Geoscience Reference
In-Depth Information
perceived hydrological processes in the catchment, so can be considered as an extension of a lumped
storage model. It has an advantage over lumped storage models that the parameters may be easier to
estimate in an optimisation exercise. An extension that includes explicit routing between grid elements
(the Grid to Grid, G2G, model) was also described.
•
The rainfall-runoff model, TOPMODEL, makes use of an index of hydrological similarity based on
topography and soils information that allows the model predictions to be mapped back into space
(see Box 6.1). Calculations are made based on the distribution of the index, which greatly reduces the
computer resources required. The TOPMODEL concepts are not, however, applicable everywhere,
particularly in catchments subject to strong seasonal drying when the basic assumptions underlying
the index break down.
•
The simplicity of the TOPMODEL calculations have allowed the interaction between grid resolution
of the topographic analysis and calibrated parameter values to be studied in a number of applications.
A similar interaction between scale of discretisation and effective parameter values should hold for
more complex models, including physically based fully distributed models, but may not be so readily
apparent.
•
A number of recent attempts have been made to improve the theory of TOPMODEL, while retaining
its advantages of simplicity.
•
Models based on the definition of hydrological response units (HRUs) derived by overlaying different
soil, geology, topography and vegetation characteristics in a geographical information system are
becoming increasingly popular (the USDA SWAT model (Box 6.2) is particularly widely used). These
models differ widely in the representation for each of the HRUs defined in this way and in the routing of
flows to the catchment outlet. Such models do allow the mapping of the distribution of responses back
into the catchment space using the GIS. With fine discretisations of the catchment and routing of flow
between HRUs, such models may be considered as simplified fully distributed models. Where similar
HRU elements are grouped together for calculation purposes, they may be considered as distribution
function models.
•
HRU models might involve specifying hundreds or thousands of parameter values. Default values
might be available in associated databases for different models but we should expect that effective
values required to get good predictions might vary with discretisation scale and model structure. The
question of how to define hydrologically good performance given uncertainties in input data, model
structure and discharge observations recurs in the remaining chapters.
