Geology Reference
In-Depth Information
inclusion of these processes would make a model
too cumbersome.
●
Spatial scale
: Modelling approaches exist at
the (i) slope-scale, representing processes occur-
ring mostly at the field and hillslope scale, such
as splash and rill erosion; and (ii) catchment-
scale, also representing processes operating in
regions of accumulated runoff, such as gully ero-
sion and channel sediment processes (e.g. Lane
et al
., 1997; de Vente & Poesen, 2005); there is
often some juxtaposition between scales. It
should be noted that this distinction is between
the simulated processes rather than the extent of
the area of application; slope-scale models can be
applied to catchments without representing other
processes. As an example, Mantel
et al
. (2003)
applied the PESERA model to large areas in north-
ern France and Belgium and southern Iberia, but
the model did not represent gully erosion and
channel processes within these regions and there-
fore the study was classified as slope-scale.
●
Climate change scenario
: A common problem
in climate change studies is the mismatch
between GCM results, with higher quality at
coarse spatial scales and for annual and seasonal
values, and those required for model application,
at fine spatial scales and for daily averages (Xu &
Singh, 2004). Therefore, GCM results are usually
taken as a starting point to generate climate
change scenarios using two processes: (i) downs-
caling of GCM results to the desired spatial and
temporal scale; and (ii) hypothetical, where GCM
results are used to provide a range of possible
changes to climate variables and scenarios are
artificially built within these ranges. The most
common downscaling methods referred to by Xu
and Singh (2004) are (a) dynamic downscaling,
where GCM results are used to force regional
simulations of climate change at finer spatial and
temporal scales using regional climate models
(RCM); and (b) statistical downscaling, which
uses a statistical relationship between GCM 'con-
trol' runs (for current conditions) and the observed
climate patterns in a given location to provide
future climate scenarios at the desired spatial and
temporal scale. The choice of downscaling
method can have significant impacts on the
results given by erosion models (Zhang, 2007).
In contrast, hypothetical scenarios are usually
perturbations of current climate conditions with
several degrees of change, aiming to obtain a
response function of soil erosion to changes in
climate parameters, in effect studying the sensi-
tivity of soil erosion to changes in climate given
a reasonable interval (Xu & Singh, 2004).
Selecting between the modelling approaches sum-
marized in Table 15.1 depends upon the overall
objectives of the study. It should be taken into
account that increasing the complexity of a model-
ling study - in terms of both process description
and spatial and temporal discretization - does not
lead to improved results, due, to a great extent, to
the uncertainty associated with the input parame-
ters required by complex models which often lead
to a greater uncertainty in the results without pro-
viding additional predictive power (Jetten
et al
.,
1999, 2003; see also Section 6.4). Therefore, the
complexity of the selected approach should match
the questions which the modelling study wishes to
answer. Two examples of this selection process
can be taken from the studies in Table 15.1:
●
studies with a continuous modelling approach
focus on interactions between climate, vege-
tation growth and soil erosion at longer tempo-
ral scales, while those with an event-based
approach focus on non-linear processes such as
gully erosion and peak discharge/sediment yield
relationships;
●
studies at the catchment scale usually focus on
within-watershed erosion patterns (e.g. gully ero-
sion and sediment deposition) and channel proc-
esses, while studies focusing on soil erosion in
agricultural fields constrained simulations to the
slope scale.
Other factors should also influence the selection
of a modelling approach, such as the dominant
erosion processes in the study area or the availa-
bility of data for model parameterization and val-
idation (Jetten
et al
., 1999, 2003; Section 6.4).
In some cases, a multiscale modelling framework
could be selected, using different models to study
different problems with the required degree of
complexity; this approach can be exemplified
by the work of Nunes (2007), who studied the
