where the water body being looked at was not as
well defined, but this technique has the advan-
Upper left
Lower right
#cols
tage of being relatively simple and has results
corner
corner
#rows
that are easy to interpret.
Full scene
E
71563.5 m
E
152,218.5 m
7853
In developing this classification technique, a
N 1220712.0 m
N
1,020,015.0 m
7043
determination of which of the seven Landsat TM
Reservoir
E
23199.0 m
E
46,882.5 m
2460
spectral bands to use was required. The distinc-
subarea
N
1127631.0 m
N
1,072,369.5 m
1940
tion between water and other surface types varies
Pixel size = 28.5 m 28.5 m
among the bands, as can be seen in Figure 3,
which shows the east end of the reservoir on 7
March 1993. Band 5 (mid-infrared) was selected
however, with pixels that include part land and
as the basis for the water classification because
part water, along the shoreline of the reservoir. It
water is spectrally uniform and distinct from
is not appropriate to assign the whole area of any
land. Band 4 (near-infrared) was added to elimi-
such "mixed" pixel to either water or land; rather,
nate some confusion between vegetated areas
the area should be divided between water and
and shoreline areas, which look similar in Band 5
(dark, but not black), but which are distinct in
ter in the pixel. There are enough mixed pixels
Band 4 (shoreline is dark, vegetation is bright).
that it is not appropriate to ignore them. The
The visible bands (TM Bands 1, 2, 3) were not
water area classification technique described be-
used because they had significant spectral vari-
low (using TM bands 4 and 5), which is referred
ability within the water class, caused by varying
to as the "Band 5 threshold" technique, was de-
amounts of sediment load. TM Band 6 (thermal
veloped to accommodate the mixed pixels. Other
infrared) was not used because of poor spatial
techniques might work as well or indeed be more
resolution. Band 7 (mid-infrared) was not used,
appropriate for other situations, in particular
Figure 3. Landsat Thematic Mapper spectral bands.
6
Back to contents