WATER & SOIL - These are not the droids you are looking for.

6 Summary
In New Zealand little progress has been made in flood
estimation techniques over the last 25 years despite an upsurge
in the amount of streamflow data that has been col-
Iected over this period. This study has attempted to improve
this situation by
al flood frequency
analysis procedu
the available
annual and historical flo
I catchments'
The procedure, known as the Regional Flood Estimation
(RFË) method, is applicable to both gauged and ungauged
iural catchments which in general are greater than 20 km'
in area. Since the method was developed by averaging the
sampling variation that exists in individual flood records, it
should provide a more reliable design flood peak estimate
than that determined by fitting a frequency curve to a relatively
short record.
The RFE method comprises a set of eight regional flood
frequency cu
vs T, and a set of
niné regionat
when there is little
or no flood ¡
S the T-Year flood,
and Q is the mean annual flood. The most important independent
variables in the equations are catchment area
and an index of the catchment rainfall.
The regional curves may be used up to the 200-year return
period to estimate a design flood peak, except in the
Otago-Southland region where the upper limit on return
period is restricted to 100 years because of the limited data
in this area that were available for analysis. The curves are
defined by the straight-line extreme value type I (EVl) distribution
for all but two of the regions the Bay of Plenty
-
and South Canterbury regions, where the extreme value
type 2 (EV2) tlistribution was found to give a better definitión
of the regional trend in the data. Although the general
extreme value (C
gional curves, th
of the log-Pears
tion tests carried
that the LP3 distribution may well have given an equally
good description of the curves.
It is evident, both from the regional mass probability
plots and from the standard error equations derived for the
iegional curves, that the variability in the regional Qr/Q
data is well within acceptable limits. A quantitative indication
of the confidence that may be placed on values of
Q1/Q estimated fiom a regional curve is obtainable from
the standard error equations which give estimates comparing
very favourably with those given by the equivalent
NERC (1975) equation.
A feature of this study is the dependence of the results on
climate. This is illustrated by the regions, which are partially
consistent with recognised climatic boundaries, and
by the difference in slope of the western and eastern regional
curves. The latter curves have greater slopes, which
ãre uttributable to the greater variability in the flood peak
data for the eastern regions where the climate is drier and
the antecedent conditions more variable. Further indication
of the climatic influence is given by the regional equations
for estimating Q . ¡.lo physical characteristics, other than
catchment area, are included in the equations, the only
other important parameters being catchment rainfall estimates.
This suggests that climate may be the dominant factor
affecting flood peaks with magnitude equal to or
greater than the mean annual flood. Other factors often
considered important, such as geology and topography,
have been accounted for to some extent in the regionalisation
of the country.
The country is divided into two sets of regions' one set
for estimating Qr/Q and one for estimating Q. These are
very similar
t
tempts were
e
purposes, b
t
regions is a
a
Q1/Q and Q.
The application_of the method to a catchment for the
estimation of Q1/Q and
ted
in Chapter 5 with four
the
advantage, when only a
of
combining the estimates
uation
and the flood record to obtain a weighted average
"best" estimate of Q . t¡e precision of each equation is expressed
of record and,
àepend
estimate of Q
from a
he error of estimating
od record. Thus
when an important waterway project is being considered, a
recorder should be installed as soon as possible to record
the flood peaks.
In addition to the regional curves of Q1/Q which extend
up to a maximum of 200 years, generalised curves, one for
the west and one for the east, are given. These curves were
derived from all the flood peak data collected for this
study, excluding four extreme flood events, and they can be
applied from beyond the limit of the regional curves up to
the 1000-year return period. Of interest is the marked similarity
of these curves with those derived by Stevens and
ing too many regions. lt is envisaged that as more flood
p.ãk dut" become available, revisions and refinements will
te made to the RFE method' especially for the estimation
of Q.
In all cases, we recommend that other methods for estimating
design flood magnitude also should be used and the
results compared before a final figure is selected'
Water & soil technical publication no. 20 (1982)
83