Conference papers

How central are local centres? Testing archaeological hypotheses of the Dutch part of the Roman limes through spatial analysis and network science
by Mark Groenhuijzen and Philip Verhagen
Presented at AIAC2018, Bonn, 22-26 May 2018

The integration of the Dutch Rhine-Meuse delta in the Roman Empire during the first centuries AD is thought to have led to an increased structuring of the settlement pattern and an increased level of interaction between the rural population and the military population that resided in the forts along the Rhine.It has been argued that the Dutch limes zone can be characterised by a dendritic settlement system, particularly for the (re)distribution of goods, wherein most interactions moved up and down this hierarchic system and fairly little ‘horizontal’ interaction took place.
Through the modelling of transport movements using least-cost path analysis, the reconstruction of transport networks and the application of network analysis, this study presents a test of this archaeological hypothesis of a hierarchic settlement structure. The proposed ‘intermediary’ sites in the dendritic system are analysed to see whether or not they actually ‘central’ for their subsidiary areas, and by extension we can see if there are possibly other settlements that may have functioned as intermediary sites, that we did not know about previously. On a more general level, this study hopes to show that by expressing our archaeological ideas into relatively simple testable hypotheses, we can use methods from spatial analysis and network science to further our archaeological understanding in a more quantifiable way.

Modelling the cultural landscape of the Dutch Roman Limes: approach, results and prospects
by Philip Verhagen, Mark Groenhuijzen and Jamie Joyce
Presented at Archaeological Colloquium, University of Cologne, 13 June 2017

Continuity and Discontinuity in Transport Networks of the Dutch Limes Zone
by Mark Groenhuijzen and Philip Verhagen
Presented at CAA2017, Atlanta, 16 March 2017

Comparison of Different Network Construction Techniques to Model the Distribution of Rural Surplus Production in the Dutch limes zones
by Mark Groenhuijzen, Philip Verhagen and Jamie Joyce
Presented at CAA-NL/FL-DE, Gent, 24 November 2016

Identifying and using uncertainty in archaeological inventories for modelling settlement pattern development in the Dutch Roman limes zone
by Philip Verhagen, Mark Groenhuijzen and Jamie Joyce
Presented at WAC-8, Kyoto, 29 August – 2 September 2016

The ‘Finding the limits of the limes’-project aims to apply spatial dynamical modelling to reconstruct and understand the development of the cultural landscape in the Dutch river area during the Roman period. In this paper, we will address the issue of uncertainty of dating and interpretation in the archaeological dataset that is the basis of our studies. We will propose formalized procedures to identify uncertainty, and to include this in our further analysis and modelling, such as site location analysis, network analysis, agent-based modelling and predictive modelling. We will present examples of applications in both model building and testing.

Experimenting with theory: how modelling can cover the middle range. A case study in modelling surplus production in the Dutch Roman limes
by Philip Verhagen and Jamie Joyce
Presented at the 23rd Archaeology and Theory Symposium, Amsterdam, 18 April 2016

In the call for papers for AD2016, the organizers mention that Binford ‘spoiled’ middle range theory (MRT) – but perhaps archaeologists have not always understood the nature of MRT. We suggest that the main problem with Binford’s definition of MRT is found in its emphasis on replicability: MRT was explicitly confined to processes that can be observed and/or replicated in the present. Where it concerns socio-economic processes, we are then confronted with a real problem, since there is no laboratory in which to replicate past societies – or is there? The current rise of computer modelling approaches to socio-economic questions – and in particular agent-based modelling – seems to suggest otherwise. These modelling techniques are now used to simulate the dynamics of social, economic and natural processes, not just in isolation, but in conjunction, and at different levels of interaction. And they make us think not just about the resulting patterns, but even more about how to build theories.

In this paper we will discuss these issues using the example of surplus production in the Dutch Roman limes. Recent research – contradicting earlier studies – has suggested that the Dutch River Area possessed sufficient resources to produce a food surplus that could be traded with the Roman occupiers. However, we still don’t know how the change of the supposedly self-sufficient pre-Roman economy to a more market-based economy came about. Was a drastic redistribution of labour necessary, did agricultural practices change, were changes in settlement pattern and territorial division required for a more efficient allocation and redistribution of resources, and why do we seem to witness population growth despite the increased pressure on resources by the Roman administration? In trying to come to terms with these questions, we had to use and develop middle-range theoretical concepts that would allow us to set up plausible models and scenarios of surplus production, and compare these with the archaeological record. In this way, we have tried to build theory bottom-up, instead of relying on (too) grand theories to explain the archaeological record.

