Multivariate calibration of near infrared spectra for predicting nutrient concentrations of solid moose rumen contents

Abstract

This study aimed at establishing calibrations to predict nutrient concentrations of solid moose ( L.) rumen content using near infrared spectroscopy (NIRS), as an alternative to expensive chemical analyses. NIR reflectance spectra of 148 dry pulverized samples were recorded. The scanned samples were then analyzed for crude protein, available protein, microbial nitrogen (N), ash, acid-detergent fiber (ADF), neutral detergent fiber (NDF) and lignin contents following standard chemical analysis procedures. The calibration models were derived by Orthogonal Projection to Latent Structure (OPLS) and validated using external prediction sets. The calibration models accurately predicted crude protein, available protein and ash contents (R = 0.99, 0.96, and 0.92, prediction error = 0.39, 0.72 and 0.53% dry matter, respectively) while NDF (R = 0.92; prediction error = 2.23% dry matter) and ADF (R = 0.89; prediction error = 1.94% dry matter) were predicted with sufficient accuracy and that of microbial-N (R = 0.81; prediction error = 1.25 mg yeast-RNA g dry matter) and lignin (R = 0.84; prediction error = 1.05% dry matter) were acceptable. The ratio of performance to deviation values were > 3.0 for crude protein and available protein, between 3.0 and 2.5 for ADF, NDF and lignin, and 2.32 for microbial-N; attesting the robustness of the calibration models. It can be concluded that NIR spectroscopy offers a quick and inexpensive procedure for prediction of nutrient concentrations of solid rumen contents in wild herbivores.Alces alces2222–12