235 
a decrease in visible reflectance through the rise in photosynthetic absorption, and an increase in near-infrared 
reflectance. 
At the beech site, temporal fluctuations in visible reflectance are also broadly the inverse of canopy cover, while 
for near-infrared this relationship is directly proportional, but again, some temporal discrepancies exist. While 
canopy cover rises rapidly between May and June and remains high through to December, visible reflectance 
exhibits a gradual decrease between May and August, then a smaller decrease through to November. This can 
be explained by the fact that photosynthetic pigments accumulate gradually as the beech leaves mature 
(Rinderle et al., 1991). Therefore, despite the rapid increase in canopy cover after bud-burst, the young beech 
leaves contain low concentrations of pigments and thus are inefficient at absorbing visible radiation. During 
maturation, leaf pigmentation increases causing an increase visible absorption, and decreased reflectance, 
despite canopy cover remaining relatively constant. Canopy cover decreases rapidly between December and 
January, due to the synchronous tree leaf abscission, which is matched by a drastic increase in visible 
reflectance. This indicates that during the period of leaf senescence, prior to abscission, the total pigment 
content does not decrease. Observed changes in the colour of beech leaves during senescence suggest that 
different pigments, such as the anthocyanins, come to dominate the absorption, as the role of the chlorophylls 
decreases (Boyer et al., 1988). Although the colour of beech leaves change, mean visible reflectance remains 
constant, and only immediately before abscission are pigments withdrawn from leaves, and the visible 
absorption decreases. Between December and May, the remaining pigments are destroyed as the leaf litter 
decays, and visible reflectance increases. 
Near-infrared reflectance parallels the rapid increase in canopy cover as the tree leaves develop, and remains 
high until October, whence it decreases rapidly, despite canopy cover remaining high. Two mechanisms may 
account for this. Firstly, cell breakdown during leaf senescence could reduce the internal scattering and 
reflectance of near-infrared radiation. The vertical canopy photographs confirm that between October and 
November, a significant proportion of the beech leaves began to show visible signs of senescence. Boyer et 
«/.(1988) have found that such visible changes are often associated with significant alterations in the internal 
geometry of leaves as a result of progressive desiccation. The second possibility is that although canopy cover 
remains high between October and December, the number of layers of leaves in the tree canopy decreases as 
leaves are gradually lost, thereby reducing near-infrared reflectance. Due to the nature of the measurements of 
canopy cover used in this experiment, such changes in the number of layers of leaves may not have been 
detected as a change in canopy cover. However, observations made during this period suggest that very few 
beech leaves were lost from the canopy, and an increase in leaf litter at the woodland floor was only observed 
between December and January Thus the results favour the first hypothesis. A linear (least squares) regression 
analysis was performed on the time series for reflectance and percentage canopy cover, using a 95% confidence 
limit, for each site in turn. This indicated that at the ash site, visible reflectance has a strong negative 
relationship to percentage canopy cover (r 2 = 0.87), while for near-infrared reflectance this relationship is less 
strong and positive (r 2 = 0.70). At the beech site, such relationships are much weaker (r 2 = 0.63 and 0.56 
respectively). 
3.2 Seasonal variation in NDVI 
Figure 4 shows the seasonal variations in NDVI and percentage canopy cover, for the two study sites. The ash 
canopy shows gradual variations in NDVI over time, but does appear to reach a plateau of around 0.9 between 
July and November, while the beech site exhibits rapid shifts between two levels of NDVI. Figure 4 clearly 
illustrates that at both sites, there is a close temporal correspondence between NDVI and percentage canopy 
cover. Linear regression between these two variables produced very high correlation coefficients for both study 
sites (r 2 = 0.96 (ash) and 0.93 (beech)). The relationship between NDVI and canopy cover is much stronger 
than those between either visible or near-infrared reflectance (individually) and canopy cover, especially for the 
beech site. Figure 5 shows that a single regression model can be fitted to the combined data from the ash and 
beech sites.