Geography Reference
In-Depth Information
Figure 40.4
Spectral
signatures.
have information about the intensity of reflected
radiation back from the Earth's surface for each
of the spectral bands recorded by the sensor. This
way of storing information is known as the
raster
format.
Different land cover types vary in their ability
to reflect the Sun's incident radiation in different
parts of the spectrum. Living vegetation, for
example, reflects strongly in the green part of the
spectrum and in the near infrared. Dry bare soil,
by contrast, does not show the peak in the green
or near infrared. We can therefore use the
information recorded on the spectral bands of TM
or SPOT, for example, to identify different land
cover types according to their
spectral signatures
(Figure 40.4).
The simplest way of using TM or SPOT data
to make a classification of land cover is to
interpret the images visually. This was the method
used within the CORINE programme, which
made use of SPOT data to classify land cover in
various European countries (e.g. Cruickshank
and Tomlinson 1996; EEA 1998; and see
CORINE web site). In the case of the Land
Cover Map of Great Britain (LCMGB), by
contrast, automated, computer-based routines
were used to classify TM images (Fuller
et al
. 1994
a, b). The computer is 'trained' using areas whose
cover type is known. The routines then find all
the other areas on the image with similar spectral
characteristics. In order to improve classification
accuracy, a
multi-temporal
approach to
classification was used that combined winter and
summer images. The spectral signatures of
vegetated surfaces will change in different ways
over the course of the growing season. For the
production of the LCMGB, this property could
be used to improve the ability of the classification
routines used to distinguish between different
cover types. Further general information on the
methods of image processing can be found in
Mather (1999).
Maps of land cover such as that described above
can be used as baseline data against which change
can be measured. CORINE data, for example,
have been widely used in Europe to map land
cover change and its environmental consequences
(see CORINE web site). The LCMGB formed
part of Countryside Survey 1990 (Barr
et al
. 1993),
which was itself a major baseline survey of the state
