to about 25,000 square kilometers. But it's not whole within protected areas.
It's kind of diffused how do you define an ecosystem, right? There's no boundary. Yeah. And there's thousands of animals kicking about.
Yeah, I think. Yeah, exactly. So the question was, how do all these animals sort of coexist? And one of the interesting things about being a herbivore is actually it's quite difficult.
You know, I mean, because you've got to break down cellulose, and that's incredibly hard thing to do. And avoid predators. And avoid predators, right? And really in order to break down those things, you kind of need a very complicated physiology, gut physiology.
And that depends on having, well, in the ruminant case, four stomachs, you know, these four stomachs or in the equids, you have this hind gut from the patients. So we started looking at how the gut physiology might determine what an animal could eat. But also the other interesting thing is that the smaller you get, the more susceptible you become, like you said, short of a few lots of predators. But also, I didn't really thought about this before, but smaller animals have a tougher time digesting their food if they're a little herbivore.
Yeah, right. Right. And that's because the food doesn't stay in their gut long. Yeah.
Well, the thing is, in order to get enough energy out of things like grass and cellulose, you need to ferment it for quite a long time. And in order to do that, you actually need quite a big space in order if you're going to get enough quality out of it. Whereas if you're the size of a small dog, like a Thompson gazelle is only 20 kilos. You're simply just don't have the space.
So the only way that you can get enough energy out is by eating the high-quality. And when I read that, I was thinking to myself, like, these small ruminants are really kind of snobs. They only eat the finest quality of grass. Oh, yeah.
The food quality is what limits them. Yeah, absolutely. Whereas you're talking about this paper, this is a paper that was in Journal of Animal Ecology in 2011 that we're basing most of this conversation on. And you're also talking about smaller ruminants are limited by food quality.
It's the larger ones that are more limited by abundance. Yeah, right. And in fact, they become very unconstrained. So the smaller, essentially, they're like the poodles.
I mean, if you're going to think of this in terms of dogs across Glasgow, the little guys, all the little Thompson gazelle and the herbivores, they're the dainty poodles of Serengeti. And then you've got the big buffalo and stuff that can essentially survive on anything. Well, it's pretty convenient, too, that being large leaves you unconstrained in the types of food you can eat and basically eat anything. But also protects you from predators, to make sense.
So if you're a huge elephant, you can kind of go anywhere. What you need to do, basically, is you're only constrained by how much food you can get. So you're just kind of... Exactly.
So in order to get enough food, you just have to wander far and wide to get enough, right? So you basically, for this paper, and I guess a lot of your work involves going down to the Serengeti, surveying areas. And so in this paper, you're surveying landscape features, all kinds of different characteristics, like food quality or grass quality and abundance, moisture. Yeah, it's different thicknesses of the vegetation, and things like that.
So it's kind of, and of course, on top of that, we have to get like abundances of animals, in other words, where are these animals actually occupying? And these are issues that Jason was addressing in his paper as well, right? And that's a bit of a problem in itself. So essentially, what we were trying to do is combine sort of metrics of, like you said, grass quality, so you have a whole bunch of grass vegetation sites that you go out and clear grasses on, measure the nitrogen.
You're looking for thickness of different types of species of trees and stuff like that, that might be associated with cover distances to rivers or this sort of stuff. And then you overlay that with essentially where these animals are found, getting back to our sort of spatial ecology. And how do you quantify where the animals are? Are you out in trucks, drive around saying, there's a little bit of a...
There's a little bit of a waste. Is it you doing aerial surveys or tracking data? Or how do you quantify? So we do a bit of both.
But for this paper, we based mainly on aerial surveys because it's the best sort of synopsis across the entire ecosystem. You can imagine if you're trying to count animals across 25,000 square kilometers, you're probably not going to get very far, right? So essentially, what's done is there's trans-sex that we fly, and you're looking outside of the aircraft on a strip, and you count the number of animals in that strip, and you quantify that by that. So the strip is on the window of the airplane?
It's actually, it's very high-tech. We grew on some fishing rods onto the wing store on the aircraft. And they're calibrated so that when you're sitting comfortably, you, at a certain elevation, you're counting essentially 200 meters. Wow, that's pretty much the aerial surveys for those of course.
Or seals and marine mammals and stuff like that. So proving once again that gaffer tapered, definitely, is the most organized haversing. It's scientific. Right behind cable plastic cable ties.
