310 
3.3 Spaceborne data 
The day on which the field experiment was carried out (7/9/93) coincided with Landsat-5 passage TM over 
the study area (track 199, scene 24). Analysis results of this image will be included in a later stage. 
3.4 Derivative spectroscopy 
In fig. 2 two spectral measurements of the same sediment sample are presented. The upper curve is a 
measurement of the dry sediment, whilst the lower one is a measurement of the same sample with a 5 cm 
layer of water on top of it. The sample without water layer exhibits a monotonically increasing reflectance 
spectrum, which is the usual pattern for bare soils. There is a small trough evident around 680 nm, which 
could be associated with chloropyll-a absorption. As can be seen from the figure, the presence of a small 
water layer (water plus possible constituents) influences the spectrum rather dramatically. This results in a 
’darker’ overall spectrum, a more pronounced through around 680 nm and the introduction of some 
additional (water) absorption features. 
Some researchers identify distinct classes for dry and wet sediment (e.g. Yates et al, 1993), which is 
probably a sensible thing to do when satellite imagery is being used. Hobbs and Shennan (1985) report that 
even then distinction between those classes can be cumbersome. 
Figure 2. The effect of a water layer on the reflectance of a sand sample from the test site. 
a. Reflectance spectra of sand with and without water layer + diffemce spectrum 
b. First order derivative spectra 
High spectral resolution data offer the possibility of analysis of narrow spectral features. Derivative 
reflectance spectroscopy is one of the techniques suited to do so. Phenomena like troughs, peaks and 
shoulders can be identified more precisely on the basis of derivative analysis. Chen et al. (1992) showed the 
applicability of derivative reflectance spectroscopy to estimate suspended sediment concentration. Results of 
Goodin et al. (1993) show that first order and second order derivative spectra can be used to discriminate 
between the effects of water, algal chlorophyll and suspended sediment. They indicate that spectral effects 
caused by water can be minimized by taking the first derivative. 
Of all the measured sediment reflectance curves first order derivative spectra were calculated using a 5-point 
derivative calculation algorithm. First the reflectance data were ’smoothed’ by using the simple moving 
average principle, which tracks noisy trends quite well but implicates that the windowed median is nt 
smooth in the sense of differentiable. 
3.5 Results 
In figure 3 reflectance spectra of 6 sediment samples are presented. The figure includes the spectra 
corresponding to the samples with minimum and maximum clay percentage. The spectral correlation of 
reflectance data with clay percentage is very poor (Pearson ’s correlation coefficient N < 0.17). This is 
partly caused by not taking into account the presence of water in the sediment samples. 
The effect of using the derivative spectra for the analysis can be seen in in figure 4. The spectral correlation 
of the first order derivatives with clay percentage of the associated sediment samples has improved (Pearson 
’s correlation coefficient r amounts to -.62 for a wavelength of 627.5 nm.)