Beruflich Dokumente
Kultur Dokumente
Chapter 10
Splash and wash processes on soil in experimental
trays exposed to natural rainfall
Abstract: The complex interactions between rainfall-driven erosion processes and rainfall
characteristics, slope gradient, soil treatment and soil surface processes are not very well
understood. A combination of experiments under natural rainfall and a consistent physical theory
for their interpretation is needed to shed more light on the underlying processes. The present
study demonstrates such a methodology. An experimental device employed earlier in laboratory
studies was used to measure downslope rain splash and splash-creep, lateral splash, upslope
splash, and rainfall-driven runoff transport (wash) from a highly aggregated clay-rich oxisol
exposed to natural rainfall in West Java, Indonesia. Two series of measurements were made: the
first with the soil surface at angles of 0-, 5-, 15- and 40-; and the second all at an angle of 5- but
with different tillage and mulching treatments. A number of rainfall erosivity indices were
calculated from rainfall intensity measurements and compared with measured transport
components. Overall storm kinetic energy correlated reasonably well to sediment transport, but
much better agreement was obtained when a threshold rainfall intensity (20 mm h
-1
) was
introduced. Rain splash transport measurements were interpreted using a recently developed
theory relating detachment to sediment transport. Furthermore, a conceptually sound yet simple
wash transport model is advanced that satisfactorily predicted observed washed sediment
concentrations. The lack of replication precluded rigorous assessment of the effect of slope and
soil treatment on erosion processes, but some general conclusions could still be drawn. The
results stress the importance of experiments under conditions of natural rainfall.
Parts of this chapter are published as: Van Dijk, A.I.J.M., Bruijnzeel, L.A. and Eisma, S.E. A
methodology to study rain splash and wash processes under natural rainfall. Hydrological
Processes 16 (in press)
10.1. Introduction
Rain splash - soil detachment and transport by impacting rain drops - is an important
first step in soil erosion. Unconcentrated (sheet) runoff usually does not have enough
power to actively detach and entrain soil particles (Rose, 1993), but particles detached by
rain splash may subsequently be transported by the flow (Hudson, 1995; Kinnell, 1990).
On short steep slopes (e.g. bench terrace risers) rain splash may be the dominant
transport mechanism (Chapter 9 and 13). Physical understanding of the processes causing
A.I.J.M. van Dijk (2002) Water and Sediment Dynamics in Bench-terraced Agricultural
Steeplands in West Java, Indonesia. PhD Thesis, Vrije Universiteit Amsterdam
CHAPTER 10 - SPLASH AND WASH UNDER NATURAL RAINFALL
178
detachment and transport by falling rain drops has improved considerably over the last
two decades, mainly by laboratory experiments (Poesen and Savat, 1981; Ghadiri and
Payne, 1986, 1988; Poesen and Torri, 1988; Sharma and Gupta, 1989; Salles and Poesen,
2000).
Unfortunately, the results of laboratory studies are not readily translated to field
situations. For example, rainfall kinetic energy has been related to rain splash in both
natural and laboratory studies (Kneale, 1982; Salles and Poesen, 2000), but the manner in
which kinetic energy was varied and estimated differs between the two experiment types.
