Geology Reference
In-Depth Information
greater exposure to raindrop impact. It is possible
that the presence of sheep on the field during the
simulation period may have significantly changed
the soil structure between the two events.
However, it is worth noting that none of the cur-
rent generation hydrological models is fit for sim-
ulating the transient nature of some of the
hydraulic processes at the field scale and beyond.
Another area which warrants further discus-
sion is the difference in identified slope length for
both events. Slope length controls the catchment
area and therefore the volume of water falling
on the catchment, as well as influencing the
sediment concentration (Smith
et al
., 1995). As
EUROSEM does not simulate the flows of water
in the subsurface, soil hydraulic properties across
the field are uniform until water begins to flow,
so runoff is generated at all points across the ele-
ment at the same point in time. It is likely in the
simulated storms that we have variable source
areas which cannot be simulated internally by
EUROSEM, and that by using different slope
lengths we are able to provide a representation of
this. For event 2, the identified
XL
parameter
region around 400 m (Fig. 5.6), together with the
assumed field width of 125 m, matches the actual
area of the simulated field of 5.44 ha (Fig. 5.3). In
contrast, for event 1, the identified
XL
in excess
of 800 m (Fig. 5.6) yields a contributing area
beyond the 6.81 ha which could be considered an
extended contributing area when large storms
may result in an overflow of an otherwise discon-
nected area upslope (Fig. 5.3). However, by com-
promising model performance, it is possible to
match this extended area with a slope length of
around 550 m, which is still below the Q perform-
ance threshold of
-
||
In contrast to the reasonable hydrological simu-
lations (at least for individual events and with a
high level of model flexibility in adjusting slope
length), EUROSEM failed to simulate the meas-
ured sediment concentration data within an order
of magnitude of the observed data. One possible
explanation for the model's difficulties may lie
with the processes operating at Den Brook.
Although runoff volumes in the catchment are
comparable with those measured at other UK sites,
sediment concentrations observed for the events
simulated in this study are at the lower end of sed-
iment concentrations measured from similar field
areas in the UK. In work on unbounded hillsides at
three sites in the UK, Deasy
et al
. (2008) reported
mean annual sediment concentrations of between
132 and 7226 mg l
−1
. The lower concentrations of
this range were associated with clay soils under
arable agriculture, but on steeper slopes than those
at Den Brook. It is possible that EUROSEM does
not simulate low-energy erosion processes well,
or that processes are occurring, such as the mobi-
lization of soil particles by physio-chemical dis-
persion, that EUROSEM does not simulate. If this
were the case it would be right to question whether
applying EUROSEM to Den Brook represents a
fair test of the model, and it might be concluded
that it was not. However, it does illustrate how
physically-based models force us to think criti-
cally about the system that we are trying to simu-
late and suggests that if we wish to simulate
erosion in areas such as Den Brook, we will need
to develop new descriptions of erosion processes
that are better suited to such environments.
Whilst this chapter was not principally about
model uncertainty, it does illustrate that a large
number of different parameter sets, from often dis-
tinctly different regions in the sampled parameter
space, are able to simulate the observed hydrolo-
gical response, but not the observed sediment
dynamics. The authors therefore echo the call of
previous authors (Beven & Binley, 1992; Quinton,
1997; Brazier
et al
., 2000) for erosion modellers to
adopt uncertainty representation techniques,
which are now prevalent in hydrological model-
ling. Without exploring the model simulations in
detail, by sampling the possible parameter space,
0.4 (Fig 5.6). It was valuable
for this study to allow the slope length to vary in
order to analyse the site characteristic of a poten-
tially highly variable contributing area. However,
we do not recommend using slope length as a
means for compensating for model structural
error without carefully considering the field situ-
ation (see Chapter 13 for a further example of the
importance of using field evidence for determin-
ing the effective contributing area when parame-
terizing erosion models).
=
