 Right, yes. It's very much a team effort this one. I'm going to be presenting the work of the modelling team. And I just want to give a very brief introduction to the modelling team and who's done what. An overview of the role of modelling within the project and whiz through three of the more complete models that we have produced. And then a little bit of talk about the future of computer simulation of submerged landscapes, because I think there's a lot of research of submerged landscapes still yet to do. And there's a lot of computer simulation still yet to do. So a little bit of a look at the future. As Ben said, the modelling teams, Tabitha, Eugene, Miho and me, Tabitha did an awful lot of the research behind the first two models that I'm going to be presenting here, sea level change and duration and deposition. Eugene has done work on an advanced piece of infrastructure for doing large, large models in a high performance environment. Miho has been working on the forest growth dynamics and I've been doing the rest of the stuff, which is I suppose mainly the design and programming of models one and two based on Tabitha's data. And as far as the role of modelling within the project goes, and one of the key elements is that of data integration as you've seen from today, there are a lot of disparate sets of data with their own strengths, their own limitations. And the computer simulation environment was seen as suitable wanted to bring all those together. And because it's not just incorporating the data from the project, which incorporating research data from other projects, and also processes that are available from other projects, and this is one thing that computer simulation allows us to do is it allows us to treat data and the processes behind that data equally. And, and that's particularly important for submerged landscape research because the data is really difficult and expensive to get and sometimes it's under 10s of meters of marine sand and sometimes it's been eroded away. And so the difficulty of getting primary data is to a certain extent compensated by the other things that you can bring in with a computer simulation environment, and more than any other archaeological tool certainly it treats a four dimensional dimensional environment in a four dimensional way dogland to land change from the last place you maximum through to the inundation is constantly changing. And as a result, we need a tool that can deal with change over time and computer simulation is a lot better in that way than traditional tools such as GIS. Also, one of the things one of the things I particularly an enthusiastic about with computer simulation is asking the question, what would that look like. And it's important to draw the distinction between asking that question about dog land what would dog land look like. And because that recreation is not what we're trying to do. Basically, it allows us to look at, for instance, the data in the course to produce a hypothesis regarding how that data got there whether it's taffanomic or whether it relates to the environment around the core. And, and when we produce a mental model to explain that we can use computer simulation to ask a question. What would that mental model look like. And if it was a simulated system that we could play around with that way what are the implications of our hypotheses when it comes to the landscape as a whole. And that's a very important point with the computer simulation is we're simulating hypotheses, we are not recreating dogland as a complete environment. So, and just to illustrate that, and the first model that was produced and deals with sea level change, and there are a number of long term sea level change models out there produced by a variety of different research projects for a variety of different purposes based on a variety of different data. And what they do is they provide a model of how the sea level changes over time. And our question, what would that look like is what would the results of those look like. Those models look like if we applied them to the Southern River Valley, and if we applied them to the timescale of that would be noticeable by humans. And what would those long term sea level change models look like what the implications of them to the inhabitants of dog. And so we created a model that produces two main inputs, one of which is kind of a simulated inundation history of a particular point in the landscape, and the other is a look at the landscape as a whole. And this is the kind of input data that we have. This is a model of sea level change over the last 21,000 years produced by Sarah Bradley and her collaborators within and without the British ice project. Sarah has been very helpful to us over the process of putting this the data together for this model. And as you can see here you can you can look at this graph relative to sea level change over 21,000 years, and you can see flatter areas you can see steeper areas whether the sea level rises relatively quickly. But I mean, what does relatively quickly mean on a humans guy what do do the inhabitants of Dogland know that they are living in an inundating landscape. And, and there have been people who have expressed opinions about this, but actually we have with us within computer simulation a tool to provide a part of the, of the picture that might lead us to an answer it certainly won't give us an answer, but it will give us an extra bit of data that we couldn't elsewhere. And this is, I'm not sure how well this is going to come over on zoom but this is a little animation of this model in action. And this is the Southern River Valley and they see it just plots where the sea level would be based on the long term sea level curve and plausibly modeled short term processes such as ties and whether so you can see that that kind of pulsing the tides, that kind of thing, and it lines up quite well with with modern sea level data. And we can produce things like this, which is a, it's a model of the inundation of a particular