 dedicated to the Saint-Egar. Sandy is a professor of global paleoclimate and biogeochemical cycles in the School of Archaeology, Geography, and Environmental Sciences at the University of Reading in the UK. She's also a distinguished visiting professor at Tsinghua University in Beijing, China, and a co-director of the Liverholm Center for wildfires. Sandy joins us today, I just learned, from a 14th century house in Devon. She studied geography in Cambridge, did a master in geomorphology in Macquarie University. That is in Australia, I learned. And earned her PhD in Sweden in culinary science from Lund University. Reading through Sandy's bibliography is like a time travel through milestones of climate research in the late, starting somewhere in the late 80s, when modeling still was really new in climate research. And it reflects a long list of her contributions to ever more precise assessments of past climate change. In brief, one could say that her main research focus is on the interactions of climate and the terrestrial biosphere, both for past times and for the present, from regional to global scales. And she's using both models and data reconstructions. And this frame, a particular also renewed or more recent focus, is on improving land use and land cover patterns and reconstructions, both for models. And here is where also archaeology comes into play. Who, if not archaeologists, can inform us about how people managed land on demographic variability in the past? Yeah, well, detailing on Sandy's rich record and many, many key contributions would be a feature length enterprise for itself. You are not here to learn about Sandy, but to her keynote. The title of her presentation today is Interactions between Climate Change, Natural Ecosystems, and Disturbance Regimes, and Anthropogenic Activities during the Holocene. Please send me. Thank you, Mara. I'm very happy to be here today to talk to you about interactions between climate change, natural ecosystems, and disturbances. I'll now try and make this slide live. So what I want to do today is to talk about some new developments that I think will help us to understand the interactions between climate change, natural ecosystems, and disturbance regimes, and anthropogenic activities during the Holocene. So I don't have to tell this audience that the Holocene was characterized by significant increases in human population and by the transition from hunter-gatherers to farming lifestyles. These changes would have had an impact on natural ecosystems and landscapes, and changes in landscapes could have affected the climate system through changing land surface properties, particularly changing land surface albedo, surface roughness, and evapotranspiration rates. Changes in climate and vegetation would in turn have affected the resource base for humans. So I'm arguing today that to gain a full understanding of human-environment interactions during the Holocene, we need to understand the role of climate, and we need to find ways, clever ways, of distinguishing between climate change, natural ecosystem change, and anthropogenic activities during the Holocene. So nevertheless, human changes were set against a background of changing climate. I'm putting a very old diagram up here because it has some human periods written on it. We know that climate has been changing through the Holocene, and these changes would have had profound impacts on natural ecosystems and disturbance regimes, and the changes in climate and vegetation would in turn affect the resource base for humans. The difficulty that we face in unraveling cause and effect here is that all of these elements interact with one another. So the climate determines vegetation, natural vegetation patterns, and natural vegetation patterns determine the disturbance regimes, including the natural fire regimes. But on shorter terms, from seasonal to inter-annual to multi-decadal, climate determines the kinds of weather that influences fire and fires themselves then all to the vegetation structure, and so there's a feedback loop there. And then on top of this, we have climate and vegetation and fire regimes determining the resource base for people, and yet also people affecting climate, vegetation, and fire. So we need to be clever, and what I'm going to argue is that we need independent sources of evidence for each of these elements that we can then compare, and we need statistical or possibly modeling tools that will help us to investigate the mechanistic causal linkages between them. And in this talk, I want to highlight some relatively new approaches that people in my group are making that will enable us to make stronger reconstructions that are independent of one another. So I'm going to talk a little bit about climate reconstructions, about vegetation reconstructions, about fire reconstructions. I'm going to make an advert about population reconstructions that seems to have taken up quite a lot of this conference so far. And then I'm going to talk briefly about the kind of statistical or modeling tools that we might use that would help us to come up with mechanistic links between these records. So let's start, first of all, with climate reconstructions. The most widely based tool for making quantitative climate reconstructions is to use pollen records, which are extremely abundant on all continents and can be dated using radiocarbon. And there have been