To Top
of Page
There are two ways to use computers for
predicting reservoir inflow from rainfall. Many engineering firms use generalized computer programs, such as the U. S. Corps of Engineers’ HEC-1 Program.
Pre-programming of the computer makes it possible for an engineer or technician to
estimate stream flow in multiple situations by describing to the computer conditions, such
as elevation, distance, grade, area, topography, vegetation and so on. Once this has been
done, the computer is able to tell the difference between what happens in flat areas, such
as the Delta, areas with steeper grade such as Jackson and mountainous areas such as the
foothills of Tennessee. Even though programs such as these can be modified to reflect
actual stream data, generalized computer programs such as these are not always the most
accurate. This is because, as a basis, programs such as these tend to remain less specific
than some jobs require. The advantage provided by generalized computer programs is that
they allow a technician to provide results without the need to perform functions that
control flow. Functions are done for him by the computer.
Our reservoir has thirty seven years of hour by hour record showing rainfall and
reservoir inflow. This provides information with which to create a customized computer program for the Barnett Reservoir alone. A specific program such as this does not
need to compute the absorption ratios of the topography or the acceleration of flow
velocity with increasing watershed volume and grade. The Barnett Reservoir programmer can
simply know these things from thousands of direct observations. With the vast amount of
record available, a program such as this can be refined and re-refined as long as the
programmer is willing to continue applying and comparing his results to this extensive
base of data. The inflow simulations shown on the following graphs detail a degree of
consistency and accuracy which can, at this time, assure us that additional flood relief
is to be expected from the reservoir. We will continue to work with the minor flood record
in order to improve the program.
To Top
of Page
To Top
of Page
The graph below shows what would happen with the
two-lakes in place (the unconstrained procedure). Actual discharge in 1979 (the black
line) was as high as 134,000 cfs for a period of 12 hours. Flood crests, however, were not
controlled by this peak discharge. It did not last long enough. The crest was determined
by the average discharge over 48 hours. This was 130,000 cfs.
The discharge procedure takes advantage of this principal to move a
large amount of water out of the reservoir during early stages. This allows a large volume
of water to move downstream, ahead and away from the damaging floods that will occur 3
days later. This is what is happening on the 12th and 13th of April
as shown by the green and orange lines representing discharge on the graph opposite.
Remember that with the two-lake plan in place, this amount of discharge will not flood
structures in Jackson until the floodplain south of Jackson has filled up. Since this fill
up must occur concurrently with a large amount of volume passing thru (to flood plains
farther south) the lake discharge rates in the early stage of the unconstrained discharge
procedure are permissible.
Following this procedure, peak flooding will occur on April 17th at near
steady state levels for a discharge rate of 109,000 cfs. This would effect flood levels
south of Jackson that would be 1.3 feet lower than would have occurred at 1979 discharge
rates. Because of the loss of capacity in the north Jackson floodplain, the net savings would be 0.9 feet downstream. With provision for error, 0.7 feet less flooding
could be depended upon.
To Top
of Page
The graph below shows what would happen (following the constrained
discharge procedure and inflow simulation) when the inflow occurs six hours sooner than
expected. The orange lines show what was supposed to occur and the green lines show what
would actually occur.
This graph below shows what would happen (following the constrained
discharge procedure and inflow simulation) when the inflow occurs six hours later than
expected. In both cases where timing was in error, little effect occurred on the outcome.
To Top
of Page
The graph below shows what would happen following
the constrained discharge procedure and inflow simulation when 5% less inflow was expected
than actually occurred. (Inflow was underestimated). The green line overlying the orange
line shows that the procedure called for discharging 102,000 cfs after the prerelease
(April 15, 1979). The orange line (at top) shows that the reservoir was supposed to fill
to 299.7 feet. However, because the volume of inflow was underestimated, the reservoir
overfilled (green line).
In actual practice, the error would become obvious when significant tributary flow data
becomes available during the late hours of April 14, 1979. Reservoir inflow data during
the early hours of April 15, 1979 will also provide data which will result in a correction
in discharge to approximately 114,000 cfs. This amount (approximately 6,000 cfs more than
would have occurred with perfect data) will mean that the procedure will take one foot off
of a flood similar to the 1979 flood (not 1.3 feet which could have been obtained with
perfect simulation).