That's right again. It's interesting to see the proportion of tape in many papers. So that's pretty cool. So basically what you found then is that you found more or less what you expected.
When you surveyed the abundance of animals, look at the habit or landscape characteristics, you found that the smaller animals were drawn in areas where there was a quality of quality. Well, exactly. I mean, so this, first of all, this is taking a 21 years of survey days. Yeah, that was amazing.
They've been doing these surveys for a long time. In fact, Tony Sinclair and people like that started it in 1964 or 65. Focusing on five particular herbivores, so too large or too small. No, in fact, the surveys do a lot more than that.
So they're trying to account essentially all the major herbivores. Plus they're looking for impacts of human stuff, human impact. So they're looking for signs of illegal events, for instance, poaching events, and also agricultural areas because they're trans-extend beyond the area, the protected area. So they're going out into community areas as well.
So they're looking at changes and things like that. So one of the things that are counting are these five species of herbivores. And essentially what we found was that, so you can imagine a buffalo which is about 650 kilos right now. That's a big animal.
And then there's an animal called the Toby and a kongoni which are both relative to the wildebeest at about 160 kilos. Then you're getting down into the smaller gazelle, the pomps gazelle. How do you get a gazelle named after you? Yeah, who's pomps in anyhow?
Sorry. You're real. What's the difference? What's the difference?
You're trying to play 100 kilometers an hour and it's like you're trying to pick out something that's 20 kilos and 50 cents a meter at the shoulder. You do have to be able to see anything like that. No, that's right. Yeah, dick-dicks in fact, we had to exclude.
Those were out all together. Not going to make any jokes. You get the jokes obvious. Yeah, why two?
Jason's wondering what the fuck is going on here. I guess when we talk about like space use and we talk about how animals are selecting for food or finding resource patches and having this kind of struggling for this trade off between foraging for as much as they can, especially in the case of small deer when their gut isn't very large so they can't ferment very quickly. So they've got to be as much as they can but at the same time mitigate for prediction risk. We've got this trade off between different traits if you like and it's desperately trying to leave us into Jason.
Well, actually I'll tell you. I think one of the interesting segues is that some of the work that Jason does is looking at sort of the assumptions that we're making when we're making resource selection functions. And so my paper was based entirely on a resource selection function where we assume that the animals are what we would detect on that one count, 21 counts, one data time is applicable to the entire ecosystem. And it's entirely likely, you know, that's kind of the data we have but it's entirely likely that that might not be.
So the concern and if I'm understanding correctly is that the days that you're sampling on the day might be something unusual about those days that are drawing the animals particular parts of their habitat. Was that season or so we always have the same season but the concern is that it's essentially it's 21 years of data. We're pulling it in truth. It's 21 days of data over 21 years.
So so an actual factor in is incredibly small actually. And how do you get around that? Yeah, that's interesting. So there's no within repeatability.
No, so you only have one survey day for it's well, the surveys actually take 10 days. Okay, but you only do the one second location. Okay. So you're working on the data.
Yeah, so one of the biggest realizations, I mean, just looking at other people's data has been how, you know, the specificity of the inferences you draw from these analysis on these precise conditions. How do you extrapolate the data? You collected the data. So I'm not just talking about random variation.
I'm talking about circumstances conspiring to make you draw conclusions that are not necessarily going to be the same if you change the circumstances under which you collect your data. So I'm mostly talking about statistical models or empirical models here as opposed to the much more mechanistic flavor of models that a lot of people talk about but very few people actually develop. So I'm not thinking in terms of models, individual-based models where people make simulations right from the integrative energetics behavior all the way up to population dynamics. We think about these models, but actually what most of us do is regression models.
So we basically draw lines on more complicated lines or curves through a class of data. And you know, what my stats, lectures used to tell me is that correlation is not causation. And that was a mantra that we were forced to, it's a memorized yet. The vast majority of the literature looks at these correlations and tries to draw some biological inferences because that's our job.
That's what we're here to do. So I guess one of the problems is that you, so maybe you track an individual through a landscape or you go out and you survey an individual through a landscape. And what you try and do is correlate the locations or something about that kind of space use bit, the speed or some kind of behavioral proxy. And then you look at the underlying habitat and you say, so when some of them are traveling slowly is it because it's foraging?
