Chemometric models for high-throughput biomass grading of the kelp Ecklonia radiata, using mid-infrared (MIR) and near-infrared (NIR) spectroscopy

Abstract

The kelp Ecklonia radiata is a target species for aquaculture in the southern hemisphere, but the biomass quality varies both spatially and temporally, making high-throughput chemometric methods essential for biomass grading. We built partial least squares regression models and partial least squares discriminant analysis models using mid-infrared and near-infrared spectroscopy, based on 162 calibration samples and 24 validation samples. The near-infrared models generally outperformed the mid-infrared models based on both predictions and sensitivity to outliers. Near-infrared partial least squares regression models were highly accurate for predicting glucose, alginate, phlorotannins, and carbon content (relative prediction error < 8 %) and the near-infrared partial least squares discriminant analysis model could accurately predict region of origin for Wellington samples (zero false negatives and one false positive). Overall, our results demonstrate that near-infrared and mid-infrared spectroscopy are valuable tools for biomass grading of raw seaweed samples and can be used for establishing provenance.