Prediction of chemical, nutritive and agronomic characteristics of wheat by near infrared spectroscopy

Abstract

Near infrared spectroscopy (NIRS) is widely used in the flour milling industry for rapid determination of moisture and protein in wheat. However, these measurements give little indication of the nutritive value of wheat when fed to poultry or pigs. Accurate estimates of nutritive value require specialist facilities and are time-consuming and costly. Accordingly, prediction from chemical or NIRS measurements would be of some considerable benefit. In the current study 160 samples of wheat, representing 24 different varieties, were used to generate NIRS calibration equations for chemical, nutritive and agronomic characteristics. Predictions of chemical constituents in wheat were very accurate. Coefficients of determination (r2) were 0·94 for dry matter, 0·90 for crude protein, 0·97 for ash, 0·78 for starch and 0·98 for oil. True metabolizable energy in broiler chickens was predicted more accurately (r2 = 0·52 for adult birds, 0·74 for young birds) than apparent metabolizable energy (r2 = 0·45). Digestible energy (r2 = 0·17) and nitrogen digestibility (r = 0·22) in pigs were not predicted very accurately on a smaller subset (n = 33). Agronomic characteristics were predicted very accurately (r2 = 0·98 hardness, 0·80 bushel weight, 0·99 thousand-grain weight). Predictions of nutritive value of wheat from chemical or agronomic characteristics are very inaccurate, since coefficients of determination vary from zero to 0·25. It is concluded that NIRS can accurately estimate the chemical composition of wheat, but accurate prediction of nutritive value is reduced by animal variation. Nevertheless, NIRS is potentially more reliable for assessing nutritive value than chemical composition or agronomic characteristics.