Geology Reference
In-Depth Information
actual flow distance, had great effects. After rec-
alibration, the adapted model performed slightly
better than the original version.
that one should be cautious when applying the
LISEM model to predict runoff for future events.
Such predictions might be possible when these
events are similar in size to the ones used here.
Even then, initial conditions might well be differ-
ent, so that it would probably be necessary to do
simulations with different initial moisture con-
tents. At present LISEM might be more suited to
evaluate certain land use and management sce-
narios for their effects on erosion, since in that
case all scenarios will use the same rainfall data
and the same initial conditions.
Rill erosion intensity was mapped in the field
and compared with LISEM simulations of erosion
distribution. This comparison shows that the
general appearance of simulated and mapped ero-
sion patterns is similar, but also that the patterns
are very different in detail. Many explanations
are possible for the discrepancies between
observed and simulated erosion rates and pat-
terns inside the catchment (see e.g. Takken
et al
.
1999). For the topographically complex Danangou
catchment, the following factors are likely to be
important:
●
DEM inaccuracy. The DEM determines the flow
direction in the model, which, in turn, determines
where erosion will occur according to the model.
On relatively flat areas, the tillage direction can
determine the direction of water flow (Ludwig
et
al
., 1996; Takken
et al
., 1999; Van Dijck, 2000), but
also in steep terrain, the flow direction is influ-
enced by small topographic features (like path-
ways, gullies, cut-off drains, local escarpments)
which are smaller than the pixel size, and there-
fore cannot be captured by the DEM. For topo-
graphically complex areas such as the Danangou
catchment, it seems questionable whether it would
be possible to obtain a DEM with sufficient detail
to extract flow directions accurately.
●
Limitations of LISEM. The pixel-based approach
used in LISEM has some consequences for the cal-
culation process. The most important is probably
related to inertia. All processes are calculated per
pixel, and only the kinematic wave distributes
water and sediment. There is no momentum trans-
fer between cells. Thus a change in slope may cause
abrupt changes in streampower and a large amount
12.7.2 Calibration
Calibration of LISEM showed that the model can
simulate runoff and soil loss from the catchment
for each event separately. However, it was not
possible to find a general calibration set that could
be used for all measured events. Validation indi-
cated that it might be possible to find calibra-
tion sets for low-magnitude and high-magnitude
events when data for more storms become avail-
able. LISEM gave the worst results for small
events with low average rainfall intensity. Several
authors (e.g. Nearing, 1998; Jetten
et al
., 2003)
also found that erosion models have difficulty in
predicting small events. They ascribed this to spa-
tial variability and uncertainty in the input data.
The problem with small storms in the Danangou
catchment is probably caused by several factors:
(1)
For smaller storms it is much more important
to get the initial conditions right. For large storms,
a smaller or larger initial loss of water to infiltra-
tion probably does not matter much for the total
amount of runoff, while for small storms it might
be a large percentage of total runoff.
(2)
Spatial distribution of rain in the catchment,
such that rainfall measured at the gauges might
not represent the actual spatial distribution of
rainfall, and maximum rainfall intensities might
not have been measured.
(3)
If runoff was caused by heavy rainfall in part
of the catchment, it is also crucial to get the other
input data for this region right. For the events of
980823 and 000811, for example, heavy rain
occurred in the gullied northern part of the catch-
ment. Most LISEM parameters were not measured
there, but were extrapolated from elsewhere.
(4)
Finally, for smaller storms, storage in pools
will have a larger effect.
These results indicate that the usefulness of
LISEM for predictive purposes is limited to events
that are large enough to cover the entire catch-
ment. Our finding that a separate calibration is
necessary for small and for large events means