multiple approaches, but all of these approaches are based on some kind of an analog technique between, so in other words, finding analogs between the modern pollen assemblages and specific climate variables. And there are two basic issues with this. The first is that if you apply modern analogs, you need to have analogs in the past or the modern day has to provide an analog for the past. And one of the things that we don't have analogs for today in the past is the fact that we had lower CO2, not only during the glacials, but also remember that CO2 during the Holocene was about 100 ppm lower than it is today. So if you're looking for analogs for today and you don't take CO2 into account, you have a problem because CO2 has a direct impact on water use efficiency of plants. And that can mean that the vegetation looks drier under those CO2 than it actually was. Now, I'm not going to talk too much about this because we do have modeling tools, plant physiological modeling tools that enable us to take account of the CO2 changes. And that's what that rather complicated diagram at the top is showing. And I'm happy to talk about that if anybody wants to later on. The other thing I want to talk about now are the fact that all of the techniques that we have for making quantitative reconstructions statistically tend to compress those reconstructions towards the center of the range, the climate range that they're sampling. So I'm showing this diagram at the bottom here where I'm showing a very widely used technique. So weighted average partial these squares and showing reconstructions of winter temperature on the left and summer temperature on the right. The black line that goes through zero would be if there were no biases in your reconstruction. So the points of the reconstructions showing the biases in the reconstructions and the red line shows the strength of that bias. And what you see is that you have an overestimation at the cold end at warm end and an underestimation at the cold end, for example, in temperature, which means that you are compressing the climate towards the middle of its range. Now, why is that important? Well, one of the reasons that's important is it means that these kinds of reconstructions are going to underestimate the real climate variability. What I want to showcase here is some new work by my student Meng Meng Liu who has been developing a new method to make climate reconstructions that weights the climate reconstructions by the tolerance of the pollen taxa that are present in it and also to allow for over sampling of the ranges uses a frequency weighting. So this is called FXTWA PLS and you can see the reference to the description of the technique at the bottom there. And just to make the point that this method produces more reasonable reconstructions, at the top we have reconstructions based on the standard WA PLS method and the bottom shows the reconstructions made with this new technique and you can see that there is much less compression towards the center of the range, much less bias at the extremes. The diagram on the right here shows the impact of that. It shows two reconstructions, one with WA PLS and one with TWA PLS. And you can see that the FXTWA PLS method produces much more variability in climate and the histograms on the side there just quantify that. You can see that we're getting almost twice as much variability in, for example, winter temperatures using this new technique. So I'm arguing that this is something that we should take seriously if we really want to understand how climate has changed in the past. Now, Mung Mung has applied this to a set of individual pollen records from Iberia, so I'm putting this up specifically for Mara to have a look at. So these are individual pollen records. It's the anomaly from present in winter temperature, summer temperature and a moisture index for each individual record going from the early Holocene at the bottom of each diagram towards the present. Those are a little bit difficult to interpret, but you can then look at these patterns averaged across the Iberian Peninsula through time. What we see here is that there's a gradual increase in winter temperature, MTCO through time, which mirrors January insulation changes. However, the changes in summer temperature are not strongly correlated with July insulation. In fact, they're relatively flat compared to the declining July insulation. And that's because they are highly correlated with changes in moisture in the Iberian Peninsula. And you see that in the diagrams on the right, both for the modern day patterns and for the fossil percentages. One of the interesting things that this analysis shows is that there was a significant change in atmospheric transport into Iberia during the early to mid Holocene. So here we're showing a plot. So today, remember, there's a very strong gradient from west to east in terms of moisture and precipitation across Iberia with wet on the coast and dry inland. What you see here is that that gradient was very much changed in the mid Holocene from about 8.5 to 3.5, so significantly different from present, such that it was wetter in the east and slightly drier in the west. What we're seeing here is a much stronger advection of moisture into the continent. And this would have had profound effects on vegetation and on resources for humans. So let me now move on to the second type of reconstruction. I