Environmental Engineering Reference
In-Depth Information
analytical approach in which the storage in the
hillslope soils is assumed to be distributed as if it
was in steady state with a distributed recharge
equivalent to the subsurface discharge from the
slope. Within this approach, the level of saturation
deficit or water table in the soil at any point in the
catchment can be related to a topographic index
involving the area draining through that point
from upslope and the slope angle at that point.
A similar approach was developed independently
by O'Loughlin (1986), while more recent variants
taking account of the unsaturated zone and
perched water tables have been suggested by Liu
and Todini (2002; Topkapi) and Scanlon et al.
(2000). The important feature of Topmodel is that
the predictions of the model can not only be
simplified by discretizing the distribution func-
tion of the topographic index in the catchment,
and then mapped back into the catchment if the
map of the topographic index is known. This
results in a computationally efficient model that
can still produce distributed predictions that can
be evaluated for realism (e.g. Beven and Kirk-
by 1979; Seibert et al. 1997; Guntner et al. 1999;
Blazkova et al. 2002) in addition to evaluating
discharge predictions. The major limitation of
Topmodel lies in its simplifying assumptions,
which will only really be appropriate in humid
catchments with relatively impermeable bedrock
and moderate topography. They will certainly
not be appropriate simplifications in many other
catchments.
The final class of simplified distributed models
includes those that take account of surface and
subsurface transfers of water on hillslopes by
explicit routing between the spatial elements of
the catchment discretization. The discretization
might be based on triangular irregular networks
(Delauny discretization) or square grids. The for-
mer generally allow a better representation of
hillslope topography (e.g. the model of Ivanov
et al. 2004); the latter can more easily use the
raster grid, which is used to store many types of
data in geographical information systems, includ-
ing remote-sensing data. Examples of grid-based
models are the LISFLOOD model of De Roo et al.
(2000); the ARCHydro model (Maidment 2002);
discussion of this class of simplified distributed
models may be found in Beven (2001b).
Within the first class are models based on
'hydrological response units' (HRUs), in which
each response unit is treated as an essentially
one-dimensional element from which any pre-
dicted surface runoff or subsurface drainage
reaches the stream channel network directly. An
example is the SLURP model of Kite and Kou-
wen (1992), which made use of soil and vegetation
maps to divide the landscape into a number of
functional classes, or HRUs; this approach has
been perpetuated to the present day in a large
number of 'land surface parameterizations' (LSPs)
used as the lower boundary conditions of numer-
ical weather predictors and general circulation
models of the atmosphere. Recent examples of the
latter are the NCAR Community Land Model
(Bonan et al. 2002) and the Jules model in the UK
(see http://www.jchmr.org/jules/science/science.
html). Both rely on a similar strategy of functional
class units to represent the hydrology within
a global circulation model (GCM) calculation
element, neglecting any exchanges between units.
This is an important problem with this type of
model based on spatially distinct HRUs. In any
landscape with soil-covered hillslopes, topogra-
phy is important in controlling land surface
hydrology in terms of both evapotranspiration
and runoff generation processes. The lower parts
of hillslopes tend to have more water available
for evapotranspiration as a result of flows from
upslope. Since they are generally wetter, they are
also more likely to act as contributing areas for
surface and subsurface runoff. Within any rainfall
regime, these types of responses are structured by
the topography and soil (and, inmany catchments,
the underlying geology). Thus it would also
be useful to be able to reflect these controls in
a simple way.
One way of doing so is tomake use of analytical
approximations to more complete mathematical
representations of the flow processes, but in such
a way that the results can be mapped back
into space. One widely used example of such
a model is Topmodel (e.g. Beven and Kirkby 1979;
Beven 1997, 2001b). This makes use of a simple
