1. Rectify the data spatially, i.e., orient the pix-
ber of reasons. The 1993 data give a consistent pic-
el rows and columns to a known geographic
ture of the reservoir at that time; by 1995 the reser-
coordinate system such as latitude/longi-
voir could have changed. The 1993 data also had
tude, Universal Transverse Mercator
less cloud cover than the 1995 data, and depict a
(UTM), or Albers, and specify pixel size.
greater range of reservoir elevations. Finally,
2. Classify the data. Each pixel is assigned to a
ground survey profile data available were from
surface cover-type based on its spectral
1993, so it is more valid to compare them with
characteristics. For instance, those pixels
Landsat data from 1993 rather than 1995.
that are dark in the visible portion of the
Rectification to Albers projection
spectrum and bright in the near-infrared
would be assigned to a vegetation surface
USGS 1:100,000-scale DLG data were used to
type, while those dark in both visible and
rectify the 7 March 1993 Landsat scene. The Albers
near-infrared would be assigned to water.
conical equal area projection was used for the out-
3. Verify accuracy. Compare classification re-
put projection of the rectified image because it is
sults with reference information (ground
the same as that used in the Los Angeles District's
truth) for verification.
implementation of the CorpsView software. It is
The Landsat data were processed on a Sun
an equal area projection, which makes each pixel
SparcStation 20 UNIX workstation. Software
the same size, thus giving more accurate surface
used was ERDAS IMAGINE version 8.3, and Re-
area estimates. Table 4 lists the Albers projection
search Systems' ENVI, version 2.6.
parameters.
In this project, surface area was determined by
Table 4. Albers projection parameters for Painted
classifying the Landsat data; storage capacity was
Rock Reservoir.
computed from the surface area values; sediment
depth estimates were derived from the updated
Projection
Albers conical equal area
area and capacity values; sources of error in the
Spheroid
Clarke 1866
Datum
NAD27
area and capacity estimates were identified and
Units
meters
quantified; and an estimate was made of what the
elevation vs. area and elevation vs. storage capac-
Albers projection parameters:
1st standard parallel
32 36 00 N
ity curves would have looked like if 500,000 acre-
2nd standard parallel
38 00 00 N
ft of storage capacity had been lost.
Central meridian
113 00 00 W
Although the goal of the project is to estimate
Latitude of projection's origin
23 00 00 N
the capacity of the reservoir, it is noted that the
False easting (meters)
0
Landsat data contribute only area estimates to
False northing (meters)
0
this process, and only for selected reservoir eleva-
Once the 7 March 1993 scene was rectified, it
tions. Different approaches can be taken to derive
was used to rectify the remaining scenes. From 28
the complete area and capacity curves from the
to 73 ground control points were selected in each
Landsat area estimates.
scene, using a second-order polynomial warp,
with a root-mean-square (rms) error ranging be-
Surface area procedure
tween 15.3 m and 46.7 m. The cubic convolution
The procedure to estimate the surface area of
resampling technique was used; output pixel size
the reservoir was to select the Landsat scenes, rec-
was 28.5 m 28.5 m, which corresponds to 0.2007
tify each one to the Albers conical equal area
acres. Each full scene was rectified, and then a
projection and subset the image, classify for wa-
subarea including the reservoir and some area
ter, mask the classification results by hand, and
around it was extracted. Table 5 lists the Albers
tally acreage. A multitemporal water classifica-
coordinates of the rectified full scene and subset
tion map was created from the individual
areas. These areas were the same for all 19 Land-
classification maps, and was compared with the
sat scenes.
ground survey profile points for verification.
Water classification
Landsat scene selection
As mentioned above, water and land are spec-
Landsat scenes from both 1993 and 1995 were
trally distinct, making it relatively straightfor-
used for the capacity estimate in the preliminary
ward to separate water from land with a number
report (Pangburn et al. 1998), but for this final re-
of classification techniques. A challenge arises,
port, only scenes from 1993 were used, for a num-
5
Back to contents
Previous Page
