Predictive Modeling of Archaeological Sites

1
Predictive Modeling of Archaeological Sites

• Lower Adirondack GIS Users Group Meeting
• September 14, 2005

2
Why Create a Model?

• To narrow focus of field investigation to
  locations most likely to contain prehistoric
  sites.
• Contract surveys only test a small sample of
  total project area
• Outliers may reveal interesting distinctions
  between cultures and time periods

3
Considerations

• Terrain diversity
• Resource availability
• Known sites (i.e., sample pool)
• Time Period/Cultural Sequence
• Practical Concerns of Prehistoric Peoples
• Food, Water, Shelter, Trade, Protection,
  Transportation

QUESTION Is all human settlement practical
(i.e., cost/benefit analysis)?
4
Environmental Variables

• Slope
• Elevation (absolute, relative)
• Landforms (valleys, ridgeline)
• Aspect (e.g., north-facing)
• Raw materials (e.g. stone for tools)
• Water (lakes, streams, confluences)
• Soil type

5
Sample Set

• Accuracy of locations
• Points versus Polygons
• Biased pool of known sites
• Do sites count twice (or more times) if multiple
  time periods are represented?
• Is the sample set large enough for statistical
  accuracy?
• Narrow set based on time period, culture, site
  type?

6
Variables to GIS Data

• Need measurable GIS data layer to represent each
  variable
• Catagorical/Qualitiative coding
• Raster representation

7
Model

• Random Selection of Locations
• Equal number of known site cells and assumed
  Non-site cells
• Logistic Regression
• Stepwise, backward
• Eliminate variables with little causality
• Generate an output GRID
• Cell value between 0-1
• Closer to 0 low probability of finding sites
• Closer to 1 high probability of finding sites

8
Does the Model Work?

• Does the final model appear skewed?
• Test the model in the field

9
QUESTIONS?