362 
where Rmax is the reflectance value at the high point at llOOnm and Rmin is the reflectance minimum 
between 1100 and 1264nm. 
The relationship between both of these relative depth indices and wet% are summarised in Table 
3. The relative depth parameter derived by Method 2 was found to be more closely associated with the 
wet% than that derived by Method 1. Figure 4 shows the index calculated by Method 2, plotted against 
wet%. The other features characterisation parameters derived by the Method 1 procedure (asymmetry, 
FWHM and position of the absorption maximum) were found to have no correlation with wet%. The weaker 
correlation for the relative depth parameter resulting from Method 1 may indicate that the addition of the 
second high point confuses the relationship between reflectance and wet% or that the linear interpolation 
between the two high points is not appropriate for characterising the shape of the reflectance curve. 
Table 3 
Wet % 
Method 1 
-0.66 ** 
Method 2 
-0.73 * 
Significance level: * = .0001, ** = .001 
Figure 4. Scattergram of Relative Depth Index against canopy wet%. 
4.4. Sensitivity of spectral indices to processing 
Experiments showed that the level of smoothing influenced the correlation between the first derivative and 
wet% (Figure 5). For the visible/NIR region the effect of increased smoothing is evident on the correlation 
plot, but does not result in a shift in the wavelength position of the peak correlation at 950nm with wet%. 
However, for the 1150nm water absorption feature in the SWIR region, different levels of smoothing result 
in shifts in the wavelength position of peak correlations in addition to some variation in the peak correlation 
factors. In particular, the increased smoothing from a cutoff value of 2 to 5 points causes the peak correlation 
to shift from 1124nm to 1142nm. The proposed explanation of this sensitivity is that the two variables are 
generally well correlated across quite a broad spectral region (30nm) which spans the water absorption 
feature, but minor processing differences emphasise different spectral features, and this alters the balance 
of the correlation and shifts the peak correlation. In this case the strongest correlation shifts down the left 
limb of the water absorption feature, from an approximately mid-way down the slope to a point at 
approximately the base of the slope. Figure 6 shows the reflectance and first derivative spectra for one