Geology Reference
In-Depth Information
process-based in that they effectively describe the
water balance of a catchment whereby a propor-
tion of the incoming rainfall passes into runoff
whilst the rest is held in various stores, such as
interception by the plant cover, soil moisture and
groundwater. An erosion component can be built
on to such a model as illustrated by the Stanford
Sediment Model (Negev, 1967), which is linked
to the Stanford Watershed Model (Crawford &
Linsley, 1966). Since most practical applications
require knowledge of the way sediment is moved
over the landscape so that protection measures
can be targeted either at areas of sediment source
or along pathways of movement, there is a
demand for models which can describe what is
happening within a catchment, and lumped mod-
els are unable to do this. Lumped models are
therefore of value for predicting soil loss from
relatively small areas such as an agricultural field,
road embankment or construction site. They can
also be used to assess erosion over large areas, as
illustrated by the PESERA model (Gobin
et al
.,
2006) which estimates mean annual erosion over
1 km
2
size units. A process-based approach is used
to generate infiltration-excess and saturation
overland flow from daily rainfall. The calcula-
tions are integrated across the frequency distribu-
tion of daily rainfall events. Sediment transport is
estimated according to the runoff, soil erodibility
and slope of each cell. Both runoff generation and
sediment transport are modified by land cover,
surface roughness and soil crusting.
Where the need is to determine where erosion
and deposition take place within a catchment,
distributed models are used. These operate by
dividing the catchment into discrete land units
and use mathematical procedures to route water
and sediment from one unit to another. Such
models are necessarily process-based and, in so far
as they use input data that can be measured physi-
cally in the field and use continuity equations to
ensure the conservation of mass and energy as
water and sediment are moved in space and over
time, they are often considered as physically-based
(Beven & Kirkby, 1979). Most of the recently
developed erosion models, like WEPP, GUEST,
LISEM and EUROSEM, are distributed models.
They are suitable for analysing the effects of
changes in land use in different parts of a catch-
ment, as well as the effects of variations in rain-
fall, soil type, slope and land cover within a
catchment.
Some distributed models, like EUROSEM and
CREAMS, require land units in a catchment to
be identified in terms of similarity in soils, slope
and land cover. For most practical purposes,
the land units are similar in nature to the land
facets identified in terrain analysis (Christian &
Stewart, 1968; Webster & Beckett, 1970). These
can be grouped into larger units or land systems
(usually between 100 m
2
and 10,000 m
2
in size)
which have been shown to be significantly differ-
ent from each other in terms of both erosion sta-
tus and the rate of change in erosion over time
(Morgan
et al
., 1997). The art in setting up these
models is to identify the land units so that they
are internally as uniform in their characteristics
as possible, and then to determine the likely pat-
terns of water flow from one unit to another
(Auzet
et al
., 1995). Although this can be done
using aerial photographs, topographic maps or
digital elevation models to determine the low
points in the landscape along which water will
concentrate, there is often the need for field
observations to identify where flow paths deviate
from those which would occur naturally, for
example as a result of installing diversion ter-
races or ditches to take water across the slope to
a safe outlet rather than allowing it to flow
downslope.
In recent years, with the advent of geographi-
cal information systems, model users have
moved away from defining land units in relation
to the natural variations in the landscape, in
favour of dividing the catchment into grid cells
of uniform size. Although the movement of
water from one cell to another is still based on
the local topography, the units themselves are
less likely to be internally consistent in their
soils, slopes and land cover. Unless the grid cells
are extremely small, say 10 m × 10 m, it is likely
that they will be crossed by boundaries between
soil types or slope breaks. The larger the grid
cells used, the more likely that each one will
