Abstract
Monitoring of the spatial distribution of vegetation characteristics (e.g., forage yield and quality, botanical composition) using aerial remote sensing would be useful for the management and utilization of grasslands. The present study assessed the potential application of remotely sensed hyperspectral data (observed by an AISA Eagle hyperspectral imaging sensor on-boarded a Cessna plane) to grassland management. Specifically, hyperspectral data were used to identify plant species and estimate forage yield, quality, and other grassland vegetation characteristics. Cluster analysis of the hyperspectral data extracted from the images revealed that the plots could be classified into four groups. Principal component analysis of the hyperspectral data accentuated the hyperspectral signal of the near-infrared region as the first principal component. Broad-leaved dock (Rumex obtusifolius L.) was found in the group A classified based on the cluster analysis with the strongest hyperspectral signal in the near-infrared region. However, grass species in the other three groups (B, C, D) classified based on the cluster analysis could not be distinguished using the hyperspectral data. The present results suggest that hyperspectral data in the near-infrared region were effective for distinguishing broadleaved dock in a mixed sward. In addition, the linear discriminant function of hyperspectral data could distinguish the existence of broad-leaved dock at a correct answer rate of not less than 69%.
