(PDF) Developing a multidimensional GIS framework for archaeological excavations

________CIPA 2005 XX International Symposium, 26 September – 01 October, 2005, Torino, Italy________ DEVELOPING A MULTIDIMENSIONAL GIS FRAMEWORK FOR ARCHAEOLOGICAL EXCAVATIONS S. Merlo, A. Colin Shell, Department of Archaeology, University of Cambridge, Downing Site, Cambridge. United Kingdom

,

KEY WORDS: 3D, excavation, exploration, model, GIS ABSTRACT Geographical Information Systems (GIS) are currently used in the area of archaeological excavation as mainly a data storage and simple output toolkit. The heterogeneous and complex nature of spatiotemporal datasets involving many variables needs to be addressed if GIS is to be of more use for the understanding and interpretation of excavations. Since archaeological objects are to be treated not only as static but dynamic (when considered as the result of temporally evolving processes), they can be referenced to one, two, three or four dimensions. The traditional 2-D data models are insufficient to support such data description and analysis. The diverging approach from the traditional planar (2-D) data models is multidimensionality, involving 3-D geometry and time. Although existing GIS software lacks the ability to support complex analysis involving space-time and further dimensions (attributes), establishing a framework for supporting analysis in such domains should be a priority of archaeologists wanting to contribute to the advancement of the use of GIS for intra-site analysis. This paper presents the development of a possible framework, where exploratory analysis and simulation using different types of data are used for knowledge discovery and enhancement. The designed concepts are applied to the Hoge Vaart database (collected in the Netherlands in the early 1990s during a rescue archaeology operation, using three-dimensional data collection criteria) and results are discussed in terms of applicability and assessment of the analytical framework. 1. INTRODUCTION to integrate theories of individual behaviours, interaction and space/time constraints in order to model spatial micro-level The spatiotemporal nature of processes is an indisputable dynamics. Another difficulty is to generate and analyse paradigm in the discipline of archaeology. This paradigm is at empirical data of micro processes in order to specify and the core of archaeological analysis and, due to the particular calibrate dynamic micro models. purpose of GIS (Geographical Information Systems) of A framework is proposed that aims to integrate archaeological handling spatiotemporal data, it lead to a very rapid integration excavation data in all its phases, from on-site recording to of such systems in archaeological practice. Nonetheless the specialists processing and overall interpretation of the peculiarity of archaeological spatiotemporal data, as different considered site. The case study used for the intellectual from geographical data for which GIS was initially designed, construction of the framework is limited in terms of chronology has remained almost unexplored. A theoretical discussion of and geographical scenarios, but the framework aims at these themes, although certainly due, goes beyond the scope of generality and its application is by no means confined to these this paper. The main aim is to present a practical way of specific circumstances. Applicability and suitability of the approaching the spatiotemporal problem through the design of a framework as a methodological tool transcends a purely framework that that bridges archaeological and computational quantitative and case-specific approach towards excavation data concepts in order to achieve a platform for knowledge integration and analysis. In fact, any system operating under construction. circumstances that could account for both short- and/or long- Proprietary GIS for a long time has remained essentially storage term phenomena must provide flexible analytical support. The and display systems for two-dimensional spatial data formed fundamental themes of space and time provide a common from an association of geometric features and attributes. GIS, in thread that runs through the various perspectives that relate to particular in the field of intra-site archaeology, has been any given spatiotemporal system. These perspectives can be similarly used. It is difficult to say if this has happened due to observed as micro-, meso-, and macro-levels of the application software limitations or incapability of archaeologists to see both in time and in space. In this way the intellectual effort beyond the immediate data storage potentials of the tool. At moves away from direct descriptions of excavations (and in present, the initial rudimentary analytical functions of the general of archaeological contexts) based on geometric and software are developing into increasingly sophisticated tools for numeric spatiality (simple models with no analysis potentials) spatial modelling and analysis. Moreover, the interest of to complex models, with uncertain spatial parameters, varied archaeologists in the application of spatial analysis for intra-site effects and inputs where full system description might become data is increasing. This encourages the incorporation of the impossible, but exploration and explanation could find a place. available tools in archaeological research. However, the design The paper presents a GIS based framework for exploratory data aim of most widely used GIS packages has not been analysis and dynamic modelling of archaeological processes archaeologically driven. As a result, translation of the unique (deposition and post-deposition). It outlines the