Geology Reference
In-Depth Information
applications that will lead to erroneous results
because there are fundamental reasons why the
model is not applicable to the conditions/envi-
ronment under study. Thereafter we will give
attention to other misconceptions about erosion
models that may not necessarily lead to errone-
ous model applications. Such misconceptions
may, for instance, hamper a correct interpreta-
tion of application results and/or may have
important implications in directing future
research on soil erosion modelling.
rates: the model is conceived to represent an 'aver-
age annual' situation based upon average annual
values for the temporally varying input parame-
ters (most notably rainfall erosivity, soil erodibil-
ity and soil cover). Using such a model to simulate
what will happen over a short time period or dur-
ing a single event will inevitably lead to poor
results if the system under consideration is char-
acterized by a high temporal variability.
Soil erosion rates do indeed vary significantly
over short time-scales, even if all controlling
factors except climate (crop type, tillage tech-
niques, soil type, etc.) are kept as constant as pos-
sible. Figure 7.1 shows the variation of mean soil
erosion rates as measured on three plots under
continuous corn at the USDA-ARS research sta-
tion of Clarinda, illustrating the enormous varia-
bility in annual erosion rates. This variability
results in a large uncertainty in average erosion
rates at a given location, even when measure-
ments are maintained for a long period of time.
Coefficients of variation of the mean value decline
somewhat with average soil losses but are gener-
ally between 0.2 and 0.5, indicating that, even
when measurements are maintained for as long
as 10-27 years, the 90% confidence interval on
the mean is between 40 and 100% of the meas-
ured average value (Fig. 7.2).
Evidently, this variation is partly due to yearly
variations in climatic conditions at the station's
location, and this variation might be accounted
for by using annual rainfall erosivity values rather
than an overall average annual value. However,
climatic variability is only one component of
overall variability: variability is also due to varia-
tions in plot conditions that cannot be accounted
for by the factors included in the original USLE,
and one therefore cannot expect that this varia-
tion can be simulated or predicted using a model
that was designed to predict correctly average
values only, such as the USLE.
The implications of this simple observation
are perhaps more profound than we usually think.
First of all it means that a meaningful calibration
of a model of the USLE-type using field data
requires the collection of erosion rates under
standardized conditions over a sufficiently long
7.2
Misapplications of Soil Erosion Models
The issue of misapplications of erosion models
can, in principle, be dealt with at a very concep-
tual level: models can only work when they are
applied to conditions for which they have been
calibrated and, if possible, validated. The term
'conditions' has, in this context, a very broad
meaning: the model should be applied at the cor-
rect temporal and spatial scales, and care should
be taken to ensure that the model accounts for
the erosion processes under consideration. Only
model applications where these conditions are
met can be considered to be scientifically valid: if
these conditions are not met, model applications
may lead to the wrong conclusions about the
model's performance, the erosion risk in a certain
area and/or the efficiency of erosion control meas-
ures (e.g. Nyssen
et al
., 2006).
7.2.1 The time-scale
The classic example of model misapplication due
to an inappropriate time scale is the use of the
USLE at time-scales for which this statistical
model was not intended. In their original paper,
Wischmeier and Smith (1965) clearly stated that a
measurement period of over 20 years was neces-
sary in order to establish a true mean soil erosion
rate that could be compared with the value pre-
dicted by the USLE. As discussed by Wischmeier
and Smith, one of the basic reasons for this is that
statistically based models such as the USLE are
not capable of accounting for the considerable
temporal variation that occurs in soil erosion
