Chemometric analysis of EDXRF measurements from fossil bone

Bone chemistry is an important source of biological and environmental information. Elemental compositions of archaeological and fossil bone have granted insight into the diets of ancient humans and other animals, as well as informing about the burial conditions of bone. Chemical studies of ancient bone can be performed non‐destructively with portable energy‐dispersive X‐ray fluorescence (EDXRF) spectrometers, which is particularly advantageous for on‐site analyses of museum specimens. Portable EDXRF instruments carry some analytical disadvantages, however, which may result in reduced precision or accuracy. Analytical shortfalls may be overcome by analysing inter‐sample trends in EDXRF spectral data instead of reported concentration measurements. We investigated the utility of statistically treating handheld EDXRF spectra from fossil bone and teeth, specifically the normalisation and mean centering of spectral data before principal component analysis. Fossil bone and tooth specimens were sourced from two Pleistocene localities in the Western Cape of South Africa, Swartklip 1 and Elandsfontein Main. Samples from the two localities could be distinguished using principal component score values, and coefficient loadings allowed chemical interpretation of the score clusters. Swartklip 1 samples were associated with elevated concentrations of Ca, indicating an additional Ca‐bearing mineral (i.e. calcite), whereas Elandsfontein Main samples were associated with elevated Fe and Sr concentrations. Fossil bone chemistry could be related to groundwater percolation through the sedimentary matrices of each locality. The methodology behind the case study presented here could readily be applied elsewhere and would be particularly useful to handheld EDXRF studies of museum specimens. Copyright © 2011 John Wiley & Sons, Ltd.