Although IKONOS multispectral imagery does not
appear to be cost-effective for coarse-level habitat mapping,
would the data be effective for other applications? The
analyses presented here assessed the ability of remotely
sensed data to predict the correct habitat type (e.g., for
baseline habitat mapping). However, while IKONOS data
cannot identify many habitats spectrally, the boundaries of
habitat patches are mapped with much greater accuracy than
with other satellite sensors. Furthermore, colour composites
of IKONOS imagery compare not unfavourably with colour
aerial photography for visual interpretation. If the identity of
habitat patches was determined independently (e.g., by field
survey), IKONOS data would allow relatively small
changes in patch location and boundary to be measured.
For example, if IKONOS 1 m panchromatic data were used
in conjunction with multispectral data, and geometric registration
errors were approximately 1– 2 pixels, changes of
several metres could be monitored (note that such imagery
would cost US$53 km 2
). Such modest changes in habitat
area or location could not be detected with either SPOT or
Landsat TM/ETM+ which would probably only detect
changes of 40 – 60 m at best. Thus, for small areas (much
less than 500 km2
) in which habitat patches were identified
independently, IKONOS would be a more cost-effective
source of high-resolution information than CASI (compare
costs but not accuracies in Fig. 6).