Computer Applications and Quantitative Methods in Archaeology

advertisement
Temporal uncertainty and artefact chronologies
Computer Applications and Quantitative
Methods in Archaeology:
Session: Embracing uncertainty in
archaeology
Southampton, 28 March 2012
Andrew Bevan
(with contributions from many
colleagues)
UCL Institute of Archaeology
Chronology and
Periodisation
•
•
Lots of work emerging in other
disciplines about space-time
analysis and how to handle
uncertainties associated with the
timing of events
Archaeological time poses its own
peculiar challenges… a native
archaeological tradition associated
with radiocarbon dating (and
mapping) but the handling of
categorical dates now increasingly
considered as well
Aoristic Analysis
•
A probability of presence of an
event, site, artefact is assigned to a
series of time-block, usually based
on defining start and end dates, with
a uniform distribution in between.
•
Works well for existing
archaeological datasets and
supports Monte Carlo simulation
•
Typically, done long after data
collection, when some information is
already lost …
•
Aoristic time-blocks are often equal
and absolute …
…mixes up the uncertainty of
attribution to a time period (a
relative scale) with the fixing of this
period in absolute units (e.g. to a
range in years BCE)
•
Ratcliffe 2000 Int. J. GIS
Johnston 2003. CAA
Crema, Bevan, Lake 2010 J. Archaeological Science
Crema in press J. Archaeological Method and Theory
Intensive Surface Survey
on Antikythera
•
•
•
Directed by Andrew Bevan (UCL) and James Conolly
(Trent University, Canada), in collaboration with Aris
Tsaravopoulos (Greek Archaeological Service)
Stage-one survey by walkers spaced 15m apart
Stage-two survey of certain localities on a 10x10 grid.
Uncertainty in Assigning
an Artefact to a Period
Jar handle
Initial impression of date:
“possibly Hellenistic…more likely
Late Roman…but not in between”
Perhaps recorded in a traditional
database as:
“Late Roman”, “Late Roman?” or
“Late Roman/?Hellenistic”
An alternative is “percentage confidence”… 30% Hellenistic, 70% Late Roman
Bevan, Conolly, Hennig, Johnston, Quercia, Spencer, Vroom in press. Archaeometry
Diagnostic Confidence
by Period
First or Second Palace
Late Roman
Mapping Diagnostic Confidence
Sherds with ≥70%
confidence of being
Middle Byzantine
(c.1000-1200 AD)
Sherds with ≥20%
confidence of being
Middle Byzantine
(c.1000-1200 AD)
Sherds with ≥70%
confidence of being Middle
Byzantine to Early Venetian
(c.1000-1400 AD).
Mapping Diagnostic Confidence
All possible candidate sherds
for a Middle Roman date
(i.e. >0% confidence),
Close-up (diameter 600m) with
percentage confidences of Middle
Roman date shown as graduated
colours overlain on all other sherds (in
grey)
Overall Uncertainty
n
Uj = ∑ min( Pij ,maxk ≠j ( Pik ))
i =1
€
n
∑P
i =1
ij
Example for period 2:
(40+10+0+20+0) / (40+10+0+20+0) = 1
and for period 3:
(40+10+0+20+0) / (50+10+0+20+100) = 0.389
Pairwise Uncertainty
n
Ujk = ∑ min( Pij ,Pik ) × 2 ( Pij + Pik )
i =1
€
Example for period 2 and 3:
(40+10+0+20+0) x 2 / (90+20+0+40+100)
= 140/250 = 0.56.
First or Second Palace
Late Roman
Local Uncertainty
Relative chances of defining Middle Roman
dates compared to Early and Late Roman
Wider Implications for
Archaeological Fieldwork
•
Observer Variability – ability to explore intra- and inter-observer variability
•
Permanent Collection – ability to demonstrate added value of the collection
and long-term storage of artefact assemblages.
•
Physical Re-investigation – can be used to design strategies for new
fieldwork
•
Regional Profiling – supports construction of a regional uncertainty profile
and exploration of case-by-case departures from it.
Thank you….
…and to those involved in…
The Antikythera Survey Project
www.ucl.ac.uk/asp
…and on behalf of ASP to…
The Greek Ministry of Culture
The Canadian Institute in Greece
Datasets available as ADS Collection 1115
http://dx.doi.org/10.5334/4f3bcb3f7f21d
Download