application spatial concepts, relationships and processes typical of scenario which inspired an alternative mode of data integration archaeology, formed independently of GIS, is not obvious and and will concentrate on the temporal side of the without misapplications. Therefore, although the current multidimensional approach. The main objective of the work is limitation of commercial GIS in providing a dynamic to question traditional concepts of space, time, scale and representation of spatiotemporal phenomena is only one modelling in archaeological theory and practice through the use obstacle to the development of spatiotemporal modelling, the of a tool that has been introduced to the discipline without much major challenge is to formulate the conceptual framework and appraisal and evaluation. ________CIPA 2005 XX International Symposium, 26 September – 01 October, 2005, Torino, Italy________ 2. APPLICATION SCENARIO The Hoge Vaart project in Holland (Figure 1) researches the Mesolithic and Early Neolithic of the Flevoland polder province (Hogestijn and Peeters 2001, Peeters forthcoming). The investigation was conducted by ROB (Netherlands National Service for the Archaeological Heritage) and it took place between 1994 and 1997 within the framework of the completion of A27 motorway between Blaricum and Almere. Figure 2. Hoge Vaart, Almere. Compisite map showing the extent of the excavation site, the gridding systems of 2x2 m and the finer resolution excavation areas (50x50 cm). The map presents also the location of the bore holes of the 1996 campaign and the features recognised during excavation Data were collected using the so-called arbitrary excavation (Lucas 2001, 163) based on a system of 50x50 cm grid-cells, as it is usual in the Netherlands for Neolithic sites (Hans Peeters, Figure 1. Geographical location of the Flevoland polder area, p.c.). The underlying idea was the consideration that three- the Netherlands. The excavation site is located in the dimensional recording of individual objects is time consuming municipality of Almere and total stations are expensive. At the same time the deposit presented ecologically very fragile material. As sieving recovers The project, funded by the Executive Ijsselmeergebied of the ecological material in good conditions and gridded spatial Ministry of Transport and Water, can therefore be considered a research is appropriate for finding structures where the limits of s rescue archaeology. The main interest of the research team units are invisible, a combination of grid and spit systems was was in the formation processes and landscape dynamics in the used to gather the data. Methods of assessment and adjustment transition between the Mesolithic and Early Neolithic, which of vertical grid resolution were constantly used and led to the are barely represented in the regional literature. The project was decision of using a vertical resolution of 4 cm. No use was conducted in two stages. At first, in 1996, exploratory research made of the Harris Matrix diagram. As the majority of the was carried out. This aimed at collecting material to understand features was located within 15/20 cm and the character of the the main principles of landscape development and soil sediment was chaotic, this would have required an enormous formation of the Holocene coversand, underneath the peat. The investment of time. Although the principles of arbitrary area was cored at 61 locations along 12 lines and the cores were excavation were employed during the excavation, when then analysed using a combination of geological, pedological, particular features were recognised, these were recorded by line micromorphological and palynological research techniques, drawing (mainly in plan but also in section) and a description completed with a range of C14-dating. to obtain an overview of and interpretation was given. The research strategy gave a the geological development of the coversand and of the changes fundamental importance to the use of automation in order to the climate and vegetation of the area underwent in the course improve data quality, efficiency and monitor errors in real time. of the Holocene. Moreover, the coring allowed the identification During the fieldwork an effort was made to collect, when of the excavation area, on the base of the charcoal content. The possible, all data digitally. To achieve this, two parallel systems site (approximately 8600 m2, fig. 2) was located on a coversand were used: ridge along a low area, dissected by various tidal gullies. The • ‘automatic’ 3D location definition by means of a excavation revealed a large body of occupation remains and tacheometer (total station) with a radio link to a hand- anthropogenic features. Data on paleo-environmental conditions held computer and by means of altitude measurements were collected and interpreted in terms of landscape dynamics. with laser theodolite or water level in parallel with the Identified features consisted of nearly 120 surface hearts and assigning of work pit and section numbers. The find several hundreds small stake holes. Approximately 100 deep numbers were recorded by means of a bar code hearth pits (essentially containing charcoal) were also scanner, the feature data input through a menu on the recognised. hand terminal. • A part of the field record was written and drawn in analogous form. This applied particularly to the recording and descriptions of the anthropogenic features. The field drawings were made in the ________CIPA 2005 XX International Symposium, 26 September – 01 October, 2005, Torino, Italy________ traditional manner ‘by hand’, without recourse to attached attributes. To conclude, the complexity