A Rapid Method for Predicting Alfalfa Quality in the Field

Abstract

Predictive equations for alfalfa quality (PEAQ) based on the tallest stem and the most mature stem of a sample were developed to estimate the neutral detergent fiber (NDF), acid detergent fiber (ADF), and crude protein (CP) concentrations in alfalfa (Medicago sativa L.). The objective of this study was to evaluate these equations in a preharvest setting in the field. Separate equations were developed for use on samples obtained from (i) all cuttings (original equations), (ii) first cutting only, and (iii) second and third cuttings only. Prediction equations accurately estimated alfalfa fiber composition of 308 samples taken in 1992 and 1993. Crude protein was not accurately predicted using PEAQ. Using the original PEAQ on all samples resulted in r 2 values of 0.72 for NDF and ADF and 0.37 for CP while root mean square error (RMSE) values were 3.02, 2.47, and 2.12% of dry matter (DM) for NDF, ADF, and CP, respectively. The relationship between PEAQ estimates and observed values did not improve using the first cut equations on first cut samples when compared with estimates obtained using 114 the original equations on first cut samples. Equations developed from second and third cuts and the original equations predicted NDF and ADF similarly in second and third cut alfalfa. A nearly perfect relationship between predicted and observed values would be shown by a slope of 1.0 and a y-intercept of 0.0. Twelve of the 15 equations had slopes statistically different from 1.0 and y-intercepts different from 0.0 indicating bias in these models. The biological differences may be insignificant, however, since many of the values for slope and y-intercept are relatively close to 1.0 and 0.0, respectively. Quality estimates were obtained for all samples using commercial near infrared reflectance spectroscopy (NIRS). Root mean square error and r 2 were similar for NDF and ADF using either PEAQ or NIRS. Near infrared reflectance spectroscopy was much more accurate than PEAQ in estimating CP. However, all NIRS regression equations were biased to some extent because slopes were different from 1.0 and y-intercepts were different from 0.0. For field evaluations, the PEAQ method shows promise as a rapid and inexpensive method of estimating preharvest alfalfa fiber quality components, but not crude protein.