Percolation Analysis – Archaeological Applications at Widely Different Spatial Scales

M Simon Maddison,Sophie C Schmidt

doi:10.5334/jcaa.54

M Simon Maddison, Sophie C Schmidt

Open Access

https://doi.org/10.5334/jcaa.54

Copy DOI

Abstract

This paper provides a summary of the percolation method of cluster analysis, and then describes two detailed archaeological case studies where the technique is applied at significantly different geographical scales. This demonstrates not only the potential for the technique within archaeology, but also how it can be applied at different spatial scales with different objectives appropriate to the specific problem in question. The technique, originally developed in physics and more recently adopted in geography, is a way of identifying groupings or clusters, purely based on spatial separation using Euclidian distance. The paper includes a discussion comparing percolation analysis with other clustering techniques already established within archaeology. The first case study uses percolation analysis as an exploratory tool for investigating the distribution of hillforts in Britain, identifying clusters and groupings at a national scale. The purpose is to identify possible socio-political entities for further investigation. In the second case study, the technique is applied to excavated features at a sub-regional level, in Saxony-Anhalt, central Germany, with the different objective of identifying settlement sites along a 13 km strip excavation. The aim here was to arrive at estimates of settlement sizes, which in turn can inform landscape archaeological surveys for deciding on the attribution of single finds to an already known site or registering a new site. Percolation analysis is shown to be an effective tool for analysing archaeological data sets at widely different geographical scales and should become a standard part of the archaeologist’s spatial analysis toolkit.

Highlights

Percolation analysis has been recently applied in archaeology for identifying clusters and groupings at a national scale, for example investigating the distribution of hillforts in Britain and Ireland (Maddison 2019) and Domesday sites in England in 1086 AD (Arcaute et al forthcoming)
This paper provides a summary of the percolation analysis method, compares it to other known cluster algorithms in archaeology and provides two detailed case studies where the technique is applied at significantly different geographical scales
This will demonstrate the potential for the technique within archaeology, as an exploratory tool, and how it can be applied at different spatial scales with distinct objectives appropriate to the specific problem in question

Summary

Introduction

Percolation analysis has been recently applied in archaeology for identifying clusters and groupings at a national scale, for example investigating the distribution of hillforts in Britain and Ireland (Maddison 2019) and Domesday sites in England in 1086 AD (Arcaute et al forthcoming). This approach interprets the archaeological record as a continual distribution in space, with spots of higher density, the so-called ‘sites’, and areas of lower density (Wobst 1983: 39) At this point the continuously excavated data set of the road and the percolation clustering analysis can be fruitfully combined: As percolation works with radii around points it may assist in quantitatively assessing which distances between features are common and at what distance features do not seem to belong to each other anymore. This question needs to be answered for each archaeological culture separately, as we cannot assume similar settlement distributions through time (Schirren 1997: 30). Features inbetween sites may exist, such as traps or temporary storage facilities, places such as working areas or fields which may be detected by off-site surface find distributions and will not be included

Percolation analysis

Interpretation

Conclusion