Geoscience Reference
In-Depth Information
to properly account for the land use change.
To fully account for this type of land use change
it is necessary to use a hydrological model that
explicitly accounts for antecedent soil moisture,
soil infiltration characteristics and rainfall inter-
ception as distinct hydrological processes.
The CN approach has been used extensively to
make runoff predictions based on a time series of
rainfall (Ponce and Hawkins, 1996). It has also
been incorporated into more sophisticated models
such as the Soil and Water Assessment Tool
(SWAT; Cao
et al
., 2006). In this case the rainfall-
runoff relationship is derived from the CN
approach and combined with other hydrological
processes such as evaporation estimation and
river flow rates.
normally carried out by comparing predicted flows
to measured values and adjusting (or 'optimising')
the parameters until the best fit is obtained. There
is considerable debate on this technique as it may
sometimes be possible to obtain a similar predicted
hydrograph using a completely different set of
optimised parameters. It is certainly true that the
optimised parameters cannot be treated as having
any physical meaning and should not be transferred
to catchments other than those used for calibration.
Lumped conceptual models offer a method of
formulating the hydrological cycle into a water
budget model that allows simulation of streamflow
while also being able to 'see' the individual processes
operating. This is an advance beyond black box
modelling, but because the processes are represented
conceptually they are sometimes referred to as grey
box models (i.e. you can see partially into them).
Lumped conceptual models
Lumped conceptual models were the first attempt
to reproduce the different hydrological processes
within a catchment in a numerical form. Rainfall
is added to the catchment and a water budget
approach used to track the losses (e.g. evaporation)
and movements of water (e.g. to and from soil water
storage) within the catchment area. There are many
examples in the literature of lumped conceptual
models used to predict streamflows (e.g. Brandt
et
al
., 1988).
The term 'lumped' is used because all of the
processes operate at one spatial scale - that is, they
are lumped together and there is no spatial dis-
cretisation. The scale chosen is often a catchment or
sometimes sub-catchments.
The term 'conceptual' is used because the equa-
tions governing flow rates are often deemed to be
conceptually similar to the physical processes
operating. So, for instance, the storage of water in a
canopy or the soil may be thought of as similar to
storage within a bucket. As water enters the bucket
it fills up until it overflows water at a rate equal to
the entry rate. At the same time it is possible to have
a 'hole' in the bucket that allows flow out at a rate
dependent on the level of water within the bucket
(faster with more water). This is analogous to soil
water or canopy flow but is not a detailed descrip-
tion such as the Darcy-Richards approximation or
the Rutter model. The rate of flow through the
catchment, and hence the estimated streamflow, is
controlled by a series of parameters that need to be
calibrated for a given catchment. Calibration is
Physically based distributed models
The rapid advances in computing power that have
occurred since the 1970s mean that numerical
modelling has become much easier. Freeze and
Harlan (1969) were the first to formulate the idea
of a numerical model that operates as a series of
differential equations in a spatially distributed
sense, an idea that prior to computers was unwork-
able. Their ideas (with some modifications from
more recent research) were put into practice by
several different organisations to make a physically
based distributed hydrological model. Perhaps the
best known of these is the Système Hydrologique
Européen (SHE) model, built by a consortium of
