359 
3.3. Smoothing 
Spectra were then smoothed using an inverse Fourier transform of the Blackmann window, still treating the 
two spectral regions separately. This procedure automatically adjusted the smoothing weights at the 
beginning and end of the data array to avoid edge effects. This procedure removed features with a period 
shorter than the defined cut-off level. Three different cut-off levels were tried: 2,5 and 10 and the smoothed 
data were compared with the unsmoothed data. This suggested a cut-off level of 5 data points was 
appropriate to removing noise but retaining spectral information and for succeeding analysis this level of 
smoothing was selected as optimal. However, the effect of different levels of smoothing on the final 
correlations was also examined. Findings from this are presented in section 4. 
3.4. First derivative 
The first derivative was calculated from the smoothed data by interpolating a 3 point polynomial through 
consecutive data points, then differentiating the polynomial. The derivative of reflectance was used because 
of its capacity to remove the effect of leaf structure on reflectance (Danson, et al, 1992). 
4. ANALYSIS AND RESULTS 
The analysis focused on the relationship between reflectance and the first derivative of reflectance and three 
biophysical variables. Data from two of the 30 sample points were discarded either because the biophysical 
data were incomplete or because the quadrat included a high proportion of flowers. At the remaining 28 
points a single reflectance spectrum related directly to the area of vegetation sampled for derivation of the 
biophysical data. Spectral data were compared with each variable in turn by means of a Product Moment 
Correlation function, which produced a correlation coefficient for all channels of spectroradiometer 
measurement. In addition, correlation coefficients were calculated between the three biophysical variables. 
4.1. Correlation between biophysical variables and reflectance 
Figure 1 shows the correlation coefficients between spectral reflectance (resampled and smoothed with a 5 
point cut-off) and three biophysical variables: total chlorophyll, fresh biomass and canopy wet%. The strong 
association between all three biophysical variables is evidenced by the similarity of the plots and is 
summarised in Table 1. 
Figure 1. Product Moment Correlation Coefficients between spectral reflectance and three 
biophysical variables.