Geology Reference
In-Depth Information
model outputs are sensitive to soil properties, it
has to be recognized that the model predictions
will also have an error of similar magnitude asso-
ciated with them.
The situation would seem to be better with
respect to data on slope and land cover because
these can be obtained from several sources, rang-
ing from field measurement to the use of aerial
photography and satellite imagery. Height survey
data and aerial photographs can be used to pro-
duce digital elevation models from which slope
information can be derived. Unfortunately, few
models can use this information direct. Crop or
land cover usually needs to be expressed as per-
centage canopy cover or ground cover, or both,
alongside parameters such as plant height. Even
where the model requires a single value such as
the
C
-factor in the USLE, this, as seen earlier, is
derived from more detailed information linking
crop cover and height to rainfall. Whilst slope
steepness data can be used directly, slope length
data cannot. Some models require a measure of
the actual length of a slope segment, but others
need only the length over which runoff occurs.
Since the latter is highly dynamic, changing both
between and within storms, an average or typical
value usually needs to be chosen based on local
experience.
The more detailed process-based models like
WEPP and EUROSEM are more data-hungry than
grey-box models like the USLE or MMF. Such
models require data on, for example, inter-rill
erodibility, rill erodibility, soil shear strength,
soil cohesion, surface roughness, Manning's
n
(a roughness coefficient affecting flow velocity)
and plant density. In recognition of the fact that
these data do not generally exist and that many
model users have neither the time nor the
resources to obtain the data from measurement
or experiment, model developers have produced
various tables containing guide values for differ-
ent conditions. Whilst using these often works
reasonably well, there is little information avail-
able on the errors associated with or arising from
using these guide values. Also, their use makes
an assumption that the conditions at the field site
fall within the range of the values given, and are
not abnormal or extreme. Experience with
the MMF model indicates that using locally-
measured values of the input parameters rather
than the guide values yields better results
(Morgan
et al
., 1984).
In addition to data required as model inputs,
data are also needed to compare model outputs
with observed values. Only by such comparisons
can models be demonstrated as successful or oth-
erwise. The field data need to be appropriate to the
scale and objectives of the model. Unfortunately,
these requirements are often ignored, usually
because the data are not available. As an example,
the objective of many models and model users is to
predict how much sediment is removed from small
catchments, ranging in size from a few hectares to
one or two square kilometres. This is exactly the
scale at which measured data are deficient. Lots of
data exist for erosion plots, particularly those of
40 m
2
, and for larger river systems. Thus models
like WEPP and EUROSEM are tested on their abil-
ity to predict soil loss from small plots, although
they are designed to operate on small catchments.
Some long-term data on erosion rates in the land-
scape are available for small catchments based on
studies of the spatial distribution of caesium-137
(Ritchie & Ritchie, 2001) and other radioactive
tracers. Process-based models can be run as con-
tinuous simulation models for hundreds of years,
and the output averaged to compare with erosion
rates obtained from tracer studies. However, even
if the comparisons are good, they are unreliable
because the models generally only consider sedi-
ment movement by rainsplash, unconcentrated
overland flow and rills, whereas, in reality, assess-
ments based on tracers incorporate the effects of
all processes operating in the landscape including
translocation of soils by tillage, gully erosion, wind
erosion, mudflows and landslides.
Surprisingly few studies assess the ability of
models to predict the timing and location of ero-
sion, rather than the quantity. Yet field data on
time and place are usually readily available. Aerial
photography and field observations will identify
where, in the landscape, rills and sediment deposi-
tion occur, and talks with the local population,
especially farmers, will usually provide details of
