ibul 2004
1 Remote
ssor, IRS-
% {5.2 m)
1e spatial
1e first of
d'in 1999.
arge parts
prising |
tispectral,
dly more
0, EROS-
chromatic
has also
where. A
aunch of
| imagery
resolution
ly). Since
including
ions (two
1 and 5 m
al sensor),
struments
ectral
= Indian
servation
y.)
been used
has been
me of the
cploitation
il. (2001)
ved from
forest age
rests, and
ited as an
lark et al.
However,
areas, and
| analyses.
benefit of
ecological
extensive
1s such as
d marine
uet ef al.
/er coarse
n is fairly
(becomes
tified and
International Archives of the Photogrammetry, Remote Sensing and Spatial Information Sciences, Vol XXXV, Part B1. Istanbul 2004
characterised increases. Indeed, there are various examples of
studies where fine spatial resolution imagery has been
compared favourably to coarser spatial resolution imagery. For
instance, IKONOS imagery has been found to be more accurate
than SPOT HRV and Landsat Enhanced Thematic Mapper Plus
(ETM+) imagery for monitoring forest storm damage (Schwarz
et al. 2003) and mapping coral reefs (Capolsini et al. 2003).
However, fine spatial resolution imagery is not always
appropriate for ecological studies. Hochberg and Atkinson
(2003) found IKONOS imagery less accurate than hyperspectral
imagery for distinguishing coral, algae and sand, and Asner et
al. (2002) concluded that IKONOS imagery was insufficient for
accurate estimation of tree crown dimensions. Further, Sawaya
et al. (2003) demonstrated IKONOS and QuickBird imagery to
be useful for resource management, but suggest that these data
are uneconomic for large area studies.
It should be noted that airborne remote sensing may be a
suitable alternative to fine spatial resolution spaceborne
imagery, given that the relatively low altitude of airborne
platforms enables the generation of very fine spatial resolution
data. However, airborne remote sensing is limited in that data
are acquired on a piecemeal basis (compared to continuous
satellite sensor image acquisition), they may be expensive and
they are particularly susceptible to geometric distortion (Goetz
et al. 2003). Overall, the key factor determining successful
applications in ecological remote sensing, and in remote sensing
in general, is to match project goals to technical capabilities
(Sawaya et al. 2003). Therefore, the level of detail required in
any individual study will determine whether or not fine spatial
resolution imagery is required. Sometimes, fine spatial
resolution imagery may be useful only as a supporting data
source, in combination with other resources (Quinton ef al.
2003). For instance, Palandro ef al. (2003) describe the use of
IKONOS imagery to assess the accuracy of Landsat TM and
ETM+ image classification.
4. AFRICAN APPLICATIONS
Fine spatial resolution spaceborne imagery has been tested
fairly extensively for a range of ecological analyses in North
America, for tropical forest studies in South America and for
coral reef projects at locations throughout the world. However,
relatively little such work has been conducted throughout
Africa. Thenkabail (2004) describes a major study conducted in
Nigeria, Benin and Cameroon to compare the capabilities of
IKONOS and Landsat ETM+ imagery for representing
rainforest and savanna ecoregions. NDVI analysis was used to
determine the vegetal component of a range of land cover
classes and ecological units. Thenkabail (2004) concludes that
IKONOS data provide a more detailed depiction of vegetation
and related factors such as biomass than Landsat ETM+ data.
This is due partly to the finer spatial resolution of IKONOS,
and partly to the greater (11 bit) dynamic range of IKONOS
data than (8 bit) Landsat ETM- data. In another project,
Thenkabail et a/. (2004) conduct an exhaustive comparison
between IKONOS imagery and various other sources of
multispectral and hyperspectral imagery for calculating
rainforest biomass in Cameroon. In this case, IKONOS and the
other multispectral data sources were markedly less accurate
than hyperspectral Hyperion imagery.
In a Zambian study, IKONOS imagery has been processed to
define LAI and forest canopy roughness, used in a wider
experiment to monitor energy fluxes between vegetation and
327
the atmosphere (Scanlon and Albertson 2003). Elsewhere in
Zambia, Hansen ef al. (2002) describe the use of IKONOS
imagery to generate an accurate tree crown cover map, used to
validate a global percent tree cover data set, generated by the
Moderate Resolution Imaging Spectroradiometer (MODIS). In
fact, IKONOS imagery is conducted for MODIS validation
elsewhere in Africa, including Botswana (Morisette et al.
2003).
4. HABITAT MONITORING IN SOUTHERN AFRICA
Kruger National Park (KNP) represents a managed, semi-
natural environment. The park operates as a preserve for
endemic flora and fauna, and is a major visitor attraction. It is
believed that management practices throughout the twentieth
century have fundamentally altered vegetation distribution
throughout the park (Van Wilgen er al, 1998). In particular,
numerous artificial water resources were created to attract
wildlife to specific locations for viewing by tourists. The
increased water resources may have led to an overall increase in
vegetation abundance, and may be contributing to structural
homogenisation of the park's vegetation (Eckhardt et a/., 2000.
Related processes include infestation of alien plant species and
dramatic growth in certain animal populations, notably elephant
and rhinoceros (Figure 1).
N ® Black rhino
0 km 50 : :
White rhino
& Elephant
Figure 1. Recent growth in large herbivore distributions in
Kruger National Park.
There is a strong need for accurate ecological monitoring in
KNP to inform management practices, thereby maintaining
biodiversity. Remote sensing provides an excellent source of
data for such ecological investigation and, in fact, has long been