want to talk about pollen-based vegetation reconstructions. Now, again, there are multiple methods that are used to make vegetation reconstructions, but the most common approach is called biomization. And that's been because it has very low data demands. You have your pollen assemblages. You have an allocation of the pollen to a specific plant functional type and that plant functional type to biomes. And that enables you to reconstruct, as you see on the right here, the vegetation type at any time in the past. So here we have modern 6,000. And it's supposed to be the last glacial maximum, but you can tell how old this is because at the time we put it down as 18,000 GSBP because it was before we knew about radiocarbon calibration. However, there are, unfortunately, some very clear known problems with this kind of reconstruction of vegetation. There's a great deal of subjectivity in allocations of pollen taxa to PFTs and biomes. One of the things about the biomization procedure is that it has a large element of expert manipulation and is iterative in order to produce the best possible reconstructions of the modern, which you then would apply to the past. So this subjectivity is rife and needs to keep it in mind. It's also a difficulty in the sense that pollen is transported by long distance transport from forested regions to open vegetation. And so it becomes very difficult to reconstruct open vegetation using this method. And finally, if you want to actually look at changes in vegetation through time, so down a core, the method is extremely sensitive to very small changes in pollen abundance. So this has led to a problem that's often referred to as the flickering switch problem where you shift between different biomes when you only have a relatively few changes in the pollen abundance. So maybe a few percent is enough to shift you from one biome to the other. So we really need a more objective way of making such reconstructions. And what I want to suggest to you now is that a new method that's being developed by Esme Cruz-Silver, which essentially characterizes vegetation, not in terms of expert opinion of what should be present in the vegetation, but in terms of what actually is present in the vegetation today. So it uses pollen abundances in particular biomes rather than going via plant functional types. And it uses the observed distribution which can vary across a biome quite substantially. Just to illustrate this, here's two different vegetation types. On the left, it's cool mixed forest and on the right, it's gaminoid and four steps. And what you see is that when you characterize these two biomes using modern observation, it is true that characteristic species are dominant in the assemblages. So in the cool mixed forest, for example, you see that Pinus and Phagus and Piscia are all extremely abundant and other tree species, again, these are all shown in blue here, are represented, but two things. First of all, the relative abundance of those can vary quite considerably, as is shown by the boxes here. So if you're looking at the Pinus, it can go from less than 10% to nearly 50% within this one biome. But it also shows you that there are other contaminants or other plants that are present that you wouldn't think of as typical. So here, the yellow cells here show open, ground-story vegetation, grasses, and things like cycloracy. And you can see exactly the same in the step vegetation here. The yellow show you the things that would be considered typical of this biome and the blue show you contaminants, such as trees that are coming in. Okay, taking this into account, we can now characterize biomes in terms of the observed variability in the assemblages. And then we can use that observed variability with a dissimilarity measure to actually calculate what the biomes should be. And I have to say that this technique without any intervention, any subjective decision-making produces a very accurate simulation of modern vegetation patterns here. On the top, you see a kind of quantitative confusion matrix estimate of how well the predicted biome does against the observed biome. And at the bottom, you see a map that shows the distribution in space, which if you're used to the vegetation of this area, you would say is a reasonable representation of the biome distribution in this region. The other nice thing about this technique is that you can apply it down core. And when you do that, you can get an estimate of the affinity of a particular record here. We're just showing six specific records from the Eastern Europe Mediterranean Black Sea Corridor area. So six specific records and you can show the affinity of the pollen to a number of different biomes. So you can see exactly which biomes are important in this region. You can see interesting transitions in all of these cores, for example, from more open vegetation shown in the lighter colors, the reds, the yellows and oranges here towards more forested vegetation. And you can see that specific biomes do persist through time. They're not switching on and off as they were in standard biomization method. So again, a new technique that I think will enable us to make more robust reconstructions of vegetation through time. Okay, now I want to move on to fire reconstructions. And again, I want to talk about what's wrong with what we're doing at the moment and then a new method that might be used to make