Visibillity and GIS

Here is the final report from Archaeological Applications of GIS from the University of Southampton

 Intervisibility of Megaliths and Pile-Dwellings along the edge of Lake Neuchâtel in Switzerland

Introduction

Beginning with the chance discovery of the Ober-Meilen prehistoric lacustrine village on the Lake of Zurich in 1854 by a curious school teacher, pile-dwelling sites around lakes and bogs have yielded incalculable knowledge on prehistoric cultures living in the foothills of the Alps. Dendrochronological analyses of well-preserved timbers from numerous sites, along with a wealth of plant and animal remains, suggest a beginning for the Alpine foreland Neolithic around 4300 BC spanning until the seventh century BC (Menotti 2004). Spread across six countries and listed as World Heritage Sites by UNESCO since 2011, the pile-dwellings are hailed as “among the most important examples of archaeological heritage in Europe” (Hafner and Schlichtherle 2008).

As studying nearly one thousand sites throughout the Alpine foreland is beyond the scope of this brief research, Lake Neuchâtel in Switzerland (Figure 1) was selected for study to simplify the necessary computations, as well as to present a straightforward example of using GIS to answer specific archaeological questions. The GIS provides the tools to systematically analyse vision in prehistory, although methods may be the same as manual methods decades ago (Wheatley and Gillings 2000). It should be noted that the scope of this research could easily - read tediously - be adjusted to include all known sites in the surrounding countryside.

Figure 1.  Map of Switzerland with study area. (Copyright: ©2013 Esri, DeLorme, NAVTEQ)

Undoubtedly, the pile-dwellings have been part of a persistent and respected research tradition for over a century and a half. However, standing solemnly nearby, megalithic monuments are consistently overlooked in the context of the cultural landscape. Although archaeological evidence in associated strata around the megaliths is sparse, they were certainly a part of the landscape throughout the 4000 years of documented Neolithic settlement around the lake, notwithstanding the possibility of their construction during this period. The principle aim of this study is to describe and analyse intervisibility between pile-dwelling and megalithic sites to assess whether the megaliths might have played a significant role in the cultural landscape of the Alpine foreland Neolithic. As Wheatley and Gillings point out, “the suggestion is that it is through viewshed analysis that the GIS makes its most unique and valuable contribution to landscape study” (2000). It has thus been chosen to determine the importance of these mysterious megaliths in context with the pile-dwellings. The Kolmogorov-Smirnov one-sample goodness-of-fit test is used to assess the significance of the relationship.

The procedure of georeferencing source maps and digitizing a workable dataset is documented, and the technical and analytical choices made in the interpretation are described. ESRI's ArcMap and ArcCatalog 10.1 were used throughout the data process.

The Context

Lake Neuchâtel is the largest lake in Switzerland at 38 km long and 8 km at its widest point. It was formed through the uplift of the Alps and subsequent glaciations and lies primarily in the modern canton of Neuchâtel, but also shares its border with the cantons of Vaud, Fribourg and Bern. The elevation of its surface is 422 m above sea-level. Visibility from the lake is very high and the assumption that these Neolithic cultures used the lake as a means for transportation and subsistence is not farfetched.

Pile-dwelling sites were settlements built on stilts, or piles, on dry land near the water’s edge. Due to seasonal flooding of the lake, the houses were raised upon piles in such a way that the intermittent overflow did not cause inconvenience. Some of the piles around the lake have been observed to have been taller than 5 m and as short as 0.4 m. There is little evidence to determine what the specific architecture of the dwellings was like, but most research suggests that they were probably single-story houses accessed by ladders or stairs (Leuzinger 2004). Settlements were situated every 2.5-5 kilometres on the shores of Lake Neuchâtel (Rouff 2004), although in this particular study, only the UNESCO qualified sites, of which there are thirteen, are being researched.

The age of the megalithic sites is largely hypothetical due to the lack of archaeological evidence. Therefore, it is not possible to establish precise chronological connections between the megalithic monuments and the pile-dwelling sites, a grim obstacle in my research. Discovering any information of any kind about the megaliths was a challenge in and of itself. However, this problem can cautiously be resolved by suggesting that most megaliths were erected in the Neolithic and were probably contemporaneous with or predated the pile-dwelling settlements. Within the scope of this research, several types of megaliths are being studied: menhirs, large tapered erected stones sometimes found in groups but primarily alone; stone rows or alignments, which include dozens of menhirs in alignment with each other; and dolmens, which are stone slabs assembled as burial chambers. Some of these stones are as tall as 4.5 m (Figure 2). The megaliths must have been important places within the landscape, marked as they were by such imposing heights. For purposes of the research, megalith type is not considered an important factor in intervisibility.