spot in the landscape and the spot in our virtual landscape corresponds to the location of core and left up by which is in the Southern River Valley. And basically, we used the outputs of nine different models for long term sea level change, which are all plotted here. And the, the, the, the kind of wiggliness of each line of the graph as it goes from 0% inundated before a wave has, has lapped over that particular point to 100% inundated which is kind of the last time he sees the sky as it were. And as you can see, the old nine of these are different in some way. And, and there's a wiggliness to it that represents the random nature of the effects of the weather. But from, from the inundation process as a whole from starts to finish within this model, that's an important caveat that we'll get back to soon, within this model takes about 400, 450 years from the first wave to the last one. And we can include incorporate other kinds of data. So for instance, we have two bars here, bar a is the, is the last radio carbon date from core e left of low five from a pre inundation context, and Barbie is from intertidal models. So, were these long term C level curves to be an accurate representation of the inundation of this location you would expect, and they a to be before the start curve that B to be during the curve. And we can also, and this is the output of the landscape as a whole based on just one of these curves. And we can split the landscape up into a series of different categories that people areas or areas with zero percent inundation the blue areas or areas with 100% inundation. And the different types of green are different percentages of inundation over a certain period of time so that the green areas are basically the intertidal zone. And they're listed on the pie charts as mud flats, low marsh, mid marsh and high marsh, but don't take those categories to literally, and other factors are involved obviously in those environments. And as you can see, you can start to look at the southern river valley area and you can, you can start you can maybe even draw conclusions about how you think the inundation may have progressed over a period of time and over 1500 years it goes from very little marine to almost completely marine dominated. And it's important to point out what this model isn't and what it is, it's certainly not a recreation of the inundation of the southern river valley there are very important processes in the real world that aren't in the model. And it's not a conclusion in and of itself you can't present one of these maps as the answer to anything. And although we can do this, we can model this very precisely. And that doesn't mean say it doesn't give you any indication of accuracy. However, it is something that we can compare to all the other types of data that we have seen throughout the course of this day. It is its own thing. It is what happens when a static landscape surface interacts with the output of sea level change models, combined with plausibly modeled short term processes, it's no more or less than that. And, and has to be taken as part that is one of the tools for interpreting landscape, not the answer to interpreting landscape. And it can also use it for examining what's called data downscaling. So, for instance, the curve that sea level curve is, is only a curve because of the resolution of the data points mainly the data points every 500 1000 years. And it may well be that the actual process is more of a stepped processing which you get periods of much quicker inundation followed by periods of slower inundation or maybe there's inundation and regression that kind of thing. Maybe it's more of a weekly one. And we can't give the answer to that, but we can model each of the alternatives and we can start to look for indicators that may be present in the landscape that would point towards one of one or other of those options. Model one has a very static landscape doesn't change in any way shape or form it's like a like a like it was carved in stone as it were. And model two looks at processes of erosion and deposition and it takes the same seismic data as a base. And it adds a rainfall model based on modern rainfall data that you can you can basically on anything you like. And it allows us to introduce tokens into the system. These are simulated bits of 10 represent environmental proxies, for instance, basically what they do is they move with whatever thing they're in, whether it's water or whether it's sediment. Whether they can they can flow over the landscape they can end up in sediments they can be and they can be buried by other sediments that could be eroded out shifted so it uses some way of simulating environmental proxies. And this model produces three different types of outputs. It gives us the opportunity to pick a location in landscape and do a virtual core so it's and we can call a particular location this landscape in the same way that we can call the real landscape. We can produce a map where it tells us which areas are lower than they started which areas are higher which have net erosion which have net deposition, and we can also track the tokens as they go through the system. And just like the last one this is a video of this model in action, which doesn't do anything for a short while, but the blue areas are where water is but there's no indication of death. And then at some point it will rain and that rain will flow downhill. You notice there's there's no sea involved in this one and there's no explicit modeling of the southern river, although if you leave it running long enough the river kind of creates itself. And as a result of that we can get virtual cores. And one of the interesting things about this model is that the processes involved in landscape are really simple. It rains the water runs downhill and based on how that how that war moves either stuff gets eroded or it gets really positive. And however, and that leads to really quite a large variance in in in output of virtual cores. And you don't get much