Prediction of low flows in ungauged basins - Runoff Prediction in Ungauged Basins: Synthesis across Processes, Places and Scales - page 173

Figure 8.8. Average number of days
per year without runoff (a) and
average length of the no-flow period
(b) plotted versus catchment area for
streams in the Lake Eyre basin,
Australia. The trend is mainly due
to catchment storage on the surface.
For photos of two catchments see
Figure 8.9 . From Knighton and
Nanson ( 2001 ).

1000

a)

500

Boulia

Camooweal

Roxborough

200

Birdsville
Currareva
Callamurra

Nappa Merrie

100

500

b)

Camooweal

200

Roxborough

Boulia

100

Nappa Merrie

Birdsville

Currareva

Callamurra

50

1000 10000 100000 500000

Drainage area (km 2 )

their approach was that they interpreted the estimates for
each ungauged basin hydrologically in the regional context
of the gauged basins.
In arid places, Q 95 or Q 7,10 may no longer be a mean-
ingful low flow index if river flow ceases for an extended
period of time during the year. Alternative indices are the
duration and the frequency of dry spells (or no flow). An
example of this is shown in Figure 8.8 from the Lake Eyre
basin, Australia where the average number of days per year
without runoff (top) and average length of the no-flow
period (bottom) are plotted versus catchment area. As the
catchment area increases, the days without runoff become
less frequent, and the periods of no-flow become shorter.
The Lake Eyre basin is arid, so the channel transmission
losses can be substantial and runoff volumes tend to
decrease downstream. There is also a rainfall gradient with
annual rainfall varying from 600 mm/yr in the upstream
areas to 200 mm/yr in the downstream areas. However, the
inverse relationship between catchment area and the no-
flow period mostly reflects the combination of increasing
area of the contributing catchment (i.e., surface storage)
and flow attenuation over the very long flow paths in the
Lake Eyre Basin catchments. Much of the flow is gener-
ated in the upper to middle reaches of the Lake Eyre Basin.
Large rivers and the flows become increasingly attenuated

as they flow downstream over very long and low gradients.
So the lower number of no-flow periods in the downstream
(large catchment areas) reaches are a reflection of channel

-
floodplain storage and very low gradients. After the last
tributary junction there is typically little inflow from the
arid surrounding catchment and the rivers experience large
transmission losses (McMahon et al., 2008 ). The photos in
Figure 8.9 illustrate the low gradient landscape. Figure 2.1
in Chapter 2 shows a satellite image of part of the catch-
ment highlighting the complex nature of the flow paths.
There are a number of concerns with the application of
regional regression models. If data are far from the regres-
sion line (i.e., outliers) regression coefficients can be
inappropriate for the remainder of the catchments. Lever-
age statistics have been developed to identify data points
that have an inordinate influence on the model (Rousseeuw
and van Zomeren, 1990 ). It is important to check whether
outliers are due to data problems or due to real unusual
behaviour of catchments. Geology can be highly heteroge-
neous over short distances and this can be reflected in
particularly high or low numbers for low flow values.
Typically, the analyst examines the unusual catchments
in more detail by exploring the hydrological processes,
e.g., through examination of hydrogeological data and
maps and, ideally, visiting the catchment. Another possible