more robust reconstructions in the future. So at the moment when we're making fire reconstructions, we use charcoal records in sedimentary settings, lakes, bogs, soils, and we reconstruct the abundance of charcoal down the core through time. And then if you want to know what's happening to regional fire regimes, you can put those things together and over a global or a regional area, which is what's shown in the map at the bottom there. And then you can composite records from a specific region in order to come up with trends of fire through time. So what we see on the plot on the right, and again, this is a relatively old diagram that makes the point is that you can composite, for example, all the records of the northern extra tropics, shown in the second panel down here on the right, you can see that fire was low during the glacial and early part of the deglaciation and increased as we moved into the Holocene. And that superimposed on that general increase, there are wiggles times when there was less fire, times when there were more fire, and do note this time of high fire around 2000 years BP, which I think is an interesting phenomenon, which we still haven't explained. And just to make the case that this is not an artifact, if you look and contrast that with what's happening in the southern extra tropics, you see a very different pattern. So that's the fourth panel down here, a very different pattern of fire history in the southern extra tropics. Well, this is all very well and good, but there are problems with charcoal. So one of the problems with charcoal is that there are many different types of measurements and so that are made on individual cores. And so we have to find some way of standardizing those measurements across sites. So that technique has been developed, but of course it means processing the data and that is always problematic. The second problem with charcoal is that it's unclear whether charcoal actually represents burnt area, burnt biomass or some other aspect of the fire regime. And it may well be that in different sites and different regions, it reflects different things. And finally, it only provides a qualitative estimate of relative changes. We'll talk about there being more fire or less fire. And as yet it's been very difficult to calibrate charcoal records, except at individual sites to come up with any kind of quantitative estimates. But if you're going to compare fire with climate or fire with people or fire with vegetation changes that are quantitative in nature, you need also to find a way of making charcoal quantitative. And especially if you want to compare it with model results, you need to have a quantitative estimate. Now, one of the nice things about fire regimes is that the vegetation itself reflects or has traits that vary with the fire regime. So here, one of the things that we've been looking at in the modern is the abundance of trees and woody species that resprout after a fire versus the abundance of those that don't resprout. And this diagram here shows you that there is in fact general tendency for resprouters to be much more abundant. So that's the panel on the bottom labeled A here, much more abundant when the fire regime, the fire return time is short than they are when the fire return time is long. This means that there is information in poll and assemblages about fire return time and we can exploit that in order to make reconstructions, quantitative reconstructions of fire return times or conversely the converse of fire return times which is burnt area. So that's what I'm going to show you now. The way we do this, and this is some work done by my student, Yicheng Shen, is we derive a relationship between fossil pollen and fossil charcoal. We then calibrate the modern charcoal against modern burnt area. And in this case, we actually use modelled burnt area to reduce the noise because of the short observational record. And then we apply this calibration to pollen records which then enables us to have a quantitative estimate of burnt area through time. So again, we're applying this to the Iberian Peninsula. Bit biased here, but there you go. Let me first show you the modelled relationship between burnt area and climate vegetation and human activities. So this is a GLM here where we're using a number of climate variables including diurnal temperature range and dry days per month. We're using a number of vegetation variables such as gross primary production and non-tree cover and we're also using human indicators like cropland or grazing land and population density. And essentially this enables us, this model enables us to predict the large scale patterns of fire across the continent. So we can use this in our calibration. When we do this calibration, we come up with nice quantitative reconstructions not this time of more fire or less fire but of burnt area fraction across the Iberian Peninsula. And the nice thing about this is that you can see that this burnt area fraction does mimic the climate changes, broad scale climate changes across the region. So we see the Burling Allerad, we see the increase in fire at the end of the younger dryers. We see the low interval of fire during the wetter mid Holocene. We also see an increase in fire towards the end which might be a reflection of climate but could also be a reflection of human activities on the landscape to be discovered. So again, we have