Erosion and Sediment Transport: Finding Simplicity in a Complicated Erosion Model - Environmental Modelling: Finding Simplicity in Complexity - page 254

at the Kent site (as might be provided by plot stud-
ies, for instance), it is hard to pinpoint which hillslope
combinations within the model are overpredicting or
underpredicting.
Average observed erosion rates more typically fall
towards the lower end of the distributions of hillslope
erosion predicted by WEPP within MIRSED for all hill-
slopes at each site. The lack of explicit consideration of soil
loss from interrill areas by the Evans (1988) observations
may to some extent explain this disparity, as significant
contributions to soil loss and the consequent surface low-
ering of the hillslope can result from interrill processes.
Thus, magnitudes of soil erosion from the study sites may
well be underestimated by the methods of Evans (1988),
or the results may point to deficiencies in the model
representation of the sites.
It has been shown that, at best, absolute predictions
from MIRSED can only bracket the observed data. How-
ever, this limitation is perhaps not surprising given the
temporal and spatial disparities with the observations dis-
cussed above. Improvement of predictions with the use
of slope distributions within a grid cell is not significant
in terms of average results, though better prediction of
extreme observations from the rill survey methods may
be made.
It is not always the sites with the highest maximum
erosion rates that have the most significant soil erosion.
Attention to the extent of the contributing area of erosion
over critical thresholds must also be made in order to
quantify fully which areas within a site, or which sites
within a region, have the worst erosion 'problem'.

erosion. It is shown that hillslopes within catchments,
regions or ultimately the whole of the UK can bemodelled
to identify the potential high erosion risk sites, or those
combinations of soil, slope and land-use that should be
avoided in the future. These results, though complex in
their origin, can be easily mapped at the 1-km 2 grid-cell
scale, which will enable the end user to visualize sensitive
areas. It is recognized, however, that MIRSED results
must be treated as hillslope results and not a map of
total erosion or sediment yield from an area, a task that
is beyond the scope of this modelling procedure.
Further improvements inWEPPor theMIRSEDframe-
work can be incorporated with reruns of the model to
provide an updated matrix of output values. It is hoped
that any model improvements will also be accompanied
by improvements in both the quality and the quantity
of observed data that are appropriate to evaluate model
predictions.
High-resolution, hillslope-scale data, collected across
England and Wales, are now needed to describe onsite
soil erosion rates at a national scale. Such a dataset would
enable a more meaningful evaluation of this type of
modelling exercise and would also enable model uncer-
tainty and error propagation through the WEPP model
to be assessed, with the ultimate goal of mapping uncer-
tainty (see Chapters 2 and 28), which accompanies model
results, and therefore providing an explicit consideration
of the true accuracy of this approach, at the national scale.

References

Boardman, J. (1990) Soil erosion on the South Downs: a review, in
Soil Erosion on Agricultural Land (eds J. Boardman, I.D.L. Fos-
ter and J.A. Dearing), John Wiley &, Sons, Ltd, Chichester,
pp. 87-105.
Boardman, J. (1993) The sensitivity of downland arable land to ero-
sion by water, in Landscape Sensitivity (eds D.S.G. Thomas and
R.J. Allison), John Wiley & Sons, Ltd, Chichester, pp. 211-28.
Boardman, J. and Evans, R. (1994) Soil erosion in Britain: a
review, in Conserving Soil Resources: European Perspectives , (ed.
R.J. Rickson), CAB International, Wallingford.
Boardman, J. and Favis-Mortlock, D.T. (1993) Simple methods
of characterising erosive rainfall with reference to the South
Downs, southern England, in Farm Land Erosion in Temper-
ate Plains Environment and Hills (ed. S. Wicherek), Elsevier,
Amsterdam.
Brazier, R.E., Beven, K.J., Anthony, S. et al . (2001b) Implications
of complex model uncertainty for the mapping of hillslope scale
soil erosion predictions. Earth Surface Processes and Landforms ,
26 , 1333-52.
Brazier, R.E., Beven, K.J., Freer, J. and Rowan, J.S. (2000) Equifi-
nality and Uncertainty in physically-based soil erosion models:

15.9 Conclusions

The potential of a simplified versionof a complicated, soil-
erosionmodel to predict andmap hillslope soil erosion for
a wide variety of scenarios throughout England andWales
has been demonstrated. Given the constraints of the avail-
able observed data, MIRSED performs reasonably well in
predicting both average and maximum erosion rates at
each of the 11 sites studied. Predicted rates of erosion gen-
erally agree with those observed from the existing surveys
and data describing extent of erosion in excess of pre-
determined 'tolerance' levels provide a useful measure of
the significance of the erosion problem in an area.
Comparison of different sites is made, based on
available observations, though equally MIRSED can be
employed as the basis for assessment of areas within a
catchment or region that are sensitive to hillslope soil