of the data digital measuring apparatus. Nonetheless they were collected at sites such as Hoge Vaart can only be accounted for digitised and stored in dxf vector format at a later by a concept of multidimensionality that incorporates space as stage. volumetric geometry, multiple spatiotemporal resolution and The dataset is characterised by an enormous potential in terms linked attributes and takes into account data inhomogeneity of quantity and quality of multidimensional (as intended in arising from the excavation pragmatic decisions on the way the section 3.1) data. It presents a combination of data sources site is excavated and recorded. The issue here concerns the collected at different scales. It was therefore the ideal candidate accurate reconstruction of various phenomena, where the for testing the ability of GIS to respond to the needs of complex complexity occurs in all the dimensions involved in archaeological datasets. spatiotemporal analysis. Time, location, weight, densities are examples of what a dimension can represent. 3. THE CONCEPTUAL FRAMEWORK 3.2. Rationale of the system: exploratory analysis and knowledge discovery Archaeological excavation records are characterised by large heterogeneous spatiotemporal data sets involving many Archaeological analysis and interpretation happens at different variables. The spatial data infrastructure is the combination of a levels, touches several sub-disciplines and is carried out by a subject-oriented framework and field data used for a variety of variety of experts. Nonetheless it is characterised by a common purposes (rescue archaeology, research oriented to answer thread: the search for interesting patterns, which, in computing specific questions, palaeo-landscape studies). Moreover, terminology corresponds to the concept of knowledge inhomogeneity of space and time and of time represented by discovery. The main concern is to provide the archaeologist space (underlying the principles of stratigraphy) must be taken with a general capability for specifying the patterns and into account too. Very often, for example, periods of intense aggregates that may be required across several dimensions to activity are of greater relevance to the analysis than periods of produce such knowledge. Generally, different analytical relative inactivity. Similar requirements for context dependent techniques and base models are needed of a particular spatio-temporal rules and user-defined semantics for the spatiotemporal dataset. Moreover, views of such dataset are manipulation of variables are also to be noted. often particular and different. In addition to this, in a In these circumstances, a computer framework is used to multidimensional problem space the constraints and conditions underpin spatiotemporal research in an exploratory and intuitive that apply to each dimension can only be expressed in the manner. Among the requirements of such framework Lee and context of the specific analytical process carried out. The aim of Kemp (1998) list: designing a prototype as a framework for archaeological • The user to deal with natural representations of the information system is to inform the choice of data collection phenomena and relationship of interest and storage strategies and of data structures to be used and • The application of statistical functions to the raw integrated for an informed exploration and knowledge discovery dataset so that flexible partitioning of the problem within GIS terminology and capabilities. It provides a set of space is achieved approaches to the design of patterns discovery in the dataset, • The intuitive and interactive visualization of selected allowing the user to make quantitative and qualitative subsets to enhance exploration and analysis considerations on the methods used and results obtained. It is a framework for reasoning within a reasonable position between 3.1. Multidimensionality explained the data and its existence within a computerised system. Conceptual design becomes here an exercise for alternative The definition of multidimensional data is here used to express interpretation. a concept that goes beyond the simple representation of space The conceptual framework is based upon an exploratory and time in a Cartesian world. The major limitations of intra- analysis approach, strictly connected with the two principles site archaeological analysis have been linked to the inability of exposed by Kemp (1993), as follows: current GIS systems to represent a three-dimensional space, in 1. it is both desirable and possible to separate the mathematical particular in terms of a vector structure where a triplet of operations which will be performed on data about spatial coordinates represents the entity point (Harris and Lock, 1996). phenomena from the form of discretisation used to represent In fact, even when attempts are made to achieve such those phenomena in the computer representation, the complexity of archaeological data and the 2. this separation allows issues about the implementation and richness of the aspatial attributes is frequently lost. In particular manipulation of these digital representations to be dealt with the direct connection often made between stratigraphic automatically, without external control, in such a way that they excavation, single contexts’ identification, their recording and can be considered extraneous to the modelling task. representation as surfaces is a dangerous oversimplification of The aim is to provide the archaeologist with a general capability other aspects of the excavation process and the excavated required for identifying and specifying the patterns and record. The Hoge Vaart excavation presented in