another way of getting an independent and quantitative estimate of fire oceans. Now I was originally gonna talk about population changes but then there have been so many sessions so far on this that I'm not gonna talk about reconstructing population changes from radiocarbon dates or anything else but I will make an advertisement for my student, Lee Sweeney who is talking on Saturday, trying to relate reconstructions of population with fire history. So please do look in on that if you want some more information about putting these independent variables together and trying to make sense of them. Okay, so I hope I've made the case that we can actually make independent reconstructions of many of the elements that we need to do to look at these interactions between people, climate and the environment. What I want to do now is to give a case example of using a model. Now this will only be applied to the present day but I am looking towards applying this in a paleo context so watch this space. What I want to do here is to look at crop modeling and here we've developed a very simple light use efficiency model that predicts gross primary production very well globally and then we use this prediction of gross primary production with empirically derived equations that relate that to above ground biomass and to seed yield and what we're actually applying this to at the moment is wheat but it can be applied to any other kind of crop both C3 and C4 but I'll give you an example for wheat first. So just to make the case that this model makes a good prediction at the top here we're showing a global prediction of potential wheat yield in response to climate, CO2 and management and we're comparing that with the observed potential yield from the EarthStat dataset which I think was for the year 2000 and you see that the broad scale patterns here are very similar to one another. Now that's fine so we can predict wheat yield but we can observe it too so why would we want to do that? Well, the whole point about this is that we now have a model that allows us to calculate the gap between potential yield and actual yield or observed yield and more importantly we can diagnose whether the change in the yield gap is due to changes in climate or changes in CO2 fertilization or changes in management and what we've done here is we've run a couple of different simulations in one we've held climate constant over the period 2000 to 2015 and the other one we've held CO2 constant and allowed climate to change and this enables us to diagnose the fact that CO2 has in many regions the increase in CO2 has had a positive effect on yields and offset negative effects on climate. So here we're able to use the power of the modeling work to diagnose causality in terms of changing properties. The other thing that's nice about having models of this sort is that they can be used to what if games and I think these are the kinds of games that we need to play during the Holocene. So here as an example of this is the question what is the optimal time to plant wheat? We have the model, the PC model in the middle here. We can do a massive set of experiments where we basically change the planting date through time it could be we plant in winter, we plant in spring, we plant on a different day during the winter, we plant on different day during the spring. You do thousands of runs of this sort and you calculate what the yield would be for each of those planting dates and you would find the optimal planting date. Now this is making the assumption that people will want to plant at an optimal time to maximize their yield. And we can look at this and we can then see how well our predictions of what the optimal planting date would be correspond to the actual observed planting dates in the model. So at the top here, we've got spring wheat and we're looking at the census data of when spring wheat was planted versus our predictions of when wheat was planted. And this in the bottom right hand corner, we're looking at winter wheat again, comparing the census data of planting date with our predictions. And again, you can see that this model is able to choose to find an optimal solution for when farmers should have been planting wheat and indeed to make the distinction between whether it makes sense to plant spring wheat or plant winter wheat. And for those people who were wondering about the transition of wheat growing in Europe, this offers us a tool where we could actually work out why we have a transition from spring wheat to winter wheat during the whole season. Okay, so I'm going to finish there, but I want to finish with three important for me take home messages. The first of these is that I do believe that disentangling the interactions between climate environment and people during the whole scene is possible, given the fact that we can have independent reconstructions of each of these elements and increasingly powerful statistical approaches to separate them out. I'd also like to make the point that we can use existing modeling tools not just crop models, but also climate models and vegetation models and fire models, all of which exist. We can use these tools to look at how climate and the resource base could have affected humans or conversely what the role of humans is in affecting this resource space. So we can play games with these models which would I hope enable us to get a better handle on the