The aim of this study was thus to test the relevance of statistical methods to detect the variations in spectral signature of oil-palm trees correlated to Ganoderma disease, a fungus responsible of high loss of yield and trees in palm groves. The objective is too discriminate infected palm trees and to establish a ranking in the degree of infection. Some previous studies (Lanore, 2006; et Brégand, 2007) revealed that it is feasible, but the number of individuals was too small to lead to statistically reliable models; thus, it is still to confirm and validate. More especially, the present study focuses on the possibility of infected palm-tree discrimination in accordance to four sickness degrees: Healthy, Low, Medium and High infection. It will test this potential at several scales: the leaflet, the canopy, and by remote sensing.