 Our final keynote speaker is Don D'Angeloz, who will be talking about ecological applications of agent-based models. Thank you very much. Especially for James for inviting me. I have to confess, before this, I didn't know anything about serious dirtiness, and I've worked a lot in the last few days, and I'm very happy to have. So, I think my purpose is to give an overview of agent-based models. I know a lot of you know quite a bit about them. So I'm going to give you an answer. And I'm going to do it as follows. I slide over to you in a new direction. And then I'll talk about agent-based models modeling in the following years, which I think will be kind of relevant to you, which you're interested in. First of all, I'd like to tell you some serious concerns. We're really excited. You know when I talk about the importance of spatial-efficient AVMs, what particular systems are they? For example, when I talk about, of course, spatial dynamics, in particular the problem with story-up is the world problem that people are trying to deal with. There's a little bit about vegetation change along with ingredients, climate change, interaction with vegetation, short vegetation boundaries, regime shifts, if I get to these, and they're the leading spaces we don't control about risk. If I can talk fast, I'll get through all of those. I'm not sure. I think it's pretty big there. There you go. So let me tell you something. What are agent-based models? There's a number of populations in communities. They're being composed and discreet in their activities. So those are wages. And in economics, for instance, it's a very general type of model. It could be households, firms, social groups, a lot of other things. Right? And you know, this is a trait made very among the agents. So again, it's going to range from very simple to highly complex. It can include things like details of the world. It was a life factor. It's a loss of a state, adaptive behavior, learning, et cetera. And it's mostly spatial locations on a homogeneous landscape, but they're animal movement. Look at interactions in space are important. They may be more important than the overall interactions with the human population. Very essential for chastity. These are monoclonal simulations. You do a million replicates of simulations. And you can incorporate genetics into evolution of these generations. So if you really want to learn a lot, I recommend two books. One is by Bruno Railsbach, an individual based model. He used to call this individual based model. He was engaged in this. One art professor came. And then they gave, later on, they adapted as you please. This is a how-to handbook, whereas this is a more philosophical interview in the examples of some courses. So it's a very good pair of books. And they based on the main points. There's a good TS vector here. This is a bottom-up rather than top-down. There's properties of population, community, and ecosystem that were merged from interaction with individuals. Look first. That was a big difference. I thought I'd slip this in. In 2005, I did kind of a study of papers using the search individual based. You just to see what was happening. And I noticed back in 1990, two things were merged from this. Back in 1990, a version of Bill papers looked at individual based. They just went up very slowly. And you also notice that there are a whole lot of different topics. A lot of different taxa have been covered by individual based models. Nowadays, this is starting about 11 years later, and I wouldn't even attempt to see the literature. And I think I'll read for these topics and give away the whole book on the models of individual based. In any case, I'm going to start out by talking about applications and registrations, succession dynamics. And so the idea I'm going to deal on what it calls, he goes back to Darren Barton in the early 1970s. He wrote computer simulations of two dynamics in the first gap, this way you could have gap models. So this tree was modeled. You start here with a gap. It was created by either big tree dying or some sort of disturbance. New syndromes moved in. His growth, competition, dark winning, continued to get back to maybe one or two individuals and died. And he starts over again. Well, that's kind of boring. But I'll tell you, at that time in 1972 when I saw Dan Scott, he was pretty electrified because we haven't thought about modeling things this way. We used state-of-the-art models and we thought that was all we could do. Darren had the great idea of modeling trees individual by individual from the interactions. So there's Dan. New syndromes were very small. It was called Zabala. It basically assimilated processes since we're showing a zero point, zero one technique. Using translysis characteristics and environmental conditions. The main processes were really growth, mortality, and production. And there's a package you can buy Zabala to if you can do these things. It's very nice. They're going to give you a little more detail. The input data, site data like climate, weatheration, soil type, nitrogen, moisture conditions. And these are all modifiers of the maximum growth rate of trees. Then the tree, these are the tree species, maximum growth rate, maximum diameter, maximum height, maximum weight, also the second year production per year. And the tolerances to growth for these new tree plants. You know, there's more data, but for a lot of tree species, this steps approximately well. Approximately. Output in a simulated forest plot and for each species, the number of trees per hectare, basal area, coda biomass, and soil structure. We get a lot of information out of that. The way it works, and this is from this competition point, called tree shade disorders. The degree of shading, influence in growth, running water's revenue growth is checked in the model. The probability of mortality of the tree increases if it's running average on growth is low. So there's a lot more to it than that, but that's kind of the basics. A couple more things I want to say. It's not really spatially structured. The trees, even though it's 0.01 hectare plot, both have the location. They're kind of mixed. Appetitions, though, have similar biodegrading shading effects of all trees in the