Geology Reference
In-Depth Information
as saturated hydraulic conductivity, are often
estimated using rainfall simulation experiments
at a scale of ca. 1 m
2
, while models are applied at
the field or even the catchment scale.
The above does not imply that erosion models
can only be applied at the scale that was used for
parameter estimation or erosion rate assessment
during model development. What is necessary,
though, is that scaling is carried out appropriately
and that the results obtained at different scales
are interpreted correctly.
Making sure that scaling is appropriate is of
tremendous importance when average erosion
rates over large surface areas need to be estimated.
The advent of GIS software has made the manip-
ulation of large spatial datasets much easier, and
over recent years several global estimates of soil
erosion have been produced (Yang
et al
., 2003; Ito,
2007; Van Oost
et al
., 2007); the accuracy of such
estimates depends critically on the use of correct
scaling factors. For simple, topographically-driven
models the length or area factor that is present
within the model can be considered as a scaling
factor. If the erosion rate per unit surface area
increases with increasing slope length, average
erosion rates measured on small plots will be
lower than average regional soil erosion rates. If,
on the other hand, the erosion rate decreases with
increasing slope length/contributing areas, aver-
age regional erosion rates will be lower than aver-
age soil erosion rates. The slope length effect may
vary considerably, depending on local conditions
and the type of erosion process. In RUSLE calcu-
lations it is assumed that erosion scaling with
slope length depends on the ratio of rill to inter-
rill erosion, which is primarily controlled by
slope gradient but may also depend on other fac-
tors such as soil type and soil conditions (Loch,
1996; Renard
et al
., 1997). If inter-rill erosion is
dominant, the length factor can be small (the
length exponent is close to zero), which implies
that erosion rates are more or less constant as a
function of slope length. If rill erosion is domi-
nant, the length exponent is assumed to be much
higher (approaching 1), which implies that ero-
sion rates increase with increasing slope length/
contributing areas.
There is a significant, albeit relatively small,
amount of experimental data backing up the slope
length factors that are generally used for arable
land, especially for cases where rill erosion is
dominant (Govers
et al
., 2007). However, the sit-
uation may well be different on hillslopes under
natural vegetation: under these conditions, ero-
sion rates per unit surface area may well decrease
with increasing slope length, the fundamental
reason being that total runoff amounts do not
increase, or increase only slowly with increasing
slope length, as the spatial variability of hydrau-
lic conductivity is very high and increasing water
depths promote increasing infiltration (Dunne
et al
., 1991). Net erosion rates on such slopes will
then become limited by the amount of sediment
that can be transported by overland flow off the
slope. If time-averaged transport capacity does
not increase at least linearly with slope length,
erosion rates per unit area will decrease with
increasing slope length as proposed by Parsons
et al
. (2006). Slope length factors used in (R)USLE
applications on non-agricultural land should be
adjusted accordingly, but experimental data on
the variation of sediment fluxes with slope length
on surfaces with natural vegetation is at present
very scarce.
Process-based models generally model runoff
generation and transfer explicitly: they are there-
fore in principle capable of simulating erosion
at larger scales without an
a priori
specification
of a length or area factor. However, this evidently
requires that the mechanisms controlling the
scaling of runoff generation and transfer are
explicitly accounted for in the model. If, for
instance, a model is applied in an environment
where infiltration rates are increasing with
increasing water depth while the model descrip-
tion of infiltration and runoff generation does
not allow for this, the model may be expected
to perform badly when applied to slopes of greater
length than that of typical erosion plots
(1-20 m).
Another issue that may cause misapplication
at large scales is that the erosion response of a
larger catchment is not only controlled by the
amount of rill and inter-rill erosion, but also by