the previous aggregates that occur across several dimensions and at different section is but one example of circumstances in which these levels of analysis. This is an important issue that has been methods are not only inapplicable but most probably neglected up to now in the development of GIS applications, misleading. During excavation practice the recording is often with the consequence of preventing a methodologically correct carried out under extremely difficult circumstances that require and practically useful use of GIS. Given the importance of immediate and pragmatic reaction. Features may or may not be building a framework of study, the fundamental requirements of visible, time management may or may not allow to carry out a the conceptual framework are identified and explained in the painstaking operation of single context planning. Data may be following section. collected in gridded bulk samples and at random locations that are identified to represent the processes characteristic of the site. These data can not and are not to be represented exclusively by surfaces, rather by interpolated and extrapolated volumes with ________CIPA 2005 XX International Symposium, 26 September – 01 October, 2005, Torino, Italy________ 3.3. Underlying principles create micro-, meso- and macro-base models of the excavation data. Generally, different views of a particular spatiotemporal dataset are expressed by different requirements. Moreover, in a multidimensional platform the constraints and conditions to apply each dimension can only be expressed in the context of the analytical process being carried out. Therefore different ranges of requirements can be summarised for the spatial and temporal dimension. Despite the seeming difficulty in identifying connections across dimensions, two properties can be recognised: • Multiple resolution (issues of scale in time) • Multiple characterisation (issues of aggregation of space and time in units different from the ones used for recording) The two identified properties are inspired by the Analytical Abstraction Layer (AAL) discussed by Lee and Kemp (1998), used amongst others for the management of marine fisheries. • Multiple resolution is a concept that encapsulates issues of scale in space and time and is one of the parameters that determines the outcome of the analysis. When a dataset is explored at various spatial scales, corresponding degrees of detail are revealed; similarly, temporal information can also be presented at different levels of resolution (e.g. decennia, centuries, millennia), with each level showing different patterns and trends. When data is examined at a coarser spatio-temporal resolution than it was Figure 3. The spatiotemporal framework. Conceptual scheme recorded, it needs to be aggregated. The same of an integrated archaeological information and modelling operations can also be used to generalise information system that is considered less important for a specific level of analysis. If spatial resolution needed to be finer than This combination of raw data and their manipulation in terms of the level at which it was recorded, decomposition and scale and characterisation implies possibilities of aggregation, disaggregation could be achieved using, for example classification, generalisation, specification and partitioning to kriging. enable pattern and anomalies to be elicited. The principle is that • Multiple characterisation stems from the need of of a nested hierarchical structure where analytical operations archaeologists to classify and categorise data in a (map algebra, extrapolation, interpolation, etc.) are used to meaningful manner. In addition to this, produced new datasets suited to proceed to further phases of the characterisation enhances the dynamism and analysis and visualisation. The elements of the system are expressivity of such operations by allowing connected by a double direction flow, to avoid the determinism conceptual mobility and multiple belonging. Data sets typical of some hierarchical structures. As indicated above, the may belong to one or more characterisation classes or framework considers data multidimensionally: time, location, groups of classes. Each characterisation level or group aspatial attributes, inheritance from previous levels of can have many members. Categorisation classes allow interpretation are examples of what a dimension can represent. the preparation of higher level data for further Data as such are stored in the multidimensional database. The analysis. Multiple characterisation also allows the dimension of interest can be one or many at a time and data are same data set to be portrayed in different graphical explored using the pathways of multiple resolution and multiple forms that emphasise different aspects of data. characterisation, as shown in figure 3. These properties, Moreover they allow the representation and common to all archaeological data, are the foundations of the compatibility of data collected with different criteria system and between them they provide support for a wider (single contexts transformed in digital vectors as range of analytical procedures. Multiple resolution and multiple opposed to voxel interpolated samples). Certain characterisation become the axes of the base model building. categorisation methods will reduce the number of Changing them can change the interpretation of the dimensions or the complexity of a data set. Data sets spatiotemporal phenomenon and can create different from different classes can only be compared subject to interpretative scenarios. These operations are used both for time a mapping function, used to reconcile not directly and space to suppress or enhance