gap, so it's kind of sheared out. It's kind of spacial, but it's not real spacial. We'll get to that later. There's an external angle in here for the input of species. So what's good out of this is that simulations of succession started with this assignment. You can see that the simulations have been started really far. Really, we have succession species, rare birches. This is the rare elevations in the model. They have competed by more shade-tallered beaches and city makers. Higher elevation, it's a white birch, which is visually-actuated by this bridge, which becomes dominant. Well, it's pretty simple. This is a pretty simple demonstration, but very interesting. And one of the people that myself was electrified by it was Hank Schubert. He did that. He developed the FORA model, and I did all of it. He's got it everywhere in the world by now, but this is a secondary succession in the southern Appalachians. And he'd get input output like this, which is actually percentage of the species of various types through time, over 500 years. Mostly by our pioneer species, dominated, but through time, mostly oak species, and then mostly beaches come in. So this stuff can be calibrated in some data. There's a lot of data on forests of different ages, spores, grass. You can calibrate and you can test these models. They've been pretty well tested all over the world. So that's kind of the gap models. We're going to see how these are extended into bigger spatial areas. Before I do that, I want to talk a little bit about Savannah systems. Just to give you a reason why I think these models are really important. So I want to contrast the top-down model approach with these spatially explicit ABMs. So the top-down models are sometimes called spatially there. We often use these spatially implicit. They're really useful. They're really useful. I did a series talk myself, but you don't tell the whole story. So give an example. We support the best day for inside the van. Modeling of a Savannah system with four components, grass, Savannah saplings, Savannah trees, and forest trees. There's four differential equations. And to review two things like reproduction rates, mortality, precipitation. And these things, all of these rate coefficients are such as precipitation. So you can solve these things. You can solve these things through study study, or either on a weekly or on a weekly. And then you can look at the study state which you get as a function of a rainfall gradient. We did it for six cases which have different assumptions on the effect of rainfall on tree-like history. And what you get is something like this. We're only graded for very low rainfall. Grass is going to dominate. Or as you get higher, though, you get into cases where you can get best of quality cases where you have alternative stable states such as there could be either grass or savannah trees plus apples plus grass. And in particular, there are assumptions you get with these various things. And high-left rainfall we're going to get forest trees plus a little bit of grass. That's just a lot. You can really learn a lot from these things and see what you can expect in a savannah system across a rainfall gradient. And then the problem is this is the most possible effect of local interactions that are not really incorporated. Spatial interactions are not incorporated in these models. But so, at the ends, when they really incorporate these microscope mechanisms, what top-down approaches are the local spatial interactions such as possible feedback facilitation that can promote clustering. And so this is one of the main mechanisms that can cause these changes in systems. Now, for example, the question is what maintains a savannah system? Well, they have a tendency to go either to a pure system or pure grassland than savannah. So we started doing simulations starting with that. Well, I mean, scattering when I'm scattering trees. And we included this possible feedback facilitation. If you get a small cluster of trees together, what they'll do is they can facilitate recruitment of new trees. Or we can protect the solutions from fire because they will compete out the local grass. So if you do that, when you start to get real clusters, those clusters can grow and in time dominate the system. That can happen if there's a low frequency of fire. These are fire-dominated systems. So if the frequency of fires is low enough, the trees and shrubs can take over. The other thing, if these fires are very frequent, the system can go in the other way and you can get the conversion to complete grassland. So you realize the sensitivity of this system and the basic mechanisms that can cause it to grow in the way or another. All right. Now, what you found is that by high enough, frequency of fires, the right frequency of fires, to cure some of the seedlings that are in some distance from these crops, you can maintain a savannah with these grass shrubs and trees. When it comes to trees and shrubs, they're going to make roots of grass. And since other conditions also, it takes a very relationship of the trees to complete or facilitate the treatment. So there are a lot of things involved, but basically with the right balance of fire, you can get this, you can get a maintained savannah. All right. Now, this is stated further about calibration using the specific of your operation. The same model with including both tree-tree competition and local facilitation showed that no fire frequency in this case creates a regular distribution of trees, medium frequency creates a random distribution. Fire frequency essentially as a high fire frequency creates a account distribution. So we can get into more of these details. Another is that life history also matters. Trees have different life histories and different strategies for surviving in the savannah system. One is fire resistance because the fire avoidance keeps the water very close to less fire than the strategies these buildings can less valuable underground components. At the time, I would all model the savannah system so in which we had trees that were neither of these too tight simulated the fact that they're able to evade the savannah system, I did it in some of the different ways. These show a two-dimensional