Geology Reference
In-Depth Information
predictions of runoff and sediment yield for single
storms with their associated hydrographs and
sediment graphs must be the best. For many
problems, numerical predictions at a storm level
may not be needed; indeed, they may not even be
helpful. Further, the data may not be available to
run such models. As shown by the case studies in
this volume, many problems can be dealt with by
relatively simple models, which may even be
more reliable in their predictions (Jetten
et al
.,
2003). Assessing the nature of the match between
the conceptual framework of the problem and
that of the model is vital for avoiding the problem
of choosing an inappropriate model or, worse,
misusing a model by attempting to use it for
something for which it was not designed.
Sometimes, where there is no obvious match
between the problem and the model, it is neces-
sary to choose the model that is the nearest
approximation to what is required. The more dis-
tant the match, the greater is the likelihood of
misuse and of the model not working. Users then
need to be honest in their evaluation of the model
results. If they are useable, that is a bonus; if they
are not, it is not the fault of the model.
Having selected a model, the user must dem-
onstrate that it is applicable to the conditions of
the study area and to the temporal and spatial
scale of the problem. Ideally, the model should be
validated, at least for the outputs that are going to
be used, with data from the area. Only where this
is not possible should data from similar condi-
tions elsewhere be used. The user needs to gain a
good understanding and experience with the
model, seeking advice from the model developer
and from other users, in order to learn how best
to set the model up to simulate the local
conditions. This may be particularly important
with distributed process-based models where
knowledge of the flow paths taken by runoff and
sediment over the landscape can inform the iden-
tification of the land units (see Chapter 13).
Although the user may operate with single
predicted values as model output, partly because
developing, for example, soil conservation designs
using multiple values is too complicated, the
user should be aware of the likely levels of
uncertainty around, or the accuracy of, that value.
The user should also be aware that there are vari-
ations in the accuracy of predictions depending
upon the value being predicted. Studies with
WEPP have shown that it has the tendency to
overpredict at low values and underpredict at
high values (Brazier
et al
., 2000).
As more research takes place into the design of
soil protection measures, it may be that users
will appreciate that designing to deal with a mean
or average condition is not the most appropriate.
Users may well find they need models that give a
mean and standard deviation, and that designs
need to be made for the event which is either one
or two standard deviations above the mean. Users
should learn to treat the data produced from mod-
els in the same way as they would use data derived
from observation and measurement. They need
to ask the same questions about representative-
ness, reliability, frequency and probability. The
more that users take advantage of what models
can offer, the more they will appreciate that the
thought processes and methodologies of users
and model developers are very similar. This
should aid collaboration. It might also encourage
model developers to provide more information of
value to users, for example, the objective of the
model, the conditions for which it operates, the
level of accuracy required in the input parameter
values, and the errors associated with its predic-
tions. Unfortunately, too few User Guides at
present provide this basic information.
Acronyms
Many models are known by acronyms. Those
mentioned in the text are:
AGNPS (Agricultural Nonpoint Source Pollution
Model)
ANSWERS (Areal Nonpoint Source Watershed
Environment Response Simulation)
CREAMS (Chemicals, Runoff and Erosion from
Agricultural Management Systems)
EUROSEM (European Soil Erosion Model)
GAMES (Guelph model for evaluating the effects
of Agricultural Management systems on
Erosion and Sedimentation)