Figure 2.  An example of a Swiss megalith. The Menhir de Clendy of Yverdon-Les-Bains, part of a stone alignment, this is the tallest at 4.5 m.    (Source)

Methodology

Fortunately for the author, a Digital Elevation Model, the foundation for any visibility study, was already available for the study area, obtained from the NASA Shuttle Radar Topography Mission which has an estimate height error report of 5 m and is accurate to 3 arc seconds, which equates to 90 m on the ground. Of course, the quality of the DEM determines the accuracy of the findings. Although this was not the most accurate of DEMs, the author, with financial constraints, had to opt for data that was free of cost.

This particular DEM uses the World Geodetic System 1984 coordinate system, one of the main reference systems for cartography. However, the author deemed the European Datum 1950 Universal Transverse Mercator projected coordinate system to be better suited to the study area because the UNESCO data utilizes this system. Thus, the elevation data was exported within this datum and all subsequent analyses use ED 1950 UTM Zone 32 North.

Individually georeferenced pile-dwelling maps from a UNESCO nomination file were used to digitize the locations of each pile-dwelling site around the lake. There were inherent problems with the precision of this method because the measured grid around each of these maps was imprecise. This made itself evident in the digitizing process because the edges of the lake did not match the base map. The imprecise maps were individually adjusted to match with the base map which was accurately georeferenced. The viewpoint taken is static and situated in the relative center of the settlements or megaliths. The above factors have clear implications for accuracy and precision of the subsequent data.

The locations of the megaliths were obtained from coordinates from an online database at www.megalithic.co.uk. The coordinates were manually entered into ArcMap and display as central points to the location of the megalith(s). Lake Neuchâtel was also digitized to eventually obtain a cumulative viewshed of all the points within the lake. The end map, created for display, (Figure 3) contains contour lines, both types of sites and the lake itself.

Figure 3. Map of Lake Neuchâtel displaying megaliths and pile-dwellings with surrounding landscape contours made entirely from digitized data.

Now, the process of creating relevant viewsheds was possible. Ideally, the visibility of features should be considered according to factors of the nature, density and height of the vegetation, the weather and season (Wheatley and Gillings 2000). The DEM is a smooth surface and does not take into account the presence or absence of trees or any other vegetation. The area for which visibility was calculated was restricted to a buffer zone of 10 km around the edge of the lake. Because viewshed does not readily extend beyond a radius of about 5 km (Loots 1997), this is an overestimation of perception but, for the sake of the study, it ensures minimal edge effects within the area of study and that the viewshed is not subject to artificial truncation (Wheatley and Gillings 2000, Conolly and Lake 2006).

The modern heights of seven of the megaliths were recorded in the database. However, some of the megaliths have toppled in modern times or are so obscure that the database had no height data, in which case an average of the known megalith heights was calculated (2.7 m) and was used as the height data for the four unknown megalith heights. The heights of the pile-dwellings also had to be interpolated. According to Leuzinger (2004), the piles themselves range from 0.4 m to over 5 m tall. In order not to skew the data, the author elected again to assign the average height of the known piles (2.7 m) plus the average height of a person (1.7 m) as the offset in the viewshed (4.4 m), to simulate a person standing on the floor of a pile-dwelling. As for the heights of the pile-dwellings themselves, a conservative height of 6 m was selected to represent the maximum height of the pile-dwellings.

The author utilized the work of Kvamme (1990) and Wheatley (1995) from which to base her model. To initiate the Kolmogorov-Smirnov one-sample goodness-of-fit test, a ‘population’ of all possible pile-dwelling locations was generated from the megalith viewshed by creating a 600 m buffer around the lake to ensure all known pile-dwelling locations were within the buffer zone (Figure 4). This is the statistical population, while the pile-dwelling sites are the statistical sample from that population, from which to conduct the goodness-of-fit test. Given these, the author has constructed a pair of hypotheses from which to test as follows:

• H₀ – The pile-dwelling sites are distributed regardless of the number of megaliths which are visible.
• H₁ – The pile-dwelling sites are NOT distributed regardless of the number of megaliths which are visible.
  
  

  

  Figure 4. Map of 10km buffer zone and viewshed map from the megaliths. The thin sliver around the lake is the 600m buffer zone from which the statistical population is derived.

A separate Kolmogorov-Smirnov test was also conducted to test the significance of the intervisibility between the megaliths and the viewpoints on the lake. A buffer of 2000 m around the lake was generated to include all known megaliths sites (Figure 5). In the same manner as before, this buffer is the population and the known megaliths sites are the sample. Given these, the author has constructed a pair of hypotheses from which to test as follows:

• H₀ – The megaliths are distributed regardless of the number of lake-points which are visible.
• H₁ – The megaliths are NOT distributed regardless of the number of lake-points which are visible.