Testing the validity of network analysis results in research on local transport networks
by Mark Groenhuijzen and Philip Verhagen
Presented at CAA2016, Oslo, 29 March – 1 April 2016

A video of this presentation can be viewed on the Recording Archaeology YouTube channel.

Computational archaeology provides valuable tools for the reconstruction and analysis of transport networks. One such approach is a combination of a network constructed using least-cost paths and network analysis, which can potentially provide valuable information regarding settlement location choice, site hierarchy, the role of settlements in transport networks and so on. However, testing the validity of the network analysis results and the archaeological interpretation thereof has so far been largely neglected. One of the key questions is thus: how reliant are the results of network analysis and their interpretation on nuances and uncertainties in the methodology and the dataset? This paper aims to test the robustness of network analysis results by measuring and analysing the development of local network statistics in randomly emerging transport networks. It is applied on a case study involving the Dutch part of the Roman limes, an area which is particularly interesting for research on local transport networks in the light of social and economic relations between the local rural population and the Roman military population, and an area for which a large amount of archaeological and palaeogeographical data is available.

Factors of production: Investigating land and labour as limiting factors in agricultural production in the Dutch Roman limes zone via agent-based modelling
by Jamie Joyce and Philip Verhagen
Presented at CAA2016, Oslo, 29 March – 1 April 2016

In this paper, we investigate the role of land and labour availability in the rural economy of the Dutch Roman limes region via agent-based modelling. The availabilities of land and labour pose limits on agricultural production and are regarded in economics as two of the primary factors or inputs in the production process determining the quantity of output. Although recent research has now prompted a departure from the previously held view that surplus production in the region was not possible (see Kooistra et al. 2013; van Dinter et al. 2014), we still don’t know how the shift from subsistence farming to surplus production occurred. The likely methods of surplus production undertaken by local farmers, the limiting factors in agricultural production within different temporal and geographic scales and possible mitigation strategies have hitherto only been dealt with in generalist terms. To investigate these topics, we have produced an agent-based model in NetLogo to simulate the rural economy of the region with land and labour costs as primary outputs. We have simulated methods and strategies in the three most significant elements of this economy: animal husbandry, arable farming and wood-fuel acquisition. By comparing the model results against archaeological data of the natural and cultural landscapes in the region, we are able to test the scenarios for plausibility. In addition, the model has enabled us to investigate the elements of the rural economy not only as separate activities but in combination, reflecting the mixed agriculture practiced in the region in this period. Lastly, we are able to simulate a dynamic economy both temporally and geographically by imposing on agents conditions known from the region. We present here therefore the initial results from the model and our conclusions in defining the limits of production in the Dutch Roman limes zone.

Overview lecture: Finding the limits of the Limes
By Philip Verhagen, Jamie Joyce and Mark Groenhuijzen
Lecture presented at Vrije Universiteit Amsterdam, 10 Nov 2015

N.B. the original contains videos, these are not retained in the pdf!

Introducing dynamics in transport network research using computational approaches
By Mark Groenhuijzen and Philip Verhagen
Presented at the CAA-NL/FL conference, Amsterdam, 22 Oct 2015

With the ever-growing pool of archaeological data in Dutch archaeology and the resulting fragmentation in archaeological research, there is an increased need for archaeologists to produce more synthesising results. One way to deal with large amounts of data is by introducing computational approaches. The ‘Finding the limits of the limes’ project aims to apply spatial dynamical modelling to reconstruct and understand the development of the cultural landscape in the Dutch part of the Roman limes, with a focus on modelling the spatial and economic relations between the local population and the Roman military population. The vast amounts of archaeological data as well as palaeoenvironmental data make the Dutch limes an ideal testing ground for developing new models of the cultural landscape. In our project we aim to construct scenarios for resource management along the limes, and finally test these scenarios against the archaeological evidence in order to gain more insight in the dynamics of the cultural landscape.

