16 A GIS investigation of site location and landscape relationships in the Albegna Valley, Tuscany Philip Perkins The Open University, United Kingdom 16.1 Introduction The Albegna Valley/Ager Cosanus Survey started torians of Italy {e.g., Barker 1972, 1985; Malone & work in southern Tuscany in 1979 following on from Stoddart 1994, pp. 81-93). The technique has been the excavation of the villa at Settefinestre (Carandini applied by these scholars because none of the draw- 1985b). The survey has systematically sampled over backs listed above are held to have been important in 275 km^ of the Ager Cosanus and the Albegna Valley simple farming systems without complex land tenure (Figure 16.1). The detailed publication of the sur- (Malone & Stoddart 1994, pp. 81-93). Attempts have vey work is now in press but various interim reports been made to elaborate the technique to incorporate and associated studies have already appeared in print social as well as economic factors by employing site (Attolini et al. 1982; Cambi et al. 1991; Cambi & Fen- catchment techniques within the context of a settle- tress 1988; Carandini 1985a; Celuzza & Regoli 1982). ment hierarchy (Gent & Dean 1986). However, the This paper will concentrate upon the analysis of the technique seems to be inappropriate in this context settlement pattern in the Etruscan period, and in par- because this study does concern complex agricultural ticular the relationships between settlement location and economic systems. Furthermore, this study con- and landscape. cerns a complex society capable of manipulating its own natural environment, and one which was in con- tact with other regions of the Mediterranean. A final 16.2 Landscape and settlement reason that site catchment analysis was not employed is that this study concerns a settlement system as a Site location may be viewed in two fundamental ways: whole, and not individual settlements. from the viewpoint of an individual site in the land- This point introduces the alternative view of the scape or from the viewpoint of the landscape which settlement pattern as a landscape inhabited by archae- is partially occupied by sites. The first approach is ological sites. In this approach the settlements are characterised by site catchment analysis {e.g., Vita- considered with reference to the entire landscape with Finzi & Higgs 1970) where all the territory within the purpose of generahsing about the locations of sites. a certain distance (often taken as the distance of a This is achieved by considering the distribution of sites one hour walk) of a settlement site is considered as within the landscape in relation to the distribution of conditioning the subsistence strategy and economy of other elements of the landscape. Thus a more general the settlement site. Land within this territory is clas- landscape analysis is performed and the locations of sified according to its suitability for pasture, arable archaeological sites are considered with reference to purposes or wasteland and the proportions of each that landscape (Kvamme 1992, p. 127). Economic, class of land is used to determine likely subsistence and perhaps social interpretations may then be devel- strategies of the occupants of the sites (Dennell 1987, oped from an understanding of which parts of a land- pp. 73-5). The technique has its critics, the major scape were occupied by a settlement pattern. This drawbacks are potential change in the natural envi- approach, favoured in environmental studies, is now ronment since antiquity, the indirect relationship be- becoming more common, particularly with the devel- tween land area and the economic importance of a opment of computer based Geographical Information particular land use, its ignorance of non-economic fac- Systems (Gis) capable of handling large spatially refer- tors and the fact that it considers the site catchment enced data sets (Burrough 1986; Cliff & Ord 1981). A area in isolation from the rest of the landscape and recent archaeological application of this approach has indeed the world {e.g., Gaffney et al. 1985; Hodder & illuminated the Greek conquest of the island of Hvar Orton 1976, pp. 229-36). Despite these criticisms the in Croatia (Gaffney & Stancic 1992) and another, Ho- technique has achieved some popularity in the analy- hokam agricultural systems (Kvamme 1992). sis of field survey results, particularly among prehis- 133 Figure 16.1: Location of study region and sampled transects 16.3 Methodology atively associated with a particular variable, and the strength of the association is given by #^ (Shennan The approach taken here considers the locations of 1988, pp. 78-81). The technique is simple and will de- sites identified within the sampled areas of the survey tect statistically significant associations between dis- with reference to a variety of natural criteria: alti- tributions of variables in the landscape and the distri- tude, slope, aspect and solid geology. The purpose bution of the archaeological sites, which