detail, differentiate or compatible datasets. Characterisation involves using generalise the components both for display and analysis. as aggregation, inter- and extrapolation, classification, The described framework is not limited to pure intellectual generalisation, categorisation and partitioning to exercise. It is the conceptual thinking behind a prototype which interactively elicit patterns and anomalies alike. exploits the capabilities of GRASS 5.3 raster engine for volume analysis and GRASS 6.0 new topological 2D/3D vector engine 3.4. The framework illustrated for vector network analysis. GRASS is a free Software/Open Source released under GNU General Public License (GPL). Taking into consideration the issues discussed in the previous sections, the framework proposed (figure 3) is a system in which multiple resolution and characterisation are adopted to ________CIPA 2005 XX International Symposium, 26 September – 01 October, 2005, Torino, Italy________ 4. RELATED ISSUES analysis resembling a process of knowledge discovery. Knowledge discovery is fundamentally the search for The research presented in this paper is aimed at combining areas interesting patterns. However, in order to support knowledge of GIS, spatial database systems, modelling, simulation, discovery, it is important for different data types to be able to visualisation. The work should be viewed in the context of the exchange information in a standardised (meaningful) manner. complete framework for modelling and knowledge-discovery. The user-mediated mechanisms for organising the search By their nature, these activities are iterative and evolve in process ensures that any discovered knowledge during analysis stages. The research conducted so far identifies other areas of is consistent with the user’s requirements. This also helps investigation that necessarily follow the previous presented constrain the search process with domain knowledge not easily ones.Visualisation, closely linked to multidimensional available within the system. We believe this framework will approaches, is one of them. The system should provide a play an important role in helping to develop the next generation presentation mechanism that enables users to determine the of GISs tailored for excavation and the thinking behind the combination of display and visualisation required – application of GIS to intra-site analysis. cartographic, graphic and tabular. Another related issue concerns the capture and persistent REFERENCES storage of any derived multidimensional partition. The problem is obviously of importance in the area of archaeological data Harris, T. H. & Lock, G. R. 1996. Multi-dimensional GIS: where datasets are large and distributed, and analysis progresses exploratory approaches to spatial and temporal relationships in stages. within archaeological stratigraphy, Analecta Praehistorica Leidensia, 28, pp. 307-316. Hogestijn, J. W. H. and Peeters J.H.M 2001. De Mesolitische en 5. CONCLUSIONS Vroeg-Neolitische vindplaats Hoge-Vaart A-27 (Flevoland). Almere: RAM 79. This paper is the presentation of work in progress. It aimed at Lee, H. & Kemp, Z. 1998. Supporting complex spatiotemporal showing the problems and intricacies of analysis in the analysis in GIS. Innovations in GIS 5, pp. 151-161 spatiotemporal domain by an account of the problematic and Lucas, G. 2001. Critical approaches to fielwork. Contemporary specific nature of such data in archaeology and proposes a and historical archaeological practice. Routledge: London. framework. An important element of the framework is the Kemp, K. 1993. Environmental modeling with GIS: a strategy modularity of the system, which provides better management for dealing with spatial continuity. NCGIA Technical rep. 93-3 and understanding of spatiotemporal problems. Data in the http://www.ncgia.ucsb.edu/Publications/Tech_Reports/93/93- nested (level) system are multidimensional, where this 3.PDF (accessed 13 June 2005) multidimensionality is not only a measure of space and time Kemp, Z. and Lee, H. 2000. A Multidimensional Model for components but also multiplicity of attributes and Exploratory Spatiotemporal Analysis. Geocomputation 2000 characterisation of layers of analysis. This means that we can http://www.geocomputation.org/2000/GC033/Gc033.htm view data at various levels of detail and in a multitude of (accessed 01 June 2005) representations and elaborations. In turn, the information we Peeters, J.H.M., forthcoming. The Mesolithic to Early Neolithic retrieve can be used for complex analysis and for constructing occupation at the Hoge Vaart (Almere, the Netherlands): higher or lower order patterns of spatiotemporal processes. forager land-use dynamics in the context of a drowning With the focus on spatiotemporal information systems as landscape. In Acts of the XIVth UISPP Congress (Liège, 2001). opposed to spatiotemporal databases, we have taken a step Oxford (BAR International Series), Archaeopress. towards realising the enormous potential of such systems. The framework is particularly effective at exploratory analysis whereby different types of data are used interactively and ACKNOWLEDGEMENTS collaboratively by archaeologists from different areas of expertise and wanting to write a story of the excavation site Stefania Merlo is funded by the University of Milan, the taken into consideration. Although analytical techniques have University of Cambridge and Newnham College. Infinite thanks been traditionally associated with statistics, the complex and to Hans Peeters of ROB in the Netherlands for providing the diverse nature of archaeological data calls for an approach to case study data and giving never ending support and advice.