A Simple(R) Model to Predict the Source of Dietary Carbon in Individual Consumers.

Quantitative individual human diet reconstruction using isotopic data and a Bayesian approach typically requires the inclusion of several model parameters, such as individual isotopic data, isotopic and macronutrient composition of food groups, diet-to-tissue isotopic offsets and dietary routing. In an archaeological context, sparse data may hamper a widespread application of such models. However, simpler models may be proposed to address specific archaeological questions. As a consequence of the intake of marine foods, individuals from the first century ad Roman site of Herculaneum showed well-defined bone collagen radiocarbon age offsets from the expected terrestrial value. Taking as reference these radiocarbon offsets and using as model input stable isotope data (δ13C and δ15N), the performance of two Bayesian mixing model instances (routed and concentration-dependent model versus non-routed and concentration-independent) was compared to predict the carbon contribution of marine foods to bone collagen. Predictions generated by both models were in good agreement with observed values. The model with higher complexity showed only a slightly better performance in terms of accuracy and precision. This demonstrates that under similar circumstances, a simple Bayesian approach can be applied to quantify the carbon contribution of marine foods to human bone collagen. [ABSTRACT FROM AUTHOR]