Changes in simulated rainfall kinetic energy are normally achieved by using (uniform)
drops of different size or fall height, while rainfall intensity is often high (Free, 1960;
Quansah, 1981; Poesen, 1985). By contrast, natural rainfall kinetic energy depends
primarily on the rain drop size distribution (although moderated by wind) and in field
studies is often not measured but estimated using empirical relationships with rainfall
intensity (see Chapter 7 for a recent review). A comparable difference relates to the
apparent existence of a rainfall kinetic energy threshold observed in both laboratory and
field experiments. Individual water drops have to overcome a critical kinetic energy
threshold before they are actually able to dislodge and transport soil particles (Sharma
and Gupta, 1989; Salles et al., 2000). At first sight this appears to agree well with
findings that use of a threshold intensity (or corresponding kinetic energy) correlates
better to sediment transport measurements than does gross kinetic energy (e.g. Hudson,
1965; see Section 10.4.1). The moments of natural rain drop size distributions increase
with rainfall intensity in a gradual manner, however (despite considerable temporal
variation, see Chapter 7). No sudden increase in the proportion of drops with higher than
threshold kinetic energy occurs, therefore, and even very low intensity rainfall will
contain some erosive rain drops. This renders the existence of a threshold kinetic energy
for natural rainfall difficult to explain in physical terms. Furthermore, laboratory
experiments by Riezebos and Epema (1985) indicated that the average distance over
which particles are splashed depends on the height from which artificial drops fell. The
amounts of splash transport measured in most rain splash experiments depend on both
detachment (expressed on an area basis) and splash distance (Chapter 8). As drop fall
heights in the laboratory are different from natural rainfall, this further complicates
extrapolation of laboratory experiments.
A similar case can be made with regard to rainfall-driven sediment transport by
runoff, denoted here by the term wash. Laboratory experiments have greatly improved
the physical understanding and mathematical description of wash transport (Moss, 1988;
Kinnell, 1990; Proffitt and Rose, 1991; Heilig et al., 2001), but most involved artificial
rainfall at a constant rate and uniform soil material. For obvious reasons, natural rainfall
and field soils present added complexity whereas, in addition, experimental boundary
conditions (rainfall rate, water inflow or outflow rates, flow depth and velocity) can no
longer be controlled outside the laboratory and, in fact, become closely related.
The problems outlined above clearly indicate that to understand rainfall-driven
erosion processes experiments under natural rainfall remain indispensable. The current
study demonstrates a methodology to study rain splash and wash under such conditions.
This was done using experimental soil trays modified slightly from the design used by
Wan et al. (1996) in laboratory studies. The soil trays were filled with a highly
aggregated clay-soil exposed to the natural rainfall regime prevailing in West-Java,
Indonesia, during two consecutive periods covering 17 and 15 storms, respectively. After
each event, different components of sediment transport were measured separately.
During the first period, the devices were placed at four different angles to investigate the
CHAPTER 10 - SPLASH AND WASH UNDER NATURAL RAINFALL
179
effects of slope on splash on wash processes, whereas during the second period, the
effect of mulch cover and frequent tillage were studied.
Various rainfall erosivity indices proposed in literature were tested as predictors of
splash and wash transport for this specific combination of soil and rainfall regime. A
mathematical theory relating splash transport to detachment (Chapter 8) was employed to
interpret splash measurements. Furthermore, a simple conceptual model is proposed to
interpret wash measurements. Given the high variability of erosion processes expected in
(semi-) natural conditions, the lack of replication for the various treatments precluded a
rigorous statistical assessment of the effects of different slope gradients and treatments.
Some interesting observations could still be made, however, and an assessment of the
usefulness of the splash and wash model theory was possible.
10.2. Materials and methods
10.2.1. Site and soil
Experiments were conducted within the framework of the Cikumutuk Hydrology and
Erosion Research Project (CHERP). This project studies erosion dynamics in the small
(125 ha) but steep agricultural Cikumutuk catchment, situated about 60 km East of
Bandung in the volcanic uplands of West Java. All experiments were carried out between
19 February and 9 April 1999 near the project laboratory (7-03S; 108-04W) at an
altitude of 580 m a.s.l.. The area receives an annual rainfall of ca. 2650 mm, with a drier
season (average monthly rainfall less than 60 mm) generally extending from July until
September.
The soil used was a Red Latosol according to the Indonesian Soil Classification
System (Soeporaptohardjo, 1961) or a Haplorthox according to USDA nomenclature,
which had developed in Pleistocene and Holocene andesitic tuffs with kaolinite as the
dominant clay mineral (P.P. Hehuwat, pers. comm.) and a median dispersed particle size
of 1.2 m. Material used in the experiments was collected from the top 10 cm of the beds
of reverse sloping bench terraces. Soil texture was silty clay with very little sand (ca.
