363 
representative sample point, to illustrate where the peak correlations are occurring. This sensitivity to 
processing emphasises that the exact wavelength of peak correlation may be misleading - it is possible that 
this will different for different data sets. This will have serious implications for the use of standard indices 
based on reflectance or derivative at fixed wavelength positions. The absence of such sensitivity in the 
visible/NIR region may be due to the higher spectral resolution of the raw data in this wavelength region 
or the fact that absorption feature here is an isolated trough compared with the broader feature around 
1142nm. 
The strength of the relationship between wet% and the Relative Depth Index (as calculated by 
Method 2) was found to be relatively insensitive to the degree of spectral smoothing. 
Figure 5. Variation in Product Moment Correlation Coefficient between first derivative of 
reflectance and canopy wet% for different degrees of data smoothing. 
Wavelength (nm) 
Figure 6. Réflectance and first derivative across the 1150nm water absorption région.