The goal of this paper is to present some examples of how computational approaches can be of added value when applied to our extensive datasets and existing archaeological research, by focussing on the modelling of transport networks in the Dutch Roman limes. Firstly, we demonstrate a model of randomly emerging transport networks using our existing archaeological data to characterise the development of settlement locations and answer questions related to the position and role of settlements in transport networks. Secondly, we show an example of how the chronological information associated with our archaeological data can be reinterpreted to create a more detailed and dynamic dataset that can be used to study transport networks in more specific time slices.

With these examples we hope to demonstrate how computational approaches can adequately deal with the rising problems of handling the large amounts of data currently available in Dutch archaeology, and at the same time how they can help us tackle archaeological questions in a relevant and innovative way.

Through hell and high water: a multimodal transport network in the Dutch Roman limes
By Mark Groenhuijzen and Philip Verhagen
Presented at CAA2015, Siena, 30 March – 2 April 2015

Computational studies of movement in archaeology thus far largely focus on unimodal land-based movement or transport systems, in which natural terrain, visibility, social and cultural factors are seen as limitations or opportunities that determine the character of movement (Murrieta-Flores, 2010). Fairly little attention has been paid to water-based movement, and even less computational work has been performed on multimodal transport systems, which can potentially utilise both land-based and water-based transport modes.

As a dynamic fluvial landscape, the Dutch river area is an excellent example of an environment where people would often require multimodal transport systems to exchange information, distribute resources and maintain social contacts. During the Roman period, the Dutch river area formed part of the Roman limes and witnessed the shift from subsistence to surplus agricultural production to at least partially supply the Roman military population, meaning that transport on local to regional scales must have intensified. Furthermore, the Dutch river area possesses extensive datasets from both archaeological as well as physical geographical research, which makes it an ideal testing ground for spatial computational approaches.

This paper aims to study the role of multimodal transport systems as part of the interconnections within the local population as well as between the local population and the Roman military population in the Dutch river area. The research forms part of the NWO-funded “Finding the limits of the limes” project, which aims to investigate the cultural landscape of the Dutch Roman limes, especially concerning the spatial and economic relations between the local and military populations, through spatial dynamical modelling.

To achieve this, a detailed palaeogeographic map is constructed for the study area that allows for least cost modelling of both land-based and water-based connections, as well as a combination of transport modes. Cumulative cost path networks (Verhagen, 2013) are then modelled for different scenarios that represent factors such as the (non-)availability of certain transport modes. Problems that arise and have to be overcome when modelling multimodal transport connections are the heterogeneity of energy consumption in land- and water-based transport, as well as the possibility that water-based transport has the capacity to carry multiple persons and goods, which can be remedied for instance by applying directional cost maps for water transport and using varying access costs for water transport modes. The resulting networks can be compared using network analysis (Verhagen et al., 2014), which can identify sites of central importance in (mass) transport networks and explain the position of military sites in transport networks. Accurate dating of archaeological sites also allows the study of the evolution of transport networks chronologically.

Keeping the home fires burning: spatial dynamic modelling of the wood-fuel economy of the Roman limes zone in the Netherlands
By Jamie Joyce and Philip Verhagen
Presented at CAA2015, Siena, 30 March – 2 April 2015

Wood in the Roman world was the most important and, in many areas, the sole source of fuel for rural, urban and military settlements (Veal & Thompson, 2008). It has been estimated that fuel in the form of wood or charcoal comprised half of the per capita energy consumption (Malanima, 2011).  It is surprising therefore that, in spite of fuel’s importance within ancient economies, few studies have attempted to engage with it as a primary resource. Commendable exceptions exist (namely van Dinter et al. 2013; Veal, forthcoming 2015), yet many deal with the palaeoeconomy in general terms or model supply to, and consumption in, specific localities. Modelling the fuel supply within the macro-regional scale is lacking.

