Early landscape studies using GIS have been criticised for their environmental determinism (Gaffney and van Leusen 1992). In some landscapes the environment may severely constrain human activities, and simple topographic measures (e.g. Kvamme 1992) or cost surfaces may be useful (Verhagen et al. 1995). However, in the Carnac area, with its subtle topography, it is almost impossible even to attempt an environmentally deterministic approach, should one wish to do so.
The monuments are not distributed with respect to simple topographic characteristics (Roughley 2001). However, although monument location was not determined by the environment, this does not mean that topography was not important. It has been suggested that some sites were located so as to create a specific visual experience (Le Roux 1999b). As the landscape is difficult to study in the field, digital techniques can provide significant benefits for the investigation of the locations of monuments, and elucidate the ways in which their topographic settings affect human experience of the monuments.
The post-processual critique of the study of archaeological landscapes has shown that it is important to consider landscapes not just at a regional scale, for example as distribution maps and trade routes, but also at the scale of an individual human within that landscape. Any digital study which aims to provide insight into the ways in which archaeological landscapes were engaged with and perceived by people requires a very high-resolution digital terrain model. Small topographic features can have significant impact on the view of an observer whose eye level is less than 2m above the ground, and must be included in the surface model if a representation of a human view of the landscape is to be generated. The Carnac area has many low convex rises, which have a significant impact on visibility. These slight topographic features mean that many passage graves are visible from long distances, disappear from view when approached, only to reappear again suddenly (Roughley 2001; Roughley forthcoming).
Until recently, most digital approaches to landscapes have had to rely on low resolution topographic data. While informative on a regional scale, such data are rarely adequate for representing topography at a human scale, and has hindered otherwise excellent applications of visibility analysis and visualisation (e.g. Exon et al. 2002). It is therefore imperative that archaeologists make full use of developments within remote sensing. It is now possible to obtain highly accurate digital surface models from LIDAR (light detecting and ranging - airborne laser scanning; Holden et al. 2002; Shell and Roughley 2004) or, as is the case here, through using software photogrammetry.
Software photogrammetry allows digital surface models to be generated from stereo pairs of photographs. In this case, 66 photographs at 1:20 000 were taken by the NERC Airborne Remote Sensing Facility (http://www.nerc.ac.uk/arsf/), and georeferenced using differential GPS. Photogrammetry produces a digital surface model, rather than a digital terrain model; careful choice of collection parameters can remove many of the unwanted features (trees and buildings, but not large areas of woodland etc.). The model can then be further edited if required. The photographs were then orthorectified (to remove lens distortion and relief displacement), and mosaicked together (Roughley and Shell 2001; 2002; forthcoming).
Views form an important part of a locale. The use of visibility analysis has been questioned (e.g. Gillings and Goodrick 1996) because vision is only part of the wider human sensory experience of a place. The potential of determining the areas from which a noise can be heard (Mlekuz this volume) is an important contribution. However, while it is important to be aware of the tendency of GIS analysis to privilege visual analysis because of ease of use, visibility analysis can still contribute to a wider understanding of prehistoric landscapes. Although Llobera (1996) and others have strived for 'post-processual' GIS using viewshed analysis, visibility analysis on its own (like any technique) cannot lead to theoretically aware archaeology.
Visibility analysis has allowed landscape archaeologists to investigate the extent to which sites of interest were constructed in locations with particular visual properties. GIS does not provide the only means of conducting such research (see Bradley et al. 1993), but in many cases it is the most practical. The strength of visibility analysis is its quantitative character, which allows comparison of site and non-site locations (e.g. Wheatley 1995), and can enable sites in locations with different visual characteristics to be identified, and subsequently interpreted (e.g. Lock and Harris 1996). The possibility also exists for interpretations to be revisited and developed, particularly where data and interfaces are provided, such as in the study of the Stonehenge landscape by Exon et al. (2002).