mechanistic links between climate environment and people. And finally, this is a plea from my side. I'd be very lucky during my career, my long career as Mara pointed out to be able to work with modelers, with paleo environmental scientists and increasingly I'm working with archeologists. And I think that there is real benefit to be gained by engineering a closer collaboration between these communities if we really want to make reliable reconstructions of human impacts on the environment and on climate. And so I'm asking, making a plea for us all to work together more closely in the future to address this issue of interactions. Thank you very much. Thank you very much, Sandy, for this absolutely fascinating, very dense presentation. Sorry. I love the yield gap. And it brings me straight to a question because I was asked to do the first question. I'm supposed to guide the discussion here and it's predicted that all of you need some time to sort themselves and to write down your questions and comments. So I take the chance to do the first one about this yield gap or what time scales is that applicable for, if I think there's so much inter-annual variability to predict an ideal planting time. So it is working on inter-annual variability. I think in general, farmers take a few years to shift. And so I think they're not gonna be changing the day massively. They're not gonna go from winter week to spring week within a year, but they might and they do change by a few days each time. So I think that they're responding to try and get optimal conditions. One of the things I'm putting this out as a question, a question to archaeologists, I understand that there was a significant pause in the spread of week growing as you go north in Europe. And I don't know exactly when it was, but somebody will remind me. And one of my questions is, is this because you have to have a transition from spring week to winter week in order to get optimal yields? So this is the kind of tool that would enable us to look at that question and see whether this is an optimality issue that people were being sensible and saying, we can't grow this because we don't have the appropriate cultivar. Yeah, thank you. I see quite some comments. Didn't detect all your question, yes. Lots of nice comments, but please some questions. Oh, what's the safe point to? Oh, that's a question, sorry. Is that a question or just a question? Or Katarina, is that when the hiatus happened? Is that a comment or a question? I think that might be when the hiatus happened. And there's another one one could interpret in terms of a question. It's the expression by Linda Scott Cummings to know more about this modeling. Which one, please? Ah, yes, Linda. Could you specify this question a little bit, please? So Katarina, thank you for telling me when this hiatus was, I will get on to it. Is Linda with us? Seems not to be the case. I just, maybe one thing I should make a comment on here. I was showcasing this crop modeling because I think it's probably more of more relevance, direct relevance to the archeological community and they could play with these kinds of models which are quite straightforward to run. But I think we should not forget that you can actually also model vegetation and you can model fires and you can model climate and you can play the same kind of games with those. In other words, you can say, you know, what happens if I don't have any humans on the landscape? What does that do to my fire regime? And in fact, in the modern day, the answer is very interesting in the sense that there are some environments where if you take humans out, you don't actually increase the amount of fire or decrease it. And there are other environments where you either increase it or decrease it depending on whether they were adapted to fire or not. We cannot say, Linda is now satisfied that she's, she cannot unmute her, she's typing in the chat. Okay. Okay, but there's another question by Doris Barboni. And she's asking whether European, when European farmers should start planning to cultivate millet in southern Europe, any predictions related to the IPCC predictions? We're running it Doris. Thank you for that question Doris. Good to see you. We, so we do have a version of this model that does C4 plants and we have run both models with IPCC futures. So we are going to be shortly looking at when would be the best time to translate to either millet or maize, which is I think what we probably will transfer to. But yeah, it's in the future. And of course the other side of it is that there could be shifts in terms of moving to more winter wheat. As the winter's warm and you go to crops or cultivars with shorter vernalization requirements. So there will be changes in the kind of cultivars that we're using I think as well. Okay, so meanwhile we have a specification of Linda's question. And she would like to know more specifics about FXTWAPLS. Yes, I'm very sorry about the acronym. Okay, so the FXTWAPLS, which you have to practice saying has two components. The first component is the tea part of it. So essentially this is the sensitivity of every individual taxon to climate. So you actually make an adjustment that allows things that have a very tight climate range to count more in the calculation than ones that have a very loose climate range. So you can imagine there are some plants, polygonum, that go everywhere and they really