Determination of botanical composition in multispecies forage mixtures by near infrared reflectance spectroscopy
Résumé
To improve on the time-consuming manual method often used to determine botanical composition of grasslands, near infrared reflectance spectroscopy (NIRS) was used to predict the proportion of each family (grass vs. legume) and also the proportion of each species in grass or grass-legume forage mixtures. Samples were collected in multispecies sown swards ("real mixtures") and artificial mixtures were produced from samples collected in monospecific swards. The mixtures were composed of perennial ryegrass (Loliutn perenne), cocksfoot (Dactylis glomerata), tall fescue (Festuca arundinacea) with or without white clover (Trifolium repens). Plant samples were dried and ground, and a total of 3607 spectra were collected on a NIRS Foss 6500 machine. An independent validation set was available. The proportion of white clover in the mixtures was very well predicted (SEPC = 3.3), as a consequence of differences in biochemical composition between grasses and legumes. Prediction of each grass species was not as good (SEPC between 5.4 and 9.1) but still satisfactory, particularly considering that their morphological and biochemical characteristics are close. These equations can be used to rapidly determine species composition in multispecies grasslands.