Spectral feature analysis for assessment of water status and health level in coast live oak (Quercus agrifolia) leaves

Abstract

Reflectance and relative water content (RWC %) were measured from a total of 306 coast live oak (Quercus agrifolia) leaf samples. The maximum (or minimum) first derivative (1D) and its corresponding wavelength position (WP) were extracted from 10 spectral slopes along each spectral reflectance curve. A correlation analysis was conducted between these spectral features (1Ds and WPs) and the corresponding RWC of oak leaves. An analysis of the variance of spectral features and RWC was also carried out between two health levels of oak leaves, healthy and infected. The results indicate that high correlations exist between some derivative spectral features (1Ds and WPs) and the RWC of oak leaves. With all 306 leaf spectra covering a wide range of RWC of oak leaves (including healthy, infected and newly dead), maximum (or minimum) derivative values at both sides of the absorption valleys near 1200, 1400 and 1940nm and the WP of the right side of the 1200 nm valley and of both sides of the 1400 nm and 1940 nm valleys were found to have high correlations with RWC; and with a selection of 260 samples of only green and green-yellowish leaves (including only healthy and infected levels), the WPs at the right side of the 1400 nm valley, at the left side of the 1940 nm valley and at the red well each have a stable relationship with leave RWC. The result of an ANOVA of 20 spectral features suggests that the difference of some spectral features corresponding to two health levels of oak leaves is significant at a 0.95 confidence level, especially for the above-mentioned wavelength position features. © 2004 Taylor and Francis Ltd.