70% clay, 30% silt). The soil was highly aggregated with crumbly sub-angular
aggregates sometimes attaining a diameter of more than 8 mm and a median dry-sieved
size of 3.2 mm; median wet-sieved aggregate size was ca. 1.1 mm. More details about the
site and soil properties are given in Chapter 3.
10.2.2. Experimental design
A combined splash and runoff collecting system slightly modified from Wan et al.
(1996) was used. The device consisted of a central soil tray with dimensions of 0.60 x
0.30 x 0.10 m (LxWxH), with sediment collectors attached to all sides, to separately
measure sediment transport by wash and the respective directional components of splash
on a sloping surface. The original device is described in detail by Wan et al. (1996),
while the modified version is shown in Fig. 10.1.
Runoff flowed over an apron attached to the soil tray and into a small runoff guiding
trough at the downslope end of the soil tray (Fig. 10.1b). Unlike the design by Wan et al.
(1996), the present collecting system had a separate piece of flattened U-shaped metal to
CHAPTER 10 - SPLASH AND WASH UNDER NATURAL RAINFALL
180
Fig. 10.1 (a) Overall and (b) longitudinal cross sectional view of the combined splash
and runoff collection system used in the present study as modified from Wan et al.
(1996). Drawings are not to scale, for full explanation see text.
CHAPTER 10 - SPLASH AND WASH UNDER NATURAL RAINFALL
181
runoff collector lid and intercept any splashed sediment. The downslope splash board
was roofed to prevent the later from being splashed back onto the soil. The cover of the
runoff trough rested on the apron of the wash collector with four small extensions,
leaving a slot of 5 mm height for the runoff to enter. During preliminary trials, substantial
amounts of coarse aggregates were splashed or pushed through this slot by rain drops and
ended up on the runoff apron. A small portion of these aggregates was eventually washed
into the runoff collecting trough, thereby obscuring the distinction between washed and
splashed sediment. To separate most of this splash-creep sediment (Moeyersons, 1975)
from washed sediment proper, the runoff collecting trough was covered with 2 mm nylon
mesh. Any sediment remaining in the trough was added to the splash-creep sediment
fraction. This might result in an over-estimate of wash transport, because fine sediment
not initially washed ended up in the container, or an under-estimate, because some
sediment did not reach the runoff container but remained on the apron or in the runoff
guiding trough (Fig. 10.1b).
The upslope, lateral and downslope splash transport components were collected in
separate collectors equipped with splash boards. In the original design of Wan et al.
(1996) the boards extended 0.30 m above the soil surface, but the boards were extended
to 0.50 m in the present study to avoid sediment loss from devices placed at angles of up
to 40-. The splash collectors were attached to the soil tray in such a way that the boards
could always be positioned vertically and were detachable from the splash boards to
facilitate sampling (Fig. 10.1b). Cover strips were added to guide all sediment from the
boards into the collectors. Unlike the device used by Wan et al. (1996), the lateral and
upslope splash boards were not equipped with roofing, as this might interfere with rain
falling on the soil. Small perforations were made 1 cm above the bottom of the splash
collectors to allow drainage of excess rain water, while at the same time maintaining a
water layer preventing water and sediment from splashing back onto the soil or moving
in suspension.
10.2.3. Experimental procedures
Two series of experiments were conducted with four devices each; the first series
was carried out between 19 February and 20 March 1999 and covered 17 storms, whereas
the second series was conducted from 23 March to 9 April 1999 during 15 storms. Before
each series of measurements, a 4 cm thick layer of coarse (8-30 mm) gravel was placed
on the bottom of each soil tray to facilitate drainage through holes at the downslope end
of the soil tray (Fig. 10.1b). On top of the gravel, a wet piece of cloth was placed and the
tray was filled with soil to the apron of the wash collector and up to 2-3 mm below the
rims of the splash collectors. Used soil material was not dried or sieved (although large
litter was excluded) and had an estimated initial moisture content of ca. 55% in both
measuring series, remaining well-aggregated. The soil was not artificially wetted prior to
the experiments, but after the first substantial storm some soil material was added to re-
attain the original level, after which the measurements were started.
