3. Methodological Issues

The context of this study is a Holocene, alluvial plain composed of sedimentary deposits carried by the River Arno where a complex system of lagoons and channels formed the original palaeo-environment. Over one hundred sites, dating from the late Palaeolithic to the end of the Middle Ages, have been detected in this environment by means of different diagnostic techniques (remote sensing analysis, boreholes, historical cartography, literary sources) (Pasquinucci 2000). Nowadays, most of this territory (located between the cities of Pisa and Livorno, north-western Tuscany), is occupied by agricultural land, although an increasing percentage of it is going to be affected by new development works that could compromise the preservation of the still-undisclosed below-ground archaeological deposits. For this reason the Pisa coastal plain constitutes a good case study for applying a predictive model: for here a substantial number of sites could be hidden in the still unsurveyed portion of the study area, which is too wide to be investigated by means of traditional prospecting techniques. In this sense, the predictive model, starting from a statistical hypothesis, tries to 'foresee' which are the most suitable areas for locating new archaeological sites, giving the opportunity to exclude from the actual research, substantial areas in which the probability of new finds are fairly low.

Our research, based on the data collected over twenty years of study by the University of Pisa chair of Ancient Topography (Menchelli 1984), has started from the analysis of the archaeological record and the set-up of an inductive kind of predictive model on a GIS platform. Afterwards, the resulting model has been tested through different steps, firstly on the internal dataset, and subsequently on new, independent, external data, collected in the field. As for the methodology applied, the research project uses the Regge Valley (Netherlands) predictive model as a comparison (Brandt et al. 1992), due to a set of very similar characteristics (Table 1). First of all, the two areas are very similar in extent, around 120km2 , as well as in geology and topography - very flat land mostly composed of alluvial deposits. From an archaeological point of view, both of the models present a relatively small number of known sites (76 in the Regge model and 104 in the Pisa model) distributed through time (late Palaeolithic to 15th-16th centuries AD). Moreover, in both cases the analysis of the location of known sites led to the use of environmental variables to build up the final model. Furthermore, the number of predictors used is pretty much the same, with the Regge Valley model based upon six predictors against the five employed in the Pisa coastal plain model. But, most importantly, in both of the models the final aim as to locate new archaeological sites. In particular, the goal was 'to produce a workable model for practical application that would provide archaeologists with some advantages in locating sites' (Brandt et al. 1992). That is the purpose of the Pisa coastal plain model as well, where priority was given to the creation of a risk map, able to determine and quantify the probability of new finds in the study area.

Table 1: Characteristics of the Regge Valley project and the Pisa coastal plain project

 Regge Valley project Pisa coastal plain project
Area extent 120 km 114 km
Geology Mainly alluvial deposits
Topography Mostly flat
Number of sites 76 104
Chronology From late Palaeolithic to the modern age
Predictors Environmental
No. of predictors 6 5
Final aim Locating new sites


© Internet Archaeology/Author(s)
University of York legal statements | Terms and Conditions | File last updated: Thu Dec 1 2011