These Guidelines mention three different types of sampling strategies which may be employed to plan CHMP standard assessments (surveys) and to test predictive models when probabilistic sampling strategies are required (e.g. in the absence of known Aboriginal places in an activity area).
The following provides a brief explanation of each, with the understanding that other survey methods are available; and with the expectation that all HAs know and understand these methodologies and how to best employ them.
Random sampling
This method includes simple random sampling – where an activity area is gridded and random grid squares selected by a random number generator are surveyed. Other variants include cluster sampling – which involves dividing an activity area into random subunits and selecting some for survey.
Random sampling methods are designed to ensure statistical probability is used to predict site locations, reducing the potential for bias in archaeological findings.
Systematic sampling
This method places grid squares evenly across an activity area, and grid squares are selected for surveying at regular intervals. This allows for a selected percentage of an activity area to be sampled, again eliminating bias. Transect sampling is another example of systematic sampling which employs regularly spaced lines across the activity area instead of grid squares.
Systematic sampling helps to avoid the “clustering” of sample grid squares which may result from random sampling, and is often easier to implement in the field.
Stratified sampling
Also called stratified random sampling, this method combines both the random and systematic sampling to divide an activity area into smaller regularly spaced regions and then randomly sampling chosen grid squares within each region. An activity area may be divided into different landform zones (strata) and assigned proportionate sample units (so, for example, if a 10% sample is desired, then 2 hectares of a 20-hectare river terrace and 3 hectares of a 30-hectare floodplain might be selected for survey). These smaller subsets are selected at random for surveying. Another variation is stratified unaligned systematic sampling, which involves gridding out an area but starting at a random point within a grid square.
Stratified sampling may be employed to focus resources on areas of higher potential while still checking other areas. It ensures artefacts in low-density zones are not entirely missed, which can happen with pure random sampling.
Non-probabilistic sampling
This method is employed when Aboriginal place locations or likely locations within an activity area are already known prior to survey, and concentrates on those areas. This can be informed by prior discoveries and information about cultural values or Aboriginal intangible values relating to a place obtained from Traditional Owners prior to standard assessment.
Examples can be seen at the University of Texas website.
Updated