For the first series of measurements, the devices were placed at angles of 0-, 5-, 15-
and 40-. In the second series, all devices were placed at an angle of 5-; the soil of two
devices was partially covered with mulch, while a third was tilled after each event and
the fourth served as control. Actual mulch cover, provided by partly decomposed maize
stalks and leaves, was determined by digital analysis of overhead photographs at 35%
and 55%, respectively (both accurate within 3%). Tillage was done after each storm
CHAPTER 10 - SPLASH AND WASH UNDER NATURAL RAINFALL
182
using a small rake with nails protruding 3 cm, breaking up any crusts that might have
formed.
Sediment transport components were sampled after each storm. First, the volume of
water in the runoff container was measured to the nearest millilitre, after which the water
and sediment were transferred to a beaker. Some coagulating agent (Al
2
(SO
4
)
3
(aq)
) was
added and after the sediment had settled the supernatant water was decanted. Sediment
adhering to the splash boards or to the cover of the runoff guide (Fig. 10.1b) was
transferred into the corresponding collectors, using a small paintbrush (when dry) or a
wash bottle (when wet). Next, the collectors were taken out and sediment was washed
into crucibles, as was splash-creep sediment on the apron and in the runoff trough. The
suspended wash sediment and the splashed sediment were dried to constant mass at
80-C. Sediment was allowed to cool and attract moisture from the air (increasing mass
by 1-4%) before mass was measured to the nearest 0.001 g.
10.2.4. Rainfall characteristics and erosivity indices
Storm rainfall depth was measured using a standard rain gauge (100 cm
2
orifice),
while rainfall intensity was measured with a custom-built tipping bucket-logger system,
calibrated and resampled into five-minute intervals following methods outlined in
Chapter 11. A number of erosivity indices proposed in literature was calculated for each
individual storm using the rainfall intensity data. The indices and corresponding
equations are listed in Table 10.1 and include storm rainfall depth (P in mm), storm
kinetic energy flux (E
K
in J m
-2
), the amount of rain falling at an intensity higher than a
threshold intensity R
0
(P
Ro
in mm), the kinetic energy corresponding to that amount of
rainfall (E
Ro
in J m
-2
), rainfall intensity raised by an (optimised) power (R
b
), the USLE
rainfall or R factor (EI
30
; Wischmeier and Smith, 1978) and the AI
m
index of Lal
(1976). The USLE rainfall factor is calculated as the product of event kinetic energy and
maximum 30-minutes rainfall, whereas the AI
m
index is the product of rainfall depth and
maximum rainfall intensity for a short interval (five minutes in this case). More complex
erosivity indices involving rain drop size and drop momentum or kinetic energy have
also been proposed on the basis of theoretical considerations (inter alia, Meyer, 1965;
Riezebos and Epema, 1985; Gilley and Finkner, 1985; Salles and Poesen, 2000), but to
calculate such indices for natural rainfall would require detailed knowledge of the
corresponding rain drop size distributions.
Rainfall kinetic energy was estimated using the general relationship proposed in
Chapter 7:
( ) [ ] R e
K
042 . 0 exp 52 . 0 1 3 . 28 [10.1]
where e
K
(in J m
-2
mm
-1
) is the kinetic energy load of rain falling at intensity R (in mm
h
-1
). Measurements of rain drop size distribution made on the spot for intensities ranging
from 0.2 to 124 mm h
-1
(N=21) using the stained filter paper technique (Hall, 1970) did
produce a relationship that was not appreciably different from Eq. [10.1] (Van Dijk,
unpublished data). Calculated kinetic energy flux was corrected for the increase of drop
fall velocity with altitude by multiplying the results of Eq. [10.1] with a factor 1.06 (cf.
