Mapping weed infestations in an annual crop has implications not only for site-specific herbicide applications but also for planning future management strategies and understanding weed ecology. A controlled laboratory experiment, involving detached leaves, was conducted to investigate the potential to discriminate two crop and five weed species using hyperspectral and multispectral remote sensing. Stepwise discriminant function analyses showed that reflectance in the visible and “red-edge” regions of the spectrum were consistently required for species discrimination. The seven species were correctly identified 90 and 89% of the time using the hyperspectral and multispectral data, respectively, and the classification rules derived from discriminant function analyses. Errant species prediction with the hyperspectral data resulted in a grass being predicted as a grass and a broadleaf as a broadleaf. However, for multispectral data, incorrect classifications were more serious because errant predictions sometimes resulted in a grass being classified as a broadleaf and vice-versa. Further studies using plants at a variety of growth stages, from a variety of environments, and at the canopy level are warranted.
Additional index words: Hyperspectral, leaf reflectance, multispectral, Amaranthus retroflexus, Avena fatua, Brassica kaber, Chenopodium album, Setaria viridis, Brassica napus, Triticum aestivum, AMARE, AVEFA, CHEAL, SETVI, SINAR.
Abbreviations: SMA, spectral mixture analysis.