The NWO-funded “Finding the limits of the limes” project aims to apply spatial dynamical modelling to reconstruct and understand the development of the cultural landscape in the Dutch part of the limes zone between ca. 15 BC – 270 AD. It focuses on modelling economic and spatial relations between the Roman army and the local population, in particular the interaction between agriculture, animal husbandry and wood management.  Modelling of wood management in the region presents a significant challenge as remains of carbonised wood in the Netherlands are scant. Macrobotanical evidence, therefore, of the daily tasks of keeping warm and cooking food, as well as of less regular tasks such as pottery manufacturing and metallurgy, is missing. Furthermore, the study area comprises a combination of a fluvial dynamic landscape with pre-Roman deforestation and the development of a military infrastructure with a related shift from subsistence to surplus agricultural production. Changes and continuity in strategies of resource exploitation and woodland management in the Dutch limes zone are thus ripe for investigation within a modelling environment.

In light of the above, I will propose a spatial dynamic model of the fuel economy of the Dutch limes zone. The model is constructed in NetLogo and tested via the BehaviourSpace utility. By also utilizing the GIS extension, existing palaeogeographic reconstructions and settlement patterns can be added. As a result, this tool can be used both heuristically, to investigate hypothetical scenarios, as well as simulating scenarios grounded in archaeological data. With the aid of cases studies, resource competition between settlements, the limitations posed from both the natural environment and household decisions, and strategies of exploitation and mitigation can be tested and analysed using this tool.

Simulating the farm: a role for computational modeling of the agricultural economy of a landscape?
By Jamie Joyce and Philip Verhagen

Presented at LAC2014, Rome, 18-19 September 2014

The agricultural economy is intrinsically linked with both the natural and cultural landscapes of the past and present. It forms part of the mutual relationship between people and landscape. Decisions and responses undertaken by an agrarian population are often made owing to the limitations and opportunities created by changes in the cultural landscape within the short-term, and the physical landscape within the longue durée. In turn, the reactions have themselves an impact on landscape development. Given the political, territorial and demographic changes during the early and middle Roman period in the Dutch zone, the analysis of the development of this cultural landscape undertaken by the Finding the limit of the limes project must consider the response(s) of the local agrarian population to these changes.

Until recently, the prevailing argument was that “no local community could suddenly have started to produce the surpluses required to feed the troops. This has since been rightly called for revision. Commendable and laudable, the modeling of landscape capacity employed in this revision represents the first logical step in analyzing agricultural systems and surplus production. Yet, it concerns the ability of a positive response rather than the mechanism of such a response. Furthermore, methods of investigating agricultural systems using just the bioarchaeological record reveal only snapshots of farming activities. They show the results of responses, wherein equifinality can play a hindering role.

In light of this, I will propose a simulation of Roman-period agriculture in the Dutch limes zone using NetLogo (an agent-based modeling environment). Additionally, the various obstacles within computational modeling of agricultural systems in general will be considered. In particular, how does one simulate a complex system such as agriculture, when decisions are based on both environmental and socio-cultural factors? Here, the confrontation between applicability and the minutiae must be considered. In addition, simulations show hypothetical scenarios, in this case of agriculture on a settlement, micro- and macro-regional scale. They require testing against archaeological evidence. Yet how does one verify the validity and accuracy of scenarios given the aforementioned problems of archaeobotanical and zooarchaeological assemblages? And lastly, how can the role of the natural and cultural landscape be incorporated into such models?

With the aid of case-studies from rural settlements in the limes zone of the Netherlands, I will argue whether agent-based modeling has a role in the wider understanding of agriculture systems and agricultural production within a landscape, but also to what extent.

Exploring the dynamics of transport in the Dutch limes
by Mark Groenhuijzen and Philip Verhagen
Presented at LAC2014, Rome, 18-19 September 2014

The beginning of the Roman period in the Netherlands (ca. 15 BC) is assumed to be the start of a period of changes, with the presence of the Roman army influencing the cultural landscape and economy of the local population. Although both archaeological and palaeoenvironmental datasets in the Dutch limes region are quite extensive, it is not yet well understood how this Roman influence took effect and what factors (natural or cultural) played an important role in the dynamics of the cultural landscape. As part of the ‘Finding the limits of the limes’-project, this research proposes a methodology for analysing transport networks that were active in the Dutch limes. It also serves as an outline of how this project works within the interaction between archaeological questions, analytical methods and models and existing datasets on the interface of the cultural landscape and scientific methods.

