The Use of Near‐Infrared Reflectance Spectroscopy to Predict Species Composition of Forage Mixtures 1

Assessment of species composition of hay, pasture, and range samples is an important but laborious factor in forage quality evaluation. Near‐infrared reflectance spectroscopy (NIRS) was evaluated as a rapid technique for prediction of the species composition of hay and pasture forages and weeds. In Exp. 1, 97 mixtures were formed using eight species. Each species varied from 0 to 100% in each mixture. Each mixture of dried, ground material was scanned with monochromatic light from 1100 to 2500 nm at 2‐nm intervals, and the reflected energy was recorded. Percentage of each species was then regressed on the spectra using a modified stepwise regression procedure. Calibration equations with coefficients of determination ( R 2 ) from 0.94 to 0.99 and standard errors of difference (SED) from 1.9 to 6.8% were obtained using five to seven equation terms. Johnsongrass [ Sorghum halepense (L.) Pers.] could not be predicted (SED = 6.8%) as precisely as the others, whereas blue panicgrass ( Panicum antidotale Retz.) and sunflower ( Helianthus annuus L.) were precisely predicted (SED < 2%). One to three harvest dates of seven species of hay were used in a second experiment to determine if variability within a species resulted in additional imprecision. In general, SED were higher; however, all but ‘Midland’ bermudagrass [ Cynodon dactylon (L.) Pers.] gave SED of 10 or less, and the R 2 of the calibration data set ranged from 0.84 to 0.98. Bermudagrass predictions were more precise if samples containing eastern gamagrass [ Tripsacum dactyloides (L.) L.] cut at boot were removed from the data set. With proper calibration, NIRS could save much time normally used to separate species in hand‐clipped samples. The number of samples that can be analyzed in an experiment could be increased, and hence the accuracy of determination of species composition in a given pasture should be improved.