n 
i] 
Ng 
er 
of 
ing 
ANALYSIS OF INTERACTION PATTERNS BETWEEN VEGETATION CANOPIES AND 
SMALL FOOTPRINT, HIGH-DENSITY, AIRBORNE LIDAR 
G.S.Amable * B.J.Devereux **,, T.Cockerell *, G. Renshaw? 
"Unit for Landscape Modelling, Sir William Hardy Building, Tennis Court Road, Cambridge, CB2 1QB 
KEY WORDS: LiDAR, Canopy structure, Aerial Survey, Semi-natural vegetation, laser pulse, distribution functions 
ABSTRACT: 
High resolution, small footprint, imaging lidar systems are becoming increasingly important for monitoring forest canopies. Without 
doubt they have the potential to dramatically change conventional 
the laser beam interacts with the top of the vegetation canopy whil 
methods of forest survey. It is often assumed that the first pulse of 
¢ the last pulse is assumed to come from the ground. In this paper 
we present results from a research programme which has been monitoring vegetation canopy dynamics at the UK Woodwalton Fen 
Site of Special Scientific Interest for the last twelve months using an Optech ALTM 33 airborne lidar. The instrument generates 
33,000 laser hits per second. For each hit the first pulse, last pulse and intensity are recorded with an accuracy of plus or minus 
I5cm. Surveys have been conducted in both leaf-on and leaf-off states. In an examination of several types of canopy at the site we 
apply statistical analysis techniques to the distributions of hei 
ghts contained in the first and last pulse data. We show that the 
different vegetation and stand types are characterised by different probability distribution functions which can be used to derive 
information about the structural properties of the vegetation canopies. We will further show that appropriate processing of lidar point 
clouds can yield structural information similar to that generated by less readily available large footprint systems. The results 
presented here are part of an on-going research programme funded by the Sir Isaac Newton Trust. The consortium includes the 
University of Cambridge Unit for Landscape Modelling and English Nature. 
I. INTRODUCTION 
A significant proportion of the world's landscapes are 
characterised by complex, semi-natural vegetation 
communities. In Northern Europe the moorlands and heaths are 
good examples of the type whilst in the Mediterranean, 
garrigue, maquis and mixed woodland provide classic 
examples. Frequently these landscapes are characterised by 
high scenic value and are of considerable ecological 
significance by virtue of their biodiversity. In the case of many 
Mediterranean sites, there are very high rates of endemism. 
To date however, remote sensing has only played a very limited 
role in the mapping of these areas at the level of detail required 
for conservation and management. Notwithstanding issues of 
resolution, these areas are often hilly or mountainous, have a 
very fine grain and very complex patterns of vegetation. Often 
the vegetation types are not well differentiated by traditional 
spectral signatures. Communities are often mixed, and complex 
gradients of type and biomass sometimes occur. In 
Mediterranean areas in particular, it is well known that these 
properties lead to complex patterns of shadowing and texture in 
most types of imagery. These in turn lead to very high 
proportions of mixed pixels and in turn, this leads to problems 
of mapping from imagery (see for example, Hill et al. 1995). 
Normally robust classifiers like maximum likelihood produce 
poor results because of high class variances and mixed pixels. 
Measures of vegetation amount such as NDVI are often biased 
by virtue of mixed soil/vegetation proportions within pixels and 
  
by the fact that they only record the outermost portion of dense 
canopies. 
One possible area where substantial progress can be made with 
these problems is airborne laser scanning or LiDAR. LiDAR 
measurement involves firing short pulses of laser energy at the 
ground target and measuring the return time for the energy to be 
reflected back to the sensor. Vegetation canopies present a 
surface to the laser which is porous to energy. As a 
consequence each laser pulse potentially returns a distribution 
of energy in which the first recorded values come from the top 
of the canopy and subsequent values are returned from lower 
layers. If the laser pulse is sufficiently powerful and the canopy 
sufficiently open, the graph of return time against energy has a 
shape dependent on the vertical distribution of canopy 
components (ground, under-storey, stems, branches and 
photosynthetically active leaves in the canopy (Lefsky er al., 
2002; Lim, et al. 2003) 
Two types of system are currently in use. So called ‘large 
footprint’ systems have a wide divergence angle for the laser 
beam so that it illuminates a large (typically 10 — 30 metre), 
circular area (footprint) of the scene. Often just a few footprints 
(6 — 10) are recorded for each scan line but the entire return 
pattern of energy versus time (height) is recorded in great 
detail. Such devices are relatively specialist and confined to 
NASA based research (Means er al. 2000). 
* Corresponding author. BJD1(cam.ac.uk; Telephone +44 1223 764375 
1085