Geoscience Reference
In-Depth Information
is not a very precise one. It would be more precise if it was possible to measure or predict the timing
of inflows into the stream network itself accurately. Then the routing would only have to worry about
the flow processes within the stream which can be reasonably well predicted on the basis of hydraulic
principles (although in arid areas it may also be necessary to take account of the infiltration of some of
the water into the stream bed). Unfortunately, it is generally not possible to predict the volume and timing
of the inflows precisely, so the routing problem becomes one of the velocities of surface and subsurface
flows on the hillslope as well as in the stream channel. It may be very difficult then to separate out the
effects of the different possible flow pathways that different waters take on the timing of the hydrograph
at the stream outlet.
However, every hydrological model requires two essential components: one to determine how much
of a rainfall becomes part of the storm hydrograph (the
runoff generation
component), the other to
take account of the distribution of that runoff in time to form the shape of the storm hydrograph
(the
runoff routing
component). These two components may appear in many different guises and de-
grees of complexity in different models but they are always there in any rainfall-runoff model, together
with the difficulty of clearly separating one component from the other.
In general, it is accepted that the runoff generation problem is the more difficult. Practical experience
suggests that the complexities and nonlinearities of modelling the flow generation processes are much
greater than for the routing processes and that relatively simple models for the routing may suffice
(see discussion in Section 2.2).
1.7 The Problem of Choosing a Conceptual Model
The majority of hydrologists will be model users rather than model developers. Having said that, there has
been no shortage of hydrologists, particularly those undertaking research for a doctorate, who have set
themselves the task of developing a model. This is understandable; even now, the obvious approximation
inherent in today's models suggests that it should be possible to do better! However, given the range
of models consequently available in the literature or, increasingly, as modelling software packages, the
problem of model choice is not so different for the model user as for a researcher wanting to develop a
new and improved version. The question is how to decide what is satisfactory and what are the limitations
of the models available. We will take a preliminary look at this question in this section and return to it
in Chapter 12.
Let us first outline the “generic” choices in terms of a basic classification of model types. There are
many different ways of classifying hydrological models (see, for example, Clarke, 1973; O'Connell,
1991; Wheater
et al.
, 1993; Singh, 1995). We concentrate on a very basic classification here. The
first choice is whether to use a
lumped
or
distributed
modelling approach. Lumped models treat the
catchment as a single unit, with state variables that represent averages over the catchment area, such
as average storage in the saturated zone. Distributed models make predictions that are distributed in
space, with state variables that represent local averages of storage, flow depths or hydraulic potential,
by discretising the catchment into a large number of elements or grid squares and solving the equa-
tions for the state variables associated with every element grid square. Parameter values must also be
specified for every element in a distributed model. There is a general correspondence between lumped
models and the “explicit soil moisture accounting” (ESMA) models of O'Connell (1991) (see Sec-
tion 2.4), and between distributed models and “physically-based” or process-based models. Even this
correspondence is not exact, however, since some distributed models use ESMA components to repre-
sent different subcatchments or parts of the landscape as
hydrological response units
(see Section 6.2).
Distributed models currently available must use average variables and parameters at grid or element
scales greater than the scale of variation of the processes and are consequently, in a sense, lumped
conceptual models at the element scale (Beven, 1989a, 2006b). There is also a range of models that
