International Archives of the 
and under-storey layers can only be recorded where the laser is 
able to pass through gaps larger than the footprint. 
This observation leads to the conclusion that by thresholding 
the first pulse histogram to select only those first pulses which 
are close to the under-storey it is possible to map canopy gaps. 
Displaying the results of such a thresholding exercise overlaid 
onto aerial photography reveals a very clear representation of 
canopy gaps. These range from macro features (rides and 
clearings) to small gaps caused by fallen trees. As yet no 
detailed validation data has been collected to establish the 
accuracy of this technique. 
ations it might be concluded 
Nevertheless, from these observ 
on of canopy surface 
that the first pulse data gives a representati 
morphology without the possibility of canopy penetration. 
4.2 Combined first and last pulse observations 
involved examination of 
The next stage of the analysis 
the same sampling 
combined first and last pulse data. Using 
units as for the first pulse analysis histograms were again 
produced for each of the four classes. In addition, plots of laser 
  
Birch 1 
Heg 
Frequency 
   
: akut A LHL 
  
Saitow 2 Saïtow ? 
al. 
Meet 
frequency 
  
Height 
Mixed woodland t Mixed woodians * 
Frequency 
  
Hawthom 2 
Frequency 
Height 
  
Photogrammetry, Remote Sensing and Spatial In 
formation Sciences, Vol XXXV, Part B7. Istanbul 2004 
      
Inclusion of the last pulse component in the data opens up the 
possibility of visualising canopy 3d structure. Clearly, when 
there is a first pulse recorded near to the top of the canopy and a 
corresponding last pulse recorded below it, the difference 
reflects the degree of laser penetration for that particular pulse. 
Examination of Figure 3 thus confirms expectations about the 
canopy structure for the four woodland classes based on their 
descriptions from section 3. 
Birch reveals a significant under-storey 
consequence of the relative openness of 
e formation of a significant sub-canopy 
layer of shrubs and saplings. The variability in crown height is 
also clearly reflected in the histogram. The corresponding 
profile further highlights this variability in canopy height 
together with important boundary locations corresponding to 
the top of the under-storey and the base of the canopy. It is 
ant number of first returns from 
eflecting the degree of canopy 
The histogram for 
component arising as à 
the canopy permitting th 
interesting to note the signific 
the under-storey layer again r 
openness. 
t Sallow exhibits a much more simple structure. 
In contras 
d a far less distinct 
There is a very uniform, dense canopy an 
under-storey layer. The canopy layer is uniform in height and 
is so dense that very little first return energy penetrates to the 
under-storey. However, last return data is returned from the 
under-storey confirming the ability of the LIDAR to reveal 3d 
structure even when the canopy is very dense. Again the 
boundaries between top of under-storey, stems amd base of 
canopy form distinct thresholds in the histogram. 
The histogram for Mixed Woodland has three distinct peaks 
corresponding to under-storey, sub-canopy layer of Hawthorn 
and saplings and the outer layer of Birch/Alder. Despite the 
complexity and thickness of the canopy it is clear from the 
profile that first return energy still reaches the under-storey 
layer. This again reflects the openness of the dominant Birch 
canopy. 
The final histogram for Hawthorn provides the strongest of 
contrasts. The upper layer is relatively thin in depth but 
extremely dense. Very little first return energy comes from 
below the canopy base. The relatively open layer of stems is 
clearly visible but possibly i 
the issue raised earlier over last 
returns underestimates the amount of material in this part of the 
vertical profile. 
From all of the histograms it is clear that integration of small 
footprints over appropriately sized areas can reveal detailed 
information about the structure of vertical canopies. This 
information accords well with expectations based on qualitative 
  
  
  
  
Figure 3. Combined first and last pulse data for the four 
sample sites and corresponding height profiles. 
height against easting were also produced. The results are 
shown in Figure 3. Detailed analysis of the last pulse data 
revealed that an unexpectedly high proportion of values 
coincided with their corresponding first pulse measurements 
raising the possibility that there is an under-representation of 
last pulse returns when they are close to the first pulse. 
However, no evidence was found that the measured last pulse 
data was inaccurate or that the basic patterns in the data were 
distorted. 
description of the stands and their species composition. 
Furthermore, unlike large footprint returns, it is clearly possible 
to discriminate between canopy gaps where both first and last 
n come from the under-storey and canopy penetration 
and an associated last 
tion opens up the 
retur 
where there is a first return in the canopy 
return from lower levels. This last observa 
possibility of deriving indices of canopy openness and density. 
vident from the graphs that the cumulative height 
profiles are very different for each of the four stand types in 
question. This opens up the additional possibility of integrating 
this type of information into broader classification schemes. 
Previous attempts to classify this site based on spectral data 
alone have met with very limited success due to spectral 
It is also € 
1088 
stri 
cat 
cal 
use 
wit 
dis 
mo 
the 
can 
By 
exa 
fact 
top 
stuc 
and 
ope 
whi 
the 
retu 
laye 
laye 
first 
For: 
unde 
all c 
the 
variz 
may 
pulse 
gaps 
and 
assoc 
used 
canoj 
returi 
This 
cano] 
light 
comp 
invol 
mapp 
In ad 
heighi 
clear 
This 
classi! 
detail 
the int 
needei 
6. RE 
Blair, 
wavef( 
elevati