Geology Reference
In-Depth Information
the stream power considerably. According to
Govers (1990), the transport capacity equation
derived from stream power is valid for slopes up to
20%, and this might pose one of the largest poten-
tial problems in the application of LISEM to this
area, since erosion and deposition are modelled as
transport deficits and surpluses and are therefore
strongly determined by the flow conditions.
Soil erosion models have so far not paid spe-
cific attention to high sediment concentrations.
For most regions, concentrations in runoff will
not be very high, so that no special attention is
needed. For the Loess Plateau, however, very
high concentrations have been reported regu-
larly. These kinds of flow occupy intermediate
positions between clear water flow and debris
flow, and could have properties that differ
significantly from clear water flow. More
specifically, high sediment concentrations could
change density, viscosity, resistance to flow,
velocity profile and transport capacity (Hessel,
2006). Obviously, erosion models that deal with
high concentrations should at least consider
these effects, and some of the effects can be rela-
tively easily incorporated into erosion models.
Examples are corrections to settling velocity,
fluid density and viscosity, as most existing
equations that relate viscosity and settling veloc-
ity to sediment concentration give fairly similar
results (Hessel, 2006). For other fluid properties,
such as velocity, velocity profile, flow resistance
and transport capacity, the evidence is more
scant and even partly contradictory, so that cor-
rections will be difficult. Care should also be
taken to avoid comparing model results with
measurements without taking into account that
models use clear water values (e.g. g per litre of
clear water), while the raw measurement data
are usually fluid values (e.g. g per litre of fluid).
Gully erosion is a topic that has received more
attention in erosion modelling. Some of the
present-day erosion models do simulate some sort
of gully erosion, but even models that have been
specifically developed to model gully erosion
require that some properties of the gullies be set in
advance (see Chapter 19). Modelling changes in
gully morphology is outside the scope of the cur-
rent research, but sediment yield coming from
gullies needs to be taken into account.
12.5 Method
In 1998, the EROCHINA project started, with the
aim of finding ways to decrease erosion rates in a
small catchment on the Loess Plateau. The project
consisted of two main parts: participatory research
into farm economy, and soil erosion modelling.
The results of the participatory work have been dis-
cussed elsewhere (Messing & Hoang Fagerström,
2001; Hoang Fagerström
et al
., 2003), and this chap-
ter focuses on soil erosion modelling. Soil erosion
modelling consisted of data collection, adaptation
of the LISEM model, calibration of the model, and
simulation of the effect of land use scenarios.
12.5.1 Data collection
As a process-based distributed model, LISEM
needs a large amount of input data. During the
study period (1998-2000) most of the input
parameters needed were measured repeatedly in
the Danangou catchment. Plant and soil charac-
teristics were measured on a fortnightly basis,
except for Manning's
n
, which was measured in
two separate campaigns using small runoff plots
(Hessel
et al
., 2003b). Soil physical characteris-
tics such as saturated hydraulic conductivity, soil
moisture retention curves and the water content-
conductivity relationships were determined using
samples taken in the catchment, for the land uses
shown in Fig. 12.1. All these measurements are
discussed elsewhere (Wu
et al
., 2003; Liu
et al
.,
2003; Stolte
et al
., 2003). The field data were con-
verted to input maps for LISEM using the land-
use map as a base, so that, for a given storm, these
variables were constant within a land use, but dif-
fered between land uses. For variables that clearly
also depend on soil type (e.g. cohesion), a combi-
nation of land use and soil type was used to
extrapolate the measurements. Initial moisture
content was predicted with multiple regression
equations based on aspect and slope, and was
therefore spatially variable. The resulting mois-
ture contents were yearly averages, but these
