Geoscience Reference
In-Depth Information
4
Predicting Hydrographs Using
Models Based on Data
Some experts are fond of saying that the simplest methods are the best. This really confuses
the issue. All other things being equal, this is clearly true. Unfortunately, all other things are
not usually equal. The prime criterion should be accuracy, and until equivalent accuracy
is demonstrated, simplicity should be a second-order criterion.
Ray K. Linsley, 1986
Finally, a most important aspect of DBM modelling is the identification of parametrically
efficient (parsimonious) models, as a way of avoiding the identifiability problems that are
associated with the estimation of over-parameterized models from limited time series data
.... So, a parsimonious model, in this important dynamic sense, is one that has a low-
est dynamic order that is consistent with the information content in the data and whose
parameters are statistically significant.
Peter Young, 2003
4.1 Data Availability and Empirical Modelling
There are two very different, but both widely held, views of modelling. The first holds that all models,
however physically based their theory, are essentially tools for the extrapolation of available data in
time (to different periods) and space (to different catchments). This view of modelling as induction
is the subject of this chapter. The second view holds that models should as far as possible reflect our
physical understanding of the processes involved. Only in this way, it is suggested, can we have faith in
predictions that lie outside of the range of data available in time (e.g. in the future) and space (in different
catchments). This view, modelling
almost
as deduction, is the subject of Chapter 5. It is modelling
almost
as deduction because, unfortunately, we cannot yet get away without some empiricism in the description
of hydrological processes and in estimating model parameters, and may, in fact, never be able to do so
(see, for example, Beven, 2006b; the discussion in Chapter 9).
