Geology Reference
In-Depth Information
USLE and derivatives, to physically-based process
models. In the former category, processes such as
transport and deposition are often not included.
Temporal scale varies from individual rainstorms
to lumped annual values, with the latter making
up the majority of published results.
(II)
A catchment scale varying in size from less
than 100 ha to several 100 km
2
. Models operating
at this scale are generally process-based models or
hybrid models that are adaptations of the USLE
with added process descriptions and methods for
spatial water and/or sediment routing and accu-
mulation. These vary from explicit modelling
of detachment, transport and deposition, to the
use of runoff coefficients and sediment delivery
ratios. Temporal scales vary again from individ-
ual rainfall events to annual totals, although most
calibration is done for individual events.
(III)
A so-called 'large' scale with administrative
boundaries, from provinces and parts of countries
to continental. Models at this scale are partly
physically based, but use variables derived from a
DEM (digital elevation model) as proxies for slope
angle, transport capacity and accumulation.
Generally, sediment delivery ratios are used.
Temporal scales vary from monthly to annual
totals.
Note that the model acronyms and main refer-
ences can be found in Table 3.1.
This chapter first gives an overview of calibra-
tion, using examples from all three categories.
These are manual calibrations based on some
goodness of fit parameters between predicted and
observed runoff and soil loss. Secondly, several
examples of more complex means of calibration
are given based on Monte Carlo-type automatic
parameter estimation approaches (such as GLUE
and PEST). These are generally done to explore
model sensitivity and to deal with the problem of
equifinality (elaborated below). Finally, the role
of spatial information is investigated as a possible
way of improving model performance.
Three important remarks should be made at
this point. Firstly, we only discuss calibration of
models that predict water erosion. Wind erosion,
tillage erosion and mass movement models are
not considered, as they involve very different
processes and require different methods of cali-
bration. Secondly, it is impossible to give an
exhaustive account of all calibration attempts.
Instead, examples are given to highlight certain
results. We recognize that this choice is subjec-
tive, but fortunately many different authors agree
on the general state of the art in what we can and
cannot do with erosion models. Thirdly, high-
lighted calibration problems of certain models
are not a judgment on their quality. If anything,
we consider it very positive if authors are honest
about the performance of a model in the given
circumstances and are willing to share less good
results.
3.2 Calibration at Different Scales
In the last 15 years, several exercises have been
held whereby erosion models were compared and
tested in an orchestrated way using common
datasets for calibration and validation. The IGBP-
GCTE Soil Erosion Network and European COST
623 and COST 634 erosion networks have tested
the fitness of erosion models to predict the
consequences of climate change for erosion. The
hillslope and catchment scale comparisons are
briefly mentioned here. For evaluations, six field-
based models and seven catchment-based models
were examined using common datasets from var-
ious countries, split into a 'training set' for cali-
bration and a 'testing set' for validation. Details
about the datasets, the models and model specif-
ics can be found in Boardman and Favis-Mortlock
(1998) and in De Roo (1999). Nearing
et al
. (2005)
compared the responses of seven models in two
catchments (one in Belgium and one in the US)
for climate change, expressed as a change in rain-
fall characteristics and vegetation cover. Apart
from these, many individual model comparisons
have been conducted. In particular, WEPP has
been compared to other models (RUSLE, EPIC
and others) for several environments (e.g. in Chile
by Stolpe, 2005; Peru by Romero
et al
., 2007; India
by Pandey
et al
., 2008a; Norway by Gronsten &
Lundekvam, 2006; Italy by Pieri
et al
., 2007; and
Tunisia by Raclot & Albergel, 2006).
