Geostatistics and archaeology: a review
C. D. Lloyd
School of Geography, Queen's University, BELFAST BT7 1NN
Most archaeological data have a spatial component. As such, the way archaeological data vary in space is often a crucial part of their interpretation. One set of tools that have been developed for the analysis of spatial data is Geostatistics. At the present time, the potential of Geostatistics in archaeology has not been realised. This paper seeks to provide a basic outside of some key geostatistical techniques and to illustrate how these techniques may be useful to archaeologists. Geostatistics makes use of spatial dependence (or autocorrelation) in one or more variables distributed in space.
Geostatistics is used to widely to
(i) characterise spatial variation;
(ii) estimate at locations for which no samples are available and
(iii) design sampling strategies.
Clearly, all of these objectives have direct relevance to archaeology and associated disciplines such as palaeoecology. The core tool of most geostatistical analyses is the variogram. In simple terms, the variogram is a plot of the average distance separating paired data values (x axis, where data are divided into classes such as a separating distance of 1 to 2 m, 2 to 3 m and so on) against half their average squared difference (y axis, where again the average is obtained for all pairs of data in each class). A mathematical model may be fitted to the variogram and the coefficients of the model can then be used to inform an interpolation procedure called kriging. That is, the form of spatial variation, as characterised by the variogram, is used to determine the influence the available data have in an estimate at some location for which no observation is available. A by-product of kriging is the kriging standard error. This is a function of the form of the model fitted to the variogram and the spatial configuration of the data with respect to (i) the location of the estimate and (ii) the locations of other data. The kriging standard error is guide to uncertainty in estimates but it may also be used in the design of sampling strategies, as shown in one of the case studies presented at the end of the paper. In the paper, some published applications of geostatistics and palaeoecology were outlined and reviewed. Following this, two case studies using published data were presented to clarify some of the relevant concepts and to illustrate how geostatistics may be useful to archaeologists. The case studies focused on (i) the characterisation of the spatial distribution of early Bronze age spearheads and (ii) the mapping of soil phosphates and assessment of optimal sampling strategies. The review of published work and the case studies demonstrated that geostatistics has much potential in archaeological research and the work of archaeologists would benefit from some knowledge of its capabilities as well as its limitations.