It is also assumed, for the purposes of making
To estimate the effect of the classification
a quantitative estimate of the uncertainty due to
threshold uncertainty on the storage capacity es-
wind setup, that
timates, the water surface area was recomputed
The reservoir is an east-west rectangle with
for each Landsat scene, moving both the water/
the sides parallel and perpendicular to the
shoreline and the shoreline/land thresholds up
direction of the wind, with the dam at the
one class to get an upper water surface area
west end of the rectangle.
bound, and down one class for a lower bound.
The amount of water surface elevation dif-
The storage capacity figures were then computed
ference caused by wind is proportional to
assuming that all the scenes had the upper bound
the fetch.
of surface area, and then the lower bound. It is
The fetch is proportional to the elevation
unlikely that all the scenes would have an error in
(above empty) of the reservoir.
the same direction, so the figures derived can rea-
The maximum wind setup for a gale force
sonably be considered outside limits.
wind on the full reservoir is 1-ft elevation.
The uncertainty in storage estimates at spill-
With these assumptions, it is possible to
way elevation (661 ft) caused by classification
estimate the possible surface elevation difference
threshold uncertainty, as computed above, is
caused by wind setup for each Landsat scene/
+15,000 and 16,000 acre-ft.
reservoir elevation, and the corresponding new
surface area value (for the tilted surface). In order
Masking (#5)
to compute this, the difference in area for tilted
Part of the procedure was masking the upper
surfaces vs. level surface was estimated, using the
end of the reservoir to define where the reservoir
1985 area table, because it has more detailed
pool stopped and the flowing water started. This
elevation data than the Landsat scenes. This 1985
was done by a visual assessment of each Landsat
area difference was then applied to the 1993 area
scene. To estimate the effect of uncertainty in this
figures to approximate what the effect on area
procedure, a reasonable maximum and minimum
would be for the 1993 scenes in the presence of
mask for each scene was created and the surface
wind setup.
area and storage capacity values were recom-
As a single number approximation of the max-
puted using the revised mask.
imum uncertainty caused by wind setup, it was
The uncertainty in storage capacity at spillway
assumed that the wind was blowing at half gale
elevation caused by uncertainties in masking the
force at the time of all the Landsat passes either
upper end of the reservoir was +5,000 and 8,000
west or east. If the wind was blowing west, the
acre-ft.
Landsat area estimates (and hence capacity esti-
mates) would be too high and should be correct-
Uncertainty in reservoir elevation readings (#6)
ed downward, and if east, then corrected up-
A possible source of uncertainty is in the reser-
ward.
voir elevation readings. Storage capacity is com-
At spillway elevation (661 ft), the uncertainty
puted by multiplying average surface area (acres)
in storage capacity caused by wind setup, as ap-
by the difference in elevation between any two of
proximated above, is plus or minus about 8000
the elevations in the elevation vs. area curve. An
acre-ft.
error in the elevation reading could mean an
error in the elevation difference, and thus in the
capacity estimate. The effect is significant only for
Classification threshold (#4)
In calculation of water surface area, the 21 TM
the elevation of the top of the reservoir, because
Band 5 brightness classes were split into water,
any overestimate in depth on a lower capacity
shoreline, and land categories based on a visual
layer will be approximately balanced by an
assessment of the classification images. Although
underestimate in the volume of the layer above it.
best judgment was used in these decisions, there
Only the top layer has no layer above it.
is some uncertainty in the selection of these
To ascertain the effect of errors in depth read-
classification thresholds. There is also variation in
ings, the reservoir elevation vs. capacity curve
the reflectance of the land from one part of the
was recomputed adding 0.1 ft to the elevation of
reservoir to another (bright white sediment vs.
the reservoir at the time of the Landsat over-
darker mudflats and medium soil), which means
passes. The difference in capacity was minimal
that ideally different thresholds should be chosen
(less than 450 acre-ft) for the intermediate eleva-
in different parts of the reservoir.
tions, and varied from plus or minus 805 acre-ft
46
Back to contents
Previous Page
