Geoscience Reference
In-Depth Information
for both model structures and parameter sets as hypotheses, outlined above. We take as our starting point
that,
a priori
, all the available modelling strategies and all feasible parameter sets within those modelling
strategies are potential models of a catchment for a particular project. The aims of that project, the
budget available for the project, and the data available for calibrating the different models will all limit
that potential range of simulators. The important point is that choices between models and between
parameter sets must be made in a logical and scientifically defensible way allowing for data uncertainties
and limitations. It is suggested, however, that at the end of this process, there will not be a single model
of the catchment but a number of acceptable models (even if only different parameter sets are used within
one chosen model structure) to provide predictions.
There are clearly implications for other studies that depend on models of rainfall-runoff processes.
Predictions of catchment hydrogeochemistry, sediment production and transport, the dispersion of con-
taminants, hydroecology, and, in general, integrated catchment decision support systems depend crucially
on good predictions of water flow processes. To keep my task manageable, I have not chosen to address the
vast literature that deals with these additional topics but each additional component that is added to a mod-
elling system will add additional choices in terms of the conceptual representation of the processes and the
values of the parameters required (but see the discussion of residence time distributions in Chapter 11).
In that all these components depend on the prediction of water flows, they are subject to the types of
uncertainty in predictive capability that have been outlined in this chapter and are discussed in more detail
later. This is not only a research issue. In the UK, uncertainties in model predictions have already played
a major role in decisions made at public inquiries into proposed developments. The aim of this volume
is to help provide a proper basis for rainfall-runoff modelling across this range of predictive contexts.
1.9 Key Points from Chapter 1
•
There are important stages of approximation in the modelling process in moving from the
perceptual
model
of the response of a particular catchment, to the choice of a
conceptual model
to represent that
catchment and the resulting
procedural model
that will run on a computer and provide quantitative
predictions.
•
One particular perceptual model has been outlined as a basis for comparing the descriptions of different
models that are given in the following chapters.
•
Studies of the geochemical characteristics of runoff, and particularly the use of environmental tracers,
have resulted in an increased appreciation of the importance of subsurface storm runoff in many
catchments.
•
A basic classification of modelling strategies has been outlined, differentiating between
lumped
and
distributed
models and
deterministic
and
stochastic
models.
•
Some preliminary guidelines for the choice of a conceptual model for a particular project have been
outlined. This problem will be reconsidered in Chapter 10.
•
The problem of the calibration of parameter values has been outlined. The idea of an optimum parameter
set has been found to be generally ill-founded in hydrological modelling and can be rejected in favour
of the concept of the
equifinality
of different models and parameter sets.
•
It is expected that, at the end of the model evaluation process, there will not be a single model of the
catchment but a number of acceptable models (even if only using different parameter sets within one
chosen model structure) to provide predictions.
•
Prediction of other processes, such as hydrogeochemistry, erosion and sediment transport, and ecology,
that are driven by water flows will introduce additional choices about conceptual model structures and
parameter values and will be subject to the uncertainty arising in the rainfall-runoff predictions.
