Geology Reference
In-Depth Information
to have saved 600 million dollars by using the
WEPP model for remediation design (Clark
et al
.,
2006), which was undoubtedly by any estimation
a greater sum than had been spent in the develop-
ment of the model itself. Hundreds of scientific
studies have been conducted over the last two
decades related to applying and improving proc-
ess-based erosion models, including EUROSEM,
WEPP, LISEM (DeRoo
et al
., 1996), the Hairsine-
Rose models, GUEST, and so on, the sum of
which has resulted in a greatly improved under-
standing and quantification of soil erosion and
sediment yields.
The four major conclusions of the Nearing
et al
. (1990) research-needs review paper were
that the future of erosion modelling research
would follow the paths of advancing: '…(i) funda-
mental erosion relationships, (ii) soil and plant
parameters and their effects on erosion, (iii) data-
bases, user interfaces, and conservation system
design, and (iv) model development and analysis.'
They further stated that 'Development of proc-
ess-based erosion prediction technology has
required the delineation and description of funda-
mental erosion processes and their interactions.
Further improvement in prediction technology
will require further delineation and mathemati-
cal descriptions.' These statements generally rep-
resent reductionist science, which was the norm
at least in this area of science at the time. Both
the advantages and limitations of this approach
were discussed by Govers (1996), who concluded
in part stating:
understanding of the limitations of the process-
based soil erosion models, and many of those
limitations centre around variability and uncer-
tainties of many different types. These include
issues related to, but not limited to: (a) natural
variability in rates of soil loss from replicated
soil erosion plots; (b) temporal variability, the
importance of extreme erosion events and vul-
nerable site conditions, and the associated prob-
lems of interpreting short-term data records; (c)
extreme spatial variability of erosion on the land-
scape, our lack of measured data to quantify that
variability, and our limited ability to model the
variability correctly; and (d) the effects of input
data variability on model projections, particu-
larly relative to cumulative model output uncer-
tainty. The reader will find that variability and
uncertainty have been common themes through-
out this topic, which represents a significant shift
in thinking from two decades ago.
An eminent European scientist once stated
that if there were to be a Nobel Prize for Soil
Erosion Science, it would have to go to the study
published by Wendt
et al
. (1986). In that study
the authors reported soil erosion rates for 40
cultivated, fallow, experimental plots located in
Kingdom City, MO, in 1981. All of the 40 plots
were cultivated and in other ways treated iden-
tically. The coefficients of variation between
plots for the erosion rates measured for each of
25 storms ranged from 18% to 91%. Based on
the data from that study, they calculated that
the 95% confidence interval for quantifying the
mean erosion rate of two replicated plots for a
given storm was plus or minus 175% of the
mean value. In other words, the confidence
interval for the mean erosion from the two plots
would range from essentially zero to nearly
twice the mean measured value. Also, the results
of the study indicated that 'only minor amounts
of observed variability could be attributed to
any of several measured plot properties, and plot
differences expressed by the 25 events did not
persist in prior or subsequent runoff and soil
loss observations at the site.' The study was sug-
gesting that replicated plots may give greatly
'The selection of priority subjects for process
studies should be driven by the deficiencies
between model predictions and field observa-
tions: the construction of an alternative, more
sophisticated model to include an additional
effect is only meaningful if a strategy can be
devised which allows a validation of the model
so that its presumed superiority can be proved.'
Many of the realised advances in soil erosion sci-
ence and modelling over the last two decades
have arisen coincidentally with our increased
