FLOOD01798

• County: Statewide
• Title: Australian Rainfall and Runoff 1998, Revision of Book VI - Estimation of Large to Extreme Floods
• Date: 11/28/1998
• Prepared By: Rory Nathan, Sinclair Knight Merz
• Floodplain - Doc Type: Educational/Technical/Reference Information

UKAt'1 U

I

the depth and areal extent corresponding to an extreme
event with an AEP of around 1 in 100, and the wind speed
and temperature sequences are selected to maximise
runoff. However, such approaches are not consistent with
the AEP-neutral approach, and thus careful consideration
needs to be given to the selection of inputs to ensure that
no probability bias is introduced into the transformation
between rainfall and runoff. The magnitude of snowmell
floods is particularly sensitive to initial snowpack conditions,
and accordingly it is likely that a joint probability approach
would be required to satisfy AEP-neutral requirements.

Nathan and Bowles (1997) provide one example of a
study in which an AEP-neutral Goint probabiiity) approach
was adopted for the derivation of snowmelt design floods.
They incorporated the Snow Compaction Procedure
(USBR, 1966) into a modified version of the RORB model.
This procedure uses a water budget approach which is
based on the concept of snow compaction and a threshold
density, where the maximum potential rate of snowmelt is
derived using the sub-daily application of the US Corps of
Engineers degree-day snowmelt equations (U.S. Army
Corps of Engineers, 1960)

I
I

I

I
I

I

5.5 Consideration of Long Duration Events

I
I
I

As discussed in Section 3.8, there are some design
situations in which it appears that the critical duration of
interest may be longer than the durations for which
generalised design rainfall information are avaiiable. The
longest available design storm durations generally relate to
the meteorological limits associated with single storm
events, and thus longer duration design events will involve
the consideration of storm sequences.

While it may be necessary to consider the likelihood of
storm sequences in tropical regions, it is reasonably clear
that long duration design events (one to several days) in
southeastern Australia are unlikely to be preceded by
significant antecedent rainfalls (Section 3.8). Accordingly,
the issue of storm sequences over extended periods may
be implicitly solved by undertaking a joint probability
analysis of inflow floods and reservoir volume, as described
in Section 5.2.

There are other design situations (such as tailings
dams) in which the design objective is to ensure that the
risk of spills from the storage is minimal. These Iypes of
problems can generally be handled by undertaking mass
balance calculations of all operational inflows and outflows
for very long hydroclimatic sequences. It is usually not
necessary to use a hydrograph model to route the rainfall
excess as the surface area of the storage may be large
compared to the contributing catchment area; it thus may
be sufficient to allow for a freeboard in the storage that fully
accommodates the volume of runoff corresponding to the
required AEP of rainfall. This type of problem does not lend
itself to event-based joint probability analyses but requires
water balance computations over extended periods.
Generally it desirable to generate the long hydroclimatic
sequences by stochastic data generation techniques (see
McMahon and Mein, 1986), and an example of this
approach used for spillway design is provided by Wark,
1982. The required security against overtopping can be
achieved by using sequences of different lengll1s, as
described for example in Grayson et al. (1996, Section 5.2).

One of the major practical and theoretical problems with
the application of stochastic data generation techniques -
particularly when used in the assessment of the Rare to
Extl'E!me risks --is the t:haraclel'i$Blion of slatistical
extremes. This difficulty relates both to the tail of the
distribution, as well as to the definition of the correlation
between the stochastic inputs over a range of event

I
I

I

I
I

I
I

I

I

DOOI\ VI. t:::'UllldUVII VI Loclfge (0 l:.XUefm:~ rluuu:::;

magnitudes. These issues require careful consideration
and specialist knowledge, and should only be undertaken
by practitioners with relevant experience.

5.6 Treatment of Uncertainty

5.6,1 General

(a) Consequences of uncertainty

Uncertainties in the estimation of extreme floods have
important economic and social consequences, and thus
recognition of the impacts of uncertainty should be
incorporated into advice given to management and political
decision makers. If there are significant differences in
outcome within the range of uncertainty then the likely
range of consequences should be explicitly considered
when developing mitigation strategies and advice. An
under-estimate of the flood magnitude will lead to the
infrastructure being under-designed, thus potentially
resulting in increased flood damage costs and possible loss
of life. Conversely, an over-estimate of the flood magnitude
will lead to extra costs from the over-design of the
infrastructure.

(b) Sources and types of uncertainty

The uncertainty or flood estimation error increases with
increasing size of flood (or reducing AEP) and the relative
impacts of different sources of uncertainty also change with
flood magnitude. One can distinguish between the following
components contributing to the total flood estimation
uncertainty (Nathan and Weinmann, 1995):

. observation/measurement errors (data errors);

. errors resulting from inferences made using samples of
limited size in space and time;

. non-homogeneity in flood data sets;

. model uncertainty (in selection of meteorological,
hydrological or statistical models); and,

. parameter uncertainty.

Together these components define a total band of
uncertainty around the design flood estimate obtained from
a given data set by a particular methodology. The true flood
magnitude could lie anywhere within this band. The band of
uncertainty associated with flood estimates derived using
current best practice and the best available data could be
termed 'absolute uncertainty'.

The subjective elements involved in flood estimation
lead to differences in flood estimates obtained by different
hydrologists. This form of uncertainty could be termed
reletive (or methodological) uncettainty. There are several
components contributing to relative uncertainty: lack of
clear definitions and guidelines, different levels of
hydrologic expertise, and other factors resulting in
differences in methodology applied and in judgements
made by different investigators. Relative uncertainty leads
to a wider band of errors around the band of absolute
uncertainly.

(c) Scope for reducing uncertainties

The above categorisation into different components of
uncertainty implies differences in our abilily to both reduce
and quantify the uncertainties. In the short to medium term,
strategies for reducing absolute uncertainty are most likely
to have the most impact on the accuracy of Large and Rare
flooas, for it is in this renge of probabilities that there is the
most chance that empirical evidence or theoretical
knowledge may be discovered. A pertinent example of this
is development of the CRC-FORGE technique, as