 Right, so thank you so much for attending to this lecture. It is a pleasure for me to be here presenting this research about the spatial structure of Galician-Megalific landscapes in the August of the Iberian Peninsula. We are going to present a case study in Montepenida region. This presentation is actually part of my current postdoctoral project at the University of Santiago Compostela and the main aim of the project is to analyze the locational preferences of the boroughs in Galicia to try to understand how the communities use these monuments not only for their deceased, but also for the living communities. And as Michelle said, the work was done with professors Andrew Bevan and Mark Lake from the Institute of Archaeology from the UCL. Well, first of all, a general context just to know where we are. The study area is Galicia in the northwest of the Iberian Peninsula as you can see on this map. We currently have there a database of more or less 3,305 sites. Some estimations in bibliography raise this number to more than 7,000 sites. But I mean in an area in an area which more or less has 30,000 square kilometers. I have to say that this database has several problems. The main one is the very limited available radiocarbon dates for more than 3,000 of sites. We only have 56 radiocarbon dates. So it's a very reduced sample of radiocarbon dates. The majority part of the sites are unexcavated boroughs. Sites there are typically dated by Anderec chronology. So for example, if we have positive structures, we can feed them into the megalithic period. But I mean, the majority part are unexcavated mounds. So the other solution that I actually did in my PhD was to try to find a pattern in the landscape of these monuments just to consider them as megaliths or not. Just to see if they maintain a pattern in the landscape or not. The study region that we are going to see in this presentation is Montepenide. Montepenide is a concentration of 120 sites. And the most important cluster of sites is here in the north. I'm pointing out 56 sites here. And as well, I mean, as you can see on the slide, we have megalithic structures, for example, in the upper part of the slide, but as well unexcavated mounds. This area, I have to say it was researched in the last century, was some mounds were excavated. You can find some plans of the excavation here. Well, concerning the locational factors, previous works in this area identified relevant environmental variables, such as, for example, the proximity of barrels to transit areas or the location at prominent places in the landscape. The works actually of Felipe Criado go in that sense in Galicia. In this study, however, we want to keep things as simple as possible. And we just consider two locational variables. The elevation and the distance to major watershed boundaries. This was based on bibliography work. I mean, for example, the works of Richard Bradley in England go as well in that sense. So we wanted to know if the elevation and the distance to major watershed boundaries could explain, at some point, the distribution of sites in this area. The general map of the megalithic monuments here in Montepenide documents a non-random distribution over the study area. And we can use statistical methods to estimate if there is a trend behind much of this variation. I mean, this is something that we can informally conclude. I mean, it can be informally concluded just seeing the map. We know, for example, just seeing the map that these points are clustered. And these are alone, for example. But we can use statistics to formally establish these conclusions. So in this work, we analyzed the relation of both covariates, the elevation and distance to watershed edges, with a dependent variable. In this case, the presence of megalithic sites. And this was done in our statistics using, as you can see on the slide, a non-parametric summary of the univariate relationships between them. Well, I mean, this is statistical terms. But the most important thing that you can see on the graph is the megalithic site intensity as a function of both covariates. And the conclusion, the most important thing is the conclusion, is that the results indicate that sites, in this case in Montepenide, are more likely to occur at elevations, ranging from 400, 500 meters of sea level, and a distance really close to the watershed edges than expected by chance alone. In the last part of this research, we built a first-order logistic model that you can find on the right part of the slide, using both covariates to produce a prediction surface. This is a classical approach in the prediction modeling. And we used this in a second stage to study the clustering trend of the sites. This was the aim of the prediction surface that we built in this part of the research. Well, we have shown that the elevation and the watershed distance, to some extent, can explain the spatial pattern in distribution of sites in this region. But you can also investigate if there are additional, more social causes behind this variation. So we do this in this case by using the prediction surface that we created in the previous step to detect whether there is additional second-order clustering or regularity in the sign point pattern across multiple scales. In order to do that, we used a percorrelation function, the graph that you can find on the slide, which confirms that the points are spatially clustered at distances up to one kilometer. So you can find this conclusion, because the black line is remaining above the gray envelope in both graphs. The first graph, the A-graph, is a random model, and the second one is a first-order model. In the first-order graph, in the B-graph, which you can find is that the gray envelope is conditioned by the prediction surface and demonstrates that although the first trends are important, they do not account for the overall clustering. I mean, the gray envelope, the black line is not