Geoscience Reference
In-Depth Information
commercial software. Where these are relevant to the current text they are referenced appropriately,
including (as far as is possible) up-to-date web sites. A summary of web resources is also given in
Appendix A.
A useful background volume is the IAHS Benchmark volume of classic papers in rainfall-runoff
modelling prepared by Keith Loague (2010). The volumes edited by Vijay P Singh (1995) and Singh
and Frevert (2002a, 2002b, 2005), which collected chapters written by developers of models for large
and small catchments, are also valuable sources of information. Distributed models of various types are
discussed in the texts by Mike Abbott and Jens Christian Refsgaard (1996), David Maidment (2002) and
Baxter Vieux (2004); lumped conceptual ESMA-type models by Thorsten Wagener
et al.
(2004); and
linear systems models by Jim Dooge and Philip O'Kane (2003). There are also many relevant articles
in the
Encyclopaedia of Hydrological Sciences
edited by Malcolm Anderson (2005), including a section
devoted to rainfall-runoff modelling in Part 11 of Volume 3.
2.9 Key Points from Chapter 2
•
Any hydrological model must include functional components that account for the relationship be-
tween total rainfall and runoff generation in an event, and the routing of the generated runoff to the
catchment outlet.
•
The volume of rainfall equivalent to the generated runoff is called the effective rainfall. It has a strongly
nonlinear dependence on the antecedent state of the catchment. Once the effective rainfall has been
calculated, linear routing methods, such as the unit hydrograph, often work quite well.
•
Following the work of Robert Horton, early applications of the unit hydrograph technique assumed
that all storm runoff was generated by an infiltration excess mechanism. This is not generally true
but the methods continue to be applied successfully, despite difficulties of hydrograph and rainfall
separation, because they have the functionality needed for discharge prediction at the catchment scale.
•
With calibration of parameter values, even simple storage element (ESMA) models can produce good
predictions of streamflow hydrographs and soil moisture deficits.
•
Modern transfer function models aim to overcome some of the problems of defining the storage
elements of an ESMA model correctly by letting the data available determine an appropriate structure
and level of complexity while avoiding the problem of hydrograph separation (see Chapter 4).
•
The earliest distributed models were based on the time-area concept. Recent work based on the defi-
nition of distributed hydrological response units by overlays of different data types in a GIS system is
based on essentially similar concepts.
•
Fully process-based distributed models allow the prediction of local hydrological responses within a
catchment but have many parameter values that must be specified for every grid element. This makes
parameter calibration difficult but direct measurement or estimation of effective parameter values at
the grid scale is also difficult due to heterogeneity of catchment characteristics and the limitations of
the available measurement techniques (see Chapter 5).
•
Simpler models, based on distributions of responses within a catchment may still have much to offer
for prediction at the catchment scale and some, such as TOPMODEL, have the potential to map those
responses back into the catchment to allow additional evaluation of the simulations (see Chapter 6).
•
The accuracy of the predictions of rainfall-runoff models in application to specific catchments remains
limited by the availability of data as well as process representations. The resulting uncertainties should,
where possible, be estimated (see Chapter 7).
•
Recent concepts of the Representative Elementary Watershed (REW) and Models of Everywhere are
beginning to blur the boundaries between the different types of model. The next generation of models
will change the way in which the modelling problem is tackled (see Chapter 9).