may then be is to identify what kinds of location were associated archaeologically interpreted. with settlements and how these changed through time. The purpose of the analysis is first to identify Data for the first three variables was derived from any relationship between site and landscape and then 1:100,000 topographic maps of the Istituto Geografico to investigate how that relationship changes through Militare and the geological information from the cor- time by repeating the analysis for each century be- responding maps of the Servizio Geologico d 'Italia. tween the 7th and the 2nd BC {e.g., Figures 16.2-16.3). The method in each case is the same, the spatial The calculations and results tables are tedious and are distribution of classes of each of the variables (alti- presented in detail elsewhere (Perkins 1995).^ tude, slope, aspect and solid geology) within the whole Altitude above sea level was analysed in 50m of the sample transects are compared with the loca- bands; within the sampled area altitude rises from sea tions of the sites. If sites were randomly distributed level to above 650m, yielding 13 categories to compare through the landscape one would expect 20% of the with the settlement patterns for each of the six cen- sites to be located upon limestone, for example, if turies. The largest category of land is that below 50m limestone occupied 20% of the sampled area. The which represents nearly 30% of the entire area. The observed distribution of sites is systematically com- distribution of altitude is presented in Figure 16.4. pared with the expected distributions. The difference Slopes have been measured as an average of blocks between the two distributions is measured with a x^ of land 300m square. These have then been classi- test (Shennan 1988, pp. 65-76) which indicates if the fied into four bands, negligible for slopes 0-8%, slight difference is statistically significant. This technique for 8-16%, moderate for 16-25% and steep for slopes provides a straight forward means of analysing fea- over 25% (USDA 1951). The distribution of slopes is tures of the location (Kvamme 1992). The analysis presented in Figure 16.5. may be taken further by considering further statis- Aspect is calculated as the average direction a tics derived from the x^ statistic; Yule's Q provides a slope faces over blocks of land 300m square. The as- means of assessing whether sites are positively or neg- pect has been classified into none {i.e., level ground). 134 o 5tm O ' 5km Figure 16.2: 5th century settlements (Etruscan) Figure 16.3: 2nd century settlements (Roman) Figure 16.4: Distribution of altitude above sea level Figure 16.5: Distribution of slopes in the valley north to north east, north east to east, east to south there are many others such as soil, rainfall, frost-free east etc. The distribution of aspects is presented in days etc. that contribute in a variety of ways which Figure 16.6. have not been taken into account. Nevertheless, the The 15 geological classes are taken from the statistical analysis yields a variety of significant posi- 1:100,000 geological maps and the distribution of the tive and negative associations between landscape ele- geology is presented in colour on CAA Web site. ments and the settlement patterns from different cen- turies. So far each aspect has been considered in isolation, but each is only a part of the landscape, 16.4 Discussion and and should best be considered in combination with development the other variables. Indeed all of these criteria are closely related, slope and aspect cannot exist without The purpose of this analysis has been to identify sta- one another and both may be derived from the spa- tistically significant elements of the relationship be- tial distribution of altitude. However, Kvamme has tween landscape and settlement patterns. Only four demonstrated that in a heterogeneous landscape el- components of the landscape have been considered. evation, slope and aspect are not closely correlated 135 (Kvamme 1992, p. 130), thus we may have confidence a rather unlikely looking swathe of land in the hilly in the results of the x^ analysis. areas of the middle valley. The model suggests that The problem remains of how the different land- settlement location in this period was not particularly scape variables may be combined to produce more associated with any of the elements of landscape con- generalised models of the relationships between land- sidered here. scape and settlement patterns. Here a spatial ap- In the central period between the sixth and the proach is adopted where the observed relationships third centuries two models may be generated, both between each element of the landscape and the settle- based on all four landscape elements. The model (Fig- ment locations are used to identify the areas in the ure 16.8) of positive association highlights the Pleis- valley which have been found to be associated with a tocene terraces in the Albegna valley and along the settlement pattern.