Refining near-infrared spectroscopy for collagen quantification: A new predictive model for archaeological bone

Abstract

Collagen is a vital archaeological material, preserving biochemical signatures that provide insights into past environments, diets, and human-animal interactions. However, diagenesis can lead to rapid and inconspicuous collagen degradation. Given the variability in collagen preservation and its significance for analyses such as radiocarbon dating, stable isotope analysis, and ZooMS, researchers have developed prescreening techniques to assess collagen preservation before destructive sampling. Current prescreening approaches, including %N and C:N ratios, typically require sample destruction and access to equipped laboratories. Spectroscopic techniques such as Raman spectroscopy and Fourier Transform Infrared spectroscopy have been explored as alternatives, but they are limited in penetration depth, generalizability (at present at least), and are often still destructive, if minimally. Here, we further develop single-point near-infrared (NIR) spectroscopy as a fully non-destructive, rapid, and field-portable method for prescreening bone for collagen preservation. Unlike FTIR and Raman spectroscopic techniques, NIR light penetrates below the surface of bone, enabling assessment of internal collagen preservation without destructive sample preparation. Using Partial Least Squares Regression (PLSR) and Random Forest (RF) modeling, we trained predictive models on whole bones with known collagen yields and validated the models on an independent archaeological collection. Both PLSR and RF models, when restricted to the 2030–2060 nm range, demonstrate strong and comparable performance while avoiding wavelengths associated with consolidants in our reference library. The models outperform traditional % N-based methods in identifying suitable samples for radiocarbon dating. These models enable the high-throughput screening of large collections of bone, improving sample selection and minimizing unnecessary destructive analysis.