LAUSR

Oil palm (Elaeis guineensis) trees are an important contributor of recent economic development in Southeast Asia. The high product yield, and the consequent high profitability, has led to a widespread expansion of plantations in the greater region. However, oil palms are susceptible to diseases that can have a detrimental effect. In this study we use hyper- and multi-spectral remote sensing to detect diseased oil palm trees in Krabi province, Thailand. Proximate spectroscopic measurements were used to identify and discern differences in leaf spectral radiance; the results indicate a relatively higher radiance in visible and near-infrared for the healthy leaves in comparison to the diseased. From a total of 113 samples for which the geolocation and the hyperspectral radiance were recorded, 59 and 54 samples were healthy and diseased oil palm trees, respectively. Moreover, a WorldView-2 satellite image was used to investigate the usability of traditional vegetation indices and subsequently detecting diseased oil palm trees. The results show that the overall maximum likelihood classification accuracy is 85.98%, the Kappa coefficient 0.71 and the producer’s accuracy for healthy and diseased oil palm trees 83.33 and 78.95, respectively. We conclude that high spatial and spectral resolutions can play a vital role in monitoring diseases in oil palm plantations.