fraction of burn cells in glass density as a function of time after the trees have been brewed. We're all different. What this shows is that trees need different processes. It's not something we've been through, it's a token of account use but I'm going to skip over these things because I didn't realize time was going quite so fast. Another aspect of spatial pattern is well-known broad spatial patterns in semi-oregans. We've understood some parts of different equations during the period of one-third equations. It's very nice. That was a top-down equations that this also has been shown by taking into account local interactions with these top-down equations that they can do accounts. You can get quite different outcomes instead of these very nice patterns you get these fairly reflective-occurring patterns. Technical theory said it was the last of those two very fast models that as I said, the ball and foray are only partially spatial competition was averaged so they don't take into account actual position of trees starting a lot of value on them. The color does. In this case, these of these trees is interacting with the trees in this local area. And there are different, there are a number of different true-out models that do the same. On the right side, if you're starting with a linear spatial expressiveness we're going to make different assumptions on how plants should each other. So there are a lot of different assumptions on this. The important result of 14 predicts standing stock of forest spatial area very close to field data. The model based on new field, more horizontal variation that is low internal areas of yield, predicts only half of this biomass. So there's some important things about keeping track of the potentials. Still in larger areas, which I wanted to take more time on, I was talking about that earlier, quickly switching covers only areas below the hectare. So to scale up to larger landscape issues it's necessary to collect the most larger areas. But it's important also to keep as much of the DMGAP approaches as possible. There are models where the DMGAP visitation models cover very large areas. The problem is that they're very good. But the missed scale in the actions such as competition and facilitation that they have now in the effect. We'll all purchase the scale now from Mosaic. We're in Houston. There's this model. We're ever away from the tree up to course, it's not first years up to full reasons. That is tenduously check tears by combining these small 10 by 10 cells through only connected bikes seem dispersive. And it's by bobbing off that's the same thing. And this improves the fire module and the film not module so as you can look at things like the effects, you can model fires. The effects of the fires are on different species, makers, or orcs. Fire wise, asks us to come in. So we use those all over very large landscapes. Political force, again this is an ecosystem derogatory model by MoteGraph. The whole is scaling up very detailed physiology at the plant level. And you can put it biomass compared to biomass is at this scale. It's a one by one degree as well. And political force carbon stocks as well. Violation changes a little bit. I'm not going to talk about this, but this is a model. Rankine is another model which has been used to model tree species along elevation. One thing I do want to talk about is climate change. Large forces are very large forces susceptible to change through climate warming. This is worked by Shubit and his students. Women in the class are faced with large biogreed carnipers. This is a little very wide area. We applied the graph model for these three finds to 372 sites showing models over time. As temperature goes up according to the scenarios, large is going to be replaced by spruce in many of these sites. For a wide area, there's likely to be a change in vegetation. What's important about this is that it's ever been kind of for albedo. It's much lower than large albedo. So hopefully we're going to get a very unit possible interaction with climate. But that seems to be happening with muscle. I think I'm just going to go over that way over optimistic and get to the conclusions. That was way up to me. Okay, thank you. I was not faster than I was to read myself. Those invasions have been integrated into spatially explicit modeling to provide more mechanistically based descriptions of land cover dynamics. Scaling up from individual plants to landscape and regional scales allows applications to oceans to build crucial mechanisms to create the local scale of the divisions. Thank you. There's an introduction that the tree growing kind of won't quite I like to see. That seems to me that you're saying you cannot just say the average current. The average current is not necessarily the good current to have this vegetation landscape interaction. Do you have another thing to measure to determine for landscape enemies? Well, I think I don't know how to describe it in a lot of detail, but some of these models models the largest models actually we knew from the Z model it was always we're crying. I think doing a very good job of clearing up what they can include what's happening at this most landscape including the some of the internal variability in this landscape. We just need to we just need to get back up to to get a good estimate of by last and productivity and larger stress. I'm not sure there's any perfect way of doing it unless you're not a level tree. The next, the previous presentation show maybe before winning so very really, really great decision. It was courage. One average value of this. You seem to say something different about that. Yeah. I don't know. I don't understand quite understanding what you're asking. You can take one. Okay. Yeah. It's a very good question. I have a good review of the whole of all the models but I don't know how much the changes in the landform have incorporated into these models. It could be very important. I work in the other fields. We're doing changes in elevation with time because of the sedimentation and the rest of the changes there are a lot of issues. It's going to have an effect but we haven't tried to take that into account yet. It should be a long term. I'm sitting over the long term.