happening in some locations. These are just the core locations which feature any deposition. And you get the core location else 36 which is pretty regular over the amount of time that this takes. But the other cores they have they have periods where lots of deposition occurs and then they have areas which not much deposition occurs. And this is is purely a reflection of the local and the upstream system itself so series of kind of feedbacks that creates from a very simple system actually quite complex inputs. These are the cores that feature only erosion and this is and the three cores in this particular simulation that have both erosion and deposition. And the interesting thing about these because is we can display them as virtual cores so they're analogous to the real cause. And but of course just like the real cause we're missing data from that. Whereas if we if we show them here as the change of height that location over time, you can actually see what bits of data you're missing. And that's one of the good things about computer simulation is I believe it's very important to have a simulation produce inputs that are analogous to the real archaeological data that you have. However, we can go beyond that and we can go back to the system we can see which behaviors have created those. And this is just a map of 10 tokens in one of the simulation, the tokens start on the red dot and end at the green dot, you can see some of them move quite away, some of them don't move pretty far at all. And you can you can track them over time so you can see when they're moving those a lot of them move quite a way to start with them when they stop and then whether they move on from there depends on whether they're eroded out of the deposits where they land. Again, not a conclusion on itself, and certainly not a comprehensive model of all Tafanomic processes it's not even a sophisticated model of the Tafanomic processes it involves. And, but it's necessary beginning because Tafanomic modeling is is likely to effectively have now and, and it's in the process of modeling and the process of doing things seeing the result changing things as the endless iterative process of Tafanomic modeling, where the usefulness is what Georgian is called developmental utility, it's not the necessarily the results that are directly applicable with the process, which is the learning process. And just like the data and scaling and you, you can use this model for examining Tafanomic processes themselves you can change the Tafanomic processes and and see what the result would be. Now, this is the basis of me holds PhD work, and this is his model on forest growth dynamics and basically what this does is it takes a series of models that have been around for 40 or 50 years now called forest gap models and they used in in forest management and environmental science. And, and it changes them from stand based models to agent based models so each tree is modeled individually, and you can then incorporate and fairly a climate data with that. And so you can see how the changing climate effects forest dynamics, and you can also simulate human and animal activity and the removal of either a sense of trees as a whole or a different sense is different species. And as you can see here from the images in the in the lower left corner this is just a series of snapshots from one another simulation over nearly 1000 years. After 50 years you have actually quite a homogenous environment after 300 is starting to change up and the stand dynamics of self thinning and stand density and competition flight result in over a period of time quite a heterogeneous environment. And you can see this yourself if you've been, you've been walking through a, an area of woodland that is a plantation for the timber industry, for instance, compared with an ancient area of woodland, you can see the difference kind of instantly the make up of the trees and the, the uniformity or lack there on the trees. And what this is is it's the results of a series of simulations, but this number in the top left hand corner starts at 0.0 and that is just a an environment with three different species of trees that are all competing against each other using natural dynamics. The, there's an ever increasing amount of thinning that happens on year 50 of this 1000 year simulation and going from 10% up to 80%. So at 0.8, then a year 50 in the simulation, 8% of all trees are removed. The interesting thing about this is this obviously causes an effect in the environment, but the knock on effect, the effect on the dynamics of the growth of the trees in the environment can be seen four, five, 600 years into the future. And this gives us confidence that we will be able to working back from things like pollen analysis of archaeological sites. What we may be able to do is detect markers of human activity actually quite some way into the future, based on the very indirect traces of human activity. Again, it's not a recreation, but one of the good things about it is because it's based on forest gap models and a lot of research has gone into forest gap models and a lot of research has been done on looking at woodland to validate the forest gap models. Then, then this is a very, it's very well validated, particularly for an archaeological model. It's also expandable. I mean, the model itself has far more than the three species that we've just shown, but you can theoretically use this anywhere. And it's able to be integrated with future models. And if we combine models one, two and three, which is the plan, and it would give us a complex system that is very complex despite being based on simple individual behaviors that produces things in a format that we archaeologists can already understand and can relate to our primary data so we can produce virtual cause with virtual proxies in them. It also gives an almost infinite parameter space to explain that is that the number of potential combinations of variables that go into it. However, and this is one of the medium to