Vous êtes sur la page 1sur 15

CAA2012 Southampton, UK 26-30 March 2012

Philip Verhagen VU University Amsterdam (NL) Tom Brughmans University of Southampton (UK) Laure Nuninger CNRS, UMR6249, Besanon (F)

Frdrique Bertoncello CNRS, UMR6130, Nice (F)

PHC Van Gogh 2010/2011 - Introducing the human (f)actor in predictive modelling for archaeology
Frans-Nederlandse Academie / Nederlandse Organisatie voor Wetenschappelijk Onderzoek (NWO) Ministres des Affaires trangres et europennes (MAEE) / Enseignement suprieur et de la Recherche (MESR).

predictive modelling mainly based on environmental factors socio-cultural factors too difficult to integrate?
accessibility is a potential factor to include, but how to define it?

Llobera (2000) introduced the concept of scale-dependent accessibility


modification of Shimbel or all-node distance matrix provides the minimal distance between each node of a network

relative distances determined using cost surfaces


time-consuming calculations for larger areas

interest: determine areas where local and regional accessibility are opposed
predicting the possible location of fords and passes, or defendible

positions in the landscape understanding why certain portions of the landscape are not occupied

least cost path calculations usually applied to determine routes between known locations
exception: Whitley and Burns (2008)

paths calculated within a certain radius from a randomly or systematically selected sample of starting points
creates multiple cost paths for a region at various scale levels no preferred start or end points

cost distances calculated from starting point


paths tracked back from a sufficiently large number of points at a

specific distance repeat and overlay for each starting point

+ + + + + + + + + +

+ + + + + + + + + +

+ + + + + + + + + +

+ + + + + + + + + +

+ + + + + + + + + +

+ + + + + + + + + +

+ + + + + + + + + +

+ + + + + + + + + +

cumulative cost paths resemble networks that have


weighted edges non-directed edges a large number of nodes

few hubs (nodes with lots of connections)

can network analysis techniques say something more about the structure?
e.g. identify corridors and bottlenecks

node-based global measures


average degree, heterogeneity, clustering coefficient

local measures
closeness and betweenness centrality

import/export from ArcGIS through .dbf


needs from-node, to-node and coordinates of nodes

open source software

based on space syntax concepts


calculates local (edge-based) measures
choice (=betweenness)
both topological and metric depth topological/metric radius can be defined im- and export from ArcGIS through MIF/MID-format

freeware (UCL)

cumulative cost path based networks contain a lot of noise, need post-processing to extract the general structure
for this exercise:
select 10% and 20% most frequently chosen paths simplify network structure (expand linear raster features

with 2 cells) and thin it convert to polyline

cumulative cost path based networks are not typical


few hubs, weakly structured general structure only apparent when selecting the most frequently

chosen paths

some network analysis measures can be interesting for better understanding potential foci of movement in the landscape
in particular betweenness (choice) and control

analysis very sensitive to edge effects

node-based network analysis packages


do not include metric depth cannot calculate measures over various scales

space syntax does


but gives no information on the nodes cannot include relative distances (costs)

how to interpret the results in archaeological terms?

Vous aimerez peut-être aussi