Geography Reference
In-Depth Information
1000
1.0
a)
b)
100
0.8
*
0.6
10
0.4
1
10
10
1
100
1000
1
100
1000
T P (yrs)
T P = T Q (yrs)
Figure 9.19. Mapping of design storm to flood peak return periods. (a) Return periods of design storms (T P ) vs. return period of associated flood
peaks (T Q ) with varying runoff coefficients (yellow: low, blue: high); (b) non-exceedance probability of the runoff coefficient producing the
maximum annual events providing the match of the design storm and flood return periods for varying climate (light: wet, dark: dry). From
Viglione et al.( 2009a ).
1998 ; Sivapalan et al., 2005 ). Iacobellis et al.( 2011 )
exploited a two-component derived distribution, based on
the concept of variable source area applied to two different
runoff thresholds related to infiltration excess and satur-
ation excess mechanisms. Their regional analysis showed
that based on the a-priori information provided by several
catchment characteristics related to basin climate, geology,
geomorphology and landcover it was possible to explain
more than 70% of the spatial variability of the distribution
parameters in southern Italy.
So far, despite the considerable interest in this approach
in the scientific literature, the impact of these derived flood
frequency methods on practical flood estimation, even in
gauged catchments, has been modest. The main problem is
the difficulty in quantifying the joint probabilities of the
various controls on the flood frequency curve, such as
rainfall duration, temporal patterns, multiple events, soil
moisture and routing characteristics. In ungauged catch-
ments, this problem is exacerbated by the fact that no
runoff data can be used to infer type and parameters of
these joint distributions. Potentially, however, regionalisa-
tion of parameter procedures could be used (described in
detail in Chapter 10 ) and the derived distribution method
could then be applied to ungauged catchments.
Simpler, but statistically less rigorous methods have
enjoyed some popularity where the derived distribution
approach is combined with flood regionalisation. This
avoids some of the joint probability problems. An example
is the Gradex method (Guillot, 1972 ; Duband et al., 1994 ;
Naghettini et al., 1996 ), which assumes that beyond a
threshold return period any additional rainfall produces a
corresponding increase in runoff without losses. The
method lends itself to predicting floods in ungauged
basins, where floods of a small return period are regional-
ised by statistical methods and extrapolated to large return
periods on the basis of regionalised precipitation (e.g.,
Merz et al., 1999 ).
Within this estimation framework, important informa-
tion can be exploited by remote sensing products, such as
digital elevation models, land cover and vegetation indices,
which can contribute to providing reliable predictions in
ungauged basins. Going further, one could apply the con-
tinuous models directly and use them to simulate continu-
ous runoff and from this extract annual maximum flood
peaks with which to construct the flood frequency curve.
This is discussed in the next section.
9.4.2 Continuous models
Continuous runoff models simulate runoff processes in a
time explicit manner. Runoff models are discussed in more
detail in Chapter 10 , and therefore this chapter will only
focus on specific issues that will be encountered in more
targeted applications of continuous models for
flood
estimation.
 
Search WWH ::




Custom Search