When investigating transport and networks in the Roman Empire, most interest has gone to the Roman military road network, for which archaeological data is plentiful. However, transport occurs on various levels of scale, for different purposes and by different agents and modes. Furthermore, there are external factors which influence the behaviour of transport, examples of which include the limitations and opportunities of the natural environment and political choices. The ability to investigate the functioning of all these transport networks through traditional archaeological research is limited by the nature of the (archaeological) evidence. Therefore alternative approaches are required to fully explore the dynamics of this part of the cultural landscape.

This paper will firstly present a conceptual model for transport dynamics in the Dutch limes. It ideally includes all possible variants of networks and decisions that can be made to establish or avoid certain connections. This model is subsequently applied and analysed in a case study on a part of the Dutch limes, utilising a reconstructed settlement pattern and natural landscape based on existing palaeoenvironmental and archaeological data. The application of this method enables greater insight into the relative importance of governing factors in transport dynamics. It also sheds new light on archaeological questions such as site hierarchy within the cultural landscape, network diversity, and the choice of location of newly founded settlements or even military sites such as forts (castella).

Using an innovative interdisciplinary approach to answer such archaeological questions poses challenges as well, concerning issues such as the validity of the assumptions and methods applied, the shortcomings of the available (archaeological) datasets and the manner by which these uncertainties are taken into account in drawing archaeological interpretations. Working within the interaction between archaeology and science, this study shows how these challenges can be approached and how this integrative framework adds value to archaeological research.

The archaeological distribution map: where are the emperor’s clothes?
by Philip Verhagen
Presented at CAA2014, Paris, 22-25 April 2014

The use of automated cartography and database management systems in regional archaeological studies has in general led to better designed maps, but not to a drastically different approach to mapping. The distribution map, that has been a favourite visual and analytical aid of archaeologists for at least 50 years, has become easier to produce; and various aspects of site distribution, in particular chronology and functional type, are now more quickly mapped. In fact, with the increased use of high-precision mapping using GPS and LiDAR, the focus seems to be more and more on collecting all the dots on the map. However, the concept of the site or settlement as a fixed point in geographical space with precisely known attributes and boundaries has generally not been challenged, despite the lip service that is habitually paid in regional settlement studies to the uncertainties involved in mapping and interpreting archaeological sites. This is surprising, since geographical theory regarding uncertainty has been around for at least as long as GIS; and modern cartographic tools allow us to visualize the uncertainties of our interpretations.

The issue therefore surely is not just a technological one; there is a reluctance, or maybe even resistance within the discipline to deal with uncertainty in archaeological site data. Instead, we are trying to make uncertainty disappear by using ever more precise mapping tools. Uncertainty is however a complex topic that not just involves measurement errors, but all kinds of ambiguities that cannot be easily captured in quantitative terms; furthermore, it is a dynamic property, since new research is often carried out in order to reduce uncertainties in existing data. And while some researchers have tried to deal with this, e.g. by using the concepts of fuzzy logic (e.g. Green 2008; Nicolucci and Hermon 2010; De Runz et al. 2011), aoristic analysis (e.g. Crema 2011), or by using more complex data models (e.g. Rodier and Saligny 2008), an approach to comprehensively address uncertainty in archaeological mapping still has to be developed.

In this paper, I will approach the issue from a practical point of view. The Dutch national archaeological database ARCHIS is a major source of archaeological information in The Netherlands. Yet, its quality is considered to be below the standards needed for scientific archaeological research purposes. Using an example dealing with settlement data from the Roman period, I will highlight the problems of systematically addressing this quality issue, but also how we can use this knowledge of uncertainty to advantage in publication and analysis. I will specifically focus on the problems of dating, reliability of location and reliability of interpretation of finds.


Leave a Reply

Fill in your details below or click an icon to log in: Logo

You are commenting using your account. Log Out /  Change )

Google photo

You are commenting using your Google account. Log Out /  Change )

Twitter picture

You are commenting using your Twitter account. Log Out /  Change )

Facebook photo

You are commenting using your Facebook account. Log Out /  Change )

Connecting to %s