361 
obtained from individual leaves. 
Figure 3 shows a plot of the first derivative at 1142nm against wet%. Points are coded according 
to the nitrogen application. It is noticeable that the sample points in the zero nitrogen application plot show 
considerable variability. The range of wet% covered by this group is equal to half of the total range in 
wet%. It is also apparent that the correlation is strongly affected by three numbered points on the plot. 
Excluding these three points results in a much improved correlation of -0.820 between the first derivative 
at 1142nm and wet%. Examination of these three points revealed that for two of the points, reflectance in 
the SWIR was significantly lower than for the rest of the data set and the reflectance spectrum for the third 
point showed a marked spectral peak near the shoulder of the water absorption feature at 1150nm. The 
cause of both effects remains unknown. 
0 = 0 kg/ha 
Wet % 
Figure 3. Scattergram of the first derivative of reflectance at 1142nm against canopy wet%. 
4.3. Correlation between wct% and other relative depth indices 
The relative depth of leaf water absorption features at 1150nm should be independent of leaf structure 
characteristics which affect the level of reflectance. Two methods of characterising this feature were 
examined. Both were applied to the resampled and smoothed reflectance spectra. 
Method 1 - A semi-automated features extraction procedure was applied to derive the first measure of the 
relative depth of the feature. The procedure involved identifying two high points either side of the absorption 
feature. This was done initially using an interactive graphical display routine. The high points were found 
to occur consistently at the same wavelengths for most spectra, at the wavelengths of llOOnm and 1264nm 
respectively. A modified version of a features characterisation algorithm which fits a continuum to the 
spectral curve (Kruse et al, 1990) was then applied. This involved connecting high points on the spectrum 
by linear interpolation of a straight-line segment. For the 1150nm water absorption feature, the wavelength 
position of the maximum deviation of the actual spectrum from the interpolated spectrum was identified and 
the relative depth of the feature at that position expressed as: 
((interpolated value - actual value) / interpolated value) * 100 
The algorithm also calculated the full width of the feature at half its maximum (FWHM) and a measure of 
the asymmetry of the feature. 
Method 2 - An alternative method of defining the relative depth of the feature was also used. This 
compared the reflectance minimum as described above with the first of the high points only. A normalised 
measure of the relative depth of the feature was derived as: 
((Rmax - Rmin) / Rmax ) * 100