| Abstrakt: |
Quantifying pasture composition requires either laborious or subjective approaches. Evaluations of near‐infrared reflectance spectroscopy (NIRS) to determine botanical composition of mixed pasture swards have shown potential. In this study, characterization of botanical composition of pastures comprised primarily of bahiagrass (Paspalum notatumFlugge), aeschynomene (Aeschynomene americanaL.) and phasey bean [Macroptilium lathyriodes(L.) Urb.] by NIRS was evaluated. Three approaches (hand‐composited samples, single‐component samples, and actual pasture samples) were compared for equation development. Theoretical potential of NIRS is illustrated by high coefficients of determination (0.98–0.99) and low standard errors (1.4–2.9%) of equations for the above species from hand‐composited samples. Equations developed from the three approaches were evaluated for estimation of the botanical composition of a separate group of pasture samples. Equations developed from hand‐composited samples from a single source of each component were not acceptable for estimating composition of pasture samples despite the excellent calibration statistics. Single‐component samples approached adequate results only for composite total grass and total legume groups, even though the pasture sample composition appeared to be well represented in the calibration sample set. Equations from pasture samples provided useful estimates of sample means, although some individual samples were poorly estimated. Thus, botanical composition of these pastures may be estimated using equations from actual pasture samples, and estimates of total grass and total legume may be obtained from use of single‐component samples, which provides further labor reductions. A comparison of original software and updated software packages CAL, BEST, REG70, and partial least squares principal component regression showed none to be consistently superior. |
Plný text