And so you then look at some kind of proxy for food, bit in grants case, you look at vegetation index or in my case you look at sea surface temperature or chlorophyll, some kind of underlying oceanographic productivity. And then what you do is you basically regress that. And you do a correlation, you do some kind of model and then you try and make predictions from it. And the problem with that is the current kind of statistical frameworks that we use, be it just a general linear model, the mix model, the predictive power in space isn't very good.
I think that it's exactly as you described it. So it's an entirely legitimate question biologically to say that animals or plants are where they are because or partly because there's something there that attracts them or allows them to carry on existing at that location. So from an extimological point of view it's an entirely plausible legitimate question to ask. The question is how do you interpret the results of this model fitting?
Like any regression model, what you get are estimates of the parameters involved in these models. So if you think about the intercept and the slope of a simple regression model, for example, what we do when we obtain this fitted value is to go to the slope and look first of all at whether it's positive or negative and how big it is. So how sharply is this line going up or down? And we tend to interpret this as a avoidance of increasing values of a particular resource or condition, preference for it or indifference for it.
And the sign of the slope and the magnitude of the slope, we attach a biological significance to it. Now there's two problems with that. The first problem is that the models that we need to fit in real ecological systems need to be multidimensional. We have several attributes that help the animals or the plants decide where they're going to be.
And the second one is that any organism is subject to a unit sum constraint, which means they've got only about up to 100% of their time to devote to different activities. So these different attributes and these different types of habitats compete against each other for the organism's time. And therefore, in order to decide what the organism is going to do, where we're going to find it, we really need to take into account the entire profile of the environment. All these things are competing for the organism's attention.
And if we simply look at one particular scenario of what a population would be able to be used in a given configuration of conditions, we are not really protecting ourselves against that behavior changing completely if you change the profile of the environment. So I was really quite interested in grants paper because it does exactly that. It looks at the sciences of directions and the magnitudes of slopes in response to different environmental covariates. But it also anticipates this phenomenon called functional response in environmental conditions.
So there's one point where you say that animals might select a resource, but avoid areas where it is overabundant. No. Fundamentally, I can see no biological reason why that should be happening. But I can see how fitting a model to a particular set of data could mislead you into thinking that there's avoidance for a resource.
So if I give you a very simple example, I absolutely need my bed. I need to be able to sleep for a second. So I just wait too early in the morning. You can know that the fiber commute and it's a massive, seriously critical point of my location.
My bed is an essential resource for me. So within the environment of my home, I need to have a bed somewhere in there. Now if you put places in the house which had a tiny living room and 20 bedrooms, you would find me using those bedrooms very little compared to their overall availability. A statistical model fitted to this data would lead you to conclude that I have an avoidance towards beds.
Even though biologically it's an absolutely essential resource for me for a set amount of time. So this is the sort of biological process that can be slid into the paper that we're discussing. I've got on the table here. This was a problem that we were trying to deal with.
Somehow you have to take into account the abundance or the density of whatever characteristic of the landscape or habitat that you're using. Correct. So what we've done in this paper in one phrase is we've written every single slope parameter in the model as a function itself of the entire environmental profile that the animals find themselves in. And this has allowed us to calculate parameters about that parameter which describe how that parameter, how the response to a particular resource changes as a function of the whole makeup of the environment.
So that allows us to try and predict what the organisms will do in as yet unobserved environmental profiles. I think it's actually really interesting and clever approach I think because one of the things like you're right is it comes up all the time in conservation biology. When you're looking at we're trying to reintroduce a species. What are the habitats that it uses?
And we need to recreate those habitats. But those habitats are necessarily what you're finding at one scale is coming back to what you're finding at one scale is not necessarily applicable across the entire thing. So these reintroduction with frogs for instance is a prime example. You're taking the major habitat types, recreating them, letting the frogs go and discovering the population actually doesn't do that well because you're missing out one's critical aspect.
So it's being pushed to kind of edge habitat and it's not natural niche through competition or extinction or something like that. You then try and recreate more marginal habitat for it. So all you're doing is recreating more marginal habitat. That's another problem.
The potential problem with this model is that the absolute population size whose distribution you observe stands to change the distribution that you observe. So for example if you have very few animals living in this great huge environment you would expect them assuming they're quite efficient. You would expect them to gravitate towards a local or global peak in habitat suitability. Whereas when you're starting to crowd that environment you would expect homogeneity distribution to start creeping in as these...