don't have a lot of climate information to them and yet in the original scheme they're still counting equally. So that's the first thing is that we have this tolerance waiting that refines the climate that's being reconstructed. But then the other side of it is that you have a lot of... Well, because the pollen is not sampled evenly, you can have overrepresentation of some parts of the climate range. Because of that sampling and so that's one of the things that causes this compression towards the middle. So what this is doing is actually sampling the climate range evenly to take account of the frequency of sampling and that again is a set that reduces the compression. So both of these are, if you like, stretching the climate that you can reconstruct. And it is described in the paper that was there and if anybody wants a copy of the paper you can email me and I'll send it out. Yeah, that's good because there is another comment that Katarina is looking forward to read your paper so you want a new reader. Okay, here you see the comment on the screen and Doris Barboni comes back with a question. What about... Yeah, what about the sensitivity to the trend? Is it taking you into account as well? Yes, so yes, the answer is yes. So the three variables that I was showing in this presentation and the ones that we tend to reconstruct are winter temperature, mean temperature of the coldest month. Some measure of summer warmth usually done by growing degree days and then either alpha which is the ratio of actual to potential adaptor transpiration or some kind of other moisture index. And when you make those reconstructions you are basically taking account of the sensitivity to each of them separately. So your analyses for the alpha or for the moisture index will take into account the sensitivity to drought. And that's one of the reasons why the taxa that contribute to making these reconstructions varies between the three variables because some taxa, as you know, perfectly well Doris are very sensitive to drought and don't care about the temperature whereas others are very sensitive to winter temperature and don't worry about summer and vice versa. So this is it. I've got post, I've got post my email address. I'm not sure how to do that. Oh, okay, right. I'm relatively easy to find on the net. So I don't know how to, I don't know how to post in this chat. That's great. Is there more questions? I would have one myself but writing it down in the comments list. I'm in principle a big fan of the biome idea and was wondering recently whether it would be possible to, there's also a thing called antrobiomes and that of course could be an approach and also for archaeologists to revisit that idea and maybe to refine the whole in changes. Yeah, no, I think you're right. And this method could be used to do that. I mean, I'm not a great fan of biomes. Not any longer. No, I mean, biomes are a convenient way to divide up the world if you wanted something simple. So for a modeling point of view, you wouldn't do 15 just different types and from a reconstruction point of view, you don't want to have 5,000 different types. So that's why we kind of first went with the whole idea of biomes. But I mean, as those graphs were showing, there's considerable variability across the biomes. So, you know, going from the north to the south, the relative abundance of the things that make up that biome changes quite a bit. And I was reminded by this biome where my colleague, Laili Sasa, some time ago when she pointed out to me that we have this thing called the Boreal Forest and in Russia they have 5 different categories of Boreal Forest because they are very different as you go from south to north. So the nice thing about this technique that I'm pushing here is that it does take that variability into account and it provides you with a measure of roughly where you are relative to the centroid of that biome. So it says, yes, it's in this biome but it's really on the northern fringe or yes, it's in this biome and it's right. Now, in terms of anthromes, you can do exactly the same by we could take areas of farming and we could actually look at what pollen is represented in areas of farming and one could look at that in terms of the intensity of farming or the farming types or the crops that are being grown or whatever else. And then you might get around the problem that in many pollen diagrams you might have cereals present but you don't see that they were growing peppers or beetroot or whatever else. So this would be a way, I think, of getting a more nuanced approach to different types of agriculture. At least I do. And I'd love somebody to try it so if anybody out there wants to volunteer I'll hand over the scripts. Now that you said you are not a big fan but you will try that. Well, no. It's not that I don't think we still need to use biomes and anthromes or whatever because we have to divide the world up into categories but I just want to make the point that those categories are a continuum so to some extent we're dividing up a continuum and we do it arbitrarily based on what we think is important. They're not fixed entities in time or space. Absolutely true. It would be a chance to really go more for non-use practices or even cultural differences. Now it's like pastoralists and farmers and that doesn't help very much. So Linda's volunteering by the look of it. Okay, Linda, you're