T02_P17

Predicting the collagen yield of unknown samples

Olsen J¹,  Schrøder T¹, Philippsen B¹, Kanstrup M¹

¹Aarhus University, Aarhus, Denmark

Bone samples for radiocarbon analysis are difficult and time consuming. Often the bone preservation is unknown and frequently the bone collagen yield is underestimated resulting in small samples. If the collagen can be predicting some of these problems can avoided. At Aarhus AMS Centre we have collected FTIR spectra, %C and %N of raw bone samples of more than 200 samples. Further, we extracted collagen and conducted FTIR analysis as well as elemental and stable isotope analysis. Presented here are modelling results from which the collagen yield is predicted from measurements of raw bones. Different types of models (correlation models, neural networks, machine learning) and their performance will be presented and discussed.