So in the first case you might find a much sharper signal in terms of preference compared to the second case. So the absolute population size can change your inferences about habitat preference which is another problem that we need to start tackling before it before too long. So you're talking about introducing a new way of looking at some of these problems or these types of models you're talking about aren't as commonly used then. But there's no models really that account for these changes.
So the vast majority of papers that look at habitat preference there's a lot of them as well but take a snapshot of data because that's really what we can't get in most cases. Feed a statistical model to those and directly interpret the parameters of this model in a biological way. So there is causation in there. But it's not certain that the parameter values that you get are directly related to it and always true to what they represent.
So given an example, a slope coefficient of zero usually is subjected to some form of statistical testing, hypothesis testing. And a slope coefficient that's close enough to zero is usually interpreted to mean that this particular variable has no effect on the system that you're observing. When in fact in these models it could mean that it's provided to the organisms in exactly the perfect amounts. So they have no apparent preference.
They're not missing it so they're searching it out and they have no apparent avoidance. So it's not offered in super abundance so that we get this apparent avoidance. So this is a completely different interpretation of all these models that we need to start to allow for. That's really cool.
I think it's cool how your guys work overlaps. We're not overlaps, it's very complimentary. It is in the types of data that you collect and the ways that Jason looks at it. Right.
We're also interested in common interest in the emergence of migration which is not something that these papers tackle directly. And so that's going to be Grant your next major project is looking at migration patterns again in Africa with wildebeest and zebra. That's right. I guess we'll just quickly wrap up with talking a little bit about that if you just want to quickly sum up what we'll be doing over the next few years.
So what we're going to be working on in the next few years is trying to look at variation in roots and migratory roots and trying to become a little bit more predictable in understanding how these animals are actually moving. So what we're doing is we're putting on radio callers onto zebra and wildebeest. And these are two species that are migrating around the same ecosystem both in Serengeti and very similar migratory roots but from a completely different taxonomic background. You want to ruminate an inequity.
So wildebeest and zebra. And so what we're going to do is we're going to look at how these animals are moving in their trajectories and relate that to the landscape features that they're moving through. But the exciting thing about it I think which is as far as I'm aware maybe Jason knows better is that I don't think it's been done before is to be able to recreate physiological conditions of those animals as they're moving through these environments. And the way we're doing that is we're going to we're pulling tail hair and by sequencing the tail hair to small little slices you can get and looking for steroid hormones you can get a profile of how their physiology is a buried throughout their migration.
So now you have information about where they've moved, how they've moved to landscape and moving through and what they were kind of feeling in a loose term as they're moving through. Yeah. So this is the feeling aspect unbelievably is what the modeling is going to focus on. We've got a joint proposal for Apache which is going to try and guess what the hidden behavioral processes are that give rise to the migratory patterns that we observe in grants animals but also in other processes like cancer cells and aggregations of organisms with hopefully basic behavioral rules interacting with each other and giving rise to migratory patterns of a particular form.
So some pretty far reaching implications. I get the feeling we get to sit here and talk about this all day. There's actually a lot of things I'd like to talk to you about my own interests but I think we should probably start wrapping things up here. Just to finish things off actually to the liberty of getting a couple of jokes off the internet.
Some general linear model jokes but unfortunately don't have any of those chases. But I found some wildebeest and zebra jokes for you. There's actually a website called zebra humor dot com. Really?
That's what it is. Did you book market? No, I'll send it to you. Well you can look it up.
So what's black and white, black and white, black and white and green? Three zebra is eating a pickle. And finally what do you call a confused wildebeest? It's got to be a zebra.
A bewildered beast. A bewildered beast. And with that. Just like that.
You've been listening to the Naturally Speaking Podcast. Thanks for listening again and we'll talk to you later. Do we have a sign off that we normally do? I can't remember.
It's really lame kind of bye. Bye. Thanks for listening to the Naturally Speaking Podcast. If you've got any ideas for future shows or want to get involved in please contact us.
We'd love to hear from you. Our email address is naturallyspeakingpodcast.com. Naturally speaking is a podcast from the Institute of Bar Diversity Animal Health in comparison at the University of Glasgow. For more information you can find us on Facebook.
You can find us on Twitter at theNSpodcast and you can download future episodes and find more information at our blog at naturyspeakingpodcast.webpress.com.