on. Great, thank you. No fire questions because that was also a very sophisticated approach to really better approach it quantitatively. And I know that many Angulatists are dealing with counting charcoal particles or nobody here right now. I think one of the things that would be interesting to do is to compare this with or compare these sorts of reconstructions at a wider scale with what you might infer from the charcoal records themselves. And that's something that we will do but I think it is important that we do try and quantify these things. Absolutely. We've only done it in Iberia so we need to do it for a wider area. Yeah, and the wider area per se is also a hint that it could be natural variability and not local human activity. Just a comment. Thank you, Gail. One thing that I'd like to just comment on is how wonderful pollen is for anybody out there who's doing pollen because there's so much information in the pollen assemblage. And as you see, we've been able to tease out independent reconstructions of fire, climate and vegetation from those pollen assemblages just simply because they are so flexible and they're responding to different things different texts are responding to different parts of that. Absolutely. And there's such a wonderful data set which has grown over decades and it's there for new experiments. But it's not so easy to access because we tried ourselves and then you have often records with no age model coming together like the database and it's time-consuming too. Those treasures. Yes. Oh, there's another fire fan. No, no. It's Scottish prehistory. I think people just want to get to work. Yes, terrific. It refers to the surroundings of settlements. Oh, is that what? Great, I was wondering what. I think that we need to be very careful here because there's two sorts of charcoal data. I mean, there's the archaeological charcoal data on the site and that's telling you something about people and what they're doing and what they're burning it for. But there's also the fact that people do set fires or might set fires to create open land for farming, for example, or to improve the grass for their cows and things like this. And it's those natural, as it were, fires outside of the settlement that we can get out with the charcoal record. Now, I think many of those are small-scale but we still need to see how much of an influence people are having on fires or possibly suppressing fires compared to the live fires. So we need to make this sort of separation between charcoal in-site and charcoal in natural settings. That's often a problem come across with this archaeological data that we are seeking for site archives and of course sometimes then you have preservation problems with coal and things on the Portuguese ditches. We tried very hard and it doesn't work simply. Well, I mean, this is the problem with all of us. We have preservation issues that affect all different sorts of records and we've just got to find ways to get over it. So the ideal archives are not precisely usually along with sites. Sonja Grimm is commenting that also the different sizes of microchannel counts, of course, matter. So the interesting thing here, Sonja, in our new data set that we just put together on fire, we've got all sorts of different sizes. We've got microchannel, macrochannel and we've got different sizes of macrochannel and whatever, it's all distinguished. Whatever is being counted. We've done some preliminary analyses at regional level where we've compared whether the records you'd get using microchannel tell you a different story from macrochannel because theoretically, if you believe the story, the macrochannel is more local and the microchannel is more regional. Frankly, I'm not sure I believe that anymore. It may be true at individual sites, but if you make a regional composite, we don't really see any significant differences between macro and microchannel. There's still to be explored further, but I think this is something that having the data out there enables us to be able to go back and analyse whether some of these assumptions that we are making are actually true. A very practical question of the set of the presentations you mentioned. I'm very bad at the number of so many session numbers, but there's a session tomorrow afternoon about Iberia. It's not that one. It is definitely Saturday morning. It's about 8.45. It's Saturday morning about 8.45. And it's loop-sweeney, so you can look it up in the whatever. And it's a first step to try putting together two independent reconstructions and saying how do they look compared to one another, which is an approach I'm advocating we all need to take. Sonja again? Just to support that approach. No question contained, Sonja. Maybe if there's no more questions, spontaneous questions now, we're approaching somehow in the end of this session. Yes, exactly. I'll just say we're the same message. My name is Doris. Yeah, I'm sure there's much more, but people will contact you anyway and read all those papers prior to contact us. Yeah, thank you. If anybody wants to contact me with either questions or to get reprints or anything like that or code, please do. I'll be very happy to hear from everybody. And thank you. Yeah, thank you to everybody also for the discussion, contributions, and for Sandy once more for this absolutely fascinating, new, clever approaches we will use. Thank you. Have a nice evening.