Short communication: Long term performance of near infrared spectroscopy to predict intramuscular fat content in New Zealand lamb.

This study investigates the performance of a partial least squares regression model to predict intramuscular fat (IMF) in lamb M. longissimus lumborum developed using near infrared (NIR) data collected under a range of different conditions. A total of 26 independent NIR datasets were collected across 7 years, including 14 flocks, four devices and several measurement conditions. A model is developed and its performance is tested using a total of n = 3201 NIR spectra and intramuscular fat percentage measurements by wet chemistry. The model had a coefficient of determination by cross-validation of 0.52, which agrees with previous results using smaller numbers of animals. Overall the results show that near infrared models can be robust across many varying conditions. These models could potentially be implemented in an automated meat quality monitoring system.