Geology Reference
In-Depth Information
will necessarily hold when the model is applied
to a specific location. A parameter which overall
is rated as moderately sensitive in the general
analysis may be highly sensitive in a specific situ-
ation, either because there is little or no variation
in the values of the parameters which are generi-
cally the most sensitive, or because the values of
another parameter are within the range that
makes the first parameter sensitive. It is recom-
mended that the user undertakes a sensitivity
analysis of the chosen model with local data to
test that the parameter sensitivity matches what
is known about the local situation.
averages for their data are likely to be less affected.
Further, it is sometimes possible to obtain addi-
tional local information from records made by
farmers and other land-holders, or at schools and
research stations.
The situation is less satisfactory with respect
to data on soils. Some national soil surveys map
soil units at a soil series level but, more often,
they show only generalized categories such as
soil associations or even soil orders or groups.
Published data on soil properties that accompany
the maps usually cover grain-size distributions,
organic content and soil moisture at field capac-
ity, but are unlikely to contain information on
soil strength, erodibility, detachability or infiltra-
tion capacity. These properties need to be mea-
sured on site in the field or estimated from
statistical relationships between the property
concerned and already-measured properties. The
nomograph (Wischmeier
et al
., 1971), used to
estimate soil erodibility (
K
value) for the USLE
from measurements of grain-size distribution,
organic content, permeability and structure, is an
example of such a pedotransfer function. The
confidence levels of these pedotransfer functions
are often unknown. A further issue is that in
most models, the properties of each soil unit are
described by a single value. Yet, as already noted
above, soil erodibility is not static but changes
over time; similarly, whilst primary particle size
may be considered static over short to medium
time-scales, actual particle size in terms of soil
aggregates will vary. Spatial variability in some
soil properties is considerable. For example, infil-
tration capacity is often measured in the field at a
number of points and the values averaged to give
a single number for use in modelling; yet the
spatial variability in that number within a soil
series unit can be as high as 120% (Eyles, 1967).
Furthermore, on crust-prone soils, the mean
value around which this variation occurs can fall
by 50 to 100% in a single rainstorm (Hoogmoed
& Stroosnijder, 1984; Torri
et al
., 1999). It is clear
from all of this that input data used in models to
express soil effects are subject to considerable
error, the extent of which is usually not known.
However, if the error is as much as 100% and the
2.8.2 Data availability
A problem frequently faced by model users is that
insufficient data exist to run many of the models.
This may limit their ability to select what would
otherwise be the most appropriate model. In some
instances there is no alternative but to choose
another model which is less demanding in its
data input, even though it may not meet
all the
user's requirements. Failure to appreciate the
issues associated with data availability can lead
to misconceptions about how well a model is
likely to perform (see Chapter 7). Generally, mod-
els which require information on a time-averaged
or spatial-averaged basis are the easiest to satisfy
with respect to input data. For example, annual
or mean annual rainfall and mean annual tem-
perature data exist for most areas of the world.
Even though the nearest meteorological record-
ing station may be several kilometres away from
the field area, it is possible to use the data with
some degree of confidence. The devices used to
record rainfall and temperature have high levels
of accuracy, so that the main sources of error
relate to how representative the meteorological
station is of the local site. This will depend on
how variable local climate is with respect to the
terrain. Variability will be greatest in mountain-
ous areas where rain-shadow effects and the dif-
ferences between sunny and shady slopes are
more pronounced. Such differences are most
obvious with respect to individual storms and
short time periods, so models that rely on annual
