available for testing during the project, was not geomet
rically corrected. Figure 2 is a reduced computer output
from the identification system which was used to verify the
lake identification for the eastern half of the area proc
essed. Figure 2 shows the limits of the area processed and
the quadrangle specified for lake information retrieval.
The lakes and boundaries are skewed due to lack of geometric
correction of this test area. Figure 3 is a 31 July 1977
natural color aerial photograph over the same area and lakes
in Figure 2. The capability for computer identification
of surface water is illustrated by comparing Figures 2 and
3. Although the data being compared were collected 2 years
apart, and surface water conditions could differ, the simi
larity between water bodies illustrated is manifest.
Identification System
Lake Identification (Program LAKEID). A fully processed
computer compatible tape in the new EDIPS (EROS Data Center
Digital Image Processing System) format was used in the
processing. The data contained no ground control points;
therefore, no geometric correction has been applied to the
scene. Each sample or pixel making up a scene is 57 x 57 m
and has an intensity stored for each of the 4 bands. The
system has potential for recognizing and identifying lakes
as small as 0.5 ha. Landsat multispectral scanner resolu
tion is approximately 80 m. Water body identification was
performed by thresholding at a value of 9 in band 7 (0.8
to 1.1 jam) . All intensities < 9 are water and those > 9
are land. Both the literature (Work and Gilmer 1976) and
some simple tests indicate that a threshold of 9 is a good
reflectance value for discriminating between land and water
in band 7. The EROS Data Center's General Electric IMAGE
100 System was used to vary this threshold to observe changes
in water area discriminated for Alaskan arctic lakes. No
change in areas was observed when the reflectance
value was changed + 1, while less than 2% in area occurred
with a change of ± 3 in reflectance values above and below
the 9 threshold. Figure 2 illustrates the shape and size of
lakes defined using the 9 threshold value. The shape and
size of each discrete basin are defined by like numbered
sample elements on this printout. Each sample that is
identified as land is portrayed as a zero. The printout has
been reduced to such an extent that the characters are
illegible; therefore, the lake and pond shorelines contained
within the quadrangle have been blackened by hand to better
define each basin for this illustration.
Utilizing the 9 threshold value, 1,426 separate water bodies
were computer identified that were wholly contained within
the area processed. The Meade River in the northwest corner
was not continuous which provided some additional enumerated
water bodies. Ponds smaller than 0.5 ha may not be identi
fied and water bodies with less than 80 m separation are
merged; thus, the number of computer-identified water bodies
within this scene is probably less than what actually
exists. This area represents 4.3% of the entire Landsat
scene (Figure 1). Approximately 16% of the scene in the
upper half consists of very large water bodies, including
the Arctic Ocean with embayments and Teshekpuk Lake. The
remaining 84% of the scene is the Arctic Coastal Plain,
452