The graph below shows what would happen (following
the constrained discharge procedure and inflow simulation) when 5% more inflow is
simulated than actually occurs. (Reservoir inflow is overestimated). Because of this
error, the procedure called for discharging 117,000 cfs. Following this procedure instead
of the ideal 109,000 cfs, the reservoir will only fill to 298.0 feet. Three bcf of
reservoir capacity will be wasted. This error will become obvious when significant stream
flow data becomes available in the late hours of April 14, 1979 and when peak inflow
occurs in the early hours of April 15, 1979. The reservoir would then be cut back to
104,000 cfs.
Having discharged at a rate of 117,000 cfs for approximately 24 hours
into the north Jackson floodplain (which has restricted outflow due to levees and bridges)
will result in a loss of flood benefits. Once again, about one foot less flooding would
occur in Jackson as a result of using the procedure where 1.3 feet would result from
perfect simulation.
With the two-lakes in place, an error in discharge lasting 24 hours would be unable to
fill up the larger, more open-ended Richland floodplain to a level which would be higher
than would occur anyway, later, as a result of sustained flow at a lessor rate. In other
words, the two-lake situation is more forgiving of error.
To Top
of Page
The graph below shows that under constrained
conditions with good simulation, average discharge of 82,000 cfs can be reduced to 68,000
cfs. This would result in approximately 1.25 feet less flooding.
The following graph shows that, with the two-lakes in place (affording unconstrained
discharge ability) the 1983 flood can be reduced from 85,000 cfs (average 82,000 48 hours)
to 66,000 cfs. This would result in 1.5 feet less flooding than occurred in 1983.
To Top
of Page
The graph below shows that, at the present time, with prerelease
constrained by flooding conditions downstream in Jackson, almost the same results can be
obtained as would occur with the two-lakes in place. This is providing that simulation can
be perfected to consistently produce best results.
The two following graphs show that, under the present constrained
discharge conditions, six hour errors in timing result in about 69,000 cfs discharge. This
represents an improvement of 13,000 cfs from the 82,000 cfs 48 hour average discharge that
occurred in 1983. Nevertheless, with these errors, one foot less flooding would have
occurred using the discharge procedures.
To Top
of Page
The graph below shows what would have happened (following the
constrained discharge procedure and inflow simulation) when 5% less inflow was expected
than actually occurred. (Inflow was underestimated). Following the discharge procedure,
65,000 cfs was discharged. The reservoir (which was supposed to fill, according to the
upper orange line, overfilled (green line). Tributary stream flow measurements (which
would provide significant data in the late hours of May 21, 1983 and reservoir peak inflow
which would occur in the early hours of May 23, 1983) would make it possible to correct
discharge by increasing to approximately 71,000 cfs during the last one day of the peak
inflow period which would prevent overfilling the reservoir. With this error, the
reservoir would still be able to reduce downstream flooding approximately one foot.
The graph below shows what would have happened
(following the constrained discharge procedure and inflow simulation) when 5% more volume
was expected than actually occurred (inflow was overestimated). The prerelease procedure
lowers the reservoir to 292.2 feet. However, because inflow has been overestimated, the
peak discharge is set at 75,000 cfs. This is 10,000 cfs higher than will actually be
necessary. In this instance fill up of the reservoir (the upper green line) does not
occur. Tributary stream flow (which would begin providing useful data during the late
hours of May 22, 1983) and reservoir inflow data (occurring around noon on May 23, 1983)
would enable the reservoir to cut back on discharge to a rate of approximately 62,000 cfs.
This will utilize the fill up capacity of the reservoir and prevent the fill up of the
north Jackson floodplain to steady state levels for the previous 75,000 cfs discharge
rate. Downstream flood level benefit will be 0.3 feet for the 75,000 cfs rate and a
cumulative benefit of 0.9 feet for the 82,000 cfs average discharge rate that actually
occurred in 1983.
If this overestimation of reservoir inflow occurred with the two-lake
plan in place, the 10,000 cfs discharge error would not affect the north Jackson
floodplain. Its effect upon the Richland floodplain (which is more open ended and
unaffected by levees and bridges) would be much less. This would result in a benefit of
approximately 1.1 feet (net 0.7 feet due to capacity loss) in communities south of
Jackson.
To Top
of Page |