Chapter 7). To calculate the EI
30
index (Wischmeier and Smith, 1978; Table 10.1), storm
kinetic energy was also estimated using the set of equations proposed by these authors. A
CHAPTER 10 - SPLASH AND WASH UNDER NATURAL RAINFALL
183
number of erosivity indices involve the use of a threshold rainfall intensity or empirical
coefficient (cf. Table 10.1). These were evaluated using two different approaches. Firstly,
coefficient values for each individual series of sediment transport component
measurements were optimised by least squares using the Levenberg-Marquardt method.
For each device, this yielded five different coefficient values (i.e. for upslope, lateral and
downslope splash, splash-creep and wash transport, respectively) and, therefore, 40
values in total (two measured periods with four devices each). The median of these 40
values was subsequently used to calculate erosivity indices again. In both modes, the
coefficient of determination associated with a one-parameter linear regression between
event sediment transport amounts and storm erosivity index was used as a measure of
performance.
10.2.5. Interpretation of rain splash transport components
Upslope, lateral and downslope splash and splash-creep transport rates (all in g m
-1
)
across the boundaries of the soil tray were calculated by dividing the amount of splashed
sediment by the (projected) length of that particular side. From these, gross downslope
splash transport (q
g
) was calculated as downslope splash plus splash-creep, whereas
Description Equation
storm rainfall depth (mm) P
storm kinetic energy (J m
-2
) ( )
n
i
i K K
t R e E
1
Amount of rainfall falling at higher
than threshold intensity (mm)
( )
>
0
0 1
R R
i
i R
t R P
Kinetic energy of rainfall falling at higher
than threshold intensity (in J m
-2
)
( )
>
0
0 1
R R
i
i K R
t R e E
power function of rainfall intensity ( )
n
i
i
b b
t R R
1
EI
30
index (Wischmeier and Smith, 1978)
30 30
R E EI
K
AI
m
index (Lal, 1976)
5
PR AI
m
Table 10.1. Rainfall erosivity indices as used in the present study and their method of
calculation. P (in mm) denotes rainfall depth, R (in mm h
-1
) rainfall intensity, t (in h) time
interval length, E
K
(in J m
-2
) kinetic energy flux, e
K
(in J m
-2
mm
-1
) kinetic energy load, R
0
(in
mm h
-1
) threshold rainfall intensity, R
30
and R
5
(in mm h
-1
) maximum rainfall intensity for 30-
and 5-minute periods, respectively, and b an empirical coefficient.
CHAPTER 10 - SPLASH AND WASH UNDER NATURAL RAINFALL
184
upslope transport was subtracted from these to derive net downslope transport (q
n
).
Dividing transported amounts by rainfall erosivity yielded estimates for lateral and gross
and net downslope splash transportability (T
lat
, T
g
and T
n
, respectively, with units
depending on the erosivity index used).
In Chapter 8 an expression was derived relating splash transport (q in g m
-1
) across a
boundary to soil detachment on an area basis ( in g m
-2
). If the splashed particles are
exponentially re-distributed around their source, this relation is given by:
q [10.2]
where (in m) is the weighed average splash length. The (projected) splash distance
changes on a slope because of changes to the splash process and the effect of the slope
itself on the trajectories of splash (Ghadiri and Payne, 1986, 1988; Chapter 8). The
current experimental design unfortunately did not produce measurements of average
splash distance. Field measurements yielded values of 1 cm for all-sided splash on
sub-horizontal (2-3-) bench terrace beds and 12 cm for downslope splash on 30-40-
terrace risers, respectively (Chapter 9). If a change in slope only affects the component of
splash length in the direction of the slope gradient, and not that parallel to the slope, then
lateral splash length should equal for all slopes. Tentatively making this assumption
and using a value of =1 cm, lateral transportability (T
lat
) values were used to obtain
estimates of soil detachability (D) with Eq. [10.2].