long term aims of the computer simulation community project is if we make these models as accessible as commercial software as utilities as games that you download for the PC. That means that that parameter space can be explored by many more people. It's not just me and it means that those people can explore these models for the questions they want rather than the questions I envisage. And that drastically changes the way that simulation can contribute to archaeology and you know to facilitate this this Eugene Chung has produced an infrastructure that is able to run models that you normally have to use a high performance computer cluster on on a desktop PC that has a commercial graphics card in and this slide alone could have 25 minutes spent on it but we unfortunately we have to shoot past it. So as far as the future of computer simulation in submerged landscape goes on to highlight two things. And the first of which is, we need to deal with data integration it should be obvious at this point, and that, and even though we, we only have a fraction of the data that's available for dogland we still have more data than we can comfortably switch together very easily. And, and we need to deal with data integration that includes processes that includes change over time but that also allows us to include introduced processes from from other disciplines and and computer simulation is the best tool for that. But the second point I want to make is is possibly more important in the long term, and it's to do with the accessibility of the models, because this relates to the accessibility of the results of archaeological research, research, and of archaeological research itself. Dogland is equally inaccessible for all of us, unlike the vast majority of archaeological sites, some of us can get Stonehenge, some of us can get to various places. But dog, none of us are going to dog a land anytime soon. And, and as a result, accessibility has to be built in from the very start to the tools that we use to research dog land, and that means that people now who are currently find it hard to access both the results of archaeological research and the tools archaeological research will get the benefit of our tool building endeavors. So, we can produce these models, we can make these models accessible. And so, people who might otherwise struggle to get involved in archaeology will have access to cutting edge tools that they can ask their questions, not ours about the archaeological data, and that has significant implications for the medium to long term future of archaeology not just submerged landscape archaeology. And just like everyone else has said, wouldn't have been possible without all the different members of the team and although the modelling team is four of us, and lots of other people here have been involved with various aspects of this. And that's about it. Phil, that's some, some wonderful time keeping that extremely nice by that. Okay, so while everyone just gathers themselves. We have a few minutes for talk for questions here. I've got one here for you off Becky Bryant. Two questions. The first is about uplift in model one Martin mentioned there was evidence of uplift uplift in Southern River Valley. Are you able to include this in your model. First question. Do you want to answer that first and then we'll have the second question. Yeah, can do. And uplift not involved in any way model one. And as a result, we have to, we have to involve it in the way we deal with the results of model one. And so we use what we have as a model with no uplift at all what we have in the real world data is potentially some uplift and certainly some movement and there's a the, the, there's lots of lots of research has shown that the landscape of uplift is is is much more wibbly wobbly over the medium to long term. I don't that's probably a scientific terms that but yes uplift and other movement is a factor. And so having a model that doesn't have that is one way of trying to filter out the effect of that in the real world so you can look at individual data points. And so the, the static model says the sea should be a particular time you look at where it actually is based on the, the dates you've got and that is a potential explanation for that difference and that's where they that's where all the, I mean, I would say computer simulation is a terrible answer generating machine but great question generating machine and that's one of the, one of the questions that would naturally come out of that. Okay, I think if we can do the second part of the question and the virtual cause remind me of synthetic boreholes we create and we're looking at sediment movement in the FEN basin. This is very useful approach to integrating field and model data. Could you say a bit more about your initial conditions did you try to scale rainfall in relation to climate data. And how did you decide how much sediments on the landscape initially did this vary spatially. Do you have different particle sizes in your sediment. Though, again, very relevant to our questions and I'd be interesting in looking at the fan based stuff and the landscape is is treated as if it was a homogenous unit. So it erodes and deposits but there is there's there are no bits of it that erode quicker or slower than others and and the the rainfall is based on contemporary data. And, but like I say that's just that that is, and that's just the easiest thing that was to do at the time that's that can certainly change model tools the one that I've had least time to do work on actually model ones and and yeah the how much sediment was on the landscape initially basically the landscape is entirely made of sediment was no bedrock so it is infinitely erodable although it doesn't it doesn't in effect it doesn't particularly much and so and no no no particle sizes or anything like that but he's certainly and it's only something that can be included and is something that I'm aware isn't included and has to be taken into account with with how you use data.