Sediment Transport Concepts - Sediment Transport in Irrigation Canals: A New Approach - page 113

rough flow:

u ∗ k s
ν

≥

70 (5.41)

- according to the mentioned criteria, the flow for the selected data set
is defined as a transition flow; hence the velocity distribution will be
affected by viscosity as well as by the bed and side roughness. The de
Chézy's roughness coefficient for a transition flow is given by:

18 log

12 R
k s +

C = (5.42)

3 . 3 ν/u ∗

- next the average value of the Manning's roughness coefficient is com-
puted from the k s value from the de Chézy equation and is compared
with the derived n using different methods.

The comparison between the 6 methods is made on the basis of the
number of well-predicted values within an error band. If K is the error
factor then the error band is the range between the measured value/ K and
measured value * K . Once a predicted value lies within the above given
band then it will be seen as a well-predicted value. The accuracy of a
method is given by:

number of well predicted values
number of total values

accuracy (%) = ∗ 100% (5.43)

The accuracy of the predictability of the different methods is given
in Figure 5.12 for different error factors. The result shows that the
prediction method 6 gives the best results for the equivalent roughness in
a trapezoidal canal section. Hence, method 6 will be used for the evaluation
of the overall roughness of a canal section.
The comparison of the methods to predict the equivalent roughness
in a trapezoidal cross section with data from Krüger results in the
following conclusions:
• comparison of the overall performance of the methods with the exper-

imental data set shows that the methods can be ranked in descending
order, namely 6, 5, 2, 1, 4 and 3;

• methods 1, 2, 3 and 4 were developed for river conditions, in which

the channel consists of a main canal and two parallel flood plains.
The water depths in both main canal and flood plains do not vary in a
lateral direction;

• method 6 appears to give better results for error factors smaller than

1.2. This method predicts the measured data with an accuracy higher
than 90% for an error factor of 1.15. The minimum standard error of
the predicted values was also observed for method 6;

• methods 1 and 2 behave similarly. The assumptions for both methods

give similar results; they weigh the side parts in the same way as the