It was demonstrated in Chapter 8 that splash transport measured from confined areas
(such as the soil trays) represents an under-estimate, increasing with the ratio of average
splash length over tray size. Correction factors listed in Chapter 8 for current
combinations of tray size and expected splash length suggests under-estimations to be
small overall, except perhaps for downslope splash on the steepest slope, which may
have been under-estimated by up to 10%.
10.2.6. Interpretation of runoff and wash transport
Measured storm runoff volumes were divided by the projected area of soil to obtain
total event runoff depth Q
tot
(in mm). The spatially variable infiltration (SVI) model of
Yu et al. (1997c) was used to estimate soil infiltration characteristics from these event
runoff amounts. The model assumes an exponential spatial distribution of infiltration
rates and successfully simulated runoff from (sections of) bench terraces in the study area
(Chapter 12). After an initial addition infiltration (F
0
in mm) has taken place, runoff
equals rainfall excess and can be expressed as a one-parameter function of rainfall
intensity:
1
]
1
,
_
m
m
I
R
I R Q exp 1 [10.3]
where Q (in mm h
-1
) is instantaneous runoff rate and I
m
(in mm h
-1
) average maximum
infiltration rate when the entire area generates runoff (Yu et al., 1997c). For each of the
eight data-sets a single value of I
m
was used with resampled rainfall intensity data to
CHAPTER 10 - SPLASH AND WASH UNDER NATURAL RAINFALL
185
model runoff depth for all storms. Its value was optimised by least squares using the
Levenberg-Marquardt method. Model performance for event runoff depths was
expressed by model efficiency (ME) sensu Nash and Sutcliffe (1970).
A simple conceptual model of wash transport was based on the recognition that the
maximum possible concentration of detached particles in rain water, directly after
impact, is given by the ratio of total detachment () for the event to event rainfall (P). If
it is assumed that a constant washed fraction j of the detached sediment settles slowly
enough to be transported across the downslope end of the soil area before settling, then
event soil loss by wash (M
wash
in g m
-2
) may be approximated by:
tot tot tot wash wash
Q
P
DE
j Q
P
j Q c M
[10.4]
where c
wash
(in kg m
-3
) is the concentration of washed sediment in runoff. On a soil area
of limited extent, such as used in the present experiments, Eq. [10.4] introduces an error
because material that is splashed out from the soil area is not available anymore for wash
transport, but this error is demonstrably very small in the present case. Presumably, j will
be dependent on the size distribution of the detached particles, but may also vary as a
function of slope gradient and roughness (Moss, 1988; Kinnell, 1990; Proffitt and Rose,
1991; Heilig et al., 2001). It is acknowledged that the assumption of j being constant
regardless of rainfall pattern and intensity probably represents a simplification of reality.
The performance of the combined runoff and wash transport model (Eqs. [10.3] and
[10.4]) was tested using observed sediment concentrations. Detachability derived from
lateral splash transport were used, together with j values optimised using the Levenberg-
Marquardt method. Predicted sediment concentrations were multiplied with modelled
event runoff depths to yield wash transport amounts that were compared with those
observed. Model performance was investigated by calculating Nash-Sutcliffe model
efficiency (ME), both in terms of sediment concentration and wash transport amounts for
events. In addition, for comparison calculations were made using a single optimised
overall sediment concentration for each device.
10.3. Results
10.3.1. Rainfall and performance of erosivity parameters
Rainfall patterns during the two measuring periods are shown in Fig. 10.2, while
rainfall characteristics are listed in Table 10.2. The first series of measurements (19
February - 20 March 1999) counted 17 storms of 1-39 mm, while the second series of
measurements (23 March - 9 April 1999) included 15 storms of 4-57 mm.
