the number of discharge classes; 2) the time base for discharge averaging, and; 3) the length of the period

of record.

The selection of class interval can influence the effective discharge calculation. Intuitively, it might

be expected that the smaller the class interval and, therefore, the greater the number of classes, the more

accurate would be the outcome. However, if too small an interval is used, discontinuities appear in the

discharge frequency distribution. These, in turn, produce an irregular sediment load histogram with multiple

peaks. Therefore, the selected class interval should be small enough to accurately represent the frequency

distribution of flows, but large enough to produce a continuous distribution, with no classes having a

frequency of zero.

There are no definite rules for selecting the most appropriate interval and number of classes, but

Yevjevich (1972) stated that the class interval should not be larger than s/4, where s is an estimate of the

standard deviation of the sample. For hydrological applications he suggested that the number of classes

should be between 10 and 25, depending on the sample size. Hey (1997) found that 25 classes with equal,

arithmetic intervals produced a relatively continuous flow frequency distribution and a smooth sediment load

histogram with a well-defined peak. This indicates an effective discharge which corresponded exactly with

bankfull flow. A smaller interval, and correspondingly larger number of classes, produced anomalous

results. Particular care must be exercised on rivers where there is a high incidence of very low flows. In

sand bed rivers, the low flows may be competent to transport the sediment. Under these circumstances,

the effective discharge may be biased towards the lowest discharge class, and caution must be exercised,

therefore, when using arithmetic class intervals.

Flows within the lowest discharge class are seldom normally distributed, being skewed towards

the lower class boundary. The effect is that calculation of the sediment load transported by that class of

flow, based on the median sediment transport rate, will overestimate the actual value. If this is a problem,

it may be necessary to subdivide the lowest class interval into smaller classes.

An alternative approach for determining the class interval is to use the U.S. Geological Survey

(USGS) flow duration procedure that divides the data into 35 classes. The lowest class is zero, with a class

width of 0 to 0. The next class width is 0 to the minimum discharge value. The remaining 33 classes are

determined by subtracting the logarithm of the minimum discharge from the logarithm of the maximum value,

and dividing by 33, to define equal, logarithmic class intervals. The use of log-scale class intervals is

attractive in that it divides the low discharges into more class intervals. This is useful because, on rivers

where the flow duration curve is strongly skewed due to a high incidence of low flows, it generates

approximately equal numbers of events in each. However, the discharge class intervals at the upper end

of the distribution can be extremely large, artificially biasing the value of the effective discharge towards

these flows. Caution must be exercised, therefore, when using logarithmic class intervals.

267

Integrated Publishing, Inc. |