Geoscience Reference
In-Depth Information
8.6.2 The Runoff Generation Model Component
Given a series of rainstorms, the next stage is to model how much of that rainfall becomes streamflow.
Eagleson did this by estimating effective rainfalls using a simple
-index model (see Section 2.2), with a
constant value of
, to allow an analytical solution to the derived distribution of flood peaks. His model
did also recognise that runoff might be generated over only part of the catchment, on a contributing
area that would vary both between catchments, because of different catchment characteristics of soils,
vegetation and topography, and within a catchment, because of variation in antecedent conditions. He
modelled these variations by assuming a probability density function for contributing area of triangular
form, which “provides the bias towards small fractions of the catchment areas observed by Betson (1964)”
(Eagleson, 1972, p. 885). Random selection of a contributing area for a storm is a way of effectively
introducing the antecedent condition control on runoff generation into the predictions. Eagleson then
routes the effective rainfall produced using kinematic wave routing for both overland and channel flows
(see Section 5.5). The runoff generation component produces, for each rainstorm with
i
o
>
, a constant
effective rainfall of duration
t
r
for that storm. This very simple time distribution allows analytical solutions
of the kinematic wave equation to be performed.
More recent studies using direct numerical simulation have not been so constrained and a variety of
runoff generation models and routing methods have been used, including the PDM model of Section 6.2
(Lamb, 1999; Lamb and Kay, 2004; Calver
et al.
, 2009) and TOPMODEL of Section 6.3 (Beven, 1987;
Blazkova and Beven, 2000, 2004, 2009). Those that are based on event by event simulation, as in the
original Eagleson study, require some way of reflecting the effect of varying antecedent conditions on
runoff production. Those that use continuous simulation over long periods of time, including the drying of
the catchment during interstorm periods, account for antecedent conditions directly. Berthet
et al.
(2009)
recently compared event by event and continuous simulation approaches (in the flood forecasting, rather
than frequency estimation, context) and suggested that continuous simulation would generally perform
better in predicting peak discharges.
Sivapalan
et al.
(1990) produced a scaled flood frequency model based on the TOPMODEL concepts
and showed that catchment runoff production could be compared on the basis of eight similarity variables.
Their flood frequency curves were derived from storm by storm simulations and showed a transition
between saturation excess overland flow dominated flood peaks to infiltration excess overland flow
dominated flood peaks for more extreme events. However, they presented only a sensitivity analysis and
this similarity theory still has to be tested against real data sets, although a simplified version was shown
to have value in differentiating the hydrology of seven small Australian catchments by Robinson and
Sivapalan (1995). More on estimating the flood frequency characteristics of ungauged catchments using
rainfall-runoff models is to be found in Chapter 10.
8.7 Case Study: Modelling the Flood Frequency Characteristics on the
Skalka Catchment, Czech Republic
This study, reported by Blazkova and Beven (2009) is the latest of a number of applications of using
continuous simulation to estimate flood frequency for Czech catchments, including for catchments treated
as ungauged (Blazkova and Beven, 1997, 2004, 2009). The Skalka catchment (Figure 8.4) which spans
the German-Czech border was much larger at 672 km
2
than previous applications, with an elevation
range of 460 to 1041 m asl, a number of different gauging stations for subcatchment areas and multiple
raingauge sites with a common measurement period of 28 years. The gauging stations had records of
between 28 and 67 years and there were four snow measurement sites with records of between 10 and
20 years. This catchment has the potential for floods caused by winter rainfalls, spring snowmelt, and
summer thunderstorms.