The average coefficients of determination (r
2
) for the one-parameter linear
regression equations between event sediment transport components and storm erosivity
indices are listed in Table 10.3. The best result overall was obtained using the kinetic
energy of rainfall above an optimised threshold intensity (r
2
=0.58), giving the best
predictions for 22 of the 40 data sets. Optimised threshold values for individual data sets
were 2-24 mm h
-1
, with the most extreme values still resulting in rather low r
2
values and
being associated with low overall transport amounts (upslope splash on a 40- slope and
of splash from the mulched soil). Applying the median (and average) optimised threshold
CHAPTER 10 - SPLASH AND WASH UNDER NATURAL RAINFALL
186
0
10
20
30
40
50
60
20-Feb 28-Feb 8-Mar 16-Mar 24-Mar 1-Apr 9-Apr
S
t
o
r
m
d
e
p
t
h
(
m
m
)
Series I Series II
Fig. 10.2. Patterns of daily rainfall during the two experimental periods in 1999.
Experiment series I II
Period 19 Feb - 20 Mar 23 Mar - 9 April
Number of days 28 16
Rainfall characteristics
Number of storms 17 15
Total rainfall (mm) 322 366
Average storm depth (mm) 18.9 22.9
Max. storm depth (mm) 38.8 57.0
Maximum R
5
(mm h
-1
) 81 147
Maximum R
30
(mm h
-1
) 51 101
Average rainfall intensity (mm h
-1
) 3.5 4.5
All rainfall
Total kinetic energy, E
K
(J m
-2
) 6234 8199
Average energy content, e
K
(J m
-2
mm
-1
) 19.4 22.4
Only rainfall at intensities >20 mm h
-1
E
20
(J m
-2
, % of total) 3324 (53%) 6090 (74%)
Rainfall (mm, % of total) 134 (41%) 226 (62%)
Average e
20
(J m
-2
mm
-1
) 24.8 26.9
Table 10.2. Characteristics of rainfall during the two periods experiments were carried out at
the study site in West Java, Indonesia.
CHAPTER 10 - SPLASH AND WASH UNDER NATURAL RAINFALL
187
intensity value of 20 mm h
-1
throughout did not decrease the average coefficient of
determination much (r
2
=0.57), whereas the number of best predictions when using a
single coefficient value for each individual erosivity index increased to 26 out of 40
(right-hand column in Table 10.3). The second best performance was obtained by using
the amount of rainfall at intensities higher than a threshold value (r
2
=0.55, best
performance for 8 out of 40 data sets). Median, average and range of these optimised
threshold intensities were very similar to those found for kinetic energy, but correlation
was reduced (r
2
=0.49) when the median value of 20 mm h
-1
was used throughout. A
similar degree of association (r
2
=0.54) was found for a power function of (five-minute
averages of) rainfall intensity (R
b
), but the optimised power values varied widely (a=1.0-
3.0). Using the average and median value (1.70.4) throughout resulted in an average
correlation that was somewhat less (r
2
=0.50). Total storm kinetic energy (E
K
) as an
indicator of splash transport performed less well (r
2
=0.44, best performance for 7 out of
40 cases), as did the AI
m
index of Lal (1976) (r
2
=0.41), even though it was developed
specifically for tropical, high rainfall intensity environments. The USLE rainfall factor
(EI
30
) and storm rainfall (P) per se performed poorest under the studied conditions
(r
2
=0.36 and r
2
=0.35, respectively).
10.3.2. Splash transport
Cumulative amounts of sediment transport components for individual devices and
lateral, gross and net downslope transportability values (T
20,lat
, T
20,g
and T
20,n
,
respectively, in g m
-1
kJ
-1
) relating to the best performing erosivity index E
20
(i.e. E
Ro
with R
0
=20 mm h
-1
), are summarised in Table 10.4.
Despite the lack of replication, trends in splash transport with slope gradient
conformed with expectations: both splash-creep and downslope transport increased with
slope, whereas upslope transport decreased. Lateral splash did not show any obvious
trend. Simultaneously with the increase in downslope splash transport, upslope transport
decreased very rapidly with slope, q