inside the gray envelope. So this is actually pointing out that there is more clustering in the sites which cannot be explained by the prediction surface, by the first-order trend, by the watershed edges, and the elevation, by these factors. So there is clustering, which is more likely to be endogenous in the pattern or what is called in statistical terms second-order in character. Marking, I mean, the historical thing here is like, the presence of one site is marking a nearby site more likely. This is the conclusion that we can get from the B-graph in this case. Well, having identified this clustering, we can now fit a known point interaction model to this observed pattern just to continue understanding the pattern of the sites. And this is what we have done in Graph C. There are several point processes that we can fit here, but we used what is called in statistical terms the area interaction model because it has like a more interpretative salience. This statistical model what it does is generates inhibition and clustering patterns with reference to a buffer created, I mean, around the points, which can be interpreted as a kind of scenario in which monuments have an area of influence within which they attract or repeal other monuments. And the results are shown in the Graph C on the right part and show that, I mean, the black line as you can see now falls entirely inside the gray envelope. The gray envelope is conditioned by the prediction surface and a strong second-order clustering of sites. So in conclusion here, the elevation, the water chain boundaries, jointly with a strong clustering between sites are explaining the distribution of the megalithic sites in this region. Well, given this observed clustering of megalithic sites, we can further explore the nature of this clustering and what it might imply in terms of social organization. Thus, we use a marked correlation function to consider the spatial correlation of mount volumes, the size of the mounts. And we are able to see in this graph that there is an evidence for significant autocorrelation of these sizes spaced about 4.5 kilometer apart. Put very crowded crudely, it would appear that whatever process detains tom size, it repeats approximately 4.5 kilometer intervals, which further implies that we can use, that it might be possible to detect meaningful groups of sites by clustering the tomes using a threshold of approximately half that size. And this is what we have done on this map with DiviScan grouping analysis in RAS GIS. We generated nine groups of sites, which make visual sense, as you can see, on the map. Groups are concentrated by proximity, and we used a threshold of 2000 meters. I mean, this can be extrapolated from the marked correlation function graph. Well, I'm thinking about possible social interpretations of these groups. An obvious further question is whether the mount sizes found in each group exhibit a kind of a structure, a non-random hierarchical distribution. In which each group contains at least more or less, I mean, at least larger tomes followed by a medium-sized tomes, and finally a number of small tomes. So a year in the inside, I mean, the spatial pattern of the sites, considering the mount volumes. In order to analyze this idea, there are no methods to analyze this idea, or at least we didn't know. So we devised a Nobel test in our statistics, based on permuting the rank size, the rank or the size of the mounts. The approach is explained here in the slide. I know it's a lot of text, but it's the only way to explain this method. I mean, and the and operates as follow the method. First, we rank all the tomes in the sending order of size. So the largest is ranked one, and the smallest is ranking N, I mean, where N is the number of mounts. Next, we create a hierarchy of tomes size ranks for each group, so that the biggest tom in each group is placed in hierarchical level one. The second biggest tom in each group is placed level two and so on. And the result is a set of tom sizes ranks for each hierarchical level, which then allow us to compare the mean and the sum of the ranks with, I mean, a random sample. This is the main point of this approach. And the most important thing as well are the results. The results of this approach allow us to conclude that the largest tomes are distributed across the groups in a way that is broadly, even if not perfectly hierarchical, to an extent that is unlikely to occur by chance alone. So in a certain sense, what we have modeled here is that we can say that there is a structure behind the location of these tomes, which was not known before, a structure inside, I mean, considering the mount volumes in this sense. And some general conclusions just to finish. Well, first of all, the spatial modeling that we perform reveals clear site location patterns and interactions here in Montepenide. Sites are concentrated at the specific elevations, 500 meters of our sea level, more or less, close to ridge lines that define the main water sheaths draining to the sea. Once these major locational trends are accounted for, sites still exhibit clustering within one kilometer each other, probably implying that once some megalithic mal landmarks were established, this encouraged the construction of nearby new ones. Also, clustering analysis reveals subgroups of megaliths with non-random hierarchies of site sizes. So there is a structure behind the construction of this megalithic landscape. This can be seen as well when analyzing the spacing of mounds, which seems to reflect some kind of social partitioning of the landscape. Historically speaking, we can conclude that the people in this region were rather built-up mound clusters on the upland side of local territories on ridge tops draining to the sea. And that's all. Thank you so much for your attention. This is the work that we published. I mean, you can find all this research published in this work. Thank you so much.