Figure 5. Map of 10km buffer zone and viewshed map from the lake-points. The thin sliver around the lake is the 2000m buffer zone from which the statistical population is derived.

The Kolmogorov-Smirnov test undertaken for each case adopted a 0.05 confidence interval. Because the sample size was so small for each test, 13 for the pile-dwelling test and 11 for the megalith test, d is 0.361 and 0.391 respectively. Dmax was then obtained from the results. For each data theme used in the final research and the end maps, metadata was created in GeoDoc using the AGMAP 2.1 guidelines and saved as both PDFs and XML files.

Results

For the pile-dwelling test, it can be seen that Dmax (0.14) does not exceed d (0.361) and therefore the test does not allow for the rejection of H0 at the 0.05 confidence interval (Figure 6).

Figure 6. Right: Kolmogorov-Smirnov test for the pile-dwellings. The test compares the distribution of the population and the sites, or samples, with respect to the number of lines-on-sight to the megaliths. Dmax is highlighted. Left: Cumulative distribution, or percentages, of both population and sample cumulative viewshed values. 

Equally for the megalith test, it can be seen that Dmax (0.12) does not exceed d (0.391) and therefore the test does not allow for the rejection of H0 at the 0.05 confidence interval (Figure 7). Thus, neither the distribution of pile-dwellings not the distribution of megaliths are dictated by the location of megaliths or lake-points, respectively.

Figure 7. Right: Kolmogorov-Smirnov test for the megaliths. The test compares the distribution of the population and the sites, or samples, with respect to the number of lines-on-sight to the lake-points. Dmax is highlighted. Left: Cumulative distribution, or percentages, of both population and sample cumulative viewshed values. 

Conclusion

Although the results of this study are not exceedingly thrilling, the use of the Kolmogorov-Smirnov test did statistically proof the significance, or lack thereof, of the specific location of the sites in question. The method was relatively straightforward to implement and, with the right question, could potentially yield interesting and valuable results. This is not to say that the particular location of these sites within the landscape are not dictated by other factors, seemingly unknown at the time of study. Indeed, this result doesn’t necessarily mean that visibility was not at all a factor in location, it just means that the author cannot statistically claim that it was. With a larger sample, more conclusive results could be drawn.

Bibliography

Conolly, J. and Lake, M. 2006. Geographical Information Systems in Archaeology, 6th edn. (Cambridge: Cambridge University Press).

Hafner, A. and Schlichtherle, H. 2008. Neolithic and Bronze Age lakeside settlements in the Alpine region: threatened archaeological heritage under water and possible protection measures – examples from Switzerland and Southern Germany. In Heritage at Risk: ICOMOS world report 2006/2007 on monuments and sites in danger, edited by Petzet, M. and Ziesemer, J. (Altenburg, Germany: E. Reinhold-Verlag). pp 175-180.

Kvamme, K.L. 1990. One-sample tests in regional archaeological analysis: new possibilities through computer technology. American Antiquity, 55 (2), edited by Raymond, W.W. (Washington, D.C.: Society for American Archaeology). pp 367-381.

Leuzinger, U. 2004. Experimental and applied archaeology in lake-dwelling research. In Living on the lake in prehistoric Europe: 150 years of lake-dwelling research. (Abingdon: Routledge). pp 237-250.

Loots, L., Nackaerts, K. and Waelkens, M. 1999. Fuzzy viewshed analysis of the Hellenistic city defence system at Sagalassos, Turkey. In Archaeology in the age of the internet:  computer applications and quantitative methods in archaeology 1997, edited by Dingwall, L., Exon, S., Gaffney, V., Laflin, S. and van Leusen, M., BAR International Series 750. (Oxford: Archaeopress). pp 82 [CD ROM].

Menotti, F. 2004 (ed). Introduction: the lake-dwelling phenomenon and wetland archaeology. In Living on the lake in prehistoric Europe: 150 years of lake-dwelling research. (Abingdon: Routledge). pp 1-6.

Ruoff, U. 2014. Lake-dwelling studies in Switzerland since ‘Meilen 1854’. In Living on the lake in prehistoric Europe: 150 years of lake-dwelling research. (Abingdon: Routledge). pp 9-21.

Wheatley, D. 1995. Cumulative viewshed analysis: a GIS-based method for investigating intervisibility, and its archaeological application. In Archaeology and geographic information systems: a European perspective, edited by Lock, G. and Stancic, Z. (London: Taylor & Francis). pp 171-185.

Wheatley, D. and Gillings, M. 2000. Vision, perception and GIS: developing enriched approaches to the study of archaeological visibility. In Beyond the map: archaeology and spatial technologies, edited by Lock, G.R., NATO Science Series A: Life Sciences. (Amsterdam: IOS Press). pp 1-27.