Geology Reference
In-Depth Information
gullies start are more controlled by slope gradi-
ent, while the presence of concavities control the
trajectory of the gullies until the slope gradient is
too low and (coarse) sediment deposition domi-
nates. Such an approach can be improved by
incorporating the presence of linear landscape
elements, soil surface state, vegetation cover and
root density, and possibly rain, to the input
parameters. Souchère
et al
. (2003) presented an
expert-based model for predicting the location
and the volumes of ephemeral gullies.
Even with detailed topographic data, it is
sometimes hard to predict where exactly the
water will concentrate. Anthropogenically-
induced linear roughness elements (e.g. drill and
plough furrows, headlands) in landscapes with
gentle slopes might make this even more difficult
by altering the true catchment area (e.g. Souchère
et al
., 1998; Takken
et al
., 2001). Water often con-
centrates in furrows created by tillage or along
rural roads, which are difficult - if not impossible -
to include in topographical maps. Even with
topographic maps of high resolution, a simple
pass of agricultural equipment could disturb the
existing drainage structure of a field.
Many studies investigating critical
S
-
A
thresh-
olds for incipient gullying (see Table 19.1) have
noted the large scatter of the observation data
points. This scatter has been attributed to spatial
variations in erosional or hydrological processes
(Dietrich
et al
., 1993; Prosser & Dietrich, 1995;
Prosser & Abernethy, 1996) or land-use pattern
changes (Prosser & Soufi, 1998; Desmet
et al
.,
1999; Vandekerckhove
et al
., 2000). These studies
typically classify about 70-80% of the channel
heads correctly. Istanbulluoglu
et al
. (2002) for-
malized the description of this variability by
interpreting the topographic threshold
C
(
AS
a
advantage of the probabilistic model of
Istanbulluoglu
et al
. (2002) over a single initiation
threshold is that the latter will predict significant
erosion only in locations where channelization is
predicted on a long-term basis. The probabilistic
model then provides a way to account for the less
frequent contribution to erosion due to channeli-
zation even at locations that do not meet the sin-
gle channelization threshold. Comparing their
resulting stochastic model with field data from
gullies in the Idaho Batholith, they concluded
that a large part of the observed variation could
be attributed to grain size differences.
While of great practical use, estimating gully
location using topographical conditions alone has
its limitations, given the variety of other environ-
mental variables that control gully initiation and
development. Therefore, this simple approach
might result in high prediction errors in some
cases. Vandekerckhove
et al
. (1998), for example,
applied this topographic threshold concept to
different landscapes in Spain and Portugal, and
their results revealed that
S
and
A
were weakly cor-
related with gullying. The prediction of gully ero-
sion was considerably strengthened by including
additional information on land use, soil stoni-
ness, and soil horizon hydraulic conductivity.
Knapen and Poesen (2010) demonstrated that
ephemeral gully initiation points and dimensions
are not only topographically controlled but also
depend upon the erosion resistance of the topsoil.
Various erosion models use this information in
their water flow routing routines. Therefore,
Chaplot
et al
. (2005) tested to what extent the
direct flow velocity estimations, as obtained from
an existing surface water routing algorithm, could
be used to predict the location of what they called
'linear erosion elements', which included rills
and small ephemeral gullies. By defining a critical
velocity threshold of 0.062 m s
−1
, the location and
extent of linear erosion elements corresponded
relatively well with observations in a 0.62 km
2
watershed in Laos.
Statistical techniques have been applied with
respect to the mapping of gully location. Hughes
et al
. (2001), for example, applied a regression tree
model to the prediction of gully density on a
C
)
as a random variable. Its probability distribution
is then derived physically from the random vari-
ability of quantities involved in the erosion proc-
ess. They considered median grain size, roughness,
and excess rainfall. This resulted in a probabilistic
channel incision zone that shifted according to
the variability in these input factors. The result-
ing probability distribution of the threshold
C
was shown to follow a gamma distribution. The
=