Visibility analysis has developed from early uses (e.g. Gaffney and Stančić 1991), to include statistical testing (Wheatley 1995; Fisher et al. 1997), and a recognition of the correlation between topographic features and visibility (van Leusen 1999 and Lake et al. 1998). Preliminary investigations into directionality and distance (Roughley 2001; Wheatley and Gillings 2000) and vegetation (Tschan et al. 2000), along with work on fuzzy viewsheds (Fisher 1994) have helped to make visibility studies more critical. The difference between views to and views from is well known, as is the need to avoid edge effects. Increases in computing power mean that the creation of total viewshed maps showing the size of the area visible from every cell (or the area from which a cell is visible) is now relatively straightforward (Fisher 1996; Llobera 2003).
The application of visibility analysis is not without difficulties. Significant archaeological problems remain. Where, exactly, were all the sites, and how large were they in the period of interest? In order to ensure that as many sites as possible were included in this research, modern surveys were complemented with 19th-century records (Roughley et al. 2002). Site locations were also checked against orthophotos generated for the project. The height of many of the sites is unknown; estimates were based on early records of sites which do not appear to have been greatly disturbed at the time of measurement.
More importantly, does visibility matter, and how do we interpret the results of visibility analysis? The Neolithic monuments of the Morbihan have appeared in 19th-century French Admiralty charts and 15th-century pilot boat manuals (Ferrande 1483), suggesting that their strong visual presence was important in a much later period. Most research so far has suggested that prehistoric burial monuments in northwest Europe are in places of larger than expected visibility. However, the correlation of size of view with importance should not be accepted unquestioningly. A look at ethnography quickly shows that the opposite can also be true. For example, in Madagascar, important places may be hidden to control knowledge or reinforce taboos (Mack 1986).
There are two main methods for collecting viewsheds. A grid can be generated, with values either as a Boolean grid of visible/invisible, or whose pixel values denote the number of visible points. The grid-based approach is particularly useful for understanding the visibility characteristics of the whole landscape (see Llobera 2003). The alternative approach is to generate polygons of the visible areas. The advantage of this latter approach is that one set of polygons can be generated for each point, and then clipped to a buffer according to the viewing distance required, rather than re-running the visibility algorithm. Simple overlays using these same polygons can determine whether sites are intervisible, and assess the similarity of viewsheds from different sites.
Visual characteristics of landscapes and monument locations can be explored further from a more human perspective using virtual reality applications. Increasing availability of virtual reality techniques (and the technology through which to disseminate the resulting models) has encouraged landscape archaeologists to explore the field. There are two approaches to the use of VR – either involving the use of fully immersive technology (e.g. Forte and Guidazoli 1996) to create as realistic an experience as possible, or the alternative of using less immersive, but more easily disseminated, web-based models.
Web-based approaches include the use of photo-panoramas (e.g. Edmonds and McElearney 1999). Such models have the advantage of photo-realism and can have striking visual effects, as Jeffrey demonstrates (2001). They are easy to view over the Internet. However, the limitation on the observer location prohibits effective study of movement (Larkman 2000).
VRML provides an alternative approach for creating virtual reality models which are easy to distribute over the Internet. Initial work by Gillings and Goodrick (1996) using VRML showed that a more embodied perspective could challenge our perception of the Tisza floodplain. Although requiring a viewer (e.g. Blaxxun), VRML does provide a means of including surface models, drapes, and objects which can be navigated around. For many landscape applications, the export facilities of GIS and Remote Sensing packages are adequate (but see also discussion in Goodrick and Earl 2004).
The relationship between such models and human perception is complex. Even if the observer is standing on the ground surface, the view does not necessarily resemble a human experience of a landscape. The field of view available is important, and perception of distance may not be accurate. Even when sufficiently detailed data are available (see section 3.2), limited display resolution can obscure features which are in reality clearly discernible. Links to visibility maps, either from key points, or dynamically (as in Exon et al. 2002) therefore remain an integral part of any assessment of visual relationships. Greater availability of dedicated visualisation facilities will, in the future, increase the accuracy of visualisation-based approaches, and make full use of detailed data.
Pre-recorded dynamic visualisations (AVIs or MPEGs, in this case created using ERDAS Imagine) can be used to convey topographic detail and a surface viewing position. A VRML model, created from the same data, which allows readers to explore the landscape interactively, provides a complementary resource.
© Internet Archaeology
Last updated: Thur Nov 11 2004