Geology Reference
In-Depth Information
of deposition in a pixel can occur. The downstream
pixel then has a lot of detachment because TC C is
large, resulting in alternating detachment-deposi-
tion patterns. This problem is generally more pro-
nounced where there is more water, hence in the
channels, and has larger influence on the distribu-
tion of erosion than on total sediment yield. It is
also more pronounced for topographically complex
areas with strong alternations in slope angle, and
can therefore be only partly solved by changing
computational procedures.
Incomplete or incorrect process descriptions.
The issue of steep slopes, for example, was dis-
cussed above. All currently available erosion
models use sediment transport equations that
have been developed for slopes of no more than
20%. Much steeper slopes are, however, very
common in the Danangou catchment. Similarly,
other characteristics of the catchment might also
be outside the range of conditions for which
equations that are being used were developed.
Besides, process knowledge is usually incom-
plete. When simulation errors appear to be sys-
tematic, there might be incomplete process
descriptions in LISEM. For example, the fact that
the discharge peak always seems to arrive too
early might indicate this. Incomplete and incor-
rect process descriptions might affect both distri-
bution and amount of erosion. As Beven (2001)
points out, such errors in theory might be masked
by calibration and can therefore be hard to find.
Data inaccuracy. There can be inaccuracies in
the data used to evaluate model performance as
well as in the input data for the model. Morgan
and Quinton (2001) suggested that such inaccura-
cies are a more important cause of incorrect model
predictions than model flaws. Inaccuracies in
input data can be caused by several factors. The
first is incorrect measurements. Inaccuracies can
also be caused by non-representative measure-
ments. Input data for the LISEM model were
collected on fields that were supposed to be repre-
sentative of their respective land uses. Finally,
some parameters (e.g. soil moisture content) are
liable to rapid fluctuations, while others (e.g. sat-
urated conductivity) are notoriously heterogene-
ous in space. Simulation results also indicate that
the rainfall distribution has a large influence on
simulated runoff and erosion. Data inaccuracy is a
fundamental problem with distributed modelling.
Nowadays, distributed models can contain sev-
eral tens of thousands of pixels for which the dif-
ferent calculations are performed. Data on
saturated conductivity, soil roughness, plant char-
acteristics, and so on, are needed for all these pix-
els. Furthermore, even if such data were available,
one can for reasons of spatial variability and
upscaling seriously question whether the used
values are indeed representative for the given
pixel (see Chapter 6 for a general discussion related
to upscaling). It seems likely that the actual
amount of runoff and erosion occurring is control-
led by many variations in parameters operating on
subgrid scale. One has to face reality: there will
always be a lack of accurate input data.
Often, a combination of factors could be operat-
ing, so that it will be difficult to find out what
exactly causes an observed discrepancy between
simulation and measurement. To evaluate the
LISEM model (or any other process-based, dis-
tributed erosion model) in a spatial way, very
detailed data both on model input and on ero-
sion and deposition patterns distribution are
needed. Such datasets are very hard to obtain for
catchment-size areas, especially when topogra-
phy is complex. Data for the catchment outlet
are easier to obtain, so that calibration on the
outlet alone will often be the only possibility.
Our data, however, confirm the findings of
Takken et al . (1999) that an erosion model cali-
brated on outlet data might well predict spatial
patterns incorrectly.
12.7.3 Scenarios
The present study is one of the first attempts to
use process-based soil erosion modelling as a tool
for optimizing land use and management strate-
gies to reduce runoff and erosion rates on the
Chinese Loess Plateau. To perform the simula-
tions, several assumptions were made.
The first is that LISEM can be used for scenario
simulations. Such applicability might be threa-
tened by the need to use different calibrations for
Search WWH ::




Custom Search