Geoscience Reference
In-Depth Information
of the parameters, TOPMODEL provides good simulations of stream discharges and broadly believable
simulations of variable contributing areas.
Catchments with deeper groundwater systems or locally perched saturated zones may be much more
difficult to model. Such catchments tend to go through a wetting up sequence at the end of the summer
period in which the controls on recharge to any saturated zone and the connectivity of local saturated
zones may change with time. An example is the Slapton Wood catchment in southern England modelled
by Fisher and Beven (1995).
Simulation of drier catchment responses
A model that purports to predict fast catchment responses on the basis of the dynamics of saturated
contributed areas may not seem to be a likely contender to simulate the responses of catchments that are
often dry, such as in Mediterranean or savannah climates. However, Durand
et al.
(1992) have shown that
TOPMODEL can successfully simulate discharges in such catchments at Mont Lozere in the Cevennes,
southern France, at least after the calibration of some parameters.
Experience in modelling the Booro-Borotou catchment in the Cote d'Ivoire (Quinn
et al.
, 1991),
Australia (Barling
et al.
, 1994) and catchments in the Prades mountains of Catalonia, Spain (Pinol
et al.
,
1997), suggests that TOPMODEL will only provide satisfactory simulations once the catchment has
wetted up. In many low precipitation catchments, of course, the soil may never reach a “wetted” state
and the response may be controlled by the connectivity of any saturated downslope flows. TOPMODEL
assumes that there is connected downslope saturation everywhere on the hillslope; before such
connectivity is established, a dynamic index would be required (Barling
et al.
, 1994; Beven and Freer,
2001; Vincendon
et al.
, 2010). Such catchments also tend to receive precipitation in short, high-intensity
storms. Such rainfalls may lead, at least locally, to the production of infiltration excess overland flow
which is not usually included in TOPMODEL (but see the work of Beven (1986) and Sivapalan
et al.
(1990) for example applications including infiltration excess calculations). The underlying assumptions
of the TOPMODEL concepts must always be borne in mind relative to the pertinent perceptual model
for a particular catchment.
6.3.4 Testing the Hydrological Similarity Concept in TOPMODEL
TOPMODEL may be expected to perform best when tested against catchments where its assumptions
are met, in particular those of an exponential saturated zone store, a quasi-parallel water table and a
topographic control on water table depth. A full critique of the TOPMODEL concepts can be found
in Beven (1997). There are certainly limitations on both the geographical and seasonal validity of the
TOPMODEL concepts, but they do provide a basis for thinking about the distributed nature of catchment
responses. It has always been stressed that TOPMODEL is not a fixed model structure but rather a set
of concepts that should be modified if it is perceived that a catchment does not conform to the basic
assumptions. Ways of relaxing the basic assumptions are discussed in Box 6.1.
The main limitation imposed by the model assumptions is that of the quasi-steady-state water table
shape. This has been criticised by Wigmosta
et al.
(1999), who compare the results of TOPMODEL
with a dynamic simulation based on a subsurface kinematic wave solution. They show that although
TOPMODEL could usually be calibrated to produce a reasonable simulation of discharge hydrographs
produced by the kinematic wave model, the resulting effective parameter values tended to be high and
the steady state assumption did not produce reasonable predictions of the dynamic changes in the water
table. It is the steady state assumption that allows TOPMODEL to make use of similarity in greatly
increasing computational efficiency. This is useful for a number of purposes, not the least of which is
the exploration of predictive uncertainty (considered in Chapter 7). The approach can also be modified
to allow more dynamic calculations whilst retaining the concept of the index (see Section 6.3.5 and the
TOPKAPI variant in Section 6.5).
