Trent University, Canada; University College London, UK; Greek Archaeological Service (26th EPKA)
under the aegis of the Canadian Institute in Greece and the Hellenic Ministry of Culture
Team M tractwalking
Re-visiting the issue of ground surface visibility: (top) a plot showing the showing the noisy but significant relationship between artefacts density and the amount of heavy scrub cover (maquis) when considered at smaller spatial scales; (bottom left) a Normalised Difference Vegetation Index (NDVI) calculated from Quickbird image bands, and (bottom right) a supervised classification, following ground-truthing, of vegetation classes in the same area (maquis shown in red). Images by James Conolly.
Spatial Analysis

ASP's concern with consistent collection routines and tight integration of digital datatsets reflects a desire to use these as platforms for further spatial and computational analysis. In particular, we feel that many surveys provide quantified information that, ironically, resists most attempts to compare it quantitatively. In contrast, Antikythera is small enough for us to acquire most datasets comprehensively across the whole island, to create systematic, permanent artefact collections and to record all of our findings to a uniform resolution. This allows us to deploy spatial statistics with much greater confidence than in other landscape survey contexts where variable levels of investigation, unverifiable dating of finds by a few experts in the field, and/or ragged sample edges can all be very frustrating. More specifically, we have been keen to quantify the extent and spatial configuration of the following:

a) Inconsistent artefact recovery related to geomorphology and surface visibility. This is a topic which has received much previous attention, but we are able to consider it at finer spatial scales by using a combination of multispectral Quickbird bands (e.g. for a more reliable assessment of surface visibility based on vegetation cover), explicit measures of variable artefact diagnosticity (to explore how our overall results are afected by the varying confidence with which artefacts canbe attributed to specific chronological periods), accurate plotting of finds for each chronological phase (to allow us to stratify our understanding of landscape change), and high-resolution DGPS survey of particularly interesting locales (to consider on-site formation processes).

b) Correlations between areas of concentrated human activity (e.g. settlements) and various environmental variables. This is also a well-known topic for landscape scale analysis, but one which has traditonally been practiced by deriving global correlative statistics that are prone to mis-specification and/or over-simplification. We attempt to navigate through such issues by employing belief or probability-based statistical reasoning and analytical methods such as spatial regression that are sensitive to the high levels of spatial autocorrelation present in a variety of landscape phenomena.

c) Spatial attraction or repulsion between different types of human activity area. This includes both patterns of clustering or dispersal amongst contemporary sites and the spatial relationship between sites of one chronological phase and the vestiges of previous ones (a temporal interaction which is arguably key to the way any society develops a richer sense of place). For this we use various forms of multi-scalar point pattern and area-based analysis (e.g. modified versions of kernel density estimation, time series analysis and Ripley's K).

Spatially-weighted regression
A spatially-weighted regression of surface ceramic density with logged slope values and NDVI as independent variables. Image by A. Bevan.

d) Routes of human and animal movement. A combination of improved anisotropic cost surface algorithms (e.g. GRASS r.walk) and the perspectives offered by agent-based modelling allow us to consider both on and off-island travel in more formal ways. Some of the earliest phases of human activity on the island may well be palimpsests of repeated, seasonal visitation and we hope that more detailed modelling will elucidate of the relationship between particular points of human disembarcation, patterns of food, water and material resources on the island, and routes of human and animal travel.

e) Patterns of on and off-island visibility. Visibility studies have become a key area of potential cooperation between a) those exploring the role of sensory experience in the making of human landscapes, and b) those using computational methods. We have used cumulative viewshed anlaysis and high-capacity distributed computing to build up a reliable profile of both on- and off-island patterns of visibility, This background population then allows us to consider the possible relationship between differential patterns of visibility and the locations chosen for various farming, hunting, religious and/or piratical sites on Antikythera. We have also tried to contextualise these otherwise rather bald calculations, at the very least in their modern human context, through repeated photography of particular landscapes in different weather conditions, study of placenames and interviews with Antikythera's current inhabitants,