Geology Reference
In-Depth Information
(2) Capability to represent spatial variability
Erosion is a spatially non-uniform process and it is
likely to be the case that a large part of the sediment
yield comes from only a small part of the basin. As
an example, it has been suggested that targeted
reforestation of a small part of a basin (the critical
areas for shallow landslide occurrence) could have a
disproportionately large effect in reducing sediment
yield (Reid & Page, 2002). Similarly, a sediment
problem in one part of the basin may be caused by
activity in another (e.g. erosion triggered by head-
water deforestation might cause siltation of the
downstream river channel) (e.g. Glade, 2003).
Models intended for land management applications
or for investigating sediment yield processes there-
fore need to account for spatial variability in ero-
sion and sediment transport. Typically this is
achieved with some form of grid system.
(4) Simulation period
Generation of long-term
sediment yields requires a capability for simulat-
ing periods of several years. Generation of extreme
sediment loads requires a capability for simulat-
ing individual storm events. Both requirements
can be met by continuous simulation models.
These models also have an advantage over single-
event models in being able to generate the appro-
priate antecedent soil moisture conditions for
individual events.
(5) Data availability
Physically-based models
require meteorological input data to drive the
simulation, basin property data (such as eleva-
tions, soil characteristics and vegetation charac-
teristics), and calibration or validation data
consisting of hydrological, erosion, sediment
transport, landslide and other response data.
The input data and model parameters have to be
evaluated at every grid element, so the greater
the spatial distribution in the available data, the
more accurate or more representative that eval-
uation will be. However, it is currently imprac-
tical to take measurements for every model grid
element, and some form of estimate is therefore
required for the majority of elements. On the
other hand there is a wide range of data sources
which can be used, including soil property meas-
urements, soil hydrology maps, land-use maps,
runoff and erosion plot experiments, gully
experiments, gauging station and rain gauge
records, well logs, river channel surveys and
landslide inventories. Remote-sensing tech-
niques can provide increasingly relevant data on
a spatially distributed basis and in the future
should be able to reduce the extent of parameter
estimation. The input data and many of the cali-
bration data consist of time series. In principle,
physically-based models do not require lengthy
data records for calibration. However, there are
sufficient approximations in their design
(for example, the use of one-dimensional instead
of three-dimensional formulations) that a
short period of record (especially if containing
an extreme response) is helpful in adjusting
the final values of the model parameters (see
Chapter 3).
(3) Inclusion of the relevant processes
The
main sediment-generating processes are erosion
by raindrop impact, overland flow, rilling, gully-
ing, landsliding and bank erosion. These, and the
sediment transporting processes, are hydrologi-
cally driven, and important hydrological proc-
esses therefore include the generation of overland
flow by rainfall exceeding infiltration (Hortonian
overland flow) and by upward saturation of the
soil, generation of river flow from surface and
subsurface inputs, and variation of the soil pore-
water pressures which affect slope stability.
Different processes are likely to dominate in dif-
ferent parts of a basin and at different basin
scales (e.g. de Vente & Poesen, 2005). Model rel-
evance should therefore be clearly defined, in
terms of the erosion and transport processes that
are represented, and erosion models are likely to
retain the greatest flexibility if based on a gen-
eral hydrological model. At the same time,
attempts to represent erosion processes in ever
finer detail must be carefully balanced against
the ability to distinguish the results in the face
of uncertainties associated with model parame-
terization, data collection and the natural varia-
bility in erosion and sediment yield (Nearing,
2004; Quinton, 2004; see Chapter 4 for a more
general discussion of this issue).
