Geoscience Reference
In-Depth Information
many thousands of such elements. In addition, the process equations require many different parameters
to be specified for each element. With so many parameter values, parameter calibration by comparison
with the observed responses in a catchment becomes very difficult. Which parameter values should be
changed to try to improve the simulation? The choice is not always obvious because of the interactions
between the effects of different parameters that follow from the physical basis of the model.
In principle, of course, parameter adjustment of this type should not be necessary. If the process
equations are valid, it should follow that the parameters should be strongly related to the physical char-
acteristics of the surface, soil and rock. Techniques are also available for measuring such parameters,
although, as noted in Chapter 1, there are problems of scale in such measurements. Most measurement
techniques can only be used to derive values at scales much smaller than the element mesh used in the
approximate solution. The model requires effective values at the scale of the elements. If the soil was
homogeneous, this would not matter too much, but unfortunately soils and surface vegetation tend to be
very heterogeneous at the measurement scale such that establishing a link between measurements and
element values is difficult, even theoretically. Indeed, for the case of coupled surface and subsurface flow
processes, it has been suggested that the concept of effective values of element scale parameters may
not be valid (Binley
et al.
, 1989). This remains an important topic in distributed modelling that requires
further research.
Despite these difficulties, there has been a strong surge in the use of distributed modelling over the last
two decades. This has partly been because increases in computer power, programming tools and digital
databases have made the development and use of such models so much easier and partly because there
is a natural tendency for a model development team to try to build in as much understanding from their
perceptual model of the important processes as possible. Thus, there is an obvious attraction of distributed
process modelling. There are also very good scientific reasons underlying the effort. One is the need for
distributed predictions of flow pathways as a basis for other types of modelling, such as the transport of
sediments or contaminants. It may not be possible to make such predictions without a distributed model
of some sort.
Another reason for the surge is the use of models for impact assessment. Changes in land use, such
as deforestation or urbanisation, often affect only part of a catchment area. With a distributed model,
it is possible to examine the effects of such piecemeal changes in their correct spatial context. There is
also an argument that, because of the physical basis of the model, we might be able to make a better
assessment of the effects of changing characteristics of a catchment because it will be easier to adjust
parameter values that have physical meaning. The difficulties of specifying effective parameter values at
the element scale, however, rather undermines this argument.
Examples of distributed process-based models include the Systeme Hydrologique Europeen (SHE)
model, originally a joint project between the Institute of Hydrology in the UK, the Danish Hydraulics In-
stitute and SOGREAH in France, but now being developed separately (see Abbott
et al.
, 1986a; Bathurst
et al.
, 1995; Refsgaard and Storm, 1995). Both the Danish (MIKE SHE) and UK (SHETRAN) versions
of SHE have been extended with full three-dimensional subsurface components (but also with simpli-
fied subsurface storage elements), and sediment and water quality components. Refsgaard
et al.
(2010)
provide a summary of 30 years of experience of using the MIKE SHE model. The UK Institute of Hydrol-
ogy also developed the Institute of Hydrology Distributed Model (IHDM, Calver and Wood, 1995); in
Australia, there are the THALES model (Grayson
et al.
, 1995) and the CSIRO TOPOG-dynamic model
(Vertessy
et al.
, 1993); and there are a number of models developed in the USA including the Integrated
Hydrologic Model (InHM) of VanderKwaak and Loague (2001) and the Gridded Surface/Subsurface
Hydrologic Analysis model (GSSHA) of Downer
et al.
(2005). GSSHA was an extension of an earlier
infiltration and surface runoff model, CASC2D (Downer
et al.
, 2002). These models differ primarily in
the way they discretise a catchment and solve the process equations (sometimes with simplifications),
but all are essentially based on the original Freeze and Harlan blueprint from 1969 as a description of the
flow processes (see Chapter 5 for more details).
