Fuzzy classification of sub-Antarctic vegetation on
Heard Island based on high-resolution satellite
imagery
Arko Lucieer
Centre for Spatial Information Science (CenSIS)
School of Geography and Environmental Studies
University of Tasmania
Hobart, Tasmania, Australia
Email: Arko.Lucieer@utas.edu.au
Abstract—This study presents the first land cover classification
of sub-Antarctic Heard Island based on high-resolution IKONOS
imagery and a supervised fuzzy classification algorithm. Tran-
sition zones between vegetation types and rock are identified
and analyzed by using fuzzy membership layers and a confusion
measure as a quantification of spatial and thematic uncertainty.
This study shows that a combination of high-resolution satellite
imagery and fuzzy classification techniques is extremely valu-
able for mapping sub-Antarctic vegetation and for quantifying
uncertainty related to vegetation transition zones.
I. INTRODUCTION
Heard Island is a pristine and remote volcanic sub-Antarctic
island in the Southern Ocean, south of the Antarctic Polar
Frontal Zone (APFZ). Heard Island arguably provides one of
the most rapidly changing environments for plant growth in the
sub-Antarctic region, due to extensive and rapid glacial retreat
which has been accelerated by rising temperatures. There has
been minimal human impact on the ecosystem of Heard Island,
but warmer conditions will increase the ease for invasion
of new species. Its location, climate conditions, and pristine
nature make Heard Island an ideal site to study the effects
of global climate change [1], [2], [3], [4]. Up-to-date and
accurate spatial information is, therefore, of crucial importance
for sustainable management of the island. During previous
expeditions to Heard Island in 1987/1988 and 2003/2004
terrestrial plant ecology has been studied and vegetation maps
have been produced from field samples and aerial photography.
These field surveys are expensive, labor intensive, intrusive
and often only cover small areas. Because of the island’s
remoteness and harsh environment, satellite imagery provides
advanced and cost-effective means to map its vegetation cover
and to quantify vegetation changes. High-resolution imagery
with a pixel size of one meter is most suitable to capture the
fine-scale structure of sub-Antarctic vegetation [5].
The first step in identifying vegetation changes related to
global warming is the development of a methodology to
reliably map Heard Island’s vegetation communities. Image
classification is a suitable technique to translate the spec-
tral information in a satellite image into thematic vegetation
classes. Most classification studies apply traditional classifi-
cation techniques, such as the maximum likelihood classi-
fier, dividing the image into discrete classes based on the
spectral properties of each pixel. These classifiers model the
study area as a number of unique, internally homogeneous
classes that are mutually exclusive. These assumptions are
often invalid, however, especially in areas where transition
zones (e.g. vegetation intergrades) and mixed pixels occur.
Land cover classes can hardly ever be separated by sharp
or crisp boundaries, in feature space as well as geographic
space. Moreover, classes are often hard to define resulting in
vagueness and ambiguity in a classification scheme. Most, if
not all, geographical phenomena are poorly defined to some
extent and, therefore, fuzzy set theory as an expression of
concepts of vagueness is an appropriate model for dealing with
uncertainty in remotely sensed imagery [6], [7], [8].
This paper presents the first study that applies fuzzy clas-
sification based on high-resolution satellite imagery for sub-
Antarctic vegetation mapping on Heard Island. The main
objective of this study is to develop and apply supervised
fuzzy classification techniques to map sub-Antarctic vegetation
types and quantify uncertainty related to vegetation transition
zones. This study uses high-resolution IKONOS imagery for
classification of Heard Island’s sub-Antarctic vegetation.
II. HEARD ISLAND
Heard Island (Fig. 1) is a pristine, remote and volcanic
sub-Antarctic island in the Southern Ocean approximately
4000km southwest of Australia and 1000km north of the
Antarctic continent (53oS; 73oE, 40km by 20km). Its location
and climate conditions make Heard Island an ideal site to study
the effects of global climate change. Up-to-date and accurate
spatial information is of crucial importance for sustainable
management of the island. Because of the island’s remoteness,
satellite imagery provides advanced and cost-effective means
to map its vegetation cover and to monitor vegetation changes.
Between 1979 and 2004, a combination of field surveys
and aerial photography has been used to aid in mapping
vegetation communities [9]. These field surveys are expensive,