FLOOD01888

• County: Statewide
• Title: River Hydraulics
• Date: 10/15/1993
• Prepared By: US Army Corps of Engineers
• Floodplain - Doc Type: Educational/Technical/Reference Information

EM 1110-2-1416
15 Oct 93

are influenced by backwater. At these stations the rating
curve is modified not only by the shift but also by a
slope correction which is computed from the observed
fall to a downstream gage. Discharge records at a slope
station are seldom very good and should be used as
boundary conditions with caution.

(3) There are gaging stations whose records are not
very reliable. These are usually on streams with a flat
bed slope or a mobile boundary. At these locations, only
the actual flow measurements can be used with
confidence.

5-9. Calibration and Verification

When a model is calibrated, the parameters which control
the model's performance, primarily Manning's n and
reach storage, are determined. The key to a successful
calibration is to identify the troe values of the parameters
which control the system and not to use values that com-
pensate for shortcomings in the geometry and/or the
boundary conditions. Because unsteady flow models
reproduce the entire range of flows, they should be cali-
brated to reproduce both low and high flows.

a. Manning's n. In the unsteady flow models used
in the United States, the friction slope is generally mod-
eled using Manning's equation. Manning's n value
relates the roughoeSS of the stream boundary 10 the fric-
tion force exerted on the system. For most problems, an
initial estimate of Manning's n (it is only an educated
guess) is used at the start of the calibration. The initial
values are then adjusted to match observed stage data.
When no observed stage data exists, the estimated values
take on a greater importance since they are assumed 10
be representative of the system. See Appendix D for
detailed information on selecting n values.

b. Calibration. Calibratinn of an unsteady flow
model is a four step process. In the first step the n val-
ues are adjusted to reproduce the maximum stages of an
event. The storage in the cross sections. is then adjusted,
if necessary, 10 improve timing and attenuation. In the
third step, the flow versus Manning's n relationship is
adjusted to reproduce both high and low flow event
stages. Finally, the model is fme tuned to reproduce a
longer period which should include the initial calibration
event.

(1) The initial calibration event should be one of the
larger events which are available in the time series. The

5-16

purpose of this pbase is to adjust the initial n values 10
match the crest of the event at all stations in the model.
Figure 5-12 shows the hydrographs for the Illinois River
at Havana after the initial calibration. Note that,
although the crest stage is approximately correct, the
timing of the hydrograph and the reproduction of low
flow are defICient.

e

(2) Total storage as defined by river cross sectinns
is almost always deficient. In natural rivers, the timing
of the hydrograph is determined by storage and the
dynamics of the flood wave. Timing can be adjusted by
modifying storage, friction, and distributinn of lateral
inflows. If the timing cannot be calibrated by reasonable
adjustment of these factors, then there is some other
problem, most likely an error in the cross sections. For
the Illinois River, which is confined by levees in the
reach near Havana, an increase in overbank storage of
about 20 percent yields the results shown in Figure 5-13;
an increase in storage of about 40 percent yields those
shown in Figure 5-14. Both changes are only minor
increases in storage area because the overbanks are con-
fined by levees.

.

..

.

...

(3) By varying Manning's n with flow the reproduc-
tinn of stage is improved; see Figure 5-15. The model
still. does not reproduce the initial time steps, but the
disagreement is probably caused by the initial conditions.

e

(4) The final calibration consists of fine tuning the
flow-roughness relation and the adjustments in storage.
The event selected should be an extension of the event
chosen for the initial calibration. For the Dlinois River
example, the fmal calibration was performed for the
period from 15 Nov 1982 to 15 Sep 1983. The event
includes high flow and low flow and a second major
flood in May 1983. Figure 5-16 shows the reproduction
of stage at Havana during the period. The model param-
eters required only slight adjustment to better simulate
low flow.

.

c. Verification. The calibrated model should be
verified against two or more periods which include sig-
nificant events. The periods should be long, approaching
one year, so that seasonal effects can be detected.
Figure 5-17 shows the reproduction of